papers

Development of a predictive risk model for unplanned admissions in the Basque Country

Rev Esp Salud Publica 88:251-260. Published in Spanish.

Centro de Salud de Astrabudua,.Osakidetza (Servicio Vasco de Salud), Erandio. Bizkaia, Spain; Instituto Vasco de Innovación Sanitaria. Sondika. Bizkaia, Spain

BACKGROUND: Hospitalizations are undesirable events that can be avoided to some degree through proactive interventions. The objective of this study is to determine the capability of models based on Adjusted Clinical Groups (ACG), in our milieu, to identify patients who will present unplanned admissions in the following months to their classification, in both the general population and in subpopulations of chronically ill patients (diabetes mellitus, chronic obstructive pulmonary disease and heart failure).

METHODS: Cross-sectional study which analyzes data from a two year period, of all residents over 14 years old in the Basque Country (N = 1,964,337). Data from the first year (demographic, deprivation index, diagnoses, prescriptions, procedures, admissions and other contacts with the health service) were used to construct the independent variables; hospitalizations of the second year, the dependent ones. We used the area under the ROC curve (AUC) to evaluate the capability of the models to discriminate patients with hospitalizations and calculated the positive predictive value and sensitivity of different cutoffs.

RESULTS: In the general population, models for predicting admission at 6 and 12 months, as well as long-term hospitalizations showed a good performance (AUC> 0.8), while it was acceptable (AUC 0.7 to 0.8) in the groups of chronic patients.

CONCLUSION: A hospitalization risk stratification system, based on ACG, is valid and applicable in our milieu. These models allow classifying the patients on a scale of high to low risk, which makes possible the implementation of the most expensive preventive interventions to only a small subset of patients, while other less intensive ones can be provided to larger groups.

PMID: 24914863