He had come to watch, meaning to scold the suitors for betting their lives. Then he saw her run, and he entered. Venus told him what to do with the apples.

The trumpet sounded, and the crowd shouted his name. Atalanta passed him, slowed to look at his face, and passed him again. With the goal still far off and his breath running dry, he threw the first apple. She turned out of her course to pick up “the rolling gold,” and he took the lead. She caught him. He threw the second, and she caught him again. Near the end, he threw the last one wide, off to the side. She hesitated, then went after it. Venus made the gold heavy in her hands, and Atalanta lost for the first time.

Her record had been right about every race it had seen. A line drawn through it would have said the next suitor would die too. None of those races had apples in them.

In 2004, a line drawn through real races ran all the way to a finish in less than zero seconds. It started sensibly. Four researchers published a short paper in Nature. They drew a straight line through every women’s Olympic 100-meter winning time since 1928, and another through the men’s since 1900. The women’s line was falling faster. The two lines crossed in 2156, when a woman would win in 8.079 seconds and the fastest man would finish in 8.098.

A reader named Kenneth Rice wrote back to the journal. The authors, he noted, had left out a later race. “A far more interesting race should occur in about 2636,” he wrote, “when times of less than zero seconds will be recorded.” Picture that race. The stadium goes still. The starting gun fires, and the winner is already through the tape, catching her breath before the smoke clears. The line that made her was right about every Olympics it had seen. It had never seen a winner much under ten seconds.

At your next checkup, the cuff squeezes your arm, and the nurse says “fine,” the same word as last year and the year before. Walking to the car, you draw the line forward: ten more years like the last ten. At dinner, the person who counts on you asks how it went, and you say “fine” too. In the median country, people spend about 9.1 years of life in poor health. A line through your good years cannot say when those begin. So the follow-up reading the nurse mentioned never gets booked, and the numbers behind “fine” never get read. When someone finally checks, the cuff has been reading high for two years.

Whole countries keep records like that. Part one of this series gave every country a birthday cake with a candle for each year of life, and the gap was the candles pulled for years spent unwell. Counted every year from 2000 to 2021, the pile of pulled candles moves. The study this series rebuilds, by Garmany and Terzic (2025), carries that movement to 2100 and says the gap widens 22 percent. Three simpler lines through the same years land anywhere from 9 to 37 percent. One of them has a whole region pulling fewer than zero candles by the end of the century, a race won before the gun. Which of those lines has earned any trust?

The Pile Moves

Africa pulls the fewest candles of any region, and it is adding them the fastest.

That comes from one straight line per country. For each of 185 countries, I drew a line through its gap from 2000 to 2021. I kept its slope: how many years of gap it adds each year. A slope works like the trend on a bathroom scale. Today’s number says where a person stands. The trend says which way they are heading, and how fast. Each dot here is a whole country, not one person. Figure 5a sorts those slopes by region.

Figure 5a. Per-country trend in the gap, 2000 to 2021, by region, from the flattest median slope to the steepest. Each dot is one country. Drawn by the author from the GapYears repository, CC BY 4.0.

The bar across each box marks the typical country, and the box holds the middle half of the region’s countries. Above the dashed zero mark, the gap is widening.

Africa’s median slope is about 0.05 years of gap per year, one more candle every twenty years. South-East Asia is close behind, at 0.049. Eastern Mediterranean and Europe come next, at about 0.04. The Americas and Western Pacific are the flattest, below 0.03. The paper ranked the same two regions on top, Africa and then South-East Asia, with steeper rates of 0.07 and 0.06 over 2000 to 2019.

This arithmetic works on a patient portal too. Two years of readings are enough: blood pressure, weight, resting pulse. The cue is the moment the portal shows this year’s number beside last year’s. Subtract, divide by the years between, and ask which way it moved. A blood pressure creeping up two points a year is twenty points in a decade. Left alone, that slope keeps adding candles.

A slope says how fast the pile grows. It cannot say what fills the pile.

What Fills the Pile

Of 22 kinds of disease that could fill the pile, a model threw cancer out entirely.

Part two sorted the world’s countries by how their candles go out. It used the World Health Organization’s counts of healthy years lost to 22 kinds of disease. That sort asked which diseases set one country apart from another. A second question was left over: which diseases rise and fall with the size of the gap itself. The paper used a regression, a model that weighs several causes at once against one outcome, as a doctor weighs age, weight and smoking against one reading. That regression lumped all chronic disease into one number. With the 22 categories in hand, they can compete in one model.

An ordinary regression with 22 categories and 185 countries gives every category some weight, noise included. A LASSO regression adds a penalty. LASSO stands for “least absolute shrinkage and selection operator,” and it works like packing a carry-on when every item has a fee. An item comes along only if it is worth more than its fee. Two jackets that do the same job rarely both make it in. One gets packed, and the other stays home, though it would have kept someone just as warm. Unlike an airline, the LASSO charges every category at the same rate.

That penalty is set by cross-validation, which works like a practice exam held back from studying: the model never sees the questions it is graded on. The countries are split into ten batches. The model is fit on nine and predicts the tenth, rotating through all ten and scoring the misses. I report the steepest penalty whose misses stay within ordinary noise of the best. Twelve categories survived it. At the lighter setting with the smallest misses, nineteen did.

The raw weights mislead, because countries differ on some diseases far more than on others. A fair comparison asks how far the gap moves when a country carries one typical step more of a disease than the average country. That step is one standard deviation, the usual distance between a country and the average. Heights spread like that too: in a room of adults, a typical person stands a few inches off the average height.

On that scale, mental and substance use disorders lead, at about 0.7 years of gap, nearly three-quarters of a candle. Musculoskeletal disease, the bad backs and sore joints, comes next, at about 0.3. At the lighter setting, the two tie at about two-thirds of a year each. The troubled mind and the aching body lead together, the same two that take the most healthy years in 149 of 185 countries. Infectious and parasitic disease points the other way, at about a quarter of a year less gap. One likely reason: among countries with equal chronic burden, the one with more infection tends to have shorter lives, and shorter lives leave fewer years to spend unwell.

Cancer fell out because it travels with others. Across countries, cancer burden rises and falls almost in step with heart disease, brain and nerve disease, and bad backs. The correlations run 0.84 to 0.88, where 1 would mean perfect lockstep. Cancer is the second jacket. To check, I also ran an elastic net, a cousin of the LASSO that is more willing to pack both jackets and split the weight between them. It dropped cancer too. Neither model says cancer costs no healthy years.

The gap is partly built from these same counts. WHO computes healthy life expectancy by subtracting, age by age, the share of years lived with disability. So an ordinary regression on all 22 categories explains about 83 percent of the gap’s spread across countries. “Explains” here means “accounts for,” not “causes.”

The everyday version shows up at the end of a checkup, when the doctor asks, “Anything else?” Low mood and a sore back are easy answers to leave unsaid. In most countries, they are the two that cost the most healthy years. Saying one out loud takes ten seconds. Left unsaid, those days go into the pile. So do the days of a parent who has stopped mentioning a knee.

The paper carries this pile to 2100 as one number. To get there, it borrows a forecast.

A Borrowed Forecast

The United Nations forecasts how long people in every country will live, year by year, to 2100.

The paper’s route to 2100 runs through that forecast. For each of its 183 countries, it fit the gap against life expectancy over two decades. Then it fed in the UN’s forecast for every year to 2100 and read off the gap. A GPS arrival time works like that. It does not watch the car alone. It borrows a forecast of the traffic ahead, and when that forecast is wrong, the car idles in red brake lights past the promised time. Unlike a GPS, which rechecks every minute, this forecast was fixed in 2024.

I rebuilt that route with the UN’s 2024 forecast, the edition the paper used, fitting 185 countries on 2000 to 2019. All but one country’s line sloped upward, at about 0.16 years of gap for every added year of life. Carried to 2100, the median country’s gap grows 20.2 percent, from 9.5 years to 11.4. The paper’s 22 percent came back within two points, without its code. Its range across countries in 2100, 7.3 to 18.8 years, came back as 7.4 to 17.6. Figure 5b follows each region along that route.

Figure 5b. Regional gap projected to 2100 through UN life-expectancy forecasts, the author’s rebuild of the paper’s route. Lines are regional means. Bands are 95 percent confidence intervals for each regional mean, the same idea as a poll’s margin of error, built from the spread among its countries; they leave out the forecast’s own uncertainty. Drawn by the author from the GapYears repository, CC BY 4.0.

Every region widens. Africa stays narrowest, at about 9.8 years in 2100, while the Americas, Eastern Mediterranean, and Europe reach 12.2 to 12.6.

That route assumes the link between longer lives and more unwell years holds for eighty more years. The field does not know yet whether it will. If medicine stretches healthy years as fast as it stretches life, the link bends.

Borrowed forecasts run household money too. A retirement calculator borrows a guess about returns, and a pension letter a guess about lifespan. The cue is any screen that hands over one date or one total. The question to ask is “what forecast is this leaning on?” This piece is worth sending to whoever in the family keeps the retirement spreadsheet. If a plan’s lifespan guess runs five years short, the money runs out five years early.

Leave the borrowed forecast out, and the gap has only its own past to go on.

Before the Gun

Given nothing but its own past, the gap in the Americas falls below zero by 2100.

The plainest way to follow a past forward is extrapolation: a straight line through each region’s average gap, carried past the last year of data. Figure 5c draws it.

Figure 5c. Straight-line extrapolation of each region’s average gap to 2100. The dotted mark at 2021 is the last year of data. Drawn by the author from the GapYears repository, CC BY 4.0.

Africa’s line starts 2021 at the bottom and climbs steeply enough to pass three of the other five regions by 2100. On the paper’s route, Africa ends the century narrowest. On a straight line, it ends near the top. A straight line cannot see a trend bend or level off. It is the line through Atalanta’s record, sure of the next race because it has never seen an apple.

So I fit two more. ARIMA, short for “autoregressive integrated moving average,” forecasts each year from the years just before it and from its own recent misses. A runner paces a lap like that, going by the last few laps and how far off each one was. Exponential smoothing is a runner who averages every lap so far, with the newest counted most. Figure 5d runs all three to 2100 side by side.

Figure 5d. Three projection methods compared by region: straight line, ARIMA, and exponential smoothing. The black line is the observed gap. Each panel has its own vertical scale. Drawn by the author from the GapYears repository, CC BY 4.0.

They do not agree. Averaged across the six regions, the straight line says the gap grows 37.4 percent by 2100. Exponential smoothing says 11.3, and ARIMA says 9.0. ARIMA’s 9.0 carries a forecast for the Americas that falls to minus 0.78 years: fewer than zero candles pulled, the race won before the gun. No gap can be negative. That number is what a line does when dragged seventy-nine years past twenty-two years of data. Leave ARIMA out, and the two lines that stay above zero run from 11.3 to 37.4. These are averages of six regional averages, not the median country. So they use a slightly different ruler from the paper’s 22 and the rebuild’s 20.

The spread is the finding. Twenty-two years of candles, followed through time alone, support anything in that range, so no single line has earned more trust than the others. It is not even across regions. In Africa, Europe, and Eastern Mediterranean, the three lines end about four years apart. In South-East Asia and Western Pacific, they end within half a year of each other.

Three weather apps for one Saturday work like this. When all three say 10 percent rain, plan the picnic. When they say 20, 50, and 80, the spread is the forecast. The likeness ends on Sunday: by then the apps have been graded, and a forecast for 2100 waits until 2100. Before acting on one estimate, such as a due date or a contractor’s “two weeks,” ask for a second way of getting it. If two methods land far apart, the past has said all it can.

Every line here stops at the same edge, the last year of data.

What Came Back

At the edge of the data, after three parts, most of the paper came back without any of its authors’ code.

Africa has the narrowest gap. Africa’s gap is growing fastest, narrowly. Rebuilt the paper’s way, the 2100 median gap grows 20 percent against its 22. These three results came back from a rebuild started from scratch, which is some evidence they do not hang on one team’s modeling choices.

Two things came back different. The disease-burden sort in part two found two clusters where the paper found three. And a line through time alone lands anywhere from 11 to 37 percent, so the paper’s 22 depends on its route through life expectancy. Two steps were not rebuilt the paper’s way. Its regression used income and chronic disease burden, where the rebuild used health spending. And part one’s spatial model needed a map of which countries count as neighbors, and the paper never describes its map.

I keep a working list of twelve questions that outlast any single project. One asks what it means when a body’s biological age runs ahead of its calendar age. Would pulling it back buy healthy years, or only move the decline later? This series asked that question of whole countries, and the answer depends on which line gets believed.

That choice waits at the clinic too. A run of “fine” readings is one line, and carried far enough past its data, a line can promise a finish in less than zero seconds. A follow-up test is a second way of looking, and the place where two lines would disagree is where the next reading belongs. On the clinic table, the cuff waits for the next number, the one no line has seen.

A Closing Invitation. Atalanta’s perfect record and the sprint line that ran past zero stood for one thing: a line is true to every point it has seen and blind past its last one. When lines disagree, the spread is the answer.

1. The next time you hear yourself say “I always get sick in March” or “my back always hurts in winter,” notice the years of evidence behind it. How many points is that line standing on, and what has it talked you out of?
2. At your next checkup this month, when the nurse says “fine,” ask for last year’s number and the year before, on paper or in the portal. Is the line flat, or has it started to bend? Would a partner or a parent want to look at it with you?
3. Right now, for the hike or picnic you planned this weekend, open a second weather app beside the first. How far apart do the two chances of rain land, and which would you plan around?

Somewhere ahead of every line is a race with apples in it. When a forecast puts you through the tape in less than zero seconds, look back at the start. The gun has not fired yet.

Where This Came From

This piece closes a replication of Garmany and Terzic (2025), rebuilt in R from public data without the authors’ code. The race began as a joke. Kenneth Rice’s reply to the 2004 sprint forecast ran the authors’ own line to its absurd end, and that one sentence turned into the test this series needed: run each forecast until it breaks, and see where. The LASSO (Tibshirani, 1996) and elastic net (Zou & Hastie, 2005) were added because the paper used one number for all chronic disease. The penalty reported is lambda.1se; the lighter setting is lambda.min. The time-series fits use the forecast package (Hyndman & Khandakar, 2008). Code and tables are at github.com/gauranii/GapYears, and the project page sums up the rebuild in one place.

Intellectual Honesty Note. A line through Atalanta’s record is this piece’s reading, not Ovid’s. The race won before the gun is Rice’s joke, extended into a scene. The candle pile, the bathroom scale, the carry-on, the practice exam, the GPS, the runner pacing laps, and the weather apps are this piece’s devices. The negative Americas gap is a real output of the ARIMA fit, with no meaning for 2100. The per-standard-deviation effects, the correlations with cancer, and the 83 percent fit are the author’s own calculations from the repository’s data.

References

Garmany, A., & Terzic, A. (2025). Healthspan-lifespan gap differs in magnitude and disease contribution across world regions. Communications Medicine, 5, 381.

Gauran, I. I. (2026). GapYears (Version 1.0.0) [Computer software]. GitHub. https://github.com/gauranii/GapYears

Hyndman, R. J., & Khandakar, Y. (2008). Automatic Time Series Forecasting: The forecast Package for R. Journal of Statistical Software, 27(3), 1-22.

Ovid. (1922). Metamorphoses (B. More, Trans.). Cornhill Publishing.

Rice, K. (2004). Sprint research runs into a credibility gap. Nature, 432, 147.

Sullivan, D. F. (1971). A Single Index of Mortality and Morbidity. HSMHA Health Reports, 86(4), 347-354.

Tatem, A. J., Guerra, C. A., Atkinson, P. M., & Hay, S. I. (2004). Momentous sprint at the 2156 Olympics? Nature, 431, 525-526.

Tibshirani, R. (1996). Regression Shrinkage and Selection via the Lasso. Journal of the Royal Statistical Society: Series B, 58(1), 267-288.

United Nations, Department of Economic and Social Affairs, Population Division. (2024). World Population Prospects 2024 [Data set]. Retrieved September 25, 2026, from https://population.un.org/wpp/

World Health Organization. (2024). Global Health Estimates 2021: Disease burden by cause, age, sex, by country and by region, 2000-2021 [Data set]. Retrieved August 28, 2026, from https://www.who.int/data/gho/data/themes/mortality-and-global-health-estimates

Zou, H., & Hastie, T. (2005). Regularization and Variable Selection via the Elastic Net. Journal of the Royal Statistical Society: Series B, 67(2), 301-320.

Meleager grew up to lead the hunt for a great boar. When the beast was dead, he gave its skin to Atalanta, the huntress who had drawn first blood. Althaea’s brothers took the prize from her, and Meleager killed them for it.

Althaea brought out the brand. Ovid has her hold it over the flames four times and pull back four times. Then she turned her face away and threw it in. The wood gave “a strange groan.” Far off, in his hour of victory, Meleager felt “the flame of burning wood scorching with secret fire his forfeit life.” No blade touched him. As the log burned down, so did he.

His life ran on one stick of wood in three stretches. While it lay hidden, he was well. While it burned, he was alive and in pain. When it was ash, he was gone. The middle stretch is the gap from part one of this series: the years a person lives, but not well. That piece drew each country’s life as a birthday cake with its unhealthy years pulled out, one candle for each year. It counted the pulled candles. It never asked what lit the log.

The ancient poets did not agree on what killed Meleager. In the second century, a travel writer named Pausanias stood before a painting of Meleager at Delphi and listed their answers. Homer blamed a Fury who heard his mother’s curse. Two other poems blamed the god Apollo. A playwright blamed the brand. Pausanias wrote all three down and did not pick one. The World Health Organization (WHO) cannot leave the question open. It gives every lost healthy year a cause.

At your next checkup, the doctor reaches the last question: “Anything else?” Your lower back has been stiff on cold mornings since spring, and some weeks getting out of bed feels heavy. You say no. Neither one feels like an illness. WHO counts both. A year of mild back pain costs about a fiftieth of a healthy year, and a year of severe depression about two-thirds of one. In 149 of 185 countries, aching backs and joints, together with a troubled mind, take more healthy years than any other cause. Like the brand in Althaea’s room, they burn out of sight.

In Australia, those two take nearly half of the years people spend unwell, so the open question is whether every country is on one slope, headed that way. In Lesotho, infection alone takes a quarter, ahead of both. Sort the countries by what costs them, and each can find others to learn from. That sorting is called clustering.

A published study sorted them into three groups. A rerun of its sort, with a different test choosing the count, found two, then checked whether any groups were there at all. Between 2000 and 2019, twenty-two countries changed groups. Which way did they go?

Where the Candles Went

A year of mild hearing loss costs a hundredth of a healthy year. WHO counts illness in slivers like that. Its unit is the year lived with disability. Each year with a condition counts as a sliver of a lost healthy year, bigger when the condition takes more out of a day. This piece counts only those unwell years, not deaths.

The slivers fall into 22 kinds of illness, from heart disease to problems at birth. Each is counted per 1,000 people in 2019, the last year before COVID-19.

The 22 counts do not share a scale. Aching backs and joints swing widely between countries: the typical country sits about seven and a half healthy years per 1,000 people from the average. Hard births swing by about a seventh of one. Left raw, the big swings would drown the small ones in any sort.

So each count is restated as how far a country sits from the average, in units of that illness’s typical spread, its standard deviation. This works like the growth chart at a child’s checkup. There, the 90th percentile for height and the 90th for weight read alike, though one is in centimeters and the other in kilograms. A score of +2 on hard births and +2 on bad backs both mean “unusually high, for that illness.” The analogy stops at the comparison group. A growth chart compares a child with children of one age. These scores compare each country with all 185, so a country’s score can move when the world around it does.

At a child’s next checkup, when the nurse reads out a percentile, the question to ask is “compared with whom?”

After the restating, each country has 22 scores, which make it a point in a room of 22 dimensions. No one can picture that room.

Beads in a Dark Room

A lamp in a dark room throws the shadows of 185 hanging beads onto a wall. Each bead is a country, and its place in the room is set by its 22 scores. As the beads turn together on their threads, the shadows on the wall spread out and bunch up. Principal component analysis finds the turn that spreads them widest. That widest direction is the first principal component. The widest direction at right angles to it is the second.

Here, the first direction holds 54 percent of the spread among countries. The second holds 7.

Figure 3. Each country’s 2019 illness profile on the first two principal components, colored by WHO region. Follow the regions from left to right and look for a gap between them. Each ellipse holds about 95 percent of its region’s countries. Drawn by the author from WHO Global Health Estimates, CC BY 4.0.

In Figure 3, Europe lands at the far left, Africa at the far right, and the other regions overlap in between. The shadows smear from one end to the other without breaking into clumps.

The left-to-right axis is built from all 22 illnesses at once. Infections, hunger, problems at birth, and birth defects pull right. Aching backs pull hardest to the left, with other long-term illnesses close behind. All but two of the 22 pull by a similar amount, so no single disease drives the axis. It measures a whole profile tipping, from infections and hard births toward the slow wear of a long life. In 1971, the epidemiologist Abdel Omran named that tipping the epidemiologic transition. Part of it is age itself: a country whose children survive grows old enough for arthritis.

In one family, the tipping can span a century. A great-grandparent may have lost a sibling to a fever, while a parent manages blood pressure or a stiff knee. At the next family dinner, the question for the oldest person at the table is what took their grandparents.

A smear on a wall can still be cut. The published study cut it.

Two Piles, Not Three

Three markers drop at random among the 185 countries, one for each of the study’s piles. Each country joins its nearest marker, each marker moves to the middle of its group, and the cycle repeats until no country moves. That is k-means clustering. It sorted on the first six principal components, which hold 81 percent of the spread.

K-means has to be told how many groups to make. The study chose three with the elbow method. It plots how tight the groups are against how many there are, and looks for the bend where one more stops helping. On this data, a second group removed about half the spread. A third removed about another sixth. A fourth removed only about a twentieth. The curve bends hard at two and softly at three. Three is a fair reading.

The rerun let a silhouette test choose instead. For each country, it asks whether the country sits closer to its own group than to the next one over. A score near 1 means snug at home, and near 0 means on the border. Two piles scored 0.40 on average. Three scored 0.35. Four or more scored lower still. The test chose two.

Figure 4. The two k-means clusters from 2019, on the axes of Figure 3. Look where the two colors meet. Drawn by the author from WHO Global Health Estimates, CC BY 4.0.

In Figure 4, the left cluster holds 78 countries, more than four-fifths of them in Europe and the Americas. The right cluster holds 107, three-fifths in Africa and the Eastern Mediterranean. The sort never saw a map, and it still drew one. On the right, more candles than elsewhere are blown out: by an infection carried on a breath, an empty plate, a birth that goes wrong. On the left, more are pinched out by wear: joints that ache, a mind that struggles, a heart that tires. Every country loses candles both ways, and the clusters split by which way runs above the world’s average.

But look where the two clusters meet. The points run straight through, with no empty band between the colors. Kaufman and Rousseeuw, who wrote a textbook on this kind of sorting, read an average silhouette between 0.26 and 0.50 as weak structure that could be artificial. At 0.40, this one falls inside. The fairest picture is a slope, from infection-heavy to wear-heavy, with most of the world strung along it. Two piles is the cleanest cut this data offers, and it is still a cut.

Plenty of forms ask for a cut like this. An intake form asks smoker or nonsmoker, and the true answer is two cigarettes on a Saturday night. The question is whether a person is a kind or a point on a slope.

Shake the Hat

In effect, I put the 185 countries in a hat to see whether two kinds keep coming back. I drew 185 names, putting each slip back before the next draw, so some countries came up twice and some not at all. On that new roster, I reran the whole sort and let the silhouette choose the count again. I did that two hundred times.

The study had checked its clusters another way, with a random forest: a crowd of small decision trees. Each tree asks a short run of yes-or-no questions, such as “is the cancer count above this line?”, and the crowd votes. Each tree is checked on countries it never saw. On my two clusters, the forest sorted 98.4 percent of countries correctly.

That sounds like proof. But k-means draws a clean line even through a smear, and the forest was only asked to find that line again. A high score shows the line is easy to redraw, not that the countries fell into two kinds before anyone drew it.

In the hat, two piles won 94 percent of the draws. Three, the study’s count, won twice. A bootstrap, which is what statisticians call drawing from the hat, measures whether an answer survives a shaken roster. It cannot say whether the answer is true.

I run this test on my own bad weeks now. When I am sure what caused one, I check whether that cause turns up in the other bad weeks too. On a Sunday evening, with the week’s calendar open, the question is: if I drew a different week, would I still blame the same thing?

The hat settles how many piles. It cannot say which way countries are moving along the slope.

No Country Crossed Back

Twenty-two countries changed piles after 2000, and all of them moved the same way. The rerun sorted four years within that span, each with its own restating, principal components, and k-means. The count was held at two each year, so a country could change only which side of the line it fell on.

Each year’s scores compare a country with the world that year. If every country had tipped toward chronic illness at one pace, none would have changed piles. A move means a country traveled further than the world around it.

Each mover went from the infection-heavy side to the wear-heavy side, which grew from 56 countries to 78. Ten are in the Americas, among them Colombia, Costa Rica, Mexico, and Peru. None crossed and then crossed back.

If countries were wobbling across the line by chance, some would wobble back. If each move were a coin toss, the odds that all 22 would land alike are about one in two million. WHO’s yearly figures are modeled estimates, and models like these smooth over time. That smoothing may hide some back-and-forth a raw count would show, but it cannot explain why every move points the same way.

What no one knows yet is where the slope ends. Nineteen years of data cannot show whether a country can slide back toward infection, or what the far end of wear looks like.

Most readers of this piece live in a country on the left of Figure 4, where backs, joints, and heavy months pinch out many of the candles. Those are the two causes people leave out of the clinic room. At a checkup, the doctor asks “anything else?”, and the answer is a name or a no. A no leaves the stiff back hidden in the room, costing about a fiftieth of a healthy year for each year it lasts.

Althaea pulled back from the fire four times. When the brand went in, it stayed in.

A Closing Invitation. Meleager’s log stood for one life burning along one stick. Countries burn the same way, along one slope and in one direction, and the stretch that hurts can burn for years out of sight.

1. Before you next stand up, from this chair now or from the warm bed tomorrow, lie still for a breath and take stock of your back and neck: stiff, sore, or fine? Is it the same as a month ago, or has it slid one way?
2. At the next family dinner, ask the oldest person at the table what took their grandparents, a fever or a hard birth, and what takes people now. Which way did your family slide?
3. At your next checkup, when the doctor asks “anything else?”, say the stiff back or the heavy month out loud. Which one would you have left hidden in the room?

Althaea’s log hissed as the water hit it, and it lay charred and quiet for years. When a stiff back is named in the clinic room, the brand comes out of the fire with wood still unburned.

Where This Came From

This piece reruns the disease-burden half of Garmany and Terzic (2025). Their three clusters, and the random forest that checks them, are the paper’s design. The silhouette test, the bootstrap, and the tracking of countries from 2000 to 2019 are my own, and where the two runs disagree, the disagreement is the finding. The code and full results are at github.com/gauranii/GapYears, and the project page sets the rebuild out in short.

Part three follows the gap from 2000 and projects it to 2100.

Intellectual Honesty Note. Reading the brand’s three stretches as healthy years, unwell years, and death is this piece’s own device, as are the candles and the two ways they go out, which are not a medical grouping. The country shares and the 149 of 185 count are my own calculations from WHO’s 2019 and 2021 data. The cost of a year of back pain, hearing loss, or depression is rounded from the disability weights in Salomon et al. (2015). Reading the clusters as a slope cut in two is my interpretation of the silhouette score and the figures, not a formal test.

References

Apollodorus. (1921). The Library (J. G. Frazer, Trans.). William Heinemann.

Breiman, L. (2001). Random Forests. Machine Learning, 45(1), 5-32.

Efron, B. (1979). Bootstrap Methods: Another Look at the Jackknife. The Annals of Statistics, 7(1), 1-26.

Garmany, A., & Terzic, A. (2025). Healthspan-lifespan gap differs in magnitude and disease contribution across world regions. Communications Medicine, 5, 381.

Gauran, I. I. (2026). GapYears (Version 1.0.0) [Computer software]. GitHub. https://github.com/gauranii/GapYears

Kaufman, L., & Rousseeuw, P. J. (1990). Finding Groups in Data: An Introduction to Cluster Analysis. Wiley.

MacQueen, J. (1967). Some Methods for Classification and Analysis of Multivariate Observations. In Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability (Vol. 1, pp. 281-297). University of California Press.

Omran, A. R. (1971). The Epidemiologic Transition: A Theory of the Epidemiology of Population Change. The Milbank Memorial Fund Quarterly, 49(4), 509-538.

Ovid. (1922). Metamorphoses (B. More, Trans.). Cornhill Publishing.

Pausanias. (1918). Description of Greece (W. H. S. Jones & H. A. Ormerod, Trans.). William Heinemann.

Pearson, K. (1901). On Lines and Planes of Closest Fit to Systems of Points in Space. The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science, 2(11), 559-572.

Rousseeuw, P. J. (1987). Silhouettes: A Graphical Aid to the Interpretation and Validation of Cluster Analysis. Journal of Computational and Applied Mathematics, 20, 53-65.

Salomon, J. A., Haagsma, J. A., Davis, A., de Noordhout, C. M., Polinder, S., Havelaar, A. H., Cassini, A., Devleesschauwer, B., Kretzschmar, M., Speybroeck, N., Murray, C. J. L., & Vos, T. (2015). Disability Weights for the Global Burden of Disease 2013 Study. The Lancet Global Health, 3(11), e712-e723.

World Health Organization. (2024). Global Health Estimates 2021: Disease burden by cause, age, sex, by country and by region, 2000-2021 [Data set]. Retrieved August 28, 2026, from https://www.who.int/data/gho/data/themes/mortality-and-global-health-estimates

So Tithonus goes on living, and he goes on aging. The first gray hairs ripple from his head and his chin, and she keeps away from his bed. She still feeds him ambrosia and dresses him in rich clothing. His limbs stiffen until he can no longer move or lift them. Then she lays him in a room and closes the shining doors. “There he babbles endlessly, and no more has strength at all.” Every morning the dawn goes out into the world, and behind those doors the voice keeps going.

Aphrodite tells this story on Mount Ida to a mortal lover of her own, the herdsman Anchises. Then she refuses him the very gift the dawn once asked for.

After a few hundred years, Tithonus’s cake is a wall of flame. Take out one candle for every year he could not get out of bed, and almost nothing is left.

No country is Tithonus, but every country has a cake. It carries one candle for every year a baby born there in 2021 can expect to live. One candle comes out for every year’s worth of poor health. Those are the years of the bad knee, when the stairs get taken one at a time. They are the years of the pill organizer, seven plastic lids snapped open every morning. Australia’s carries eighty-three candles. Lit all at once, they throw enough heat that everyone at the table leans back. Twelve and a half come out. Lesotho’s carries fifty-one, and almost seven come out. The longer life loses more candles.

Researchers call the pulled candles the healthspan-lifespan gap: life expectancy minus healthy life expectancy. In 2025, a paper in Communications Medicine counted them for 183 countries from 2000 to 2019, using World Health Organization and United Nations data. The global median gap was 9.1 years. The paper projects the gap to widen 22 percent by 2100. In a recount of the same public data, the gap grows as lives get longer: roughly one pulled candle for every six or seven added. Every year added to a country’s average life brings nearly two months of the bad knee and the pill organizer.

A bet on the lit candles alone can go badly wrong. In 1965, a notary in Arles named André-François Raffray, forty-seven years old, bought the apartment of a ninety-year-old woman named Jeanne Calment. The sale was a viager, a French sale for life. He would pay her 2,500 francs a month until she died, and then the apartment would be his. He paid for thirty years and died first, in December 1995. His widow kept paying. Calment lived to 122, the longest documented life on record. She spent her last twelve years in a nursing home and her last seven in a wheelchair, after breaking her femur at 114. By the end she was nearly blind and deaf. Raffray priced the lit candles. The deal never asked how many of them she would spend needing help out of bed. Of the deal, she said, “In life, one sometimes makes bad deals.”

The same bet reaches your kitchen table this month. The October statement from your retirement account comes with a calculator, and it asks for one number: the age to plan to. You type a lifespan. A baby born in the United States in 2021 can expect 76.4 years, and 63.9 of them in good health. The form has no box for the other 12.5. The trip planned for seventy, or the promise to keep a parent at home, gets funded as if every candle were one you could still blow out. The money lasts. The knee it was saved for does not.

If twelve and a half unseen years can sink a plan, does the count behind them hold up when someone outside the study counts the candles again?

Anyone at the table can recount the candles on one cake. The paper’s 183 countries are harder. Every number in it comes from public data, but the code that did the counting was never released. So I counted the candles again myself, from the methods section and the published figures alone. I also added one number the paper did not use: how much each country spends on health. Parts two and three follow the paper further: which diseases pull the candles, and how many more will come out by 2100.

Counting the Candles Again

At the end of a dinner for eight, someone runs a finger down the bill line by line. The check works because the receipt shows every line. If it showed only the total, the checker would have to guess what each person ordered. A replication without the code works like a receipt with some lines missing. Where a line is missing, the recount has to guess, and every guess makes it less of a recount.

So I gave myself one rule going in: rebuild what the public data and the methods section allow, and say so wherever they stop. The gap itself takes two numbers per country, life expectancy and healthy life expectancy, and WHO hands out both through a public data service. The paper’s second half needs disease data that lives somewhere else, and the count of which diseases pull the candles picks it up.

To compare countries, the paper used one year, its most recent, 2019. WHO now publishes both numbers through 2021, so I used 2021. WHO revises its past estimates, so matching the paper’s 2019 to the decimal was never possible anyway.

The paper found two things. Africa had the narrowest gap of the six WHO regions. Life expectancy, GDP, and noncommunicable disease burden were the most consistent predictors of gap size. For 2021 alone, 185 countries in all, here is what I found:

• Africa’s median gap, 8.22 years, is the narrowest of the six regions.
• The six regions differ by far more than chance would allow. If the regions did not really differ, a spread this wide would turn up less than once in ten trillion tries (a Kruskal-Wallis test).
• Life expectancy alone explains about four-fifths of the country-to-country spread in the gap.
• The paper’s economic number was GDP; I used health spending as a share of GDP instead. Together with life expectancy it explains 82 percent of the spread, against 81 for life expectancy alone. The third predictor, disease burden, stays out of this model. Part two uses that data to sort the countries instead.

