OBJECTIVES: Adequate management and intervention of health risk requires proper technical tools to correctly identify and classify risks at individual level. In this work, a methodology to calculate an individual health score as an identification and classification tool, using health administrative information coming from both sociodemographic and claims data, was developed.

METHODS: A database of 168.297 members, containing sociodemographic information, as well as comorbidities and claims data, was used. A representative sample of 200 members was randomly selected and evaluated by 20 physicians, who were asked to assign a score ranging from 0 to 10 according to the health status of each member based on the provided information. To estimate the weights for each variable in the score, a linear regression of the score assigned by the physicians on the potential variables forming the score, was performed. Finally, the score was applied to the 168.297 members (build as a linear combination of the selected variables and the estimated weights) and the results were validated by regressing the total cost percentiles on their respective score quantiles in the prior year.

RESULTS: The analysis showed that the selected variables in the linear regression model explain the 53% of variability of the physician’s scores in the sample. Moreover, in the validation approach, the score measured in January 31^st, 2016 (after mapping the entire population) explained the 89% of variability of the percentiles of the total cumulated cost in the period February to December, 2016.

CONCLUSIONS: The score obtained through this methodology showed coherent and intuitive results. Hence, it can be used as a suitable tool for identifying and classifying health risks, at individual level. This new approach could be also useful to understand, how health risk influences financial risk of health insurers, by exploiting the ability of the score to predict future costs.