The post that follows was written as a result of conversations between Jacqueline and me a couple of weeks ago, based on a hypothesis she created, but which I interpreted and used Claude AI to investigate.

I stress that the data sources were checked and that the presentation of data was carefully chosen to minimise the risk of misrepresentation arising from differences in data sources and data collection methods over time. The resulting presentation is, as is made clear, experimental, but the inferences are very clear and important for healthcare debate in the UK at present.

Many perceptions about changes in medical outcomes over time resulting from advances in medical science and practice are challenged by this data, and recent provisional data on declining health spans suggest the situation may be worse than presented here. 

Three centuries of life and health: what the data really shows — and what it means for the politics of care

There is a question I often think about when it comes to the politics of care, and it is this: has the extraordinary extension of human life over the past three centuries made us healthier and, so, better off? The answer, when you look carefully at the evidence, is considerably more complicated and considerably more disturbing than most people assume.

Over the last couple of weeks, I have been working on, amongst other things,  historical data on life expectancy and what demographers call healthspan, which is the number of years a person can expect to live in good health. In the process, I have been seeking to construct a coherent picture of how both have changed since roughly 1700. This post explains what I found, the methodology needed to get there, why that methodology matters, and what the results suggest about the kind of healthcare and the kind of society we have actually built.

I should add that whilst this post has been written by me, the idea for this work and much of the thinking on it was contributed by Jacquleine Murphy, who is a retired GP with numerous postgraduate medical qualifications.

Starting with children

The first methodological decision made was fundamental. When people talk about life expectancy in historical context, including the regularly made claim that people in medieval England lived to only 35 or 40, they are almost always citing a figure that is dominated by child mortality.

In 1841, when reliable English vital statistics begin, around 14 per cent of babies died before their first birthday, and over a quarter of children were dead before the age of five. Those deaths drag the average down so catastrophically that the headline figure tells you almost nothing about how long adults actually lived. The right measure, if you want to understand adult lifespan, is then what demographers call e₅, or life expectancy conditional on surviving to age five.

Strip out under-five mortality, and the lifespan and healthspan pictures change completely. A five-year-old boy in 1841 who had survived the gauntlet of infancy could expect to live to around 60. A five-year-old girl, to around 62. That is not so very different from what the same child might expect at the equivalent point in the early twentieth century. Most of the headline gains in life expectancy between 1841 and 1950 were, then, gains in child survival, not gains in adult longevity.

This matters because it forces a reframing of the available data and the conclusions that might be drawn from it. We have not extended adult life nearly as dramatically as the raw life expectancy figures suggest. What we did, first and most importantly, was stop children dying. That was a genuine and extraordinary achievement, driven by public health, clean water, better public utility engineering and sanitation, vaccination, and better nutrition, and not primarily by curative medicine. But it is a different achievement from extending adult healthspan, and conflating the two distorts everything.

The methodology problem: measuring healthspan

If measuring lifespan is complicated by child mortality, measuring healthspan across three centuries is, as became apparent during our discussions, close to impossible in a strict sense.

The Office for National Statistics (ONS) publishes a formal Healthy Life Expectancy (HLE) series, indicating the number of years spent in self-assessed good or very good health, but has only done so from 2000 onwards for the UK, with a series for Great Britain beginning in 1981. Before that, the concept was not measured systematically.

This creates a methodological challenge for anyone trying to construct a long-run picture. This required working through several different approaches to this issue in an effort to establish longer-term trend data, each of which faced genuine limitations.

The most honest solution, based on trial and error that sought to eliminate reported distortions most likely to arise from data issues than actual changes in lifespan and healthspan, was to use modal age at death as the historical healthspan proxy. The modal age at death is the age at which most adults actually die and reports the peak of the death distribution for those aged ten or over. It has been calculated from the Human Mortality Database for every year from 1841 and has been estimated with reasonable confidence for earlier periods. Crucially, the ONS itself uses modal age at death as its preferred measure of typical lifespan precisely because, unlike the mean, it is not distorted by premature deaths.

The key insight for the purposes of this review was that in an era before modern chronic disease management, serious illness killed relatively quickly. People were largely functional until a relatively short terminal period, and then declined and died quite rapidly. That terminal period, which I am calling the terminal discount, is estimable, albeit only approximately, which constraint I explicitly accept.

In the modern era, the ONS HLE data provides this data: men today typically spend around 16 to 18 years in poor health before death; women around 20 to 22 years, although recent data suggest that these periods might be extending. Working backwards and drawing on everything known about historical morbidity from documented Poor Law records, workhouse admission ages, and occupational health literature, the terminal period in the Victorian era was probably four to six years, rising slowly through the early twentieth century as medicine began to extend survival without always restoring function.

These figures have been used as anchors to calibrate a smoothly rising discount applied consistently to the modal age at death series throughout the full period being examined. Importantly, the modern end is calibrated against ONS HLE data, meaning the methodology is not free-floating. That ONS data became the constraint that informed the historical extrapolation. The result is the creation of a single continuous healthspan series from 1700 to 2023, without the methodological cliff-edge that arises when you attempt to splice the modal age series directly to the ONS HLE series, which began at 1981.

The limitations must be stated clearly

That being said, this approach has real limitations, and I want to be transparent about them.

First, the pre-1841 estimates carry the highest uncertainty. The modal age series before 1841 is itself estimated, based on Wrigley and Schofield's Cambridge Population Group reconstructions and related historical demography. The terminal discount for that period is informed guesswork. It might be grounded in historical evidence, but it is not derived from hard data.