Some numbers moved. The paper counted 183 countries in 2019. I counted 185 in 2021. Its global median gap was 9.1 years, and mine is 9.3. It puts Lesotho’s gap at 6.5 years, and mine is 6.84. The finding stays the same: Africa has the narrowest gap, and life expectancy predicts it best. Two countries have no health-spending figure, so Figure 1 uses all 185 and every model after it uses 183.

Figure 1a. Healthspan and lifespan distributions by region, 2021. Teal is lifespan; red is healthspan. Drawn by the author.

Panel (a) draws each region’s two numbers as two hills: teal for how long people live, red for how long they live in good health. A taller hill means more countries at that number of years. In every region the red hill lies to the left of the teal one, and the space between them is the pile of pulled candles. In Europe and South-East Asia each hill has two humps, most likely because each region holds two different kinds of country under one name.

Figure 1b. Healthspan-lifespan gap by country, 2021. Darker countries have a wider gap. Drawn by the author.

Panel (b) maps the gap country by country. The darkest are Australia, the United States, New Zealand, Switzerland, and France. Four of them have some of the longest lives on earth. The United States is the exception. Its people live six to seven years less than the other four, and it still loses almost as many candles as Australia.

Figure 1c. Healthspan-lifespan gap by region, 2021, from the narrowest median gap to the widest. Each dot is one country. Drawn by the author.

Panel (c) lines the regions up, from Africa’s 8.22-year median on the left to Eastern Mediterranean’s 10.22 on the right. The line inside each box marks the typical country, and the box holds the middle half of the region’s countries. AFR, AMR, EMR, EUR, SEA, and WP stand for Africa, Americas, Eastern Mediterranean, Europe, South-East Asia, and Western Pacific.

The next headline that says how long people live somewhere is a receipt with one line showing. The missing second line is how many of those years are healthy. For the United States, 63.9 belongs beside 76.4. A receipt that adds up still leaves a question, though: which countries ordered more than their table should have?

More Candles Than Expected

At a toddler’s checkup, the nurse puts a dot on a growth chart, above the curve or below it. The curve is a prediction: how tall a child of that age usually is. A dot above it is not wrong. It is more than expected, given what the chart knows.

Figure 2 is a growth chart for countries. A regression, a line fitted through all the countries, turns the two numbers from the last section into a predicted cake for every country. The prediction says how many candles the country should lose, given its life expectancy and its health spending. Each real cake then lands above its prediction or below it.

Figure 2a. The 94 countries whose gap is larger than their life expectancy and health spending predict, 2021. Drawn by the author.

Figure 2b. The 89 countries whose gap is smaller than their life expectancy and health spending predict, 2021. Drawn by the author.

Figure 2c. Regional makeup of the larger-than-predicted and smaller-than-predicted groups, 2021. Africa leads the first group, and Europe the second. Drawn by the author.

Panel (a) colors in red every country that loses more candles than predicted. Panel (b) colors in blue every country that loses fewer. Panel (c) slices each group by region.

Ninety-four countries lose more candles than predicted, and eighty-nine lose fewer. The paper and I found the same region at the top of each group. Africa makes up the biggest share of the countries that lose more: 33 percent in my count, 41 percent in the paper’s. Europe makes up the biggest share of the countries that lose fewer: 36 percent in mine, 45 percent in the paper’s.

Several wide-gap countries from the map lose more: the United States, Canada, Australia, the United Kingdom, France, Germany, Italy, and Spain. Their gap is wide, and it is also wider than their own life expectancy and spending predict. China, Japan, and Russia lose fewer than predicted.

The shares do not match exactly, because the two charts were built from different inputs. The paper’s model had a third number, disease burden, and mine does not. Adding it could move some countries from one group to the other. The paper also sorted some of its countries, not all. Its two groups held 61 and 58 countries. That left 64 of its 183 in neither. I placed every country in one group or the other.

The growth chart comparison stops here. A child grows toward the curve over the years. A country’s dot says only how its cake compares with a line drawn from two numbers. Draw the line from three, and some dots cross it.

The same question works on any number placed against a curve. At the next physical, a lab result may come back a few points over its line. The question to ask is: whose line? A growth chart shows why it matters: girls and boys get separate curves, and a dot just above one can fall below the other. Some dots, though, cross a line together with their neighbors.

What the Neighbors Share

An appraiser pricing a house on Elm Street starts with its neighbors: the houses sold on the same block. Neighbors share a school, a flood zone, and a commute, so their prices move together. Countries share things too: a climate, diseases, and trade. If that sharing shows up in the gap, some of Figure 2’s groups could be partly geography.

Every cake misses its predicted count by some amount. A plain regression treats each country’s miss as its own, unrelated to the miss of the country next door. If neighbors tend to miss in the same direction, that assumption is wrong. The paper allows for this with a spatial error model, which adjusts the regression for where each country lies. Before building one, I checked whether it was needed.

Whether neighbors’ misses resemble each other has a standard measure, Moran’s I. To judge it, I shuffled the countries’ misses across the map 999 times and recomputed it after each shuffle. Only about one shuffle in a hundred came out as extreme as the real map. Neighbors do share their misses.

The paper does not say how it decided which countries count as neighbors, so its spatial model cannot be rebuilt, only one like it. In mine, the closer two countries’ center points are, the more their misses may move together. The pull turned out to be real. Take it out, and the model matches the real cakes far worse, by more than chance could explain. The pull fades to about a third of its strength by roughly 1,200 to 1,900 km apart. Life expectancy and spending both still matter after the adjustment, with spending at about fourteen days of gap per percentage point.

Figure 2d puts panels (a) and (b) on one map. A switch flips between the plain view and the view adjusted for neighbors. Orange countries lose more candles than predicted, and blue countries lose fewer. The ringed countries change groups between the two views.

Figure 2d. Gap against prediction, 2021, from the plain regression and from the model adjusted for neighboring countries (spatial error model). Orange is larger than predicted; blue is smaller. Drawn by the author.

Eight of the 183 countries change groups. Five are in Europe: Austria, Cyprus, Czechia, Denmark, and the Netherlands. All five move from fewer candles than predicted to more. The other three, Zimbabwe, Saint Lucia, and Somalia, move the opposite way. None of the eight moved far. Before the adjustment, each one was within about a month of its prediction, a twelfth of a candle, right on the line between the groups. A spatial model pulls neighbors’ misses toward each other by design, so borderline neighbors tipping over together is the change it is built to make.

What holds is the regional pattern. After the adjustment, Africa still makes up the biggest share of the countries that lose more candles, at 31 percent. Europe still makes up the biggest share of the countries that lose fewer, also at 31 percent. The paper reports Africa’s lead holding under its own adjustment.

Before arguing with a home appraisal, look at how far it missed. A few thousand dollars under the agreed sale price, and a different set of comparison houses can lift it over the line. A third low, and no choice of neighbors will. The neighbors explain the borderline countries. They do not explain why spending’s link changed size depending on how the countries were compared.

Where the Spending Link Lives

A blood pressure cuff tightens around an arm at a pharmacy kiosk, and the screen reads 138. Set beside a neighbor’s 120, the number says one thing. Set beside the same arm’s 126 from last spring, it says something else. Only the second comparison is about this one body changing. Spending’s link to the gap needs that second comparison too.

So far every comparison has used one year of cakes. Most countries have twenty-two, from 2000 through 2021. This section puts all of them into one model, which the paper does not do.

Comparing the United States with Somalia shows that countries with longer lives lose more candles. It cannot show whether a country loses more candles in the years its own life expectancy rises. For that, the United States has to be set beside its own cake from five years ago. A fixed-effects panel model makes that comparison. It sets aside everything about a country that stays put, such as its geography, its history, and the baseline quality of its hospitals. Then it looks at change within each country alone.

Life expectancy passes this test almost unchanged. Each added year of life comes with about 0.15 years of added gap, whether countries are compared with each other or each country only with itself. That is where the one pulled candle for every six or seven comes from.

Health spending does not pass as cleanly. Across countries, each extra percentage point of GDP spent on health comes with about nineteen more days of gap, about a twentieth of a candle. Compare each cake only with its own past cakes, and that falls to about eight days. Eight days is small, but the data leave no real doubt it is above zero. Countries that spend more also differ in other lasting ways, and part of the cross-country effect came from those differences. A standard check, the Hausman test, confirms that those differences matter, so the within-country number is the one to trust.

Neither number shows that spending causes the gap. A country might spend more because its people are living longer with illness, and the arrow would then run the other way.

The regression behind Figure 2 describes the average country. Quantile regression asks whether the countries far above the line follow the same rule as the ones far below. It lines the cakes up from the fewest candles lost for their prediction to the most. Then it fits the same model at five points along the line-up, called percentiles. The lowest is the 10th, and the highest is the 90th.

Life expectancy holds steady at all five, at about 0.16 years of gap for each added year of life. Health spending does not. From the bottom of the line-up through the middle, its effect is five to nine days, too uncertain to tell apart from zero. At the 75th and 90th percentiles it turns clear. There each percentage point of spending comes with about twenty days of gap. The extra candles that come with spending show up clearly only on the cakes that lose the most, the far end of Figure 2’s red group. With 183 countries, the data cannot say for certain that the rest follow a different rule, only that they do not show this one.

The field has not settled one question beneath every cake. Healthy years are estimated from surveys and health records, and those are much fuller in Australia than in Lesotho. Whether a year counted as poor health is measured equally well in both is still open.

The count held up when counted again: in every comparison, longer lives lost more candles. A national average is the neighbor’s 120, a saver set beside strangers. The family’s own record is last spring’s 126: the age a parent’s knee gave out, the year a grandmother stopped driving. That record belongs beside the October statement’s box before the box gets its number. This is the page to send to whoever does the retirement math in the house. The box on the form holds one age, and on it every candle still burns.

A Closing Invitation. Tithonus’s cake stood for a count that includes every year lived and none of the years lived well. A long life and a well life are two counts, and a plan needs both.

1. The next headline about how long people live somewhere is the cue, and so is the last family boast about a great-aunt who made it to ninety-six. Notice whether either one gives the partner number. How many of those years were healthy, and who carried her groceries up the stairs for the rest?
2. At the next family dinner, ask the oldest person at the table when the stairs became one step at a time, or when the pill organizer first came out. With no one to ask, think of a grandparent. Was it the birthday with the big number on the cake, or years before?
3. When the October statement arrives, pencil a second age beside the one the plan uses. Make it the last age the plan needs a good knee for: a trip abroad, a garden dug by hand. Moved five years earlier, that trip is one the knee can still take. Which years is the money set aside for?

Tithonus never needed more candles. He needed someone to count the ones he could still blow out. On a real cake, the candles that matter are the ones a person can still lean over, close enough to feel the heat.

Where This Came From

The study replicated here is Garmany and Terzic (2025), in Communications Medicine. The data is WHO Global Health Observatory data, pulled on 28 August 2026. The panel model, the quantile regression, and the Moran’s I check go beyond the paper’s design, and the rest follows its methods text as closely as that text allows. The code, the data pulls, and a list of what worked and what did not are public at github.com/gauranii/GapYears. The project page says in short what rebuilt and what did not.

The oldest surviving telling of Tithonus is the Homeric Hymn to Aphrodite, where the dawn goddess has a name, Eos. Part two of this series, a count of which diseases take each country’s healthy years, sorts the countries by that count. Part three, the paper’s forecast set against three simpler ones, takes up the projection to 2100.

Intellectual Honesty Note. The birthday cake is this piece’s own device, as are the receipt, the growth chart, the appraisal, and the blood pressure cuff. The candle counts are rounded 2021 figures. The dawn going out each morning is this piece’s image; the Hymn says only that he babbles endlessly. The bad knee and the pill organizer are illustrations, not anyone’s case. Calment’s age was challenged in 2018, and most researchers still accept 122. Distance in my spatial model is measured in degrees, so its reach runs shorter east to west. The note on survey data simplifies how WHO builds its healthy-year estimates.

References

Evelyn-White, H. G. (Trans.). (1914). Hesiod, the Homeric Hymns, and Homerica. William Heinemann.

Garmany, A., & Terzic, A. (2025). Healthspan-lifespan gap differs in magnitude and disease contribution across world regions. Communications Medicine, 5, 381.

Gauran, I. I. (2026). GapYears (Version 1.0.0) [Computer software]. GitHub. https://github.com/gauranii/GapYears

Hausman, J. A. (1978). Specification Tests in Econometrics. Econometrica, 46(6), 1251-1271.

Koenker, R., & Bassett, G. (1978). Regression Quantiles. Econometrica, 46(1), 33-50.

Kruskal, W. H., & Wallis, W. A. (1952). Use of Ranks in One-Criterion Variance Analysis. Journal of the American Statistical Association, 47(260), 583-621.

Moran, P. A. P. (1950). Notes on Continuous Stochastic Phenomena. Biometrika, 37(1-2), 17-23.

Robine, J.-M., Allard, M., Herrmann, F. R., & Jeune, B. (2019). The Real Facts Supporting Jeanne Calment as the Oldest Ever Human. The Journals of Gerontology: Series A, 74(Suppl. 1), S13-S20.

Whitney, C. R. (1997, August 5). Jeanne Calment, World’s Elder, Dies at 122. The New York Times.

World Health Organization. (n.d.). Global Health Observatory data repository [Data set]. Retrieved August 28, 2026, from https://www.who.int/data/gho

Goldberger was a federal doctor, and the South offered him three suspects. The first was a germ. A commission studying the disease in South Carolina judged it most likely an infection, spreading fastest where the sewage was handled worst. The second was spoiled corn. The third was a biting insect, the favorite of an expert in London. Each suspect would have struck anyone who shared the asylum’s wards.

Figure 1. A man with pellagra, photographed by Dr. James W. Babcock, who ran South Carolina’s state asylum. The rash on the backs of the hands is what the Jackson patients had and the nurses did not. National Institutes of Health History Office, courtesy of the Waring Historical Library, Medical University of South Carolina, via Wikimedia Commons; public domain.

The asylum’s own doctor told Goldberger of one patient who fell ill after 15 years inside and another after 20. Whatever caused pellagra was working within the walls. Yet the asylum had hired 126 nurses and attendants since 1909. Of them, Goldberger wrote, “No case, so far as I was able to learn, has developed.” “If pellagra be a communicable disease,” he asked, “why should there be this exemption of the nurses and attendants?”

The difference was on the plate. The nurses ate from the same kitchen, but they took the best and the widest variety first, and they could buy more outside. The patients had no such choice. The next year, on a prison farm near Jackson, eleven healthy prisoners were promised pardons. In return, they ate a diet built on grits, cornbread and syrup for months. By November, six of them had the rash. In the spring of 1916, Goldberger and a few volunteers swallowed pills made from patients’ scabs and stool, and took injections of their blood. “We just feasted on filth,” he said. None of them caught pellagra.

The cause was in the kitchen, and the kitchen was set by the pay. Mill and sharecropper families in the South lived on cornmeal, fat pork and molasses, often bought on credit. No one isolated the missing vitamin, niacin, until 1937, eight years after Goldberger died. By the late 1940s, laws in Southern states required niacin in bread, flour and cornmeal, and pellagra had nearly vanished. The germ hunters had searched the patients’ bodies, and the answer sat on the plate across the table.

On a visit home you watch a parent carry the laundry basket down the basement stairs, one hand on the wall, nothing to hold. Whoever lives with them has stopped seeing the bare wall. More than one older person in four falls each year. The basket raises a question, and it is not whether a fall is coming. It is what comparison would show whether a rail, a class or a check would keep them on their feet. Without one, the fall arrives as the first reading anyone took, and the couch downstairs becomes the bed.

The woman this series has followed since twenty-five is sixty-six now. At fifty-five a scan read her hip at -1.8: thin, yet not thin enough for the word osteoporosis, so no one treated it. One February evening, in socks, she misses the third stair from the bottom, the one with nothing to hold, and lands on that hip. It breaks anyway. Medicare pays for the surgery the next morning. Nothing ever paid for a rail. A year after a broken hip, only 40 to 60 percent of people walk the way they did before.

The third part gave the body’s side its methods. This part asks what a house, a clinic and a bill can do for a body, the tree’s other half, and whether a method answers each question directly. Four questions hang there. Which cheap signals warn of decline before a first fall, and does finding decline sooner change anything? After a first fall, how do we break the loop of fear, moving less and weaker legs? Which changes to a home keep people on their own longest per dollar? And how do we pay for function kept rather than procedures done? Goldberger solved his case with two groups under one roof, alike in everything but the plate. Each of my four questions needs two groups like those. Who, in her story, are the nurses?

Before the First Fall

In the spring of 1927, the Mississippi River broke through its levees and flooded farms across the South. Families waited out the water in Red Cross camps, on rations heavy in cornmeal. Goldberger toured the flooded states and found pellagra spreading. On his advice, the Red Cross added brewer’s yeast, which supplies what cornmeal lacks. Within weeks, the sick were recovering and new cases stopped. The rations changed while the outbreak was still running.

A body gives warnings too, if someone takes the reading. A stopwatch and a few meters of floor can predict who will still be alive in ten years. Stephanie Studenski and colleagues pooled nine studies, about 34,000 older adults in all, whose usual walking speed had been timed. Take men of 75. About one in five of the slowest walkers was predicted to reach 85. Of the fastest, nearly nine in ten.

