OBJECTIVE: The purpose of this study was to propose a suitable cut-off point for the serum albumin value to predict which elderly individuals are qualified for long-term care insurance certification and those at risk of death. METHODS: The Tsurugaya Project was a comprehensive geriatric assessment conducted for community-dwelling elderly persons aged 70 years or older in Sendai, Japan. Of those participating, the 832 from whom consent was obtained and who were not qualified for long-term care insurance certification were adopted as subjects. Cut-off points on serum albumin levels were set every 0.1 g/dL from 3.5 g/ dL to 4.0 g/dL. The risk of incident certification of long-term care insurance or death (calculated by the Cox proportional hazards model), the frequency of the subjects, the sensitivity, the specificity and the positive predictive value of the category below each cut-off point were calculated. RESULTS: During 3 years follow-up, a total of 130 subjects either became certificated for long-term care insurance or died. On multivariate analyses, the risk of incident certification of long-term care insurance or death showed significantly increase with all cut-offpoints. The frequencies were 1.3% at 3.5 g/dL, 9.6% at 3.8 g/dL and 29.6% at 4.0 g/dL. The sensitivities were 5.4% at 3.5 g/dL, 18.5% at 3.8 g/ dL and 45.4% at 4.0 g/dL. The specificities were 99.4% at 3.5 g/dL, 92.0% at 3.8 g/dL and 73.4% at 4.0 g/dL. The positive predictive values were 63.6% at 3.5 g/dL, 30.0% at 3.8 g/dL and 24.0% at 4.0 g/dL. When the serum albumin cut-off point was set to 3.8 g/dL and the highest tertile level of the serum albumin value was used as a reference category, the sex- and age-adjusted hazard ratios (95% confidence intervals) were 2.1 (1.1-3.9) in those with serum albumin < or =3.8 g/dL, 1.5 (0.9-2.5) in those with the lowest tertile level and 1.0 (0.6-1.7) in those with the middle tertile level. CONCLUSION: This study suggested that a serum albumin value of 3.8 g/dL is a suitable cut-off point with regard to sensitivity, specificity and positive predictive value.
OBJECTIVE: The purpose of this study was to propose a suitable cut-off point for the serum albumin value to predict which elderly individuals are qualified for long-term care insurance certification and those at risk of death. METHODS: The Tsurugaya Project was a comprehensive geriatric assessment conducted for community-dwelling elderly persons aged 70 years or older in Sendai, Japan. Of those participating, the 832 from whom consent was obtained and who were not qualified for long-term care insurance certification were adopted as subjects. Cut-off points on serum albumin levels were set every 0.1 g/dL from 3.5 g/ dL to 4.0 g/dL. The risk of incident certification of long-term care insurance or death (calculated by the Cox proportional hazards model), the frequency of the subjects, the sensitivity, the specificity and the positive predictive value of the category below each cut-off point were calculated. RESULTS: During 3 years follow-up, a total of 130 subjects either became certificated for long-term care insurance or died. On multivariate analyses, the risk of incident certification of long-term care insurance or death showed significantly increase with all cut-offpoints. The frequencies were 1.3% at 3.5 g/dL, 9.6% at 3.8 g/dL and 29.6% at 4.0 g/dL. The sensitivities were 5.4% at 3.5 g/dL, 18.5% at 3.8 g/ dL and 45.4% at 4.0 g/dL. The specificities were 99.4% at 3.5 g/dL, 92.0% at 3.8 g/dL and 73.4% at 4.0 g/dL. The positive predictive values were 63.6% at 3.5 g/dL, 30.0% at 3.8 g/dL and 24.0% at 4.0 g/dL. When the serum albumin cut-off point was set to 3.8 g/dL and the highest tertile level of the serum albumin value was used as a reference category, the sex- and age-adjusted hazard ratios (95% confidence intervals) were 2.1 (1.1-3.9) in those with serum albumin < or =3.8 g/dL, 1.5 (0.9-2.5) in those with the lowest tertile level and 1.0 (0.6-1.7) in those with the middle tertile level. CONCLUSION: This study suggested that a serum albumin value of 3.8 g/dL is a suitable cut-off point with regard to sensitivity, specificity and positive predictive value.