Addressing AI Algorithmic Bias in Health Care.

This Viewpoint discusses the bias that exists in artificial intelligence (AI) algorithms used in health care despite recent federal rules to prohibit discriminatory outcomes from AI and recommends ways in which health care facilities, AI developers, and regulators could share responsibilities and actions to address bias.

The Compelling Need for Shared Responsibility of AI Oversight: Lessons From Health IT Certification.

This Viewpoint highlights the potential for artificial intelligence (AI) health care tools to introduce unintended patient harm; calls for an efficient, rigorous approach to AI testing and certification that is the shared responsibility of developers and users; and makes recommendations to inform such an approach.

Accounting for taste: preferences mediate the relationship between documentation time and ambulatory physician burnout.

OBJECTIVES: Physician burnout in the US has reached crisis levels, with one source identified as extensive after-hours documentation work in the electronic health record (EHR). Evidence has illustrated that physician preferences for after-hours work vary, such that after-hours work may not be universally burdensome. Our objectives were to analyze variation in preferences for after-hours documentation and assess if preferences mediate the relationship between after-hours documentation time and burnout. MATERIALS AND METHODS: We combined EHR active use data capturing physicians' hourly documentation work with survey data capturing documentation preferences and burnout. Our sample included 318 ambulatory physicians at MedStar Health. We conducted a mediation analysis to estimate if and how preferences mediated the relationship between after-hours documentation time and burnout. Our primary outcome was physician-reported burnout. We measured preferences for after-hours documentation work via a novel survey instrument (Burden Scenarios Assessment). We measured after-hours documentation time in the EHR as the total active time respondents spent documenting between 7 pm and 3 am. RESULTS: Physician preferences varied, with completing clinical documentation after clinic hours while at home the scenario rated most burdensome (52.8% of physicians), followed by dealing with prior authorization (49.5% of physicians). In mediation analyses, preferences partially mediated the relationship between after-hours documentation time and burnout. DISCUSSION: Physician preferences regarding EHR-based work play an important role in the relationship between after-hours documentation time and burnout. CONCLUSION: Studies of EHR work and burnout should incorporate preferences, and operational leaders should assess preferences to better target interventions aimed at EHR-based contributors to burnout.

Development and Evaluation of Patient Safety Interventions: Perspectives of Operational Safety Leaders and Patient Safety Organizations.

OBJECTIVES: The purpose of this study is to understand how patient safety professionals from healthcare facilities and patient safety organizations develop patient safety interventions and the resources used to support intervention development. METHODS: Semistructured interviews were conducted with patient safety professionals at nine healthcare facilities and nine patient safety organizations. Interview data were qualitatively analyzed, and findings were organized by the following: patient safety solutions and interventions, use of external databases, and evaluation of patient safety solutions. RESULTS: Development of patient safety interventions across healthcare facilities and patient safety organizations was similar and included literature searches, internal brainstorming, and interviews. Nearly all patient safety professionals at healthcare facilities reported contacting colleagues at other healthcare facilities to learn about similar safety issues and potential interventions. Additionally, less than half of patient safety professionals at healthcare facilities and patient safety organizations interviewed report data to publicly available patient safety databases. Finally, most patient safety professionals at healthcare facilities and patient safety organizations stated that they evaluate the effectiveness of patient safety interventions; however, they mentioned methods that may be less rigorous including audits, self-reporting, and subjective judgment. CONCLUSIONS: Patient safety professionals often utilize similar methods and resources to develop and evaluate patient safety interventions; however, many of these efforts are not coordinated across healthcare organizations and could benefit from working collectively in a systematic fashion. Additionally, healthcare facilities and patient safety organizations face similar challenges and there are several opportunities for optimization on a national level that may improve patient safety.

Safe and Equitable Pediatric Clinical Use of AI.

This Viewpoint provides recommendations and stakeholder actions to support safe and equitable use of artificial intelligence (AI) in pediatric clinical settings.

Telemedicine and In-Person Visit Modality Mix and Electronic Health Record Use in Primary Care.

This cross-sectional study investigates the association between day-to-day changes in telemedicine share and clinician time spent on electronic health record (EHR) use.

Reading Room Interruptions are Less Disruptive When Using Asynchronous Communication Methods.

Radiologist interruptions, though often necessary, can be disruptive. Prior literature has shown interruptions to be frequent, occurring during cases, and predominantly through synchronous communication methods such as phone or in person causing significant disengagement from the study being read. Asynchronous communication methods are now more widely available in hospital systems such as ours. Considering the increasing use of asynchronous communication methods, we conducted an observational study to understand the evolving nature of radiology interruptions. We hypothesize that compared to interruptions occurring through synchronous methods, interruptions via asynchronous methods reduce the disruptive nature of interruptions by occurring between cases, being shorter, and less severe. During standard weekday hours, 30 radiologists (14 attendings, 12 residents, and 4 fellows) were directly observed for