요양병원의 중증도 보정 사망률

Abstract

A long-term care hospital applies per diem payment systems rather than fee-for-service systems for the medical services provided to long-term care inpatients. To prevent the potential underprovision of medical services due to these per diem payment systems and promote improvement in medical quality, the quality assessment program for long-term care hospitals was introduced in 2008. Despite the development and introduction of various criteria in recent quality assessments, there is an increasing demand for the development of new criteria. Among these, there is a growing demand for the introduction of criteria related to mortality rates. Therefore, the purpose of this study is to examine the feasibility of introducing a mortality rate criterion in long-term care hospitals through an analysis of the relationship between severity-adjusted mortality rates and evaluation criteria. The study conducted an examination of domestic and international evaluation criteria related to long-term care hospitals and mortality rate evaluation criteria. Based on data including health insurance claim data from 2020 to 2022, patient assessment forms from long-term care hospitals, and results of quality assessment of long-term care hospitals, models for severity-adjusted mortality rates in long-term care hospitals were developed. Mortality rates were defined as in-hospital mortality rates, and the development of the severity model considered factors obtained through Bootstrap, literature review, patient assessment forms, and whether patients had a history of COVID-19 infection as adjustment variables. To understand mortality rates according to various treatment characteristics of long-term care hospitals, institutions with treatment volumes exceeding 30% were classified based on their representative treatment characteristics (cancer, dementia, rehabilitation, others). The developed models' goodness-of-fit was assessed using Bootstrap, C-statistics, Corrected C-statistics, AIC (Akaike Information Criterion), and SBC (Schwarz-Bayesian Criterion). The final models were selected through model comparisons based on AIC and SBC, and the severity-adjusted mortality rates were analyzed. Additionally, to understand the relationship between quality assessment criteria and mortality rates, correlation analyses were conducted, and ROC curve analyses were used to determine the discrimination ability of evaluation grades of mortality rate. Evaluation criteria related to long-term care, both domestically and internationally, include pressure ulcer occurrence rates, community reintegration rates, urinary tract infection rates, improvements in activities of daily living rates, among others. However, evaluations of mortality rates related to long-term care are not conducted on a national level. Furthermore, long-term care hospital mortality rates have shown significant increases after the COVID-19 pandemic. The final selected severity-adjusted model was one that adjusted for clinical risk factors. If an institution's mortality rate significantly exceeds the 95% confidence interval of the severity-adjusted mortality rate, it may indicate factors other than severity adjustment influencing the mortality rate and necessitate monitoring of that institution. Nevertheless, before concluding that the "severity-adjusted model fits well and has good predictive power" or that "monitoring is required for the institution," it is necessary to confirm the models' representativeness. The relationships between mortality rates and evaluation criteria have been demonstrated through various papers, and statistically significant results have also been derived from this study's data, showing significance not only for activities of daily living but also for pressure ulcer presence. Analyses of the discrimination ability of evaluation grade of mortality rate revealed AUC (Area Under the Curve) values of 0.5727 for pre-adjusted mortality rates and 0.5220 for post-adjusted mortality rates. In conclusion, based on the examination of the feasibility of introducing mortality rate criteria through the severity-adjusted models and the discrimination ability of evaluation grades, it is determined that the data available for generating the severity-adjusted models are limited and the discrimination ability of evaluation grades for mortality rates is lower compared to other indicators. Therefore, the introduction of mortality rate criteria is considered difficult. Instead, enhancing the reliability of current evaluation criteria through improvements is suggested, and for the severity-adjusted models to be representative, the accuracy and reliability of the data should be validated.