Second, even for the period 1841 to 1981, the modal age at death and the ONS HLE measure are not the same thing. Modal age at death is a mortality measure in that it tells you when most people died. In contrast, the ONS HLE is a morbidity measure, suggesting how people rate their own health. The conversion between them depends on the assumed terminal discount, and that assumption, though calibrated at the modern end, involves judgment about how rapidly the terminal period has lengthened over time. That can only be based on judgment. We may have got that wrong.

Third, it is thought that the ONS HLE series is itself affected by rising health expectations. There is strong evidence that people today report themselves as in poor health at thresholds that previous generations would have considered normal ageing. A 72-year-old in 1960 who could walk a mile and tend a garden would very likely have called themselves healthy. Today, the same functional capacity alongside managed hypertension, a replaced hip, and a repeat prescription for statins might well generate a self-report of limiting long-term illness. This means the modern ONS figures most probably overstates the genuine deterioration in health relative to historical norms. The true healthspan picture is almost certainly somewhat better than the ONS numbers alone suggest,  but the direction of the story is not changed by this caveat.

What the data shows

With those limitations clearly stated, the picture that emerges is striking, and consistent across a wide range of reasonable assumptions about the terminal discount.

Healthspan, estimated as modal age at death minus the terminal discount, has risen over three centuries, but much more slowly than total lifespan.

In 1841, the typical childhood survivor could expect around 55 to 65 years of reasonably good health.

Today, the equivalent figure is around 62 to 72 years, depending on sex and the assumptions used. That is a meaningful gain: perhaps eight to ten years of additional healthy life over 180 years.

But total lifespan for childhood survivors has risen by around 24 years over the same period. The gap between lifespan and healthspan, the years lived in poor health, has roughly tripled, from around five years in the Victorian era to around 15 to 20 years today.

The period since 1981, where the ONS data is hard rather than estimated, is particularly revealing. Lifespan continued to rise steadily until the post-2010 slowdown. But healthy life expectancy barely moved and has recently fallen.

Men today spend more years in poor health than at any point in the recorded series.

Women, who live longer, spend a quite extraordinary proportion of their additional years in poor health; around a quarter of their total lifespan.

Rockefeller medicine and the machine model of health

This outcome is not an accident. It is the logical result of what I would call the Rockefeller model of medicine, the system that emerged from the 1910 Flexner Report, funded by Rockefeller and Carnegie interests, which systematically privileged pharmaceutical and procedural intervention over prevention, nutrition, and whole-person care. That model has been extraordinarily successful at one thing: keeping bodies alive. It has been considerably less successful at keeping people well.

The distinction matters enormously. Keeping a body alive is an engineering problem. You identify the failing component, intervene to repair or compensate for it, and extend the machine's operational life. That is the logic of the bypass operation, the statin prescription, the dialysis unit, the chemotherapy protocol. Each intervention, individually, can be and often is genuinely beneficial. Taken together, across a health system organised around this logic, they produce a population that lives longer but spends an increasing proportion of its additional years managing chronic conditions, attending outpatient clinics, taking multiple medications, and experiencing its body as something that requires constant maintenance rather than as the vehicle of a flourishing life.

The pharmaceutical industry has a direct financial interest in this outcome. A patient who is kept alive with managed but not resolved chronic disease is a patient who generates a lifetime of repeat prescriptions. A patient who is genuinely healthy does not. The incentive structure of pharmaceutical-driven procedure-based medicine is not aligned with producing health; it is aligned with producing managed morbidity at scale. This is not a conspiracy theory. It is simply what you would expect from a profit-maximising industry operating in a system where the commodity sold is treatment rather than health.

What a politics of care would look like instead

The politics of care starts from a different premise entirely. It starts from the recognition that people are not machines. We do not want to be kept going; we want to live well. We want to be treated as whole people with relationships, purposes, and the capacity for meaning, not as biological systems requiring maintenance. Health, in this understanding, is not the absence of diagnosable pathology. It is the capacity to do the things that make life worth living; to work, to love, to contribute, to rest, to age with dignity.

If you take that seriously, the three-century data I have described here should be genuinely alarming. We have nearly eliminated early death, which is good news. We have substantially extended adult lifespan, which is remarkable. But we have failed to extend healthy adult life in proportion, and in recent decades we have actually been going backwards on the measure that matters most. The additional years being added at the end of life are disproportionately years of poor health, dependence, and managed decline.

That is not a triumph as most people would understand it. It is a failure of a particular kind of medical model to serve human flourishing. There is, however, an important exception to note. From the point of view of the neoliberal GDP maximising economy, this might be viewed as a success. Just as a person's health has been extracted from them and they cease to be an economic unit of production, they become a generator of additional healthcare consumption, driving GDP upward over many years. They become an object on which the medical industrial complex can prey for the rest of their lives. In that model, there is no incentive for this situation to change: the outcome meets the neoliberal economy's desire.

A politics of care would reorganise health systems around the question: what enables people to live well throughout their lives? The answer points away from late-stage pharmaceutical intervention and toward early investment in the social determinants of health, which include decent housing, good food, clean environments, meaningful work, strong social connections, and genuine mental health support. It points toward funding prevention rather than treatment, community health rather than acute hospital care, and continuity of relationship with known carers rather than the episodic encounters of a system designed around procedures.

It also points toward a different understanding of what we owe each other. The choice is not between a long life and a short one. It is between a long life with twenty years of managed decline at the end, and a somewhat shorter but fundamentally healthier one, or, if we organise ourselves better, a long life that is also, for most of it, a healthy one. The data suggests we have not yet found a way to deliver that. Understanding why and who benefits from our failure to do so is at the heart of what the politics of care is about.

I stress that the methodology used here is approximate, but the direction of the story is not in doubt, and it matters for everything we think about health, medicine, and the kind of society we are trying to build.

PDF of article