Salud Pública de México
Print version ISSN 0036-3634
SALINAS-RODRIGUEZ, Aarón; MANRIQUE-ESPINOZA, Betty and SOSA-RUBI, Sandra G. Statistical analysis for count data: use of healthcare services applications. Salud pública Méx [online]. 2009, vol.51, n.5, pp.397-406. ISSN 0036-3634. http://dx.doi.org/10.1590/S0036-36342009000500007.
OBJECTIVE: To describe some of the statistical models for the study of count variables in the context of the use of health services. MATERIAL AND METHODS: We used the Seguro Popular Evaluation Survey to estimate the effect of Seguro Popular on the frequency of use of outpatient health services, using Poisson regression models and negative binomial, zero-inflated negative binomial and the hurdle negative binomial models. We used the Akaike Information Criterion (AIC) to define the best model. RESULTS: Results show that the best statistical approach to model the use of health services is the hurdle model, taking into account both the main theoretical assumptions and the statistical results of the AIC. DISCUSSION: The modelling of count data requires the application of statistical models to model data dispersion; in the presence of an excess of zeros, the hurdle model is an appropriate statistical option.
Keywords : health services; statistical models; statistics; Mexico.