Can AI Help Reduce Disparities in General Medical and Mental Health Care?

Abstract

Background: As machine learning becomes increasingly common in health care applications, concerns have been raised about bias in these systems' data, algorithms, and recommendations. Simply put, as health care improves for some, it might not improve for all. Methods:Two case studies are examined using a machine learning algorithm on unstructured clinical and psychiatric notes to predict intensive care unit (ICU) mortality and 30-day psychiatric readmission with respect to race, gender, and insurance payer type as a proxy for socioeconomic status.Results: Clinical note topics and psychiatric note topics were heterogenous with respect to race, gender, and insurance payer type, which reflects known clinical findings. Differences in prediction accuracy and therefore machine bias are shown with respect to gender and insurance type for ICU mortality and with respect to insurance policy for psychiatric 30-day readmission.Conclusions: This analysis can provide a framework for assessing and identifying disparate impacts of artificial intelligence in health care. Bias in Machine Learning ModelsWhile health care is an inherently data-driven field, most clinicians operate with limited evidence guiding their decisions. Randomized trials estimate average treatment effects for a trial population, but participants in clinical trials often aren't representative of the patient population that ultimately receives the treatment with respect to race and gender. 1,2 As a result, drugs and interventions are not tailored to historically mistreated groups; for example, women, minority groups, and obese patients tend to have generally poorer treatment options and longitudinal health outcomes. [3][4][5][6][7][8][9] Advances in artificial intelligence (AI) and machine learning offer the potential to provide personalized care by taking into account granular patient differences. Machine learning using images, clinical notes, and other electronic health record (EHR) data has been Massachusetts Institute of Technology (MIT) in Cambridge, Massachusetts. She is also pursuing a graduate education in medical sciences certificate in the Harvard-MIT Program in Health Sciences and Technology. She received a bachelor of arts degree in applied math-economics and computer science and a master of science degree in computational science and engineering from Harvard University.

Artificial intelligence in clinical and genomic diagnostics

“…Furthermore, the substructure of genomic data is correlated with population structure, which can lead to the appearance of non-causal trait associations [94]. However, tools that will help to address machine bias are being developed, and careful attention to these issues could not only help to resolve machine bias issues but could eventually lead to diagnostic systems that are free from human bias [95].…”

Substance Use Disorders and COVID-19: Multi-Faceted Problems Which Require Multi-Pronged Solutions

“…Artificial intelligence (AI), another promising technology that could be used during emergency situations, could support trained clinicians to make treatment decisions. Currently, the research on the potential use and benefits of AI in addiction care and mental health services is in early development and needs to address important scientific, legal and ethical issues (108,109). Current AI research is focused on assisting addiction care practitioners with treatment for alcohol use disorder (110), identifying and preventing relapse (111), and identifying risk factors (112,113).…”

“…Practitioners should, however, be aware that algorithms can be subject to biases (due to misclassification and measurement error, missing data, and small sample size) (108). The implication of such biases can be severe as they might create disparities in addiction care (108,109). Involving addiction care specialists and patient advocacy groups from the beginning in the development of AI can facilitate innovative, ethical, acceptable, and effective solutions.…”

Machine learning and artificial intelligence research for patient benefit: 20 critical questions on transparency, replicability, ethics, and effectiveness

“…However, developers still need to demonstrate that when using sensible thresholds, the algorithm does not create or exacerbate inequalities. In fact, several methodological developments in the area of fairness evaluation support this type of analysis,717273 and ML/AI developers and health practitioners should engage with these tools. One way in which researchers might demonstrate bias in key subgroups (eg, in minority ethnic groups, or by age) would be to explicitly present these findings so that users of the algorithm know where it has good or poor predictive accuracy.…”