When the Hospital Record Becomes a Health Measure

The French study’s subject is a health indicator built from the PMSI national hospital-discharge database, and its specific question is what becomes measurable when clinical records are used to represent healthy life. Understanding that design helps the reader judge both sides of the result: hospital data can reveal dated, repeated trajectories at a scale surveys rarely reach, but the resulting “disease-free” state is still bounded by diagnosis, admission, coding, and access to care.

A record system repurposed for population health

A hospital-discharge database is not originally a diary of wellbeing. It is a record of encounters with a healthcare system. The study turns that administrative and clinical trail into a population-health instrument by connecting several kinds of information: who the patient is, when a hospital stay occurred, what diagnosis was coded, which severe event that diagnosis represents, and whether the person remains free of a defined event during follow-up.

That connection changes the unit of observation. A small survey may ask a representative sample how health limits usual activity at a particular time. The PMSI data contain repeated discharge dates and associated diagnoses for individual patients. A researcher can therefore follow a trajectory rather than rely on one person’s memory or a single response. The record can show when a severe event first appears in the study’s coding system and how long the patient was observed before that event, death, or the end of follow-up.

The study uses hospital visits in Metropolitan France from 2010 to 2013 for adults aged 50 and older. In the age group covered by the source data, more than 75% of the general population appeared in the database at some point. That is substantial coverage, but it is not the same as a random sample of everyone. Appearing in PMSI means that a person was visible to the hospital system, not that they had the same chance of appearing as someone with no hospital contact. The distinction will later determine how the authors interpret population adjustment and model results.

Before the main observation period, the researchers use information from 2008 and 2009 to exclude people who had already been observed with the severe conditions used to define the healthy population. They also exclude certain other chronic, neurological, infectious, genetic, and mental conditions. After these exclusions and further data preparation, the analysis contains almost 30 million hospital visits and more than 10 million unique individuals—10,610,629 in the study’s descriptive table.

That sequence creates a selected starting state. The people entering the analysis are not simply “all adults over 50.” They are people who were visible in the hospital data, had no recorded study-defined severe condition in the prior screening period, and met the additional exclusion rules. The distinction matters because a large sample can be precise about a selected group while still requiring assumptions before its results describe a wider population.

The records also contain more than diagnoses. The study uses year of birth, sex, approximate place of residence, and risk factors inferred from hospital data and previous diagnoses. Tobacco use, alcohol use disorders, and obesity are categorized from coded clinical information rather than from direct self-report. Education and immigration are represented by characteristics of the patient’s commune of residence, not by an individual’s own education or immigration status. These variables make stratified analysis possible, but each one carries the meaning and limitations of the way it was recorded.