Development, validation and prospective clinical implementation of a machine learning algorithm for incident cardio-renal-metabolic diseases and cardiovascular death: the OPTIMISE study
R Nadarajah,A Wahab,C Reynolds,D C Hogg,K Raveendra,S Bennett,Y M Nakao,K Nakao,R Arbel,M Haim,D Zahger,H Larvin,J C Cowan,J Wu,C P Gale
DOI: https://doi.org/10.1093/eurjpc/zwae175.312
IF: 8.526
2024-06-01
European Journal of Preventive Cardiology
Abstract:Abstract Background Leveraging machine learning in routinely-collected primary care electronic health record (EHR) data may identify individuals at risk of cardio-renal-metabolic events to target for preventative strategies.{1, 2). Purpose We aimed to train, test and prospectively implement a machine learning algorithm to identify individuals at higher risk of incident cardio-renal-metabolic diseases and cardiovascular death. Methods We used UK primary care EHR data from 2 081 139 individuals aged ≥30 years (Jan 2, 1998, Nov 30, 2018), randomly divided into training (80%) and testing (20%) datasets. We trained a random forest classifier using age, sex, ethnicity and comorbidities (OPTIMISE). We calculated the cumulative incidence rate for ten diseases and death. We excluded individuals for the analysis of each disease who had a preceding diagnosis of that disease. Fine and Gray’s models with competing risk of death were fit for each outcome between higher and lower predicted risk. We implemented OPTIMISE in a pilot interventional non-randomised single arm study across four primary care sites. Consenting individuals aged ≥30 years at higher predicted risk received a community-based intervention of cardio-renal-metabolic phenotyping, guideline target attainment assessment and treatment. Results In the testing dataset (n = 416 228), individuals at higher predicted risk had higher long-term risk of heart failure (HR 12.54, 95% CI 12.08-13.01), aortic stenosis (9.98, 9.16-10.87), AF (HR 8·75, 95% CI 8·44-9·06), stroke/TIA (8.07, 7.80-8.34), CKD (6.85, 6.70-7.00), peripheral vascular disease (6.62, 6.28-6.98), valvular heart disease (6.49, 6.14-6.85), MI (5.02, 4.82-5.22), diabetes (2.05, 2.00-2.10) and COPD (2.02, 2.00-2.05) (Figure 1). This cohort were also at higher risk of death (10.45, 10.23-10.68), accounting for 74% of cardiovascular deaths (8582 of 11676) during 10-year follow up. After adjustment for age, sex, ethnicity and presence of any other outcomes at baseline, higher-predicted risk remained associated with excess risk for each outcome (Figure 2). Of the first 30 higher risk patients in the pilot clinical implementation (mean age 72.8 years (SD 8.1), 46.7% women) there were opportunities to reduce future cardiovascular risk. Mean BMI was 30.1 kg/m2 (SD 4.9), and 56.7% (17/30) had a systolic blood pressure >140mmHg. Of patients with type 2 diabetes 40% (4/10) had HBA1c >48mmol/mol, of patients on statin therapy 22.2% (4/18) had LDL cholesterol >1.8 mmol/L, of those who underwent echocardiography 12.5% (3/24) had newly diagnosed severe LVSD, and measurement for microalbuminuria upgraded CKD severity in 21.1% (4/19) of patients with CKD. Conclusions OPTIMISE is applicable to national EHR data and identifies people at higher risk of cardio-renal-metabolic diseases and death. On prospective evaluation higher predicted risk individuals have actionable targets for preventative therapies to reduce future cardiovascular risk. Figure 1 Figure 2
cardiac & cardiovascular systems