The walk is timed over 5 to 10 meters. Grip is measured by squeezing a handheld gauge, and a review of the measure calls it “a good simple measure of muscle strength” when the conditions are standard. Low grip predicts longer hospital stays, more trouble with daily tasks, and death. Both take a minute. Which cheap signals warn of decline in time to act, and how could they be checked as routinely as blood pressure? My questions on the tree’s other half start there (Q211, Figure 2).

A walk needs a heart, lungs, nerves, muscles, joints and balance all at once, so a slow one cannot say which of them failed. In 2009, two physical therapists called walking speed “the sixth vital sign,” a number read, like any vital sign, against a person’s own normal. Like a smoke alarm, it says something is wrong in the house, not which room is burning. That makes it a good alarm and a poor diagnosis, and an alarm is checked two ways. A risk prediction model combines several readings, each with its own weight, into one probability of falling in the next year. Gait and grip at her checkups from fifty-five on would have been its inputs. At fifty-five the checkup read her blood, and nobody timed her walk.

Discrimination asks whether the model ranks her above a neighbor who never falls. Calibration asks whether, of women it scores at 30 percent, about 30 in 100 fall. Both are checked the way the third part checked its levels, on people the model never saw. A model can rank well and still be miscalibrated for one group. The second part planted the case: in diabetes, a bone scan looks better than the bone. At any given scan reading, people with diabetes break more hips than a fracture model predicts. The scan that read -1.8 at fifty-five was reading her hip through her diabetes. In a fracture model, she is in the group for which the predicted breaks come out too few. A model wrong for her group is wrong in the direction that left her hip untreated.

The second part used a cheap signal at the end of a study, as a surrogate: a reading that substitutes for an outcome years away. A screen uses that signal at the start, and the start hides a trap. Suppose a timed walk flags her at 58 instead of 66, and nothing done in between changes the day she falls. Counted from the flag, she seems to live eight years longer with frailty, and not one of those years is new.

A fair count starts everyone at the same line, a birthday or the day a study began. Whether checking the walk routinely keeps people on their feet is a question for a trial, and one kind of trial is built for clinics. In a stepped-wedge cluster trial, clinics are assigned at random to the order in which they switch the check on. Every clinic ends up checking, and the early ones are the comparison for the late ones. Her clinic was scheduled to switch on in the spring after her fall. The clinics that switched on before hers were her nurses: a checkup like hers, with a stopwatch added.

A parent’s yearly checkup leaves a question for the drive home: was the walk timed, or only the blood read? Walking speed over a few meters is the number to ask for, and the cue is the visit itself. No one took a reading in her house, so the first one was the fall on the third stair. Would anyone count the second fall?

The Spring It Came Back

Pellagra came back in the spring, much as one fall invites the next. Someone who survived one attack often had another the next year. So when Goldberger changed the meals at two orphanages in Jackson in 1914, adding milk, eggs, beans and peas, he counted the second attack. At one, none of the 67 children past the anniversary of their last attack fell ill again. At the other, one of 105 did.

Her loop starts after the hospital. Six weeks after the surgery she is home, and the stairs frighten her, so she sleeps on the couch downstairs. Fear of falling, in the field’s words, is low confidence in avoiding a fall during the ordinary tasks of a day. It sets a reinforcing loop running. A fall leads to fear, and fear to moving less. Moving less leads to deconditioning, the strength a body loses to disuse, and weaker legs lead to the next fall. Breaking that loop after a first fall is my second question (Q212).

A count that stopped at the first attack would have missed Goldberger’s test. A loop needs a method that counts more than once, and many studies stop the clock at the first fall. A fall that happens again to the same person is a recurrent event, and every one has to be counted, over time, the way Goldberger counted the second spring. The difference shows in a review of 108 exercise trials in people over sixty living at home. Exercise cut the rate of falls by 23 percent, but the number of people who fell at all by only 15 percent. The 23 percent rests on the 59 trials that counted every fall. Fewer falls and fewer fallers are two answers, and the gap between them lies in the second and third falls, the ones a count of fallers never sees.

A multistate model follows people as they move between steady, fallen, afraid and housebound, and sometimes back, and asks which move is easiest to block. The third part’s competing risk is a small multistate model: one starting state and two exits, with death as one of them. Here the moves multiply. In the spring after her fall she comes down the stairs for the mail, in shoes this time, and falls on the second step. That second fall is her recurrent event, and the model asks whether the move from afraid to housebound made that fall more likely. Her nurses would be people sent to an exercise class after a first fall, and the model would show which move the class slowed. Few fall trials report every move between states, so which move to target first is still open.

Whoever drives a parent home after a fall can ask two questions before leaving the emergency room. Which program does the hospital send people to, and who will call in two weeks? The two weeks after discharge is the window, and the discharge sheet is the cue. Without those answers, the couch downstairs is already the next state. A rail would have changed the stair, and the stair is the next question.

One Rail, One Dollar

A rail for her stairs takes a handyman an afternoon, two brackets per stretch of wall and a length of wood. In Chicago in 1995, the surroundings’ side of the healthspan question was a machine in a window, an air conditioner lowering what the heat asked of a body. At sixty-six it is a rail on a stair.

The second part put a grab bar from a program called CAPABLE beside the money it saved Medicare. CAPABLE is a complex intervention, its handyman, nurse and therapist sent as one bundle. Its result cannot say what the rail alone was worth. My third question asks which changes to a home keep people independent longest per dollar (Q221, Figure 3).

Pulling CAPABLE’s three parts apart is a design question, and the design is older than the program.

Four arms: the rail alone, the therapist alone, both, neither. The arm with the rail alone would be her nurses, the one result CAPABLE could never show, and the arm with neither is her, on the February evening. Then the “per dollar” needs its fraction. A cost-effectiveness analysis divides the extra cost of a change by the extra benefit it buys. The benefit is counted in health rather than money, so options can be set side by side as cost per outcome. For a rail, the top of the fraction is the afternoon’s work. The bottom is the second part’s disability-free life years, counted as years lived without needing help to bathe, dress or climb the stairs. Count the bottom as years alive instead, and the rail looks almost worthless, because most falls do not kill.

The steps in a parent’s home with nothing to hold are the count to take on the next visit. A rail’s price against a fall’s is the fraction in one line. The fraction still leaves out who pays. The rail is bought from one budget, and the surgery it saves comes off another. Whose budget pays is a question about the bill, not the house.

Paid for the Plate

A family that paid for enriched flour in 1945 got the vitamin without deciding to. The flour laws decided for them, and better wages put more on the plate. Pay had set the kitchen, and pay reset it. A system that paid for her surgery and never paid for her rail is set the same way.

Most American care is fee-for-service, a fee for each service performed, and keeping someone on her feet earns no fee. Paying for the quality of care instead, which the field calls value-based payment, would reward the rail. My fourth question is how to get there (Q222).

No trial can assign half a country to a new way of paying. But in 2019 a rule let Medicare Advantage plans, the private plans many people on Medicare choose, cover home safety changes such as stair rails. Some plans began to cover them, and others did not. That split is a natural experiment, the kind the third part found in Snow’s two water companies: the world, not a researcher, sorted people into two groups.

Count hip fractures among members of plans that started paying for rails, before 2019 and after. Do the same for plans that did not. The change in the first group, minus the change in the second, estimates what the rule did. Members of the paying plans would be her nurses: the same kind of stairs, a different plan card. The trap is that plans chose for themselves, and the ones that chose may have had healthier members. The years before 2019 are the check. If both groups were moving in step before the rule, a gap that opens after it is more likely the rule’s work. I have not found a study that has run this comparison yet. The cost of a fix shows up somewhere else, and here it shows up as a rail no budget owns.

The plan’s benefits booklet, when the enrollment mail comes in October, is where the question lands: does the plan pay for a rail or a grab bar? Between October 15 and December 7, people on Medicare can switch plans, and a daughter helping her mother choose can ask that one question before the premium. A plan chosen on its premium alone goes on paying for falls.

Each of the four questions now has its nurses. Each faces the same test as the body’s questions: what data to gather, which method, and whether that method gives a direct answer once the data are in. None of the data has to exist yet.

Figure 4. The four surroundings questions, traced back to the data each one needs. “Partly collected”: some studies hold it, though no clinic gathers it routinely. Drawn for this piece.

Three versions of her close the question, and they are the first part’s figure drawn through one life. In the first, she fell on the third stair at sixty-six, and the couch downstairs was the beginning of her sick years. In the second, the fall came at seventy-two instead, and she spent the same years on the couch, only later. In the third, a timed walk flagged her at fifty-eight, she kept training, and a rail went in that spring. When the fall came it was one fall, and the years that followed were her own. The second version is delay. The third is compression.

The third version is exactly the comparison lead-time bias corrupts: counted from the flag at fifty-eight, she looks as if she lived longest with frailty. So count every version from the same birthday, as years without disability from fifty-eight on, the way the first part’s figure counts, by age and not by diagnosis. On that count the third version wins, and the count is one a statistician would sign.

A Closing Invitation. What the nurses ate stood for a cause, and a fix, outside the body: a stair, a missing check, a bill. In the versions of her, the gap closes when the sick years shrink, not when they move later.

1. Right now, picture the stairs in your home or a parent’s, and count the steps with nothing to hold. Which step would you miss in socks? Would the plan card in your wallet pay for a rail beside it?
2. If someone you love fell in the last year, ask them two things: which program the hospital sent them to, and who called two weeks later. Which of the two did no one mention?
3. At this month’s checkup, yours or a parent’s past seventy, ask out loud: “Has anyone timed my walk? When will they do it again?” Which answer would change what happens before the next fall?

Her third stair has a rail now, and the years counted from fifty-eight are hers. The next piece counts years like hers for a whole country. In Jackson the fix sat on the nurses’ plate; in Chicago, beyond a shut window.

Where This Came From

Goldberger is remembered for the filth parties. His partner went on to something quieter. From 1921 to 1924, Edgar Sydenstricker’s team visited about 1,800 families in Hagerstown, Maryland, and wrote down every illness, not only every death. The Hagerstown study was one of the first American surveys to count how often a town was sick, and it grew into the National Health Survey. The healthspan gap rests on that idea: count the years spent unwell, not only the years alive.

Intellectual Honesty Note. The woman at sixty-six is invented. The February evening, the socks, the third stair, the couch downstairs, the second fall, the rail, her clinic’s place in the stepped wedge and the three versions of her are illustrations. The parent on the basement stairs and the closing’s scenes are too. The three versions are counted from the same birthday, as years without disability from fifty-eight on, which is this piece’s own device. Her miscalibration rests on a study of older adults with diabetes, which fits her at sixty-six better than at fifty-five. The list of three suspects is this piece’s device. Goldberger’s 1914 reading of the nurses’ plates was an inference from reports and visits. The prison farm and the orphanage diets came after, and “We just feasted on filth” is quoted from a secondary account. Reading the gap between fewer falls and fewer fallers as the second and third falls is my reading, not a claim the review’s authors make. The stepped-wedge trial, the factorial trial of the rail and the difference-in-differences study of the 2019 rule are designs drawn up here; none has been run. Walking speed and grip strength appear in plain words, not as definition boxes, from two reviews: Middleton, Fritz and Lusardi on walking speed, and Roberts and colleagues on grip. The answerability table names data to collect, not studies that exist.

References

Bryan, C. S., & Mull, S. R. (2015). Pellagra pre-Goldberger: Rupert Blue, Fleming Sandwith, and the “vitamine hypothesis.” Transactions of the American Clinical and Climatological Association, 126, 20–45.

Centers for Disease Control and Prevention. (2026). Facts about older adult falls. Retrieved October 8, 2026, from https://www.cdc.gov/falls/data-research/facts-stats/index.html

Centers for Medicare & Medicaid Services. (n.d.). Value-based programs. Retrieved October 8, 2026, from https://www.cms.gov/medicare/quality/value-based-programs

Centers for Medicare & Medicaid Services. (2018, April 27). Reinterpretation of “primarily health related” for supplemental benefits [HPMS memorandum].

Clay, K., Schmick, E., & Troesken, W. (2017). The rise and fall of pellagra in the American South (NBER Working Paper No. 23730). National Bureau of Economic Research.

Collins, G. S., Reitsma, J. B., Altman, D. G., & Moons, K. G. M. (2015). Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): The TRIPOD statement. BMC Medicine, 13, 1. https://doi.org/10.1186/s12916-014-0241-z

Dyer, S. M., Crotty, M., Fairhall, N., Magaziner, J., Beaupre, L. A., Cameron, I. D., & Sherrington, C. (2016). A critical review of the long-term disability outcomes following hip fracture. BMC Geriatrics, 16, 158.

Fried, L. P., Tangen, C. M., Walston, J., Newman, A. B., Hirsch, C., Gottdiener, J., et al. (2001). Frailty in older adults: Evidence for a phenotype. Journals of Gerontology: Series A, 56(3), M146–M156. https://doi.org/10.1093/gerona/56.3.m146

Fritz, S., & Lusardi, M. (2009). White paper: “Walking speed: The sixth vital sign.” Journal of Geriatric Physical Therapy, 32(2), 2–5.

Goldberger, J. (1914). The etiology of pellagra: The significance of certain epidemiological observations with respect thereto. Public Health Reports, 29(26), 1683–1686.

Goldberger, J., Waring, C. H., & Willets, D. G. (1915). The prevention of pellagra: A test of diet among institutional inmates. Public Health Reports, 30(43), 3117–3131.

Goldberger, J., & Wheeler, G. A. (1915). Experimental pellagra in the human subject brought about by a restricted diet. Public Health Reports, 30(46), 3336–3339.

Goldberger, J., Wheeler, G. A., & Sydenstricker, E. (1920). A study of the relation of family income and other economic factors to pellagra incidence in seven cotton-mill villages of South Carolina in 1916. Public Health Reports, 35(46), 2673–2714.

Gorfine, M., Cook, R. J., Kragh Andersen, P., Therneau, T. M., Joly, P., Putter, H., et al. (2026). An overview and recent developments in the analysis of multistate processes. Statistics in Medicine, 45(10–12), e70493. https://doi.org/10.1002/sim.70493

HealthCare.gov. (n.d.). Fee for service. In Glossary. Retrieved October 8, 2026, from https://www.healthcare.gov/glossary/fee-for-service/

Hemming, K., Taljaard, M., McKenzie, J. E., Hooper, R., Copas, A., Thompson, J. A., et al. (2018). Reporting of stepped wedge cluster randomised trials: Extension of the CONSORT 2010 statement with explanation and elaboration. BMJ, 363, k1614. https://doi.org/10.1136/bmj.k1614

Kwok, C. S., Hinton, A. V., Lip, G. Y. H., Qureshi, A. I., & Borovac, J. A. (2026). Deconditioning in cardiovascular disease. Coronary Artery Disease. https://doi.org/10.1097/mca.0000000000001667

Medicare.gov. (n.d.). Joining a plan. Retrieved October 9, 2026, from https://www.medicare.gov/basics/get-started-with-medicare/get-more-coverage/joining-a-plan

Middleton, A., Fritz, S. L., & Lusardi, M. (2015). Walking speed: The functional vital sign. Journal of Aging and Physical Activity, 23(2), 314–322. https://doi.org/10.1123/japa.2013-0236

Montgomery, A. A., Peters, T. J., & Little, P. (2003). Design, analysis and presentation of factorial randomised controlled trials. BMC Medical Research Methodology, 3, 26. https://doi.org/10.1186/1471-2288-3-26

National Institute for Health and Care Excellence. (n.d.). Cost-effectiveness analysis. In NICE glossary. Retrieved October 8, 2026, from https://www.nice.org.uk/glossary?letter=c

Oke, J., Fanshawe, T., & Nunan, D. (2021). Lead time bias. In Catalogue of Bias. https://catalogofbias.org/biases/lead-time-bias/

Roberts, H. C., Denison, H. J., Martin, H. J., Patel, H. P., Syddall, H., Cooper, C., & Sayer, A. A. (2011). A review of the measurement of grip strength in clinical and epidemiological studies: Towards a standardised approach. Age and Ageing, 40(4), 423–429. https://doi.org/10.1093/ageing/afr051

Schwartz, A. V., Vittinghoff, E., Bauer, D. C., et al. (2011). Association of BMD and FRAX score with risk of fracture in older adults with type 2 diabetes. JAMA, 305(21), 2184–2192. https://pmc.ncbi.nlm.nih.gov/articles/PMC3287389/

Science History Institute. (n.d.). Joseph Goldberger’s filth parties. Distillations. https://www.sciencehistory.org/stories/magazine/joseph-goldbergers-filth-parties/

Sherrington, C., Fairhall, N. J., Wallbank, G. K., et al. (2019). Exercise for preventing falls in older people living in the community. Cochrane Database of Systematic Reviews, CD012424.

Studenski, S., Perera, S., Patel, K., et al. (2011). Gait speed and survival in older adults. JAMA, 305(1), 50–58.

Sydenstricker, E. (1925). The incidence of illness in a general population group: General results of a morbidity study from December 1, 1921, through March 31, 1924, in Hagerstown, Md. Public Health Reports, 40(7), 279–291.

Szanton, S. L., Xue, Q.-L., Leff, B., et al. (2019). Effect of a biobehavioral environmental approach on disability among low-income older adults: A randomized clinical trial. JAMA Internal Medicine, 179(2), 204–211.

Thompson-McFadden Pellagra Commission. (1914). First progress report of the Thompson-McFadden Pellagra Commission of the New York Post-Graduate Medical School and Hospital. Republican Publishing Company.

