Kidney Week

Abstract: TH-OR103

Incorporating Kidney Disease Measures into Cardiovascular Risk Prediction: Evaluation Using Electronic Health Record Data from 37 Health Care Organizations

Session Information

• Hypertension and CVD: Therapies and Predictors
  November 07, 2019 | Location: 206, Walter E. Washington Convention Center
  Abstract Time: 05:30 PM - 05:42 PM

Category: Hypertension and CVD

• 1401 Hypertension and CVD: Epidemiology, Risk Factors, and Prevention

Authors

• Stempniewicz, Nikita, AMGA (American Medical Group Association), Alexandria, Virginia, United States

Nikita Stempniewicz

Nikita Stempniewicz has eight years’ experience in population health analytics and epidemiologic and health services research at AMGA, with a focus in answering important health care related questions using electronic health record data. His interests include population health, health information technology, and chronic conditions (CKD, Diabetes, Hypertension).

• Sang, Yingying, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, United States

Yingying Sang,

• Matsushita, Kunihiro, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, United States

Kunihiro Matsushita,

• Ciemins, Elizabeth, AMGA (American Medical Group Association), Alexandria, Virginia, United States

Elizabeth Ciemins,

• Ballew, Shoshana, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, United States

Shoshana Ballew,

• Grams, Morgan, Johns Hopkins University, Baltimore, Maryland, United States

Morgan Grams,

• Cuddeback, John K., AMGA (American Medical Group Association), Alexandria, Virginia, United States

John K. Cuddeback,

• Coresh, Josef, Welch Center for Prevention, Epidemiology & Clinical Research, Baltimore, Maryland, United States

Josef Coresh,

Background

Clinical guidelines for the primary prevention of atherosclerotic cardiovascular disease (ASCVD) increasingly use absolute risk to guide decision-making, often relying on the AHA/ACC Pooled Cohort Equation (PCE) for risk estimation. Two kidney measures, estimated glomerular filtration rate (eGFR) and urine albumin-to-creatinine ratio (uACR), are CVD risk enhancers but they are not included in the PCE. We hypothesize that when eGFR and uACR data are available, their inclusion will meaningfully enhance ASCVD risk assessment.

Methods

836,047 patients in the OptumLabs^® Data Warehouse’s EHR-derived data with no evidence of baseline ASCVD and data on PCE variables plus eGFR and uACR, from 37 health care organizations (HCOs) were followed (mean (SD), 3.6 (2.4) years) for EHR evidence of ASCVD events (myocardial infarction and stroke). We developed a “CKD patch” to add eGFR and ACR data to the existing PCE.

Results

Patients were age 59 (10) (mean (SD)) years, had total cholesterol 175 (35) mg/dL, HDL 45 (15) mg/dL, SBP 128 (16) mmHg, eGFR 81 (21) ml/min/1.73 m², and uACR 10 (2-29) mg/g (median (IQR)), 51% were female, 74% had diabetes, and 4.5% were smokers. The original PCE predicted a 5-year risk of 8.4%, and observed rate of ASCVD was 4.2%. Adding eGFR and ACR to PCE improved the C-statistic by 0.022 (95% CI 0.020-0.024) overall, 97% of HCOs (36 of 37) improved by ≥ 0.01. Overall, reclassification of 5-year ASCVD risk from low (< 3.75%) to intermediate (3.75%–9.9%) risk and intermediate to high risk (≥ 10%) was 5.1% (range by HCO: 3.0–8.9%) and 6.2% (4.1 – 10.9%), respectively. ASCVD reclassification rates were higher with higher CKD risk (intermediate to high ASCVD risk: 2.6% at no CKD, 11.9% at intermediate-risk CKD, 15.8% at high-risk CKD and 20.4% at very high-risk CKD). Recalibrating PCE in each HCO prior to adding kidney measures yielded similar results.

Conclusion

CKD measures (eGFR and uACR) are often available, and their integration into the PCE is feasible and results in meaningful risk reclassification across HCOs, particularly among higher risk CKD. Implementation in EHRs should include rigorous validation, attending to limitations, e.g., EHRs are blind to events occurring outside the HCO.