Impact of health services delivery on hospital admission utilization in Brazil

OBJETIVO: Analisar o uso de servi??os hospitalares no Brasil, incorporando informa????es sobre a oferta de servi??os de sa??de. M??TODOS: Foram analisados dados da Pesquisa Nacional por Amostra de Domic??lios, realizada pelo Instituto Brasileiro de Geografia e Estat??stica (IBGE) em 1998, e tamb??m de outras fontes. Foram utilizados modelos hier??rquicos, sendo a pessoa o primeiro n??vel e a unidade da federa????o de resid??ncia, o segundo. Dois modelos foram ajustados separadamente para adultos e crian??as: regress??o log??stica, para modelar ter ou n??o ter tido uma admiss??o, e regress??o de Poisson, para modelar o n??mero das admiss??es, sendo que o ??ltimo modelo considerou apenas pessoas que tiveram pelo menos uma admiss??o. RESULTADOS: O principal fator associado ??s admiss??es hospitalares foi necessidade da sa??de. As pessoas de menor renda tiveram maior chance de serem internadas, quando controlado por necessidade de sa??de e fatores capacitantes (cobertura por seguro da sa??de e exist??ncia de servi??o de sa??de de uso regular). Somente de 1 a 3% da varia????o da utiliza????o das admiss??es hospitalares foi atribu??do ??s diferen??as na oferta de servi??os de sa??de no n??vel da unidade da federa????o. Nos modelos log??sticos, o n??mero de leitos hospitalares foi positivamente associado e o de m??dicos per capita negativamente associado ?? chance de admiss??o. Nos modelos de Poisson, nenhuma das vari??veis de oferta foi associada ?? chance de admiss??o. CONCLUS??O: Os resultados sugerem a exist??ncia de efeito de demanda induzida pela oferta no caso dos leitos hospitalares. A associa????o inversa das admiss??es hospitalares com o n??mero de m??dicos indica a influ??ncia do cuidado ambulatorial no uso do hospital.Objective To assess hospital services utilization in Brazil incorporating information about health services delivery. Methods Data from the National Household Survey carried out by Brazilian Institute of Statistics and Geography (IBGE) in 1998 and from other sources were collected. Hierarchical models having the individual in the first level and the state of residence in the second level were used. Two models were separately adjusted for adults and children: logistic regression was used when to have been or not admitted was the response variable, and Poisson???s regression was used when the number of admissions was the response variable. This last model was adjusted only for those individuals who had been admitted at least once. Results The main determinant of hospital admissions was health need. Poor people were more likely to be admitted when controlling for health needs and enabling factors (health insurance coverage and regular health service). Only 1 to 3% of the variability in hospital admission utilization could be attributed to differences in services delivery at the state level. In the logistic models, the number of beds was positively associated and the number of doctors was negatively associated with the likelihood of admission. Poisson???s models did not show any delivery variables associated with the likelihood of admission. Conclusions These results suggest a delivery-induced demand impact concerning hospital beds. The inverse association of the number of doctors and hospital admissions suggests the impact of outpatient care on hospital utilization