Tinetti, M. E., Richman, D., & Powell, L. (1990). Falls efficacy as a measure of fear of falling. Journal of Gerontology, 45(6), P239–P243. https://doi.org/10.1093/geronj/45.6.p239

Van Calster, B., McLernon, D. J., van Smeden, M., Wynants, L., & Steyerberg, E. W. (2019). Calibration: The Achilles heel of predictive analytics. BMC Medicine, 17, 230. https://doi.org/10.1186/s12916-019-1466-7

Wang, G., Hamad, R., & White, J. S. (2024). Advances in difference-in-differences methods for policy evaluation research. Epidemiology, 35(5), 628–637. https://doi.org/10.1097/EDE.0000000000001755

Watson, V., Tudur Smith, C., & Bonnett, L. J. (2024). Systematic review of methods used in prediction models with recurrent event data. Diagnostic and Prognostic Research, 8, 13. https://doi.org/10.1186/s41512-024-00173-5

Wilcox, L. S. (2005). Old black water. Preventing Chronic Disease, 2(Special issue). https://www.cdc.gov/pcd/issues/2005/nov/05_0191.htm

A doctor named John Snow went house to house to find out which company served each house. Where people could not say, he carried a little of their water home in a small bottle. A drop of silver nitrate, a chemical that clouds in salty water, turned the downstream company’s water cloudy white and left the upstream supply nearly clear. Through that epidemic, the downstream company’s houses lost 315 people to cholera for every 10,000 houses. The upstream company’s lost 37. That is more than eight times the deaths, among people who shared streets, air and weather.

Figure 1. The water supply of the districts south of the Thames, from John Snow’s On the Mode of Communication of Cholera (second edition, 1855). Blue is the Southwark and Vauxhall company’s area, red the Lambeth company’s, and purple the streets where “the pipes of both Companies are intermingled.” The color is the data, so this map is not shown in grayscale. Public domain, via Project Gutenberg ebook 72894.

“What causes cholera?” was a debate. Most doctors held that disease rose from filth as a foul smell, a miasma. A government committee wrote of Snow’s idea, “we see no reason to adopt this belief.” The debate could not be settled as asked. A smaller question could. Do houses on the downstream company’s water die more than their neighbors on the upstream company’s? Nobody chose which houses got which pipe; a landlord had, years before, for reasons that had nothing to do with cholera. The world had sorted the groups, and counting answered the question directly.

The patient portal pings after a blood test, and this year’s blood sugar sits beside last year’s, a little higher, still marked in range. More than two in five American adults have prediabetes, blood sugar above normal but not yet diabetes. Eight in ten people with prediabetes do not know they have it. You read the number on a phone in the kitchen you grew up in, a parent with diabetes at the table. Do you act on a number still marked in range, or wait a year for the next test? The portal cannot say whether a number read now arrives in time to change anything.

The woman of the Chicago heat-wave piece, who wrote out her four-part question at forty-seven, is fifty-five now. One checkup adds a second name to her polycystic ovary syndrome (PCOS), a hormone disorder that often makes the body respond poorly to insulin. Her HbA1c reads 6.6 percent. HbA1c is a blood test that reads the average blood sugar over about three months, and at 6.5, a doctor calls it diabetes. Women with PCOS carry about three and a half times the risk of type 2 diabetes. A review of 23 studies found that excess only in women whose weight was also high. Hers has been high since the chart read a body mass index of 27 at twenty-five. Thirty years of metformin, a pill that helps the body use insulin, delayed the diagnosis and did not prevent it.

The same checkup reads her hip: a T-score of -1.8, a measure of how far bone density falls below a young adult’s average. That is low, but short of the -2.5 that defines osteoporosis. Two chronic conditions in one woman, PCOS and diabetes, make the multimorbidity the second part’s trial was built for. The checkup reads her blood and her bone, and nobody times her walk. She leaves with two leaflets and the same breakfast waiting at home.

Before any data, the second part said, each of my questions has to set its comparison, its measure, and its rule for whatever happens partway through. This part asks the next thing of the three questions on the body’s side: once the data exist, is there a method that answers each one directly? Snow’s count is the standard. The handle on the Broad Street pump, the better-known half of his story, came off after the dying had slowed. First, her training, which she has kept up since forty-seven: how much of it is enough?

How Much Is Enough

Two strength days and three walks a week, for eight years. Her question at fifty-five is whether that is the least that keeps her walking on her own at eighty, or more than she needs. The second part’s computer model of bone said a higher start buys more than a slower fall. Most of the bone a body will ever have is laid down by about thirty. The hip that reads -1.8 is still spending what she banked in her twenties. What is the smallest dose of strength and aerobic training in midlife that keeps the most function at eighty? That is my first question (Q111, Figure 2).

In exercise science, a minimal dose is the smallest amount of training, below the usual guideline, that still produces the benefit. Finding it starts with a curve. Hannah Arem and colleagues pooled six studies of about 661,000 adults and counted deaths at each level of weekly leisure exercise. The standard recipe is 150 minutes of moderate activity a week. Those who exercised but fell short of it had about a 20 percent lower risk of dying than those who did nothing. Meeting it brought the drop to 31 percent. Most of the drop came with the first steps up from doing nothing, and less with each step after. A curve that bends near zero means the first ten minutes buy more than the last ten.

A trial that sets the full recipe against nothing can say whether exercise helps. It cannot say where the bend sits, because it measures only two points on the curve. Finding the smallest dose that works takes several doses, and a model that lets the curve bend where the data bend.

The bend is drawn with restricted cubic splines: short curved pieces joined smoothly into one line. The line is held straight at the two ends, where the data thin out, so it can bend in the middle without flapping at the tails. Her dose, two strength days and three walks, is one point on that curve. What matters is whether the curve has flattened by the time it reaches her.

Then comes the heart attack at seventy that the second part wrote into her question as an event the study has to plan for. Her diabetes makes it likelier: people with diabetes die of any cause at about 1.8 times the rate of people without. Someone who dies at seventy never gets measured walking at eighty. Drop that person from the count, and the study assumes, without saying so, that she would have walked like the women who lived. The first part found that deaths make poor witnesses, since each is tallied under a single cause. Here death is not a cause to be sorted but an exit that rules out the outcome. My field calls such an exit a competing risk.

For the method to fit, the outcome has to be a time, not a snapshot. It is the age at which she can no longer walk on her own, with death before that as the competing event. Function measured once at eighty is a different problem, because the women who died first have no reading at all. Jason Fine and Robert Gray gave the time version its model in 1999. Their regression keeps the women who died in the denominator instead of treating them as if they had simply left the study. Fine-Gray regression, run on a curve of doses, gives the chance of losing independent walking by eighty at each dose, with death counted as its own answer. The training is chosen, not assigned, so the curve is an association; a causal “enough” needs a trial that assigns doses, which is another split.

Sunday’s plan for the week, written on the fridge, is one week of the dose. The guideline pairs the 150 minutes with strength work on two or more days, and the strength days are what to look for on the plan. Counting them takes a pen and a minute. The dose is one question. Which level of the body the answer lives at is another.

Same Number, Different Roads

Two people can leave the same clinic with the same 6.6 percent and different futures. In 2018, a Swedish team sorted about 9,000 adults newly diagnosed with diabetes. Six measurements split them into five groups. In one group, the cells that make insulin were failing. Another group made plenty of insulin that the body could not use, and that group had the highest risk of kidney disease. All of them came from one region of southern Sweden. One number on the lab sheet, and at least five roads behind it.

Which road a person is on depends on the level at which the disease is modeled. A body is built in levels, each made of the one below: molecules, cells, tissues, organs, organ systems, the whole organism. A disease can be modeled at a gene, at a pathway that many genes feed, at a network of pathways, or at the whole person. Which level is the right one, and can the data say? That is my second question (Q121, Figure 3).

The gene level has its own design. A 2024 study pooled more than two and a half million people. It found 611 places in the genome linked to type 2 diabetes. The strongest, in a gene called TCF7L2, raises the risk by about 45 percent in people who carry one copy. Most of the rest raise it by a few percent each.

A genome has millions of places that could line up with diabetes by chance, and so does the layer of marks on top of it. So my field believes a single signal only when chance alone would give a result that strong less than about 5 times in 100 million. That line is called genome-wide significance. Then the study sorted its 611 signals into eight groups by what they seem to act on: the insulin-making cells, the way the body stores fat, the liver. Her own condition was searched the same way in 2018. A study of about 10,000 women with PCOS found 14 places in the genome past that line. Its genetic signal as a whole runs with obesity, fasting insulin and type 2 diabetes. At the gene level, her PCOS and her diabetes share ground. At the pathway level, that ground has a name, insulin resistance, and at the level of the whole person it is one woman at breakfast.

A GWAS answers only at the gene level. To choose between levels, the data have to be asked which one predicts best, and my field has a way to ask without fooling itself. Build a model at each level, gene, pathway, network, whole person, and test each one on people it never saw. The trap is that choosing a model’s settings on the same people it is then scored on makes every model look better than it is. Nested cross-validation closes the trap with two loops. An inner loop picks the settings on training data only, and an outer loop scores the chosen model on data held entirely apart. It is work I have done on genetic data, and it answers “which level predicts best” directly. It does not say which level causes the disease.

The family-history box on an intake form, ticked for a parent’s diabetes, is the same question in a kitchen. The thing to find out is the age the parent was diagnosed. The question to ask is whether the risk is shared through the genes or through the kitchen. A level that predicts is not yet a level that explains, and the third question asks what the body’s own clock has to say.

Older Than Its Owner

A model trained on thousands of brain scans learns what a healthy brain looks like from 40 to 70. Then it reads a new scan and returns a number: this brain looks 58, and its owner is 52. The number is a brain age, the age a model predicts from a scan, and the gap between it and the birthday is the thing to watch. A group of 669 Scottish adults, all born in 1936, had brain scans in their early seventies. Over the following years, those whose brains looked older than their age were more likely to die. In a 2013 study, older adults with type 2 diabetes had brains that looked about four and a half years older than their age. That study caught both at one moment, so it cannot say which came first.

Which exposures speed a body’s clock, and does the gap between a body’s age and its birthday predict the next disease? That is my third question (Q122). The first half is a question about cause, and cause is where Snow’s water companies come back.

Snow found two groups alike in almost everything but the water, so the water stood out, the way migration sorted two groups who share an ancestry. Nobody assigned the pipes at random, and Snow could only count. But a gap that large, between next-door neighbors, was hard to explain any other way.

Her weight raises the same kind of question. Did her weight bring on the PCOS, or did the PCOS bring the weight? No trial can assign weight at birth. Genes can. The variants that raise body weight are handed out at conception, by the shuffle of a parent’s genes into a child, and no researcher chose who got them. In 2003, George Davey Smith and Shah Ebrahim named the design.

It is Snow’s experiment, run at conception. In 1854 the landlord picked the water company, not the household. Here the shuffle of genes picks the exposure, not the person. Same design, a century and a half apart: the world sorted the groups, so counting can answer a question about cause. The 2018 PCOS study ran it, and its results suggested that variants for body mass index and fasting insulin “play a causal role in PCOS.” By that reading, her weight came first, not the other way round. Both designs hold only on one condition. Whatever sorted the groups has to reach the outcome through the exposure alone. For Snow, that means the water and not the landlord; here, the gene and not some other path the gene also takes. That condition can fail, and a study has to say what it did to check.

The second half of the question is prediction, and its method is older. Does a body that reads older than fifty-five get its next diagnosis sooner than a body that reads its age? Cox regression, from 1972, models the rate at which an event arrives from the measurements a person carries. Adjusted for the birthday, it asks whether the gap predicts anything beyond age alone. Fine-Gray, from the first section, is Cox’s model adapted for a competing exit. Brain-age models pull their guesses toward the middle, so young brains read older and old brains read younger. The correction takes out the part of the gap that age alone predicts. What remains belongs to the person. The second part showed a bone scan earning standing as a stand-in for the outcome that counts; no reading of body or brain age has earned it yet.

Back on the patient portal, this year’s blood sugar beside last year’s asks the same question. Prediabetes runs from an HbA1c of 5.7 to 6.4 percent, and the thing to ask is what the trend predicts beyond the birthday. The decision it leads to is when to retest, and the step is to ask for the date before leaving the clinic.

Each of the three questions now has a method. Each also faces one check: name the data to collect, name the method, and ask whether the method answers the question directly once the data exists. The data does not have to exist yet.

Figure 4. From each question back to its data. “Partly collected” means some studies hold it, but not routinely. Drawn for this piece.

A count can answer cause and still arrive late. In September 1854, on Broad Street in Soho, Snow counted the dead. He found “upwards of five hundred fatal attacks of cholera in ten days.” They clustered within 250 yards of one pump. On the evening of 7 September, he put his case to the parish Board of Guardians. They were “quite incredulous,” but they ordered the handle taken off the next morning. The famous version ends there. Snow himself said otherwise. Many families had already fled, and the attacks had “so far diminished” before the pump was shut that it was “impossible to decide” whether removing the handle saved anyone.

The cause turned up in April 1855, at No. 40 Broad Street, the house nearest the pump. It was the one case a curate named Henry Whitehead had skipped, “because it was the case of an infant.” The baby had fallen ill on 28 August, three days before the street did. Its mother had soaked its diapers in pails and poured the water into the cesspool at the front of the house. The cesspool leaked, and the house drain ran 2 feet 8 inches from the well. The baby’s father, a policeman, fell ill the day the handle came off, and his waste found its way into the same cesspool. This time no one could pump the water up, and the street had no second wave.

The second part found its comparison on a Broad Street in Philadelphia. This one is in London, more than a century earlier. On both, the people a count left out were the answer. The handle came off too late for the first wave and just in time for the second. The count was right, and it changed who drank only once the dying had done most of its work.

A Closing Invitation. The handle on Broad Street stood for an answer that arrives right and late. When the world sorts the groups, by water company or by genes, counting can name a cause, but only an answer in time changes anything.

1. Open the last lab result in your inbox and find one number with last year’s beside it: blood sugar, blood pressure. Which way did it move, and did anything at breakfast move with it?
2. On Sunday, write the week’s training on the fridge with the strength days circled: a set of squats, a bag carried up the stairs. Are there two, and which one would a busy week take first?
3. At your next checkup, when the doctor names a condition, such as prediabetes or low bone density, say aloud: “Which kind, and what should I watch for next?” What would the answer change this month?

In 1999 the heat came back to Chicago, and the city knocked on doors in time; far fewer people died than in 1995. The next handle can come off early: a breakfast changed, a walk timed in the clinic room.

Where This Came From

I first heard the physician Peter Attia during the pandemic, in a 2018 interview with Tom Bilyeu. Until then, staying healthy had meant one thing to me: not catching covid. Attia was training for what he called a hundred-year-old Olympics: squatting, carrying groceries, and playing with grandchildren at a hundred. He later wrote it up as the Centenarian Decathlon, a plan that starts from the end and trains backward. I was not married then, and I had no daughters. Now I have both, and the goal has faces. My dad takes our girls on bike rides and to the park. My mother-in-law, a stand-up comedian with a master’s degree in early childhood education, plays with the babies as if they were her best audience. I want to play with my grandchildren the way they play with ours. Research can start from the end too, and that is the test all three questions here have to pass. My research statement asks the same of a causal claim: that it be solid enough to act on.

Intellectual Honesty Note. The woman at fifty-five is invented; her hip at -1.8, her HbA1c of 6.6, her walks and strength days, and her weight are illustrations, and so is the 58-year-old brain on a 52-year-old. The 2018 PCOS genetics study covered women of European ancestry only. The father’s waste reaching the cesspool, and “no second wave,” are Whitehead’s own later reading, as Chave reports it. The Chicago 1999 count is from the city’s own study of that heat wave. Nested cross-validation is summarized from my own paper, listed below, without further citations. Figure 4’s “collected” and “partly collected” are my own reading of what large studies hold; the table names no study.

References

Ahlqvist, E., Storm, P., Käräjämäki, A., Martinell, M., Dorkhan, M., Carlsson, A., et al. (2018). Novel subgroups of adult-onset diabetes and their association with outcomes: A data-driven cluster analysis of six variables. The Lancet Diabetes & Endocrinology, 6(5), 361–369.

American Diabetes Association Professional Practice Committee. (2024). Diagnosis and classification of diabetes: Standards of Care in Diabetes, 2024. Diabetes Care, 47(Suppl. 1), S20–S42.

Anagnostis, P., Paparodis, R. D., Bosdou, J. K., Bothou, C., Macut, D., Goulis, D. G., & Livadas, S. (2021). Risk of type 2 diabetes mellitus in polycystic ovary syndrome is associated with obesity: A meta-analysis of observational studies. Endocrine, 74(2), 245–253. https://doi.org/10.1007/s12020-021-02801-2

Arem, H., Moore, S. C., Patel, A., et al. (2015). Leisure time physical activity and mortality: A detailed pooled analysis of the dose-response relationship. JAMA Internal Medicine, 175(6), 959–967.

Attia, P., with Gifford, B. (2023). Outlive: The Science and Art of Longevity. Harmony.

