Background and objective: The standardized mortality ratio (SMR) is frequently used to compare health status among different populations. However, SMR could be biased when based upon communities with small population size such as towns and wards and comparison of SMRs in such cases is not appropriate. The "empirical Bayes estimate of standardized mortality ratio" (EBSMR) is a useful alternative index for comparing mortalities among small populations. The objective of the present study was to use the EBSMR to clarify the relationships between health care resources and mortalities in 3,360 municipalities in Japan.
Materials and methods: Health care resource data (number of physicians, number of general clinics, number of general sickbeds, and number of emergency hospitals) and socioeconomic factors (population, birth rate, aged households, marital rate, divorce rate, taxable income per individual under taxes duty, unemployment, secondary, tertiary industrial employment and prefecture) were obtained from officially published reports. EBSMRs for all causes, cerebrovascular disease, heart disease, acute myocardial infarction, and malignant neoplasms were calculated from the 1997-2001 vital statistic records. Multiple regression analysis was used to examine the relationships between EBSMRs and the variables representing health care resources and socioeconomic factors as covariates. Some of the variables were log-transformed to normalize the distribution of variables.
Results: The correlation between number of physicians and general sickbeds was very high (Pearson's r = 0.776). So, we excluded the number of general sickbeds. Some of the EBSMRs were inversely associated with the number of physicians per person (all causes in males (beta = -0.042, P = 0.024) and females (beta = -0.150, P < 0.001), cerebrovascular disease in females (beta = -0.074, P < 0.001), heart disease in males (beta = -0.066, P < 0.001) and females (beta = - 0.087, P < 0.001), acute myocardial infarction in females (beta = -0.061, P = 0.003), and malignant neoplasms in females (beta = -0.064, P = 0.001)). In contrast, when there was a higher number of clinics per persons, the EBSMR was higher for all causes in males (beta = 0.053, P = 0.001) and females (beta = 0.115, P < 0.001), cerebrovascular disease in males (beta = 0.047, P = 0.002) and females (beta = 0.070, P < 0.001), heart disease in females (beta = 0.061, P < 0.001), acute myocardial infarction in females (beta = 0.048, P = 0.006), and malignant neoplasms in males (beta = 0.036, P = 0.018) and females (beta = 0.046, P = 0.005). Next, we selected the number of emergency hospitals as the variable representing health care resources. Some of the EBSMRs were inversely associated with the existence of emergency hospitals (all causes in females (beta = -0.085, P < 0.001), cerebrovascular disease in males (beta = -0.032, P = 0.031) and females (beta = -0.059, P = 0.001), and heart disease in females (beta = -0.052, P = 0.008)).
Conclusion: The results suggested that an appropriate distribution of health care resources such as physicians and emergency hospitals is an important factor associated with mortality in a community.