A personal health index based on the International Classification of Functioning, Disability and Health has shown early potential to combine different health measurements into a standardised score. A preliminary evaluation published in JMIR Human Factors used rehabilitation data from 505 people in Finland and compared the index with self-rated health and pain measures. The model follows the classification’s hierarchy and does not need training on a large population dataset. It can produce both an overall score and a detailed health profile from incomplete and varied information, while giving more weight to recent measurements and to data linked more reliably to the classification.
A Standardised Model for Different Health Data
Personal health indices reduce overall health to a single value, while health profiles show several scores for different areas. Many existing methods rely on fixed questionnaires, selected variables or population-based calculations. This can make them harder to use when clinics collect different types of data or follow different assessment practices.
Must Read: LLM Embeddings Match EHR Foundation Models
The new model uses the International Classification of Functioning, Disability and Health as a shared structure. The classification covers body functions, body structures, activities and participation and environmental factors through a hierarchy of categories. Questionnaire answers and clinical measurements are first converted into classification codes.
A recursive algorithm then moves through the hierarchy from the most detailed categories towards the top. At each level, it combines the available measurements. More recent information can receive greater weight, while measurements linked through less reliable procedures can receive less weight. The final raw value is converted to a scale from zero, representing the poorest health, to 100, representing the best health.
The model does not need information for every category. It can still calculate a result when data is missing and can produce both one overall score and separate scores for individual health areas. A coverage measure can also show how much information supports the result. This supports its use across datasets that do not contain the same measurements.
Early Results Show Predictable Performance
The preliminary evaluation used data collected between 2013 and 2019 from one private rehabilitation clinic in Helsinki. Participants received treatment for back, neck, hip, knee, shoulder, general health or other problems. Available data included disability and general health questionnaires, mobility and strength tests and pain ratings for different parts of the body. The amount and duration of information differed considerably between individuals.
The index was compared with self-assessed health scores in groups that met minimum requirements for treatment duration and number of measurement days. Across the different time-weighting settings, the index showed moderate positive correlations with self-rated health. All reported correlations were statistically significant. The strongest results appeared when older measurements still contributed but had less influence than newer ones.
Maximum pain trajectories were also compared with health index trajectories for individual participants. The correlations were negative, meaning that higher pain levels were linked with lower index values. The proportion of significant correlations generally increased when treatment sequences were longer and included more data. Pain ratings were also used as inputs to the index, so some relationship between the two measures was expected.
Sensitivity testing showed that the model behaved predictably. Greater problem severity reduced the score, recent measurements had more influence than older ones and values closer to the top of the classification hierarchy had a larger effect.
Wider Validation Is Still Needed
The index could support rehabilitation planning and progress monitoring by showing a person’s condition before and after treatment. A detailed profile could identify areas that improve and areas that continue to need attention. The proposed display combines the overall score with results for separate health areas and shows how much data are available for each one.
The model could also be used in electronic health records or patient portals. Possible applications include routine assessment in primary care, monitoring functional decline in older people and tracking functional outcomes in mental health. Standardised scores could also support comparisons between organisations and provide common variables or targets for machine learning applications.
Several limits remain. The current version uses only the first general qualifier in the classification. It does not include environmental facilitators and cannot use qualifiers for unspecified or non-applicable information. Sparse data could also make the overall score appear more complete than it is unless the coverage measure is considered.
The validation dataset came from one clinic and did not cover all areas of functioning. Several measurements were linked to classification codes through expert judgement rather than formally validated procedures. Children were not analysed separately from adults. Two contributors were affiliated with the clinic operator, which partly funded one contributor’s work, although no conflicts of interest were declared. Wider validation is still required before routine clinical use.
The International Classification of Functioning, Disability and Health offers a structured way to combine different measurements into one health index and a more detailed profile. Early results show clear links with self-rated health and pain, predictable responses to changes in input values and continued calculation when information is missing. The model may support health monitoring, rehabilitation assessment and more standardised use of health data. However, the evidence is still limited by the single-clinic dataset, incomplete classification coverage and expert-derived data linkages. Larger datasets, validated mappings and further clinical testing are needed before the approach can be used confidently in routine care.
Source: JMIR Human Factors
Image Credit: iStock
References:
Rautiainen I, Parviainen L, Jakoaho V et al. (2026) Monitoring Health Status: Development and Preliminary Validation of a Personal Health Index Using the International Classification of Functioning, Disability and Health. JMIR Hum Factors;13:e84802.
