Pencina, Michael

Moritz F Sinner, 2012 Finalist and Presenting Author

Importance Given the importance of rigorous development and evaluation standards needed of artificial intelligence (AI) models used in health care, nationwide accepted procedures to provide assurance that the use of AI is fair, appropriate, valid, effective, and safe are urgently needed. Observations While there are several efforts to develop standards and best practices to evaluate AI, there is a gap between having such guidance and the application of such guidance to both existing and new AI models being developed. As of now, there is no publicly available, nationwide mechanism that enables objective evaluation and ongoing assessment of the consequences of using health AI models in clinical care settings. Conclusion and Relevance The need to create a public-private partnership to support a nationwide health AI assurance labs network is outlined here. In this network, community best practices could be applied for testing health AI models to produce reports on their performance that can be widely shared for managing the lifecycle of AI models over time and across populations and sites where these models are deployed.

We are at a pivotal moment in the development of healthcare artificial intelligence (AI), a point at which enthusiasm for machine learning has not caught up with the scientific evidence to support the equity and accuracy of diagnostic and therapeutic algorithms. This proposal examines algorithmic biases, including those related to race, gender and socioeconomic status, and accuracy, including the paucity of prospective studies and lack of multisite validation. We then suggest solutions to these problems. We describe the Mayo Clinic, Duke University, Change Healthcare project that is evaluating 35.1 billion healthcare records for bias. And we propose ‘Ingredients’ style labels and an AI evaluation/testing system to help clinicians judge the merits of products and services that include algorithms. Said testing would include input data sources and types, dataset population composition, algorithm validation techniques, bias assessment evaluation and performance metrics.

Background Tools for the prediction of atrial fibrillation ( AF ) may identify high‐risk individuals more likely to benefit from preventive interventions and serve as a benchmark to test novel putative risk factors. Methods and Results Individual‐level data from 3 large cohorts in the U nited S tates (Atherosclerosis Risk in Communities [ ARIC ] study, the Cardiovascular Health Study [ CHS ], and the Framingham Heart Study [ FHS ]), including 18 556 men and women aged 46 to 94 years (19% A frican A mericans, 81% whites) were pooled to derive predictive models for AF using clinical variables. Validation of the derived models was performed in 7672 participants from the Age, Gene and Environment—Reykjavik study ( AGES ) and the Rotterdam Study ( RS ). The analysis included 1186 incident AF cases in the derivation cohorts and 585 in the validation cohorts. A simple 5‐year predictive model including the variables age, race, height, weight, systolic and diastolic blood pressure, current smoking, use of antihypertensive medication, diabetes, and history of myocardial infarction and heart failure had good discrimination (C‐statistic, 0.765; 95% CI , 0.748 to 0.781). Addition of variables from the electrocardiogram did not improve the overall model discrimination (C‐statistic, 0.767; 95% CI , 0.750 to 0.783; categorical net reclassification improvement, −0.0032; 95% CI , −0.0178 to 0.0113). In the validation cohorts, discrimination was acceptable ( AGES C‐statistic, 0.664; 95% CI , 0.632 to 0.697 and RS C‐statistic, 0.705; 95% CI , 0.664 to 0.747) and calibration was adequate. Conclusion A risk model including variables readily available in primary care settings adequately predicted AF in diverse populations from the U nited S tates and E urope.