Comparación de métodos para modelizar los factores asociados con los ingresos hospitalarios en casos incidentes de insuficiencia cardíaca

Abstract

OBJECTIVE: Heart failure (HF) is an important public health problem due to its increasing prevalence, and the decompensation associated with hospital admission represents an increased risk of death. The objective of this study was to compare several methods to model the variable hospitalizations and to determine the effect of factors associated with hospital admissions in incident cases of HF. METHODS: Study of a retrospective cohort of patients with information extracted from electronic medical records of PC was performed. Patients 24 year and older with at least 1 visit to PC in 2006 were included. Registered hospital admissions of HF incident cases between 2006 and 2010 or until death were analyzed and comparison of Poisson, Negative Binomial (NB), zero-inflated and Hurdle regression models were conducted to identify factors associated con hospitalizations. RESULTS: 3,061 patients were identified in a cohort of 227,984. Regarding the factors associated with hospitalizations and according to the zero inflated NB regression model, patients who presented valvular disease (OR=2.01; CI95% 1.22-3.30), or were being treated with antithrombotics (OR=3.45; CI95%: 1.61-7.42) or diuretics (OR=2.28; CI95% 1.13-4.58) had a lower likelihood of hospitalization. Factors associated with a higher rate of hospital admissions were having valvular disease (IRR=1.37; CI95% 1.03-1.81) or diabetes (IRR=1.38; 1.07-1.78), and being treated with calcium antagonists (IRR=1.35; CI95% 1.05- 1.73) or ACE inhibitors (IRR=1.43; CI95% 1.06- 1.92). Having being referred to a cardiologist had a protective effect (IRR=0.86; CI95% 0.76- 0.97). CONCLUSIONS: The regression model that obtained the best adjustment was the zero inflated NB. According to this model, the factors associated with an increase in hospital admissions were valvulopathies, diabetes and treatment with calcium antagonists.