Comparing theoretical and real complex chronic populations: cross-sectional study

Abstract

Introduction: Population stratification systems are based, primarily, on burden of disease and pattern of use of resources. They do not include other variables that, according to available evidence, determine the complexity, especially environmental and social circumstances. In Catalonia, a strategy for proactive identification of complex chronic patients (CCP) by healthcare professionals in primary care, based on clinical judgment and broad criteria, encompassing adverse psychosocial situations and potential benefit from integrated care strategies, has been promoted by the Chronicity Prevention and Care Program (CPCP) from the Department of Health. The aim of this study is to compare the characteristics of the theoretical complex chronic population (TP) identified by a theoretical construct based on a population stratification system (Clinical Risk Groups, CRG) with the characteristics of the real complex chronic population (RP) identified by professionals according to clinical judgment and broad criteria promoted by CPCP. Methods: Information about CRG, demographic variables, presence of chronic diseases, use of health services and expenditure on pharmaceutical prescriptions was collected from the population morbidity database of Catalonia in 2013 (7.753.482 inhabitants). Identification of CCP was retrieved from the Shared Clinical Record of Catalonia, a common technological platform accessible to all providers of public health network. TP is that which fulfilled the theoretical construct developed by expert consensus based on population stratification variables (CRG 5 severity 6; CRG 6 severities 5-7; CRG 7 severities 2-6; CRG 8 severities 3-6; CRG 9 severities 2-6). RP is that which was recorded by professionals as CCP in the Shared Clinical Record of Catalonia on 31st December 2013. Specific criteria for identification as CCP were not available. For comparisons, odds ratio (OR) of association (and 99.9% confidence interval) were calculated using logistic regression models. Weighted coefficient of variation, percentage of explained variability and discrimination (area under the ROC curve) were also calculated.