Screening for heart failure using simple pulse oximetry in outpatient care

Abstract Background Early detection of heart failure (HF) is crucial to reduce its clinical and socioeconomic burden, as timely identification and treatment can improve outcomes. Disparities in access to standard diagnostic HF testing along with delays in initiation of evidence-based HF medications underscore the need for a widely available, effective, and inexpensive diagnostic tool. For this purpose, we developed and cross-validated a machine learning (ML) algorithm capable of detecting HF by analyzing photoplethysmographic (PPG) signals recorded from a simple pulse oximeter. Purpose To develop a widely accessible, reliable, and cost-effective method for HF screening. Methods Patients attending routine ambulatory check-ups were included in the study. HF diagnosis was based on the latest ESC guidelines. PPG signal was recorded by a simple finger pulse oximeter. The signal quality was evaluated prior to processing, including noise removal, filtering, artifact detection, and pulse wave identification. A total of 57 features were extracted from the signal and were subsequently utilized as input for a ML algorithm to facilitate HF classification. Results In total, 386 patients were included in the study. 243 HF patients (37% female, average age 68 ± 12 – of those 49% with HFpEF, 12% HFmrEF and 39% HFrEF) and 143 non-HF patients (48% female, average age 56.03 ± 16.02). The Random Forest Classifier achieved an average cross-validated c-statistics of 0.89 ± 0.035 in detecting HF (0.86 ± 0.053% sensitivity, 0.76 ± 0.092% specificity). Conclusions Our results suggest that HF could be effectively screened using a simple pulse oximeter enhanced by ML signal analysis. This solution could be adapted into a simple, widely available, reliable, inexpensive fast-track HF diagnostic tool for primary care. Early HF patient identification and referral to a cardiologist or HF specialist with prompt etiology work-up and treatment initiation, could positively impact patient morbidity and mortality, as well as healthcare expenses.