Why Healthspan Is Not One Number
Healthspan is often spoken of as if it were the healthy counterpart of life expectancy: one number that tells us how long people live well. But the primary subject of this series is not a hidden quantity waiting to be discovered. It is a family of healthy-life indicators—HALE, HLY, and Disease-Free Life Expectancy—and the relationships among the definitions, data sources, and models that produce them. The central question is simple but consequential: when a report says that people can expect a certain number of healthy years, what exactly has been counted?
That question matters because “years” creates an illusion of direct comparability. A year without a severe hospital-recorded condition is not necessarily a year without pain, a year without a long-term activity limitation, or a year experienced as full health. Those states can overlap, but they are not identical. A number can be carefully calculated and still be the wrong number for the question a reader wants to answer.
The question hidden inside “healthy years”
Life expectancy has a relatively clear endpoint: death. Health is not so tidy. It can refer to the ability to perform daily activities, the presence or absence of disease, functional capacity, perceived limitations, or a weighted combination of health losses. Each choice changes the boundary between a year counted as healthy and a year counted as less healthy.
This is why Healthy Life Expectancy, or HLE, is best understood as an umbrella term. It covers expected durations lived in specified health states rather than one universally agreed measure. The lack of a universal definition is not a minor technical inconvenience. It means that disagreement between healthspan estimates may begin before researchers choose a formula. They may be observing different aspects of the same population.
The distinction is easy to miss when all the results are reported in years. A measure of long-term activity limitation asks something about what people can do or perceive themselves as able to do. A measure based on clinical diagnoses asks whether specified events have appeared in a medical record. A weighted measure asks how much time is lost, in health terms, when people live with disease or injury. These are not three ways of writing the same fact. They are three ways of constructing a population-level question about health.
The practical implication is not that one definition is automatically superior. It is that the definition must fit the decision. A public-health system tracking broad functional experience may need a measure that is understandable across countries and years. A clinical-data study may need a measure that can follow severe events through individual records. A broad population summary may need to combine mortality with graded health loss. The useful question is therefore not “Which number is the real healthspan?” but “Which health state and evidence are relevant to this question?”
Three answers to the same-sounding question
The three indicators at the center of this series make the contrast concrete.
HALE—Healthy Life Expectancy as defined in the WHO metadata—estimates the average remaining years a person can expect to live in “full health” at a given age, taking account of years lived in less than full health because of disease or injury. It uses the Sullivan method to combine life-table years with a measure of functional health loss. In broad terms, years lived are discounted according to the fraction of healthy life considered lost, and the resulting healthy years are averaged across survivors. HALE is therefore a weighted summary across causes, not a simple tally of years during which no diagnosis was recorded.
HLY, or Healthy Life Years, uses another boundary. In the French study’s comparison, Eurostat’s HLY represents expected years without long-term activity limitation, measured through the Global Activity Limitation Instrument in a survey. Its strength is a relatively direct focus on perceived ability and a design intended to support wide coverage and comparisons across countries and time. Its dependence on self-evaluation also matters: responses can reflect how people understand health, activity, age, and limitation, not only the clinical conditions in their bodies.
Dis-FLE, the Disease-Free Life Expectancy developed in the French hospital-discharge study, makes a third choice. It treats remaining free of a defined set of severe events as the relevant healthy state. The study uses hospital diagnoses and patient trajectories rather than asking people to rate their limitations. Its “disease-free” state is not the absence of every illness. It is the absence of new events considered severe enough to be incompatible with good health in the study’s operational definition.
The difference can be represented without ranking the measures:
The clinical study’s definition is deliberately broad within its own data system. It includes severe conditions associated with disability or death rather than following only one disease. That makes the resulting trajectory closer to a general picture of serious deterioration than a single-condition measure would be. Yet it remains a clinical construction. A diagnosis may have different effects on different people, and a condition prevented before hospital admission may never appear in the data. The measure can be useful precisely because it is specific; that specificity is also its boundary.
The same reasoning applies to HALE and HLY. A weighted functional-health estimate and a self-reported activity estimate may both be valuable, but they do not become equivalent merely because both are expressed as expected years. Their differences can reveal something about health: people may report activity limitations that are not captured by severe diagnoses, while clinical records may reveal serious events that surveys measure differently. But interpreting that divergence requires knowing what each instrument can see.
How to read a healthspan number
A healthspan estimate should be read as a compact answer to four prior questions.
First: what health state is being counted? “Full health,” “without long-term activity limitation,” and “without a new severe clinical event” are meaningful phrases, but they set different thresholds. A threshold determines who crosses from one state to another. A weighting rule determines whether all losses count equally. Without this information, the label “healthy” carries more confidence than the measure itself warrants.
Second: what evidence observes that state? WHO’s HALE metadata points to vital-registration data for mortality and health examination surveys for diseases, injuries, and disabilities, while also noting possible alternatives such as sample registration and verbal autopsy. The French study uses standardized hospital diagnoses, repeated discharge dates, and individual follow-up. Surveys, life tables, and clinical records each provide information that the others may not. They also introduce different missingness, comparability, and access problems.
Third: what quantity has actually been estimated? An average remaining duration for a population is not the same as a conditional duration among people who have reached an age without a specified event. A weighted health-loss summary is not the same as the area under a disease-free survival curve. Later in this series, the clinical study’s Dis-FLE will show why censoring, age, and the observed risk set matter. For now, the essential point is that “expected years” describes an estimand, not merely a unit.
Fourth: who does the number describe? A national estimate, a sex-specific estimate, a hospitalized cohort, and a model-based risk profile are not interchangeable populations. The French study had more than ten million unique people after exclusions, but its data came from people observed in the hospital system. Its authors therefore had to make an explicit adjustment before comparing a sex-specific clinical measure with a general-population indicator. The multivariable clinical model, by contrast, remains interpretable as a model of relative risk within the hospitalized population. Scale does not erase selection.
These questions also protect against a common overreading: treating a population association as a personal forecast or an intervention effect. If a model finds that a coded risk factor is associated with earlier severe events, it has identified a pattern under particular definitions and assumptions. It has not automatically shown how many years a specific person would gain by changing one behavior, nor isolated the mechanism behind a geographic difference. The number is evidence about a measured relationship; it is not a complete explanation of health.
That discipline gives healthspan measures their real value. They are not competing attempts to produce a single scoreboard. They are instruments built for different views of the same broad subject. HALE can summarize weighted functional-health loss across causes. HLY can describe years without long-term activity limitation as people report it. Dis-FLE can expose how severe clinical events accumulate along observed trajectories. Used together, their agreement can strengthen a broad pattern, while their disagreement can identify a difference between lived function, clinical diagnosis, and the assumptions used to represent the population.
The rest of this series follows that difference rather than smoothing it away. The goal is not to replace one misleading number with three numbers that cannot be understood. It is to make the construction visible: define the health state, inspect the evidence, identify the estimand, locate the population, and then decide what comparison the result can support. Once those conditions travel with the number, “healthspan” becomes less like a promise of one definitive score and more like a precise question about how people live.