Abstract: PO0452
Development and Validation of a Predictive Model to Identify Patients with Undiagnosed CKD
Session Information
- CKD Epidemiology, Biomarkers, Predictors
October 22, 2020 | Location: On-Demand
Abstract Time: 10:00 AM - 12:00 PM
Category: CKD (Non-Dialysis)
- 2101 CKD (Non-Dialysis): Epidemiology, Risk Factors, and Prevention
Authors
- Brunelli, Steven M., Davita Clinical Research, Minneapolis, Minnesota, United States
- Walker, Adam G., Davita Clinical Research, Minneapolis, Minnesota, United States
- Gray, Kathryn S., Davita Clinical Research, Minneapolis, Minnesota, United States
Background
Chronic kidney disease (CKD) is a common condition and often goes undiagnosed. Unmanaged CKD can progress rapidly, resulting in poor clinical outcomes and increased health care costs. Identification of individuals who may have undiagnosed CKD would allow for implementation of CKD management practices in order to slow progression, potentially improving outcomes and reducing health care costs. Here, we report the development and validation of a claims-based algorithm to identify CKD.
Methods
This model was developed using Medicare Part A and Part B claims from calendar year 2017. Data from 378,460 unique patients with no evidence of end-stage kidney disease or claims for dialysis through April 2017 were split into derivation (n = 189,203) and validation (n = 189,257) sets. The predicted outcome was the presence of a diagnosis code for CKD stages 3 to 5, which occurred in 4.4% of patients within the data. To simulate the use case, codes for kidney disease were not eligible as predictors in the model. Area under the curve (AUC) of the receiver operating curve and positive predictive value (PPV) were used to assess the performance of candidate models.
Results
The best model was a logistic regression algorithm based on 94 input terms derived from 13 clinical constructs. The model demonstrated an excellent ability to discriminate (AUC = 0.90), which was stable when tested in the validation set (AUC = 0.90). The PPV in the top 1% and top 2% of patients identified by the model was approximately 72% and 59%, respectively.
Conclusion
We developed an algorithm that uses medical claims to identify patients who are most likely to have unrecognized CKD. If the algorithm were applied to a population of 10,000 patients, it could identify the 100 patients at highest risk, among whom 72 would have CKD if tested.