Development of a tree model that allows simple estimation of the required care level using the items of the basic investigation of Long-term care insurance

Abstract

In long-term care insurance, the required care level of the disabled elderly is calculated from the results of the basic investigation. However, this calculation involves complex mathematical processes, and the estimation of the required care level at small facilities is difficult. We, therefore, developed a tree model that allows simple estimation of the required care level from the state of noticeable disabilities in daily activities. The model was prepared separately for dementia and physical disabilities. From the patients being cared for at Higashiyama Geriatric Hospital for the Elderly who had undergone primary rating, a total of 240 individuals consisting of 20 each in each of the 6 required care levels for both dementia and physical disabilities were selected, and the results of their primary rating were reviewed. “Putting on and taking off a jacket” and “care after urination”, in which the required care levels increase relatively consistently as the investigation items progressed from those for “independent” to those for “totally assisted”, were selected as the first selection items in dementia and physical disability models respectively. In a dementia model, the state of “putting on and taking off a jacket” and “care after urination” were matched for various required care levels as follows: “Independent”→assistance needed, “observation needed” and “independent” in “standing up”→required care level 1, “observation needed” and “not independent” in “standing up” →required care level 2, “partly assisted”→required care level 3, “totally assisted” and “not totally assisted” in “eating”→required care level 4, and “totally assisted” and “totally assisted” in “eating”→required care level 5. In a physical disability model, the state of “care after urination” was matched for various required care levels as follows: “Independent” and “independent” in “walking”→assistance needed, “independent” and “not independent” in “walking”→required care level 1, “direct or indirect assistance”→required care level 2, “totally assisted” and “independent” or “observation needed” in “eating” →required care level 3, “totally assisted” and “partly assisted” in “eating”→required care level 4, “totally assisted” and “totally assisted” in “eating”→required care level 5. The accuracy rate, i. e. the frequency of complete matching betweeen the estimation of the required care level using this tree and that of the primary rating, was 71.1% in those with dementia and 66.7% in those with physical disabilities. The near accuracy rate, i. e. the frequency of matching between the two estimations within one rank higher or lower was 98.3% in those with dementia and 99.2% in those with physical disabilities. From these results, this tree model is considered to be useful for clinical rating.