Kidney Week

Abstract: SA-PO763

Predicting CKD in the CARRS Study: Strategies for Screening in the Indian Population

Session Information

• CKD: Epidemiology, Risk Factors, Prevention - III
  October 27, 2018 | Location: Exhibit Hall, San Diego Convention Center
  Abstract Time: 10:00 AM - 12:00 PM

Category: CKD (Non-Dialysis)

• 1901 CKD (Non-Dialysis): Epidemiology, Risk Factors, and Prevention

Authors

• Bradshaw, Christina L., Stanford University School of Medicine, Palo Alto, California, United States

Christina L. Bradshaw,

• Zheng, Yuanchao, Stanford University School of Medicine, Palo Alto, California, United States

Yuanchao Zheng,

• Montez-Rath, Maria E., Stanford University School of Medicine, Palo Alto, California, United States

Maria E. Montez-Rath,

• Prabhakaran, Dorairaj, Public Health Foundation of India, Gurgaon, HARYANA, India

Dorairaj Prabhakaran,

• Anand, Shuchi, Stanford University School of Medicine, Palo Alto, California, United States

Shuchi Anand,

Background

Chronic kidney disease (CKD) is a significant global health issue. Much of the rise in CKD-attributable deaths has occurred in low- and middle-income countries (LMICs). Population-wide screening for CKD is not cost-effective and likely untenable in resource-limited settings. Development of a tool that identifies at-risk persons and allows targeted screening is paramount to manage this growing problem.

Methods

Using data from the Centre for Cardiometabolic Risk Reduction in South Asia (CARRS) study, we fit a logistic regression model based on 30 variables easily measured during a home visit. Final predictors were selected through a step-down procedure. We defined presence of CKD as eGFR <60 ml/min/1.73m² by the CKD-EPI equation or urine albumin-to-creatinine ratio (UACR) ≥30 mg/g. Discrimination was assessed with the c-statistic and calibration with the calibration slope. Model predictive ability was compared across different probability cut-offs. We used bootstrap for internal validation.

Results

In our sample of 8,698 participants from Delhi and Chennai, CKD prevalence was 10.9%. Our final model consisted of 19 variables (Figure 1). Development and validated c-statistics were 0.79 and 0.78, respectively. Calibration slope was 0.97 after validation. Using a probability cut-off of 0.075, the model identified 3,489 people (40% of original sample) as needing confirmatory testing for CKD with serum creatinine or UACR, giving a sensitivity of 78% and specificity of 65%.

Conclusion

Our prediction model exhibited good sensitivity and discrimination for CKD, while narrowing the pool of people who necessitate confirmatory testing. We are the first to use large-scale population data to develop a tool that facilitates targeted screening for CKD in urban India. Future efforts include external validation on a separate Indian cohort.

Funding

• NIDDK Support