Estimating and Predicting the Rate of Kidney Function Decline over 10 Years in the General Population.

KEY POINTS

This study estimated the kidney function decline rate over 10 years in the general population. We also developed a machine learning prediction model based on annual health checkup results and claims for the first 5 years. Prediction models for kidney function decline would be useful for stratifying the general population and identifying rapid decliners.

BACKGROUND

We aimed to estimate the rate of kidney function decline over 10 years in the general population and develop a machine learning model to predict it.

METHODS

We used the JMDC database from 2012 to 2021, which includes company employees and their family members in Japan, where annual health checks are mandated for people aged 40–74 years. We estimated the slope (average change) of eGFR over a period of 10 years. Then, using the annual health-check results and prescription claims for the first 5 years from 2012 to 2016 as predictor variables, we developed an XGBoost model, evaluated its prediction performance with the root mean squared error (RMSE), R, and area under the receiver operating characteristic curve (AUROC) for rapid decliners (defined as the slope <−3 ml/min per 1.73 m per year) using five-fold cross validation, and compared these indicators with those of () the simple application of the eGFR slope from 2012 to 2016 and () the adjusted linear regression model.

RESULTS

We included 126,424 adults (mean age, 45.2 years; male, 82.4%; mean eGFR, 79.0 ml/min per 1.73 m in 2016). The mean slope was −0.89 (SD, 0.96) ml/min per 1.73 m per year. The predictive performance of the XGBoost model (RMSE, 0.78; R, 0.35; and AUROC, 0.89) was better than that of either the simple application of the eGFR slope from 2012 to 2016 (RMSE, 1.94; R, −3.03; and AUROC, 0.79) or the adjusted linear regression model (RMSE, 0.81; R, 0.30; and AUROC, 0.87).

CONCLUSIONS

We estimated the rate of kidney function decline over 10 years in the general population, as well as demonstrated that application of machine learning to annual health-check and claims data, provides better predictive performance compared with traditional methods.