Living With More Than One Healthspan Number
This series has followed one question through three measurement systems: why can “healthy years” not be treated as a single natural quantity? The subject is the family of indicators—HALE, HLY, and Dis-FLE—and the final task is to choose, compare, and communicate them without dropping the health state, evidence, estimand, or population that gives each number meaning. The answer is not that numbers are useless. It is that a useful healthspan number is always a qualified answer to a specified question.
One subject, several valid measurements
Healthspan is a convenient name for a concern that contains several dimensions. People want to know not only how long populations live, but how much of that time is lived with functional capacity, without serious disease, or without limitations that affect ordinary activity. Those dimensions are related, but they do not collapse into one universally accepted state called health.
The first dimension is the health construct: what the measure calls healthy and what counts as loss. HALE treats health as a weighted summary of functional-health loss across causes. HLY treats it as the absence of long-term activity limitation reported through a survey instrument. Dis-FLE treats it as remaining free of new events from a selected set of severe conditions recorded in hospital trajectories. The difference is not a choice between truth and error. It is a choice about which aspect of health the estimate is designed to represent.
The second dimension is the evidence stream. HALE uses a life-table mortality framework together with information on disease, injury, disability, or functional-health loss. HLY uses self-evaluated responses about long-term activity limitation. Dis-FLE uses discharge dates, coded diagnoses, and individual follow-up in the PMSI hospital database. Each stream sees something the others may miss. Each also has a characteristic blind spot: incomplete mortality or morbidity data, variation in self-report, hospital access, coding practices, and clinical records that do not fully describe lived function.
The third dimension is the estimand, the quantity the calculation is trying to produce. HALE sums equivalent healthy years after a loss fraction is applied to life-table years and divides by survivors at an age. Dis-FLE estimates the area under a disease-free survival curve, conditional on reaching an age without the defined event. HLY is also an expected duration, but its health state comes from reported activity limitation rather than the clinical event transition used by Dis-FLE. “Years” names the unit; it does not identify the estimand.
The fourth dimension is the target population. A number can describe the general population, a sex-specific population, people who have reached a particular age, or a selected group visible in a health system. The French study shows why this matters. Its PMSI data contain millions of people, but the multivariable Cox results remain relative to the hospitalized population. The sex-specific Dis-FLE comparison with HLY uses an additional adjustment that assumes people absent from PMSI are alive and healthy with respect to the study definition. That adjustment is an explicit bridge, not a direct observation of everyone outside the hospital system.
These dimensions interact. A new threshold changes the health construct. A new survey or record system changes what can be observed. A new conditioning rule changes who is averaged. A new population target changes the denominator. A result is therefore not produced by the formula alone. It is produced by a chain of design decisions, data, and assumptions.
This is also why the word “healthy” should not be treated as an empty adjective. In HALE, health loss is represented as a fraction of lived time. In HLY, the relevant question is whether people report long-term activity limitation. In Dis-FLE, the relevant transition is a new severe clinical event. A year may be counted differently by each measure without any arithmetic inconsistency. The measures are describing different projections of the same population’s health.
The distinction protects against two opposite mistakes.