Outpatient acute kidney disease detection by national and institutional health data

The development of diagnostic definitions for acute kidney disease in outpatients (AKD_OPT) remains an unattained goal because of consensus gaps and data silos in health-care systems.

Our team developed the Acute Kidney Injury Detection System (AKIDS) to screen for undiagnosed AKD_OPT using National Health Insurance MediCloud data and institutional electronic medical records. The criteria for AKD_OPT were a >50% change in maximum and minimum serum creatinine levels or a >35% change in corresponding estimated glomerular filtration rate values within the 180 days before an appointed outpatient visit. In this retrospective cohort study, the associations between AKD_OPT and composite kidney outcome (CKO; e.g., end-stage kidney disease, a >40% drop in estimated glomerular filtration rate, or a 2-fold or more increase in serum creatinine), all-cause mortality, and de novo non-dialysis chronic kidney disease (CKD-ND) were evaluated using multivariable Cox proportional hazards models.

Of 79,838 eligible adult patients screened by the AKIDS, 12,510 (15.7%) have AKD_OPT. The adjusted absolute risk increases for 1-year incident CKO, mortality, and de novo CKD-ND for patients with AKD_OPT relative to those without are 79.0 (95% confidence interval 78.9-79.1), 25.3 (25.2-25.3), and 54.8 (54.7-54.9) per 1000 patients, respectively. The adjusted hazard ratios for 1-year CKO, mortality, and de novo CKD-ND are 16.2 (14.2-18.5), 2.6 (2.4-2.9), and 3.5 (3.1-3.9), respectively.

By integrating the national-local data using the AKIDS, this study comprehensively characterizes a previously unrecognized phenotype, AKD_OPT, illustrating the potential of healthcare big data in transforming global approaches to AKD patterns and prevention.