To the Editor Wang and colleagues1 describe the challenges that arise for deep learning and other black-box machine learning algorithms for medical prediction. The authors rightfully hint at the fact that reliable performance of predictive analytics in health care is far from guaranteed by discussing data quantity, data quality, model generalizability, and interoperability. Machine-learning algorithms are prone to overfitting owing to small sample size or very flexible modeling. Moreover, the performance of diagnostic and prognostic algorithms is heterogeneous.2 The characteristics of target populations are dependent on region and hospital, and algorithm performance further depends on local operational characteristics, such as biomarker assay kits, imaging machines, or electronic health record data handling.3 In addition to heterogeneity, patient characteristics, medical technology, and clinical management change continuously. Thus, the data to develop algorithms become outdated.4
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
Van Calster B, Steyerberg EW, Collins GS. Artificial Intelligence Algorithms for Medical Prediction Should Be Nonproprietary and Readily Available. JAMA Intern Med. 2019;179(5):731. doi:10.1001/jamainternmed.2019.0597
Customize your JAMA Network experience by selecting one or more topics from the list below.
Create a personal account or sign in to: