Machine Learning Improves the Identification of Individuals With Higher Morbidity and Avoidable Health Costs After Acute Coronary Syndromes

• Published in: Value in health Vol. 23; no. 12; pp. 1570 - 1579
• Main Authors: de Carvalho, Luiz Sérgio Fernandes, Gioppato, Silvio, Fernandez, Marta Duran, Trindade, Bernardo Carvalho, Silva, José Carlos Quinaglia e, Miranda, Rebeca Gouget Sérgio, de Souza, José Roberto Matos, Nadruz, Wilson, Avila, Sandra Eliza Fontes, Sposito, Andrei Carvalho
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 01-12-2020; Elsevier Science Ltd
• Subjects: Acute Coronary Syndrome - complications; Acute Coronary Syndrome - economics; Acute coronary syndromes; Aged; artificial intelligence; Cardiovascular disease; Cardiovascular diseases; Clinical variables; Coronary artery disease; Cost Savings - economics; Cost Savings - statistics & numerical data; Dialysis; Female; Health Care Costs - statistics & numerical data; Health care expenditures; High risk; Hospitalization; Humans; Learning algorithms; Low density lipoprotein; Machine Learning; Male; modifiable risk factors; Morbidity; population health management; Risk assessment; Risk Factors; Stroke; Syntax; Treatment Outcome; Validity; modifiable risk factors; machine learning; population health management; acute coronary syndromes; artificial intelligence
• ISSN: 1098-3015; 1524-4733
• DOI: 10.1016/j.jval.2020.08.2091

Traditional risk scores improved the definition of the initial therapeutic strategy in acute coronary syndrome (ACS), but they were not designed for predicting long-term individual risks and costs. In parallel, attempts to directly predict costs from clinical variables in ACS had limited success. Thus, novel approaches to predict cardiovascular risk and health expenditure are urgently needed. Our objectives were to predict the risk of major/minor adverse cardiovascular events (MACE) and estimate assistance-related costs.
We used a 2-step approach that: (1) predicted outcomes with a common pathophysiological substrate (MACE) by using machine learning (ML) or logistic regression (LR) and compared with existing risk scores; (2) derived costs associated with noncardiovascular deaths, dialysis, ambulatory-care-sensitive-hospitalizations (ACSH), strokes, and MACE. With consecutive ACS individuals (n = 1089) from 2 cohorts, we trained in 80% of the population and tested in 20% using a 4-fold cross-validation framework. The 29-variable model included socioeconomic, clinical/lab, and coronarography variables. Individual costs were estimated based on cause-specific hospitalization from the Brazilian Health Ministry perspective.
After up to 12 years follow-up (mean = 3.3 ± 3.1; MACE = 169), the gradient-boosting machine model was superior to LR and reached an area under the curve (AUROC) of 0.891 [95% CI 0.846-0.921] (test set), outperforming the Syntax Score II (AUROC = 0.635 [95% CI 0.569-0.699]). Individuals classified as high risk (>90th percentile) presented increased HbA1c and LDL-C both at <24 hours post-ACS and 1-year follow-up. High-risk individuals required 33.5% of total costs and showed 4.96-fold (95% CI 3.71-5.48, P < .00001) greater per capita costs compared with low-risk individuals, mostly owing to avoidable costs (ACSH). This 2-step approach was more successful for finding individuals incurring high costs than predicting costs directly from clinical variables.
ML methods predicted long-term risks and avoidable costs after ACS.
•In acute coronary syndrome (ACS), traditional risk scores are not designed to consider the repercussions of in-hospital therapies or the risk of minor but expensive outcomes, such as reinterventions (including recurrent coronary catheterization), dialysis, recurrent angina, and rehospitalization. Therefore, there are no available tools/scores to predict individual health expenditure and no available mechanism to predict avoidable costs.•With 35 variables derived from clinical, socioeconomic, and coronary angiography assessments, we designed a machine learning (ML) model that efficiently predicts long-term risks after ACS. The model helped to identify individuals with a large burden of avoidable costs. In individuals at the top 10% risk threshold, 21.9% of costs were attributed to avoidable costs. Indeed, we also show that individuals at high risk also presented a higher burden of uncontrolled modifiable risk factors, such as hypercholesterolemia/diabetes, than individuals at low risk.•ML could be useful to select individuals at higher risk for clinical events and who are more likely to incur higher costs in the long term. In Brazil and the United States, the avoidable cost threshold is much higher than the cost of many high-priced therapies. Thus, this ML approach could also turn into a framework to favor access of high-cost therapies to individuals predicted as high risk, including therapies that are not considered cost-effective in the general population.