Austin, P. C., Lee, D. S., & Fine, J. P. (2016). Introduction to the analysis of survival data in the presence of competing risks. Circulation, 133(6), 601–609. https://doi.org/10.1161/CIRCULATIONAHA.115.017719

Bilyeu, T. (2018, November 8). Why you need to protect your joints if you want to live to be 100 | Peter Attia on Health Theory [Video]. YouTube. https://www.youtube.com/watch?v=YY-_ux4ZXp4

Centers for Disease Control and Prevention. (2026). A U.S. report card: Diabetes statistics. Retrieved October 8, 2026, from https://www.cdc.gov/diabetes/communication-resources/diabetes-statistics.html

Chave, S. P. W. (1958). Henry Whitehead and cholera in Broad Street. Medical History, 2(2), 92–108.

Chen, Z., Boehnke, M., Wen, X., & Mukherjee, B. (2021). Revisiting the genome-wide significance threshold for common variant GWAS. G3: Genes, Genomes, Genetics, 11(2), jkaa056. https://doi.org/10.1093/g3journal/jkaa056

Cholera Inquiry Committee. (1855). Report on the cholera outbreak in the parish of St. James, Westminster, during the autumn of 1854. J. Churchill.

Clark, M. A., Douglas, M., & Choi, J. (2018). Biology 2e (Section 1.2, Themes and concepts of biology). OpenStax. https://openstax.org/books/biology-2e/pages/1-2-themes-and-concepts-of-biology

Cole, J. H., & Franke, K. (2017). Predicting age using neuroimaging: Innovative brain ageing biomarkers. Trends in Neurosciences, 40(12), 681–690. https://doi.org/10.1016/j.tins.2017.10.001

Cole, J. H., Ritchie, S. J., Bastin, M. E., Valdés Hernández, M. C., Muñoz Maniega, S., Royle, N., et al. (2018). Brain age predicts mortality. Molecular Psychiatry, 23, 1385–1392.

Cox, D. R. (1972). Regression models and life-tables. Journal of the Royal Statistical Society: Series B, 34(2), 187–202. https://doi.org/10.1111/j.2517-6161.1972.tb00899.x

Craig, P., Cooper, C., Gunnell, D., Haw, S., Lawson, K., Macintyre, S., et al. (2012). Using natural experiments to evaluate population health interventions: New MRC guidance. Journal of Epidemiology and Community Health, 66(12), 1182–1186.

Davey Smith, G., & Ebrahim, S. (2003). ‘Mendelian randomization’: Can genetic epidemiology contribute to understanding environmental determinants of disease? International Journal of Epidemiology, 32(1), 1–22. https://academic.oup.com/ije/article/32/1/1/642797

Day, F., Karaderi, T., Jones, M. R., et al. (2018). Large-scale genome-wide meta-analysis of polycystic ovary syndrome suggests shared genetic architecture for different diagnosis criteria. PLoS Genetics, 14(12), e1007813. https://doi.org/10.1371/journal.pgen.1007813

Emerging Risk Factors Collaboration. (2011). Diabetes mellitus, fasting glucose, and risk of cause-specific death. New England Journal of Medicine, 364(9), 829–841. https://doi.org/10.1056/NEJMoa1008862

Fine, J. P., & Gray, R. J. (1999). A proportional hazards model for the subdistribution of a competing risk. Journal of the American Statistical Association, 94(446), 496–509. https://doi.org/10.1080/01621459.1999.10474144

Franke, K., Gaser, C., Manor, B., & Novak, V. (2013). Advanced BrainAGE in older adults with type 2 diabetes mellitus. Frontiers in Aging Neuroscience, 5, 90.

Gauran, I. I., Ombao, H., & Yu, Z. (2025). Predictive performance test based on the exhaustive nested cross-validation for high-dimensional data (arXiv:2408.03138v2). https://arxiv.org/abs/2408.03138

General Board of Health, Committee for Scientific Inquiries. (1855). Report on the cholera epidemic of 1854. HMSO.

Grant, S. F. A., Thorleifsson, G., Reynisdottir, I., Benediktsson, R., Manolescu, A., Sainz, J., et al. (2006). Variant of transcription factor 7-like 2 (TCF7L2) gene confers risk of type 2 diabetes. Nature Genetics, 38(3), 320–323.

Harrell, F. E. (n.d.). Regression modeling strategies (Section 2.4.5, Restricted cubic splines) [Course notes]. Retrieved October 8, 2026, from https://hbiostat.org/rmsc/genreg

Hernandez, C. J., Beaupré, G. S., & Carter, D. R. (2003). A theoretical analysis of the relative influences of peak BMD, age-related bone loss and menopause on the development of osteoporosis. Osteoporosis International, 14(10), 843–847.

Kanis, J. A., McCloskey, E. V., Johansson, H., Oden, A., Melton, L. J., & Khaltaev, N. (2008). A reference standard for the description of osteoporosis. Bone, 42(3), 467–475. https://doi.org/10.1016/j.bone.2007.11.001

Lu, J., Shin, Y., Yen, M.-S., & Sun, S. S. (2016). Peak bone mass and patterns of change in total bone mineral density and bone mineral contents from childhood into young adulthood. Journal of Clinical Densitometry, 19(2), 180–191.

National Institute of Diabetes and Digestive and Kidney Diseases. (2018). The A1C test & diabetes. https://www.niddk.nih.gov/health-information/diagnostic-tests/a1c-test

Naughton, M. P., Henderson, A., Mirabelli, M. C., Kaiser, R., Wilhelm, J. L., Kieszak, S. M., et al. (2002). Heat-related mortality during a 1999 heat wave in Chicago. American Journal of Preventive Medicine, 22(4), 221–227.

Nuzzo, J. L., Pinto, M. D., Kirk, B. J. C., & Nosaka, K. (2024). Resistance exercise minimal dose strategies for increasing muscle strength in the general population: An overview. Sports Medicine, 54(5), 1139–1162. https://doi.org/10.1007/s40279-024-02009-0

Ritz, C., Baty, F., Streibig, J. C., & Gerhard, D. (2015). Dose-response analysis using R. PLoS ONE, 10(12), e0146021. https://doi.org/10.1371/journal.pone.0146021

Snow, J. (1855). On the Mode of Communication of Cholera (2nd ed.). John Churchill. Project Gutenberg ebook 72894. https://www.gutenberg.org/ebooks/72894

Suzuki, K., Hatzikotoulas, K., Southam, L., Taylor, H. J., Yin, X., Lorenz, K. M., et al. (2024). Genetic drivers of heterogeneity in type 2 diabetes pathophysiology. Nature, 627, 347–357.

Uffelmann, E., Huang, Q. Q., Munung, N. S., de Vries, J., Okada, Y., Martin, A. R., et al. (2021). Genome-wide association studies. Nature Reviews Methods Primers, 1, 59. https://doi.org/10.1038/s43586-021-00056-9

U.S. Department of Health and Human Services. (2018). Physical activity guidelines for Americans (2nd ed.). https://health.gov/our-work/nutrition-physical-activity/physical-activity-guidelines

The veterans belonged to the American Legion, and its members call themselves Legionnaires. Of the people tied to the convention, 182 fell ill, and 29 of them died. Counting the street, that made 221 sick and 34 dead. The suspects came one after another. Swine flu came first. Then a toxic metal compound, nickel carbonyl, that a person can breathe in. Then the hotel’s own pesticides and cleaning products. Through the fall, the labs tested for poisons, bacteria, fungi and viruses, and found nothing to blame.

Figure 1. Two design choices. Left: the Bellevue-Stratford Hotel, with the Broad Street Line subway sign at the bottom of the frame; inside it or outside it was the population the investigators compared. Right: Joseph McDade (left) and Charles Shepard at a microscope at the CDC in 1977; the culture under it was the measure that first hid the answer. CDC Public Health Image Library (IDs 15278 and 8232), public domain.

“What is making the Legionnaires sick?” could not be answered as asked. The people on Broad Street made a simpler one possible: was the cause inside the hotel, or in the air around it? They had never eaten the food or drunk the water inside, so nothing served at the tables could have reached them. They had shared only the air near one building. Set beside the veterans, they pointed away from the kitchen. The investigators concluded that the exposure may have happened in the lobby or on the street just outside. The street cases went into the record under a name of their own: “Broad Street Pneumonia.”

The germ itself turned up in a narrower question still. Over Christmas, a CDC scientist named Joseph McDade went back to samples from the dead. The standard recipe for growing the kind of germ he was hunting added two antibiotics, to keep stray bacteria out. He left the antibiotics out, and a bacterium grew that no one had seen before. The recipe built to keep out the wrong germs had been killing the right one. The CDC announced the find on January 18, 1977. The bacterium was named for the people it first killed: Legionella pneumophila, the “lung-loving” one. Where it came from was never proven. The hotel’s cooling tower, part of its air conditioning, later became the main suspect.

Two choices solved the case, and both were made before anyone knew the answer. The first was who to compare: the people on the sidewalk against the people inside. The second was what to measure: a culture grown one way hid the germ, and grown another way showed it. Made first, those two choices turned a mystery into a question someone could check.

You sign up for a strength class, two sessions a week, on a mat that smells of rubber. Three months in, a doctor adds a new prescription. By month six the stairs feel easier, and the question is whether the class did that, or the pill, or the summer. The cost is not the two sessions a week. It is six months of not knowing whether to keep paying for them, or whether to tell your sister to join. Compared with whom, and what counts if something changed partway? Decided afterwards, any answer fits.

The first part of this series set out the question in biostatistics I return to most: how to narrow the gap between lifespan and healthspan. Healthspan is the stretch of life spent in good health. By the end of that part, I had split the question in two: the body’s own capacity, and the demands of the place a person lives in. Split once, each half is still more than a single study can take on, the way “What is making the Legionnaires sick?” was. Each splits again, into four ways in (Figure 2). Each way in needs what the sidewalk settled: who is compared, and what is measured. In my field, that pair written out in full, before any data exists, has a name: an estimand.

The woman of twenty-five from the first part is forty-seven now. Her PCOS, polycystic ovary syndrome, is a hormone condition that often comes with trouble using insulin. Three doctors saw her before anyone named it. She has started lifting weights, two mornings a week, to keep her hips strong. Whether that works is the first way in, and the first question that needs its comparison chosen in advance.

Start Higher

Ten percent more bone at the start can put off osteoporosis by about thirteen years. Osteoporosis is bone thin enough to break easily. The number comes from a 2003 computer model of a woman’s bones, built by Christopher Hernandez and colleagues. In the same model, slowing the loss by the same 10 percent bought about two years.

That starting supply is called reserve: what a body can draw on to withstand a stressor, beyond what daily life asks of it. The first way in builds a higher peak of reserve before decline begins (Q11). Then what a body can do on its own still declines, but it crosses the line where the loss matters later. Reserve works like savings before retirement. Two people can spend at the same rate, and the one with the bigger balance runs out years later. Muscle, fitness and bone are three accounts. Unlike a bank balance, they shrink when left untouched.

The brain keeps one too, called cognitive reserve: the ability to use its networks more efficiently, or to call up other networks when one fails.

Fitness may be the largest account, and one that builds on itself. One hard ride on an exercise bike strips marks off the genes that build a muscle’s energy supply. Months of rides add capacity for the next. Her weights pay into two of the accounts, bone and muscle: resistance training, muscles working against a load, to raise the peak before the decline.

Whether her training raises the peak needs its question written out first. Thirty-nine sick people on one street gave the Philadelphia investigators a question with four parts. Who: people who had been within a block of the hotel but never went in. Exposed to what: the air around one building. What outcome: the illness. Compared with whom: the veterans inside, who had shared the food and water as well. Leave one part out, and almost any answer fits. Bacteria shared by a mother’s milk and her baby’s gut mean little until someone asks whether a stranger’s baby would share them too.

A question written out this way is an estimand. In 2019, the international council that sets rules for drug trials told trials to state theirs before they start.

The woman’s estimand reads this way. Who: women between 45 and 60 who do not train yet. Exposed to what: two short strength sessions and a few brisk walks a week. What outcome: physical function, which here means walking on her own at eighty. Compared with whom: women like her carrying on as before. The estimand is about the population, not about her alone. It asks how many more of those women are walking at eighty, and she is one of them.

Life gets in the way of a question that long. In November 2025, the FDA announced it would remove its strongest warning label, the boxed warning, from menopausal hormone therapy. More women may now start hormones during the transition. If she starts them halfway through, the hormones change her bone too. Some women quit the sessions. And some die before eighty. Her PCOS makes diabetes more likely, and a 50-year-old with diabetes dies, on average, six years earlier than one without. A heart attack at seventy is the death the question has to plan for. The guideline calls each of these an intercurrent event, and the estimand must say how they count before anyone enrolls.

One trial, two true answers. If she starts hormones halfway through, one analysis counts her as trained, hormones and all, and asks what happened to women who were handed the sessions and lived their lives. Another asks whether she would still walk on her own at eighty had she never started hormones. Both are correct on the same data, because they answer different questions. The first part of this series found one death with four true answers. Here the estimand picks which true answer a study will give, before the data exist. A death that comes first is the hardest of the events, because it ends the outcome instead of changing it. The third part gives it a method of its own.

Someone starting a strength class this month can write the four parts on the back of the receipt. Compared with whom: a friend who did not sign up. What counts as working: the stairs at month six, timed. What happens if a new prescription starts partway: it goes on the receipt, with the date. Two sessions a week for six months is the cost. A receipt with four lines on it turns the cost into an answer. A body built higher still ages, and the second way in asks whether one pill could slow what drains it.

What the Diseases Share

For twenty-two years the woman has swallowed one pill with breakfast: metformin. It is a diabetes drug, prescribed for her periods at twenty-five. A trial called TAME was designed to test that cheap pill against aging itself. The second way in slows what the diseases share at the root: the drivers that drain every reserve at once (Q12).

The field’s name for those drivers is the hallmarks of aging, a short list of processes, such as chronic inflammation and worn-out cells that refuse to die. Each one shows with age, speeds aging when pushed, and slows it when treated. The field calls the premise the geroscience hypothesis: treating what aging does to a body should hold off many diseases at once, not one at a time.

TAME’s design says how. Its authors chose not to study metformin’s effect “on each separate condition.” Instead, they wrote, “we will measure time to a new occurrence of a composite outcome that includes cardiovascular events, cancer, dementia, and mortality.” The design works like a single stopwatch for four outcomes: whichever arrives first stops it, and that time is what the trial compares. That is the outcome part of an estimand, answered for a whole way in at once.

Death is on that stopwatch on purpose. A trial that counted only dementia would lose the people who died first, and the guideline names death as the plainest case for folding the event into the outcome. That is one way to handle the heart attack at seventy from the first way in: count it, instead of planning around it.

Attia’s Four Horsemen, the first part’s list of the four diseases most people die of, are heart disease, cancer, dementia and type 2 diabetes. TAME’s composite counts three of the four as one outcome and leaves the fourth out. It leaves her out too. The trial would assign 3,000 people to metformin or a placebo at random. All would be 65 to 79, without diabetes. At forty-seven she is too young, and by fifty-five, as the third part tells, she will have diabetes. Her twenty-two years on the pill make her the trial’s most interesting case and its excluded one.

Two or more chronic conditions in one person is what the field calls multimorbidity, and holding off that pile-up is the second way in. TAME’s intercurrent event can already be named: someone given the placebo is diagnosed with diabetes and starts metformin outside the trial. The estimand has to say how that person counts. TAME has not enrolled anyone. Metformin is generic, so no drug company stands to make money from proving it works. In 2025 the trial’s lead investigator, Nir Barzilai, said it was on hold while talks with the FDA went on.

The leaflet in a pharmacy bag says a drug lowers a risk by some percent. The questions to ask at the next refill: a risk of which outcome, in whom, compared with whom? A percent with no population under it fits anyone, and so fits no one. Whether a treatment slows the drivers can only be read on a stopwatch that runs for years. The third way in asks what a scan can read sooner.

Clues Outside the Building

At forty-seven, a technician scans the woman’s hip for the first time. A break, if it comes, may be twenty years off, and no study of her training can wait for it. So the study reads her hip on a bone scan instead: a number that stands in for the fracture, the way McDade’s culture stood in for the germ. The third way in catches decline early, before one loss cascades into others (Q21). Its first choice is what to measure in place of the thing itself.

In December 2025, the FDA accepted two years of change in hip bone density as a stand-in for fractures in drug trials. The rule covers women past menopause with osteoporosis. The stand-in has a name.

An endpoint is a variable defined precisely enough to be analyzed, chosen to reflect the outcome a study cares about. A surrogate is one that arrives sooner. It is only as good as its link to the real outcome, and the link can hold for most people and fail for a few. In type 2 diabetes, the bone scan reads better than the bone is. A 2011 study of older adults found that, for the same scan result and age, those with diabetes broke more bones than those without. Her own scan at forty-seven reads normal, and at fifty-five the third part reads it again.

The scan is one clue outside the building. There are cheaper ones. Long before a woman reports any difficulty climbing stairs, she may have changed how she climbs them, one hand on the rail, or how often. Linda Fried and colleagues named that stage preclinical disability: a task modified in method or frequency without any reported difficulty. It is the early, hidden decline of Figure 2’s third way in, read from a question instead of a scan. The last part of this series asks which such clues are worth checking at every visit.

The number on a scan report, a bone-density score or a cholesterol reading, is a stand-in for something later. The question to ask when the report comes is what the number stands in for, and whether the link holds for someone like the person holding it. A reading that stands in well for the average patient and badly for this one is a number, not an answer. Where a fall happens, on a stair or beside a bathtub, is the last way in.

Questions No Lab Can Grow

A handyman in Baltimore bolts a grab bar beside a bathtub, and the drill whines against the tile. The bar works the day it goes in. The last way in designs homes, communities and care so that a body with less can still live a full, independent life (Q22). About three in four Americans over fifty want to stay in their homes as they age.

The bar is part of a program called CAPABLE, which sends a nurse, an occupational therapist and a handyman into the homes of low-income older adults. CAPABLE was tested on 300 people, assigned at random. Five months in, those in the program had about 30 percent fewer problems with daily tasks like bathing and dressing. A later comparison estimated that Medicare saved about $22,000 per person over two years. The program cost near $3,000.

Written out like the Broad Street question, CAPABLE’s reads this way. Who: 300 low-income adults in Baltimore with difficulty in daily tasks. Exposed to what: CAPABLE. What outcome: problems with daily tasks at five months. Compared with whom: adults given about ten home visits from a research assistant. The exposure is not a pill. It is a nurse, a therapist and a handyman working together in about ten visits. My field has a name for a treatment with that many parts.

A trial of a complex intervention can say whether the bundle worked. It cannot say what the grab bar alone was worth, and that is a question for the last part. Its outcome can also run longer than five months. Over a whole life, limits on daily tasks add up to disability-free life years: the years a person can expect to live with no limit on what they can do. A country’s healthy life expectancy is built the same way. CAPABLE counted five months.

The woman’s own care has many parts and no bundle. At forty-seven she has three clinics. One treats the PCOS, one the blood sugar her metformin has held down for two decades, and one the hormones of the transition. Three clinics keep three files, and no file holds her bones, her blood sugar and her hormones together. The three doctors who each treated one part of her at twenty-five have become three buildings. A better trial does not fix that.

Chicago’s knock on the door set solving a problem beside dissolving it. The missing file is a problem to dissolve: a clinic built to follow one woman through every stage. That clinic would be a complex intervention, not a drug. Who would pay for it is a question no lab can grow, and the last part takes it up with the bill.

A parent with three clinics has three patient portals, and three passwords on a sticky note. The question to ask, at the next referral letter, is which clinic holds the whole list. The decision after that question is small: ask one of the three clinics to coordinate the others, and write its name at the top of the list.

A Closing Invitation. The thirty-nine people on Broad Street stood for the estimand, the comparison chosen before anyone knew the answer. A question becomes checkable once you settle who is compared, what is measured, and what counts if life gets in the way.

1. Find the comparison. Now, or before the next checkup this month, write one count in your phone’s notes and what to compare it with. Try the seconds to climb the stairs at home against last spring’s, or this year’s bone-scan number against the last one. Which has moved, and would anyone have noticed without the comparison?
2. Name what you are building. This week, at the gym or on the walk, say out loud what reserve you, or someone you love, are putting away (bone, muscle, wind) and one thing that could change it partway (a new prescription, a surgery, a move). Would you count the months after it, or start the count again?
3. Count the files. At the kitchen table, with a parent’s appointment letters in a pile, or your own, count the clinics. Three clinics, three files: which one would you ask to hold the whole list, and what is the first thing you would want on it?

Her hip scan is the measure, read at forty-seven and again at fifty-five, when it reads low. Which of the four ways in could have changed that? On Broad Street, the sick on the sidewalk were the comparison.

Where This Came From

Russell Ackoff, in a talk on systems thinking, called a problem an abstraction pulled out of a mess. Trial statisticians reached the same worry from the other side. The 2019 estimand guideline grew from trials that reported an answer to a different question from the one they set out to ask, and the guideline is the origin of this piece’s main term. The mess this series pulls its problems from is the one my research statement names: the years a body survives against the years it survives well.

Intellectual Honesty Note. The woman at forty-seven is invented, and her weights, her hormones, her metformin, her hip scan and her three clinics are illustrations; so are the strength class, the receipt and the scenes in the closing. The hallmarks of aging, multimorbidity and the geroscience premise are given in plain words from their sources; Brian Kennedy and colleagues state the premise without using the term “geroscience hypothesis,” which is the field’s name for it. The four-part estimand for her training, the scan as a stand-in, and the CAPABLE question in four parts are designs sketched here, not studies that were run. “They had shared only the air near one building” is this piece’s reading. The four-part reading of the Broad Street question is this piece’s, not the investigators’. The guideline lists five attributes of an estimand; this piece folds them into four parts plus intercurrent events. The six years earlier is an average across 97 studies, adjusted for age, sex, smoking and body mass, and the scan-reads-better-than-the-bone finding comes from older adults, not women of forty-seven. The CAPABLE trial ran in one city.

References

Ackoff, R. L. (2015, November 2). Systems thinking speech by Dr. Russell Ackoff [Video]. YouTube. https://www.youtube.com/watch?v=EbLh7rZ3rhU

Attia, P., with Gifford, B. (2023). Outlive: The Science and Art of Longevity. Harmony.

Barrès, R., Yan, J., Egan, B., Treebak, J. T., Rasmussen, M., Fritz, T., et al. (2012). Acute exercise remodels promoter methylation in human skeletal muscle. Cell Metabolism, 15(3), 405–411.

Barzilai, N., Crandall, J. P., Kritchevsky, S. B., & Espeland, M. A. (2016). Metformin as a tool to target aging. Cell Metabolism, 23(6), 1060–1065. https://pmc.ncbi.nlm.nih.gov/articles/PMC5943638/

Binette, J., & Farago, F. (2021). 2021 Home and Community Preferences Survey: A national survey of adults age 18-plus. AARP Research.

Centers for Disease Control and Prevention. (n.d.). Public Health Image Library (IDs 8232 and 15278). https://phil.cdc.gov

Emerging Risk Factors Collaboration. (2011). Diabetes mellitus, fasting glucose, and risk of cause-specific death. New England Journal of Medicine, 364(9), 829–841. https://doi.org/10.1056/NEJMoa1008862

Eurostat. (n.d.). Healthy life years (HLY). In Statistics Explained: Glossary. Retrieved October 8, 2026, from https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Glossary:Healthy_life_years_(HLY)

Fraser, D. W., Tsai, T. R., Orenstein, W., Parkin, W. E., Beecham, H. J., Sharrar, R. G., et al. (1977). Legionnaires’ disease: Description of an epidemic of pneumonia. New England Journal of Medicine, 297(22), 1189–1197.

Gregory, P. C., Szanton, S. L., Xue, Q.-L., Tian, J., Thorpe, R. J., & Fried, L. P. (2011). Education predicts incidence of preclinical mobility disability in initially high-functioning older women: The Women’s Health and Aging Study II. Journals of Gerontology: Series A, 66(5), 577–581. https://doi.org/10.1093/gerona/glr023

Henry, R. (2017). Etymologia: Legionella pneumophila. Emerging Infectious Diseases, 23(11), 1851.

Hernandez, C. J., Beaupré, G. S., & Carter, D. R. (2003). A theoretical analysis of the relative influences of peak BMD, age-related bone loss and menopause on the development of osteoporosis. Osteoporosis International, 14(10), 843–847.

International Council for Harmonisation. (2019). E9(R1) Statistical principles for clinical trials: Addendum on estimands and sensitivity analysis in clinical trials. https://database.ich.org/sites/default/files/E9-R1_Step4_Guideline_2019_1203.pdf

Kennedy, B. K., Berger, S. L., Brunet, A., Campisi, J., Cuervo, A. M., Epel, E. S., et al. (2014). Geroscience: Linking aging to chronic disease. Cell, 159(4), 709–713. https://pmc.ncbi.nlm.nih.gov/articles/PMC4852871/

Kritchevsky, S., & Espeland, M. (2018). Trials of geroscience-based therapeutics: The Targeting Aging with Metformin (TAME) example. Innovation in Aging, 2(Suppl. 1), 823. https://pmc.ncbi.nlm.nih.gov/articles/PMC6230116/

López-Otín, C., Blasco, M. A., Partridge, L., Serrano, M., & Kroemer, G. (2023). Hallmarks of aging: An expanding universe. Cell, 186(2), 243–278. https://doi.org/10.1016/j.cell.2022.11.001

Mazin, A. (2025, August 12). Nir Barzilai: “Positive evidence for metformin is mounting” [Interview]. Lifespan.io. https://lifespan.io/nir-barzilai-positive-evidence-for-metformin-is-mounting/

McDade, J. E., Shepard, C. C., Fraser, D. W., Tsai, T. R., Redus, M. A., & Dowdle, W. R. (1977). Legionnaires’ disease: Isolation of a bacterium and demonstration of its role in other respiratory disease. New England Journal of Medicine, 297(22), 1197–1203.

Romero-Ortuño, R., Martínez-Velilla, N., Sutton, R., Ungar, A., Fedorowski, A., Galvin, R., et al. (2021). Network physiology in aging and frailty: The grand challenge of physiological reserve in older adults. Frontiers in Network Physiology, 1, 712430. https://doi.org/10.3389/fnetp.2021.712430

Royles, D. (2016). Legionnaires’ disease. In The Encyclopedia of Greater Philadelphia. Rutgers University. https://philadelphiaencyclopedia.org/essays/legionnaires-disease/

Ruiz, S., Snyder, L. P., Rotondo, C., et al. (2017). Innovative home visit models associated with reductions in costs, hospitalizations, and emergency department use. Health Affairs, 36(3), 425–432.

Schwartz, A. V., Vittinghoff, E., Bauer, D. C., et al. (2011). Association of BMD and FRAX score with risk of fracture in older adults with type 2 diabetes. JAMA, 305(21), 2184–2192. https://pmc.ncbi.nlm.nih.gov/articles/PMC3287389/

Skivington, K., Matthews, L., Simpson, S. A., Craig, P., Baird, J., Blazeby, J. M., et al. (2021). A new framework for developing and evaluating complex interventions: Update of Medical Research Council guidance. BMJ, 374, n2061. https://pmc.ncbi.nlm.nih.gov/articles/PMC8482308/

Sriprasert, I., Hodis, H. N., Mack, W. J., Rosser, M., Evans, M. L., Xu, X., & Wright, J. D. (2026). Elimination of the black box warning on menopausal hormone therapy. Obstetrics & Gynecology, 147, 642–646. https://doi.org/10.1097/AOG.0000000000006226

Stern, Y. (2002). What is cognitive reserve? Theory and research application of the reserve concept. Journal of the International Neuropsychological Society, 8(3), 448–460. https://doi.org/10.1017/s1355617702813248

Szanton, S. L., Xue, Q.-L., Leff, B., et al. (2019). Effect of a biobehavioral environmental approach on disability among low-income older adults: A randomized clinical trial. JAMA Internal Medicine, 179(2), 204–211.

U.S. Department of Health and Human Services. (2018). Physical Activity Guidelines for Americans (2nd ed.). https://health.gov/sites/default/files/2019-09/Physical_Activity_Guidelines_2nd_edition.pdf

U.S. Food and Drug Administration. (n.d.). Surrogate endpoint resources for drug and biologic development. https://www.fda.gov/drugs/development-resources/surrogate-endpoint-resources-drug-and-biologic-development

U.S. Food and Drug Administration. (2025, December 19). FDA qualifies total hip bone mineral density (BMD) as surrogate endpoint for osteoporosis drug development. https://www.fda.gov/drugs/drug-safety-and-availability/fda-qualifies-total-hip-bone-mineral-density-bmd-surrogate-endpoint-osteoporosis-drug-development

U.S. Food and Drug Administration. (2026). Patient-focused drug development glossary. https://www.fda.gov/drugs/development-approval-process-drugs/patient-focused-drug-development-glossary

World Health Organization. (2016). Multimorbidity (Technical Series on Safer Primary Care). https://www.who.int/publications/i/item/9789241511650

Most of them were old, and many lived alone on the upper floors of apartment buildings. They were more afraid of a stranger breaking the glass than of the air they were breathing. By the end of the week, about 739 more people had died in Chicago than in a normal July week.

Figure 1. Deaths the Cook County medical examiner certified as heat-related, by day, Chicago, July 11 to 27, 1995, with the heat index above them. The count rose from 49 on July 14 to 162 on July 15, two days after the heat peaked. Centers for Disease Control and Prevention, Morbidity and Mortality Weekly Report 44(31), 1995. A United States government work, public domain.

Then the arguing started, and each person who looked at the dead saw a different killer.

The medical examiner, Edmund Donoghue, saw heat. His office certified 485 deaths that week as heat-related. It counted anyone found with a body temperature of at least 105 degrees, or in a home hot enough to explain the death. Epidemiologists from the Centers for Disease Control saw sickness. They interviewed the families and neighbors of 339 of the dead, then did the same for 339 living people matched by age and neighborhood. The people who died were more often already ill, living alone, and rarely out of the house. People with a working air conditioner had about a third of the odds of dying.

A sociologist named Eric Klinenberg saw the street. He spent years walking the West Side and found two neighborhoods, side by side, equally poor, with very different summers. North Lawndale lost 19 people. Little Village, next door, lost 3. Little Village had busy sidewalks and open shops, and its people had reasons to be out and to notice who was missing.

So which one killed them? Each answer is true, and each matches the job of the person giving it. The fourth killer was in every report and in no one’s job title: the locked window. The killer was the four of them together: a sick body, in a hot room, behind a locked window, above a street where no one walked by. “What killed them?” has four true answers, so it cannot be answered as asked.

Sit for a moment in the chair at your next checkup, this month or the one after. The clipboard is on your knee, the pen is on its chain, and whoever drove you there waits by the door. The intake form gives each condition its own line, and each line has its own doctor. Nobody adds the lines up. When I reran the World Health Organization’s numbers, the years a person spends unwell at the end of life came to about 9.3 in the median country in 2021. In the United States, they came to about twelve and a half. Those are the years waiting at the end of yours. Can those nine or twelve years shrink, or can they only move later? The form will not say. The lines stay apart, and so do the doctors.

Researchers call those years the healthspan-lifespan gap: life expectancy minus the years lived in good health. The dead in Chicago were already ill, living alone and rarely out of the house: they were living their gap years when the heat came. Of all the questions in biostatistics, how to narrow that gap is the one I keep coming back to. I want to answer it with my training as a statistician, and to teach myself the biology and medicine it takes. My rerun gave a country’s number, not a person’s. No one can be followed from birth to death with their health measured all along. So each country builds the gap from a table of this year’s death rates by age and this year’s share of each age group in poor health. The table applies those rates to a made-up group of newborns, as if the rates held for a whole life. The gap is real, and no one lives that average life. My question starts where the table stops, inside one life.

A question is answerable when it is stated precisely enough to name what it is asking for. A study that could actually be run can be designed for it. The data can be collected and analyzed. And the result is the thing the question asked about, not a stand-in for it. “How do we narrow the gap?” is none of these. Like the question about the dead in Chicago, it has to be split before any study can touch it. The dead show where to split it. The sick body is one side. The hot room, the locked window and the empty street are the other, and a working air conditioner was the clearest help on that side. A life could have been saved on either side. One woman, met at twenty-five in the first section, carries that split through all four parts of this series. The first thing to settle is why the usual plan for the gap barely moves it.

Delay, Done Whole

In 1990, three researchers asked what would happen to American life expectancy if cancer disappeared overnight. S. Jay Olshansky and his colleagues did the arithmetic. Life expectancy at birth would rise by a little over three years.

Three years sounds small for a cure that has never existed. The reason is that every person faces several causes of death at once, and only one of them can come first. A person spared cancer dies of heart failure or a stroke, a little later, sometimes after the same slow years of decline.

Causes that compete work like a roof with four leaks. Patch the biggest one, and the rain still comes in through the other three. The floor still rots, only a little later. The comparison stops at one point: leaks are separate holes, and diseases are not. They share causes.

Olshansky’s arithmetic leans on the opposite assumption: that the causes of death are independent, so removing one leaves the others as they were. In 1975, the statistician Anastasios Tsiatis showed that this assumption cannot be checked from deaths alone. For any set of causes that share roots, a set of independent causes exists that leaves an identical record of who died of what, and when. Deaths make poor witnesses: each one is counted under a single cause and keeps quiet about the rest. A table of causes of death cannot tell four separate holes from one rotting beam.

That point is where the gap becomes my problem as a statistician. For a classically trained statistician, one of the quintessential questions is how to model dependence: among variables, in networks, across processes over time. Books like Martin Crowder’s Multivariate Survival Analysis and Competing Risks teach the analysis side. What I want to know is what data could do what deaths cannot. Which measures, taken while people are still well, would show the shared causes before any single disease gets its name? So the question has to move from how people die to what can be measured while they are still living.

An earlier essay gave five definitions of a system. Russell Ackoff’s fits a body best: “a system is a whole that cannot be divided into independent parts.” In grade school, the body arrived as a list of systems, and the tests asked about the parts: the four chambers of the heart, the road from mouth to colon. The immune system turns out to be a system in the full sense. The gut does its work with help from bacteria that appear on no chart of the organs. A body is a system of systems, each one changing the others.

Medicine gets better at the parts every year. About two in three people on Medicare, the US health plan that covers most people over 65, live with more than one chronic condition. Each condition comes with its own specialist and its own best drug. Doctors have a name for what can happen next. A side effect of one drug gets mistaken for a new disease, and a second drug is prescribed to treat it. They call it a prescribing cascade. Every prescription is correct for the organ in front of the doctor who wrote it. The person filling the weekly pill organizer, seven little plastic doors snapping shut, is worse off. Making each part better does not have to make the whole better.

Whoever fills a parent’s pill organizer can look for a prescribing cascade. The cue is the intake form at the next checkup, one condition to a line. Does anyone in the room add the lines up? The thing to count is the doctors’ names seen in a year. The question for the pharmacist: could this new pill be treating a side effect of that one? An earlier essay quoted the first of Peter Senge’s laws: “Today’s problems come from yesterday’s ‘solutions.’” Not every problem starts as a fix, but a cascade does. So did the locked windows in Chicago: a fix against burglars that kept the heat in.

Picture a woman of twenty-five. Her periods come every two or three months, then not at all, along with acne and hair where she does not want it. A gynecologist puts her on the pill. A dermatologist treats her skin. A third doctor weighs her, reads 27 on the body mass index chart, and tells her to lose weight. Each does the job well, the way each expert in Chicago did, and none of them names the one condition behind all three complaints: polycystic ovary syndrome, PCOS. Nearly half of women with PCOS see three or more health professionals before anyone names it. A third wait more than two years. Up to seven in ten have never been diagnosed. The name a problem gets says more about who is looking than about the problem.

The years before the name count as morbidity: having a disease or a symptom of one, before anyone has written the disease down. The gap is widest for her: women spend about a quarter more of their lives in poor health than men do. On podcasts, I have heard women with PCOS talk about what that wait did to their mental health. When the name comes, it comes with a prescription for metformin, a diabetes drug, because PCOS makes the body resist its own insulin. At twenty-five she takes the metformin for her periods.

Much of what makes the late years hard never becomes a diagnosis at all: the slow loss of muscle, the tiredness, the hearing that fades at a loud dinner table. A list of diseases has nowhere to put them. They belong to what the organs share. The physician Peter Attia puts four diseases on one list because of what they share. He calls them the Four Horsemen: heart disease, cancer, Alzheimer’s and its relatives, and type 2 diabetes. He writes that “the odds are overwhelming” that one of the four is what a person dies of, and he builds his case for a longer life on delaying them.

The delay has to beat death to count. Push a first heart attack back five years while death moves back just as far, and a person gains five healthy years. The sick stretch at the end stays just as long. In 1980, the physician James Fries named the better outcome the compression of morbidity. It squeezes the sick years into a short stretch at the very end, the way a spring gets pressed into a smaller space (Figure 2).

Figure 2. Delay against compression, on two made-up lives. Drag either curve: a decline squeezed into a short stretch at the end adds healthy years, while a curve that only slides later carries the same sick stretch with it. The curves are hand-drawn, not fit to data. Drawn for this site.

Healthspan is a person’s. The gap in my rerun is a country’s, built from healthy life expectancy. That is the years a newborn could expect to live in full health if today’s death rates and rates of poor health held for its whole life. The World Health Organization builds it by Sullivan’s method, the made-up group above. Lifespan, at that level, is plain life expectancy, that table without the health column. A country’s gap can narrow by compression or by delay, and the table cannot tell which.

Delay aimed at one disease at a time runs into the leaking roof. Attia argues that metabolic trouble feeds the other three. In his account, insulin that stops working well, fat packed around the organs, and the low, steady inflammation they bring all raise the risk of more than one Horseman. The four diseases work like four branches of one tree. Prune one branch, and the other three keep growing. Starve the root, and all four thin at once. The comparison stops here too: a tree has one root, and the four diseases share several, some still unnamed.

No trial has shown, in people, that a treatment aimed at what the four share squeezes the sick years shorter, rather than only sliding them later. Part of the reason is what trials count. A trial that stops at a first heart attack or a death never sees the sick stretch at all. Telling compression from delay means following people for years and counting the healthy years and the sick ones separately. The two counts can move apart: more life with the same sick years, or the same life with fewer. I want to know how to design a study that can tell compression from delay well before its subjects die. Deaths will not say. Only the living can show which.

What the Body Can Do

Two bags of groceries, carried up one flight in a single trip. A chair stood up from without a hand on the armrest. A name found in the crowd at a loud dinner. These are what a body does on its own, and in 2015 the World Health Organization gave the sum of them a name: intrinsic capacity. The WHO splits it five ways: moving, energy, thinking, mood, and the senses.

Capacity is what a body could do. What a person actually does on an ordinary Tuesday, the WHO calls functional ability, “the capabilities that enable all people to be and do what they have reason to value.” It comes from capacity meeting surroundings. At twenty-five, the woman from the first section can draw on her full capacity: the stairs, the chair, the dinner table. The insulin resistance behind her diagnosis does not yet show in anything she does.

That gives the healthspan question its first branch. One way to gain a healthy year is to slow the decline of what a body and mind can do alone (Q1, Figure 3). The other way starts outside the body, and the figure shows both.

The body’s side is the one medicine already works on, and the one the tree’s root belongs to. Strong legs at eighty are part of the root. The thing to look for, in a parent or in oneself, is the task that now takes two tries. The shopping that used to come up in one trip now comes up in two. The chair needs a hand. The cue is the weekly shop, something the reader may be carrying up the stairs right now. Counting the trips is a first measure of capacity, taken years before any diagnosis. Four flights up with no elevator, though, even strong legs may not carry a person to the street.

Opening the Window

Two women have equally weak legs. One lives in a ground-floor flat next to a bus stop and runs her own errands. The other lives four flights up with no elevator and has not been outside in a month.

The two women have the same capacity and very different Tuesdays. No test of their cells could tell them apart: even a map of every layer inside a body stops at its edge, where the organism meets the world around it. The psychologists M. Powell Lawton and Lucille Nahemow drew that edge in 1973. They called what the surroundings demand of a person environmental press. What the person brings to meet it, they called competence: health, functioning, and the people and means they can call on. When the two match, a person lives comfortably and does what they set out to do. They called the match person-environment fit.

The fit works like a window. Glass separates two sides, and the light that gets through depends on both. So the other way to gain a healthy year is to shrink the mismatch between what a person can do and what their surroundings demand (Q2). That means lowering the press until it meets the competence a person still has. Four flights of stairs is press. A bus stop moved two blocks away is press. An air conditioner in a window in July is press lowered by a machine.

Chicago had both sides at once. The CDC study found that the dead were more often ill, the body’s side. It also found they were more often alone and rarely went out, and that a working air conditioner went with lower odds of dying, the surroundings’ side. Families and neighbors answered for the dead, so those answers are secondhand. Klinenberg’s two neighborhoods point the same way, though a comparison of two places cannot prove the street saved anyone.

An earlier essay gave the statistician’s name for an effect of one variable that depends on the level of another: an interaction effect. The two sides of the glass interact. Weak legs cost one woman nothing and cost the other her whole street. In Chicago, the heat did its worst to people who were already ill and alone. Interactions are the part of this question I care about most, and the hardest to study. A study built to find one effect at a time can miss them. In the simplest design, pinning down an interaction as precisely as a single effect of the same size takes about four times as many people. So the two branches of the question cannot be answered one at a time either.

The surroundings’ side is easy to miss on a visit to a parent’s home, because it looks like furniture. The cue is the front door. Standing there, a visitor can count the steps between bed and bathroom and the flights between the door and the street. Then the question: which errand stopped this year, and did the body stop it or the building? A heavy door, a dark stairwell, a window painted shut: each one raises the press without touching a single organ. The decision at the end of that count is a plain one, to move the bed downstairs or to change the stairs.

Both sides of the glass can change. In 1995, Chicago met the heat with a plan on paper and five cooling centers. Few of the people who needed them came, and the warning went out late. After the dead were counted, the city rebuilt the plan around the surroundings’ side: more cooling centers, and daily contact with older residents who lived alone. City workers and volunteers called and knocked on doors. The heat came back four years later, and the third part of this series tells how the new plan did.

Ackoff drew a line between solving a problem and dissolving it. Solving finds the best answer to the problem as it was asked: a drug that pushes back a first heart attack. Dissolving changes the system around the problem until the problem has less room to happen. A knock on the door cures no one’s heart. It makes a locked window less likely to stay locked unnoticed.

“Nobody owns a problem,” Ackoff said. Closing the gap looks like a job for medicine. Half of that job belongs as much to architects, city planners, landlords, neighbors and the people who write insurance rules. Each side is still too big for one study. The next step is deciding who to compare and what to measure. The second part of this series takes it in Philadelphia in 1976, where one choice of comparison solved the case and another had hidden it. The woman of twenty-five goes with it, carrying her metformin: the same cheap pill a trial was designed to test against aging itself.

A Closing Invitation. The locked window in Chicago stood for the sick years themselves: made where a body meets its surroundings, and owned by no one. The gap closes when those years shrink, not when they move later and not one disease at a time: a healthy year is gained by raising what the body can do or lowering what the world asks, and only the living can show which.

1. Before you get up, picture the windows where someone older you love lives. Does one open with one hand, or is it painted shut? Is it locked at night, and for what fear? Which answers would you have to go and look for?
2. This week, write two numbers on one line, for yourself or someone you love: the years lived, and the years since the first symptom that lasted, such as a knee that never healed or blood pressure that stayed high. What share of the first is the second, and has any doctor ever written both?
3. This Saturday, carry the groceries in from the car or the bus stop in one trip, the way you would want to at eighty. Which bag went down first on the stairs, which muscle gave out, and could a parent or an older friend you love make the same trip?

In July 1995, the windows stayed shut, and no one outside knew it was their job to knock. A window opens from either side: a hand on the latch, or a knuckle on the door.

Where This Came From

The piece started from a question my GapYears rerun left open: if the gap is real, who is supposed to close it? It is the gap my research statement is aimed at narrowing. Klinenberg called his book on the heat wave a “social autopsy,” an autopsy done on a city instead of a body, and the phrase is the seed of this series.

Intellectual Honesty Note. The woman of twenty-five and the two women with equally weak legs are hypotheticals; she is invented, and she carries the reader’s stakes through all four parts. “The killer was the four of them together” is this piece’s reading, not any one study’s finding. “Answerable” is the series’ own premise, not a sourced definition. The 9.3 and twelve and a half years are my GapYears rerun of WHO data for 2021.

The PCOS wait and the three professionals come from a survey of 1,385 women recruited through support-group websites, so the sample is self-selected. “Up to seven in ten” is a WHO fact-sheet figure, and the quarter more of life in poor health is a World Economic Forum and McKinsey report, not a peer-reviewed study. The podcasts are my own listening, not a study. The medical examiner’s final count was 485; the CDC’s figure shows the 465 certified by July 27. Klinenberg’s 19 and 3 are counts, not rates. The air-conditioner odds come from a model that adjusts for the other risk factors. Other accounts give the 1995 cooling centers as eleven, not five. The roof and the tree are simplifications.

References

Ackoff, R. L. (2015, November 2). Systems thinking speech by Dr. Russell Ackoff [Video]. YouTube. https://www.youtube.com/watch?v=EbLh7rZ3rhU

Attia, P., with Gifford, B. (2023). Outlive: The Science and Art of Longevity. Harmony.

Buckley, T. D. (2022). A scoping review of psychological sense of community among community-dwelling older adults. International Journal of Environmental Research and Public Health, 19(14), 8395. https://doi.org/10.3390/ijerph19148395

Centers for Disease Control and Prevention. (1995). Heat-related mortality: Chicago, July 1995. Morbidity and Mortality Weekly Report, 44(31), 577–579. https://www.cdc.gov/mmwr/preview/mmwrhtml/00038443.htm

Centers for Disease Control and Prevention. (2003). Heat-related deaths: Chicago, Illinois, 1996–2001, and United States, 1979–1999. Morbidity and Mortality Weekly Report, 52(26). https://www.cdc.gov/mmwr/preview/mmwrhtml/mm5226a2.htm

Centers for Medicare & Medicaid Services. (2012). Chronic conditions among Medicare beneficiaries, chartbook: 2012 edition.

Changnon, S. A., Kunkel, K. E., & Reinke, B. C. (1996). Impacts and responses to the 1995 heat wave: A call to action. Bulletin of the American Meteorological Society, 77(7), 1497–1506.

Chen, Z., Liu, Z., Zeng, L., Huang, L., & Zhang, L. (2023). Research on prescribing cascades: A scoping review. Frontiers in Pharmacology, 14, 1147921. https://doi.org/10.3389/fphar.2023.1147921

Crowder, M. J. (2012). Multivariate Survival Analysis and Competing Risks. Chapman & Hall/CRC.

Fries, J. F. (1980). Aging, natural death, and the compression of morbidity. New England Journal of Medicine, 303(3), 130–135.

Fries, J. F., Bruce, B., & Chakravarty, E. (2011). Compression of morbidity 1980–2011: A focused review of paradigms and progress. Journal of Aging Research, 2011, 261702. https://doi.org/10.4061/2011/261702

Gauran, I. I. (2026). GapYears (Version 1.0.0) [Computer software]. GitHub. https://github.com/gauranii/GapYears

Gelman, A., Hill, J., & Vehtari, A. (2020). Regression and Other Stories. Cambridge University Press.

Gibson-Helm, M., Teede, H., Dunaif, A., & Dokras, A. (2017). Delayed diagnosis and a lack of information associated with dissatisfaction in women with polycystic ovary syndrome. Journal of Clinical Endocrinology and Metabolism, 102(2), 604–612. https://doi.org/10.1210/jc.2016-2963

Kaeberlein, M. (2018). How healthy is the healthspan concept? GeroScience, 40(4), 361–364. https://pmc.ncbi.nlm.nih.gov/articles/PMC6136295/

Klinenberg, E. (2002). Heat Wave: A Social Autopsy of Disaster in Chicago. University of Chicago Press.

Lawton, M. P., & Nahemow, L. (1973). Ecology and the aging process. In C. Eisdorfer & M. P. Lawton (Eds.), The psychology of adult development and aging (pp. 619–674). American Psychological Association.

Lien, L. L., Steggell, C. D., & Iwarsson, S. (2015). Adaptive strategies and person-environment fit among functionally limited older adults aging in place: A mixed methods approach. International Journal of Environmental Research and Public Health, 12(9), 11954–11974. https://doi.org/10.3390/ijerph120911954

National Cancer Institute. (n.d.). Morbidity. In NCI dictionary of cancer terms. Retrieved October 6, 2026, from https://www.cancer.gov/publications/dictionaries/cancer-terms/def/morbidity

Olshansky, S. J., Carnes, B. A., & Cassel, C. (1990). In search of Methuselah: Estimating the upper limits to human longevity. Science, 250(4981), 634–640.

Semenza, J. C., Rubin, C. H., Falter, K. H., Selanikio, J. D., Flanders, W. D., Howe, H. L., & Wilhelm, J. L. (1996). Heat-related deaths during the July 1995 heat wave in Chicago. New England Journal of Medicine, 335(2), 84–90.

Senge, P. M. (1990). The Fifth Discipline: The Art and Practice of the Learning Organization. Doubleday.

Teede, H. J., Tay, C. T., Laven, J. J. E., et al. (2023). Recommendations from the 2023 International Evidence-based Guideline for the Assessment and Management of Polycystic Ovary Syndrome. Journal of Clinical Endocrinology and Metabolism, 108(10), 2447–2469. https://pmc.ncbi.nlm.nih.gov/articles/PMC10505534/

Tsiatis, A. (1975). A nonidentifiability aspect of the problem of competing risks. Proceedings of the National Academy of Sciences, 72(1), 20–22. https://doi.org/10.1073/pnas.72.1.20

Tucker, D., & McCann, R. (2025, July 14). 30 years later: What went wrong in the deadly 1995 Chicago heat wave, and what has changed since. CBS News Chicago. https://www.cbsnews.com/chicago/news/30-years-later-deadly-1995-chicago-heat-wave/

Whitman, S., Good, G., Donoghue, E. R., Benbow, N., Shou, W., & Mou, S. (1997). Mortality in Chicago attributed to the July 1995 heat wave. American Journal of Public Health, 87(9), 1515–1518.

World Economic Forum, & McKinsey Health Institute. (2024). Closing the women’s health gap: A $1 trillion opportunity to improve lives and economies. https://www.weforum.org/publications/closing-the-women-s-health-gap-a-1-trillion-opportunity-to-improve-lives-and-economies

World Health Organization. (n.d.). Global Health Observatory data repository [Data set]. Retrieved August 28, 2026, from https://www.who.int/data/gho

World Health Organization. (n.d.). Healthy life expectancy (HALE). In Global Health Observatory, Indicator Metadata Registry. Retrieved October 8, 2026, from https://www.who.int/data/gho/indicator-metadata-registry/imr-details/66

World Health Organization. (n.d.). Life expectancy. In Global Health Observatory, Indicator Metadata Registry. Retrieved October 8, 2026, from https://www.who.int/data/gho/indicator-metadata-registry/imr-details/65

World Health Organization. (2015). World report on ageing and health.

World Health Organization. (2017). Integrated care for older people: Guidelines on community-level interventions to manage declines in intrinsic capacity. https://www.who.int/publications/i/item/9789241550109

World Health Organization. (2020). Healthy ageing and functional ability [Questions and answers]. https://www.who.int/news-room/questions-and-answers/item/healthy-ageing-and-functional-ability

World Health Organization. (2026, January 22). Polycystic ovary syndrome [Fact sheet]. https://www.who.int/news-room/fact-sheets/detail/polycystic-ovary-syndrome