Association Between Frailty Transitions And Adverse Health Outcomes Among Older Chinese Inpatients

ObjectiveFrailty has been considered to be associated with an increased risk of adverse health outcomes in the elderly,however,current researches on the association between frailty transitions and adverse health outcomes is relatively scarce.This study is based on a national multi-center cohort study involving six tertiary hospitals across six regions in the country,aiming to explore the association between frailty transitions and adverse health outcomes in elderly inpatients,and explore the risk prediction performance of the random survival forest model and the Cox proportional hazard regression model for death,and select the model with better performance to predict the risk of death within 2 years for the elderly inpatients.MethodsThis multi-center cohort study was conducted from 2018 to February 2021,including 8256 elderly inpatients from multiple departments of six hospitals.Quantitative data were statistically described using mean,standard deviation,median,and interquartile spacing,and group comparisons were made using the t-test and Mann-Whitney U-test;qualitative data were statistically described using frequency counts and percentages,and the x2 test was used to compare the differences between groups.The Kaplan-Meier method was used to estimate survival rate,and the log-rank method was used to test for differences in survival curves.The Cox proportional risk regression,logistic regression,negative binomial regression,and linear regression were used to assess the association between changes in frailty and adverse health outcomes.The random survival forest model and the Cox model were constructed to predict the risk of death for the elderly inpatients,and the C-index,Time-dependent Area under the Curve(AUC),and Integrated Brier score(IBS)of the two models were evaluated in the training and testing sets to compare their predictive performance,and the Cox model was used to draw a alignment diagram for a clinical display intuitively.ResultsA total of 8,256 patients were included in the study.At baseline,40.70%of elderly patients were not frail,43.04%were in the pre-frail stage,and 16.27%were frail.At the 3month follow-up,45.46%of participants experienced changes in frailty status,while 54.54%maintained a stable frailty status.Among the five groups of participants with different frailty transition patterns,the survival curve for the group of remaining non-frail was highest,followed by the groups of improving,remaining pre-frail,worsening,while the survival curve of remaining frail group was the lowest.Following a log-rank test,differences in survival curves across the five groups were significant.When adjusting for confounding factors,compared to the group of remaining non-frail,the 2-year risk of death for the groups of improving,remaining pre-frail,worsening,and remaining frail were 1.316,1.706,2.426,and 2.443 times higher,respectively.The 2-year risk of falls for the groups of improving,worsening,remaining pre-frail,and remaining frail were 1.586,1.936,2.212,and 2.711 times higher,respectively,compared to the group of remaining non-frail.The 2-year risk of rehospitalization for the groups of improving,worsening,remaining prefrail,and remaining frail were 1.325,1.534,1.563,and 2.291 times higher,respectively,compared to the groups of remaining non-frail.The scores of health-related quality of life(HRQoL)for the groups of improving,worsening,remaining pre-frail,and remaining frail decreased by an average of 1.230,3.665,4.951,and 9.762 points within 2 years,respectively,compared to the groups of remaining non-frail.The random survival forest(RSF)model exhibited an out-of-bag error rate of 22.54%in the training set and 22.29%in the test set,reflecting its good predictive capability for the death risk of frail elderly patients.In the training set,the RSF model had significantly higher time-dependent AUC and C-index between three months to two years of enrollment,and time-dependent AUC and C-index at two years of follow-up time than the Cox model,while the IBS at two years of follow-up time was slightly lower than the Cox model.However,in the test set,the timedependent AUC and C-index between three months to two years of enrollment in the randomized survival forest model,the time-dependent AUC,C-index and IBS at two years of follow-up time were not significantly different from the Cox model.Conclusions(1)Compared to the group of remaining non-frail,the risk of death within 2 years after enrollment increased progressively for the groups of improving,remaining pre-frail,worsening,and remaining frail.(2)Compared to the group of remaining non-frail,the risk of falls within 2 years after enrollment increased progressively for the groups of improving,worsening,remaining prefrail,and remaining frail.Furthermore,the average increase in the number of falls within 2 years after enrollment for the groups of worsening and remaining pre-frail increased progressively.(3)Compared to the group of remaining non-frail,the risk of rehospitalization within 2 years after enrollment increased progressively for the groups of improving,worsening,remaining pre-frail,and remaining frail groups.The average increase in the total days of rehospitalization within 2 years after enrollment also increased progressively across these groups.(4)Compared to the group of remaining non-frail,the average decline in HRQoL scores at 2 years after enrollment increased progressively for the groups of improving,worsening,remaining pre-frail,and remaining frail across these groups.(5)The discrimination and prediction accuracy of the RSF model were not significantly different from the Cox regression model.Both models effectively predicted the 2-year mortality risk of the frail elderly inpatients and could be used depending on the specific circumstances.