Predictive factors for hospitalized and institutionalized care-giving of the aged patients with diabetes mellitus in Japan.

To identify predictive factors for hospitalized and institutionalized care-giving among a group of aged patients with diabetes mellitus in Japan, retrospective chart review was performed in 288 diabetic subjects aged 65 years or older. Independent variables, based on the chart review, were age, sex, diagnosis, diabetic control and complications. Comprehensive geriatric assessment was performed to obtain information on the functional capacity and demographic variables, including physical and mental function, and socioeconomic status. 131 diabetic patients were considered as frail elderly and characterized for their higher age, longer duration of diabetes, higher frequency of insulin use, lower cognitive function, and lower QOL, in comparison with those of non-frail patients. All non-frail diabetic patients were independently treated at their homes, while 38 subjects out of 131 frail diabetic patients were hospitalized or institutionalized. Apparent clinical features of hospitalized/institutionalized patients were higher age, higher serum creatinine, and higher prevalence of stroke episodes, advanced cognitive decline and absence of key caregiver in the family members, in comparison with those of in-home frail diabetic patients. The predicted probabilities from the multivariate logistic regression analysis in predicting hospitalized and institutionalized care-giving were as follows: Log p/(1 - p) = -19.801x1 - 54.269x2 + 721.405; where x1 = cognitive function (score), x2 = social support (score). Receiver operating characteristic curve analysis revealed a satisfactory discrimination for hospitalized and institutionalized care-giving in frail diabetic elderly with 92.9% of sensitivity and 91.4% of specificity, when the cutoff point of the model was set at 0.992. We concluded that cognitive decline and low social support are the predictive for hospital and institutional care-giving, and that demographic and mental information as well as diagnostic data should be analyzed to predict the hospitalization/institutionalization among frail diabetic elderly.