Introduction: Due to the ongoing demographic and epidemiological changes, today stakeholders need to have available information, based on population stratification, in order to plan the most suitable organizational model to meet the population health needs.
Methods: The legally adult population assisted and resident in Lazio Region on 31/12/2019 was equally and casually divided in two samples: the training sample (to define the model) and the validation sample (to measure model performances). On the base of the more complex model of Lazio Region, three population strata were defined: multi-chronic population, multi-chronic population with a high clinical complexity, multi-chronic population with socioeconomic vulnerability. Hospital discharge records were identified in the previous five years prior to 31/12/2019. Through appropriate classification models, it was evaluated the level with which the simplified system "from SDO" is able to approximate the more complex algorithm developed by the Lazio Region.
Results: Model performances, which has examined only information "from SDO", results inadequate. In fact, the Positive Predictive Value (PVV) results equal to 46.3%, 16.3% and 30.3%, respectively for the three analyzed strata.
Discussion: This study demonstrated that using the hospital system as the only Health Information System reduces the possibility to stratify and predict population health needs. For this reason, the improvement of the completeness and of the quality of data from different social and health information systems and their interconnection, represent the essential starting point. In addition, updating record track is needed to simplify the workload of compilers and improve data availability. These actions must also be accompanied by supportive interventions by central bodies for the regions that show the greatest weaknesses, as well as training actions that improve the level of knowledge of compilers.