Model of cardiovascular disease in people with reduced kidney function using routine healthcare data

Conference:

SAPC ASM 2019 - Exeter

Talk Code:

4B.5b

Presenter:

Iryna Schlackow

Twitter:

@HERC_Oxford

Co-authors:

Claire Simons, Jason Oke, Benjamin Feakins, Daniel Lasserson, Richard Stevens, Rafael Perera, Borislava Mihaylova

Author institutions:

Health Economics Research Centre Nuffield Department of Population Health University of Oxford; Nuffield Department of Primary Care Health Sciences University of Oxford; Institute of Applied Health Research University of Birmingham; Blizard Institute Barts and The London School of Medicine and Dentistry Queen Mary University of London

Problem

Reduced kidney function is associated with increased cardiovascular disease (CVD) risk. To evaluate the long-term effects and cost-effectiveness of kidney function monitoring and treatments to prevent cardiovascular events, a long-term model of CVD in people with reduced kidney function is required.

Approach

We used a large routine individual patient primary care dataset (Clinical Practice Research Datalink, UK) linked with secondary care (Hospital Episodes Statistics, UK) and UK mortality registry data to model long-term cardiovascular outcomes and survival in people with impaired kidney function. A cohort of 1.13 million patients with at least two eGFR measurements <90 mL/min/1.73m2 ≥90 days apart between 2005 and 2015 was established. Information on patients’ characteristics, including sociodemographic (e.g. age, gender, smoking status) and clinical factors (e.g. blood pressure, history of diabetes, cholesterol levels, urinary albumin-to-creatinine ratio), as well as, kidney function measures and CVD events, were used to model the major CVD risks (myocardial infarction (MI), stroke, hospital admission for heart failure and vascular death) using parametric survival models separately in men and women with and without prior cardiovascular disease. The covariates in models were selected using statistical information criteria and epidemiological evidence. Missing data were multiply imputed using chained equations. A Markov state disease model, developed using the separate risk equations, was validated in categories of participants by gender, eGFR (90-60; 60-45; 45-30; 30-15; <15 mL/min/1.73m2) and prior CVD. External data informed annual cost and quality of life in the model. Relative risk reductions in cardiovascular events with CVD preventive treatments, informed from randomised controlled trials, were implemented.

Findings

The patient cohort included 509,179 men (77% without and 23% with prior CVD at entry) and 630,369 women (83% without and 17% with prior CVD at entry) followed for a median of 4.9 (interquartile range 2.3-7.8) years. At baseline 74% had an eGFR of 60-90mL/min/1.73m2; 18% 45-60; 6% 30-45; 1.3% 15-30 and 0.2% less than 15mL/min/1.73m2. Six percent of patients without and 11% with CVD experienced a non-fatal stroke or MI; and 2% and 8%, respectively, died from vascular causes. In all risk equations, age, current eGFR, and, where relevant, prior CVD, were the key determinants of subsequent CVD. The model-projected CVD risks corresponded well to observed CVD risks in categories of patients and over time.

Consequences

This new model projects long-term CVD risks, quality-adjusted survival, healthcare costs and the effects of CVD preventive interventions, and can be used to guide CKD monitoring strategies and CVD preventive treatments in people with reduced kidney function.

Submitted by:

Iryna Schlackow

Funding acknowledgement:

Funded by the NIHR PGfAR programme (RP-PG-1210-12003) and NIHR Biomedical Research Centre, Oxford. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health and Social Care.