2017 Volume 64 Issue 5 Pages 246-257
Objectives This study aimed to develop risk assessment scales for predicting the incidence of Needed Support/Long-Term Care certification, by aggregating data from the Kihon Checklist, medical assessments, and long-term care insurance certification during a follow-up period (a maximum of 4 years and 2 months) conducted in a municipality.
Methods This retrospective cohort study included 72,127 older adults aged 65 years or older living in K City (an ordinance-designated city) who responded to the Kihon Checklist in 2011. We linked their medical assessment data (examined/unexamined, blood pressure, and five blood biochemical items) from 2011 and information on the incidence of long-term care insurance certification from 2011 to 2015 to the Kihon Checklist data (the 12 essential items and seven optional items from the Needs Survey). We constructed four Cox proportional hazards models as follows: 1) age, sex, and the Needs Survey's 12 essential items; 2) model 1 plus seven optional items; 3) model 2 plus examined/unexamined at medical assessment; and 4) model 3 plus blood pressure and five blood biochemical items, as independent variables. Recent requirement for Support/Long-Term Care certification was included as an outcome with stepwise forward selection. We assigned scores for each item based on the non-standardized regression coefficients obtained (B) and the sum of those scores was used to establish the risk assessment scales for predicting Needed Support/Long-Term Care certification from each model. A receiver operating characteristic (ROC) analysis was conducted to estimate the sensitivity and specificity in order to compare predictive validity of the scales.
Results During the follow-up period, 11,039 (15.3%) individuals required a new incidence of a Needed Support/Needed Long-Term Care certification. A risk assessment scale of 0-55 was established based on age, sex, and the 10 essential items from the Needs Survey's. The incidence of certification were 3.2%, 14.7%, 31.6%, 56.7%, and 75.0% at scores of 10, 20, 30, 40, and 50, respectively. The area under the ROC curve (AUC) was 0.783, and the sensitivity and the specificity were 0.705 and 0.731, respectively (cut-off: 21/22). These values remained almost unchanged despite the addition of optional and medical assessment items (AUC: 0.786-0.787, sensitivity: 0.721-0.730, and specificity: 0.710-0.717).
Conclusion Although the medical assessment data was not aggregated, the scale developed from the Kihon Checklist's 10 items (included in the Needs Survey's essential items) is useful for predicting the incidence of Needed Support/Long-Term Care certification. The scale, which evaluates the risk of needed support/long-term care at individual and community levels, was developed using the existing Kihon Checklist data or the Needs Survey's data collected subsequently by municipalities.
