Decision support systems embodying machine learning models offer the promise of an improved standard of care for major depressive disorder, but little is known about how clinicians’ treatment decisions will be influenced by machine learning recommendations and explanations. We used a within-subject factorial experiment to present 220 clinicians with patient vignettes, each with or without a machine-learning (ML) recommendation and one of the multiple forms of explanation. We found that interacting with ML recommendations did not significantly improve clinicians’ treatment selection accuracy, assessed as concordance with expert psychopharmacologist consensus, compared to baseline scenarios in which clinicians made treatment decisions independently. Interacting with incorrect recommendations paired with explanations that included limited but easily interpretable information did lead to a significant reduction in treatment selection accuracy compared to baseline questions. These results suggest that incorrect ML recommendations may adversely impact clinician treatment selections and that explanations are insufficient for addressing overreliance on imperfect ML algorithms. More generally, our findings challenge the common assumption that clinicians interacting with ML tools will perform better than either clinicians or ML algorithms individually.

Of patients with acute pancreatitis (AP), there remains a group who suffer life-threatening complications despite current modes of therapy. To identify factors which distinguish this group from the entire patient population, a retrospectiva analysis of 519 cases of AP occurring over a 5-year period was undertaken. Thirty-one per cent of these patients had a history of alcoholism and 47% had a history of biliary disease. The overall mortality was 12.9%. Of symptoms and signs recorded at the time of admission, hypotension, tachycardia, fever, abdominal mass, and abnormal examination of the lung fields correlated positively with increased mortality. Seven features of the initial laboratory examination correlated with increased mortality. Shock, massive colloid requirement, hypocalcemia, renal failure, and respiratory failure requiring endotracheal intubation were complications associated with the poorest prognosis. Among patients in this series with three or more of these clinical characteristics, maximal nonoperative treatment yielded a survival rate of 29%, compared to the 64% survival rate for a group of patients treated operatively with cholecystostomy, gastrostomy, feeding jejunostomy, and sump drainage of the lesser sac and retroperitoneum.

Major depressive disorder is a debilitating disease affecting 264 million people worldwide. While many antidepressant medications are available, few clinical guidelines support choosing among them. Decision support tools (DSTs) embodying machine learning models may help improve the treatment selection process, but often fail in clinical practice due to poor system integration.We use an iterative, co-design process to investigate clinicians' perceptions of using DSTs in antidepressant treatment decisions. We identify ways in which DSTs need to engage with the healthcare sociotechnical system, including clinical processes, patient preferences, resource constraints, and domain knowledge. Our results suggest that clinical DSTs should be designed as multi-user systems that support patient-provider collaboration and offer on-demand explanations that address discrepancies between predictions and current standards of care. Through this work, we demonstrate how current trends in explainable AI may be inappropriate for clinical environments and consider paths towards designing these tools for real-world medical systems. CCS CONCEPTS• Human-centered computing → User centered design; • Applied computing → Health care information systems; • Information systems → Decision support systems.

As technologies such as personal health records and symptom trackers become more common, we are seeing an increase in patients actively engaging in health tracking behaviors. Patient collected data can provide valuable insight for healthcare providers, particularly in the area of breast cancer. Thus far, little work has examined whether the health information that patients are willing to track and share aligns with the information needs of healthcare providers. Our work provides a comparison between the health information sharing preferences of breast cancer patients, doctors and navigators. We identify discrepancies between stakeholders' preferences, such as patients' hesitation to share feelings of loneliness, signifying where technology can play an important role in helping patients prioritize the health information shared with providers. We present design implications from this work to guide the development of future health information sharing tools that consider the differing needs of healthcare stakeholders.