PROJECTING UTILIZATION OF HOSPITAL INPATIENT DAYS IN THE NETHERLANDS - A TIME-SERIES ANALYSIS

The object was to model and project utilization of hospital in-patient days for selected diseases in The Netherlands. We used sex- and age-s pecific standardized monthly utilization of hospital in-patient days d uring 1980-90 for lung cancer, diabetes, coronary heart disease, strok e, and pneumonia. These data were supplied by the Health Care Informat ion Centre (Stichting Informatiecentrum voor de Gezondheidszorg). We a pplied Box-Jenkins time-series analysis seasonal autoregressive integr ated moving-average (SARIMA) models. Estimated models are tested by co nsidering the Portmanteau test and the Akaike information criterion. S ARIMA models give an adequate representation of hospital-in-patient-da ys utilization for the major sex and age classes of most selected dise ases. Poor modelling results are obtained for diabetes in all sex and age groups and in elderly women with coronary heart disease or with st roke. Seasonality is an important factor in most of the models that we have estimated, particularly for utilization of pneumonia and stroke patients. The major trends in standardized in-patient days are downwar d, and projected 1995 levels of standardized utilization are below the 1990 levels for all the selected diseases. Population-based projectio ns for 1995 are lower than the 1990 projections only for lung cancer a nd diabetes. The adequacy of the SARIMA models appears to be sensitive with respect to the parameter in the Portmanteau test. We discuss two possible explanatory developments for in-patient-days utilization: (i ) developments in the provision of hospital care, and (ii) epidemiolog ical developments. The selected diseases showed a decreasing mean dura tion of stay in 1980-90. Only for coronary heart disease did a rise in discharges in the same period outweigh this trend. We assessed contra sts between published epidemiological developments and the trends in i n-patient-days utilization. Possible explanations concern shifts from in-patient to out-patient care and changes in treatment. Finally, comp lementary to our SARIMA models, the investigation of future in-patient days utilization by means of scenario analytic appoaches remains impo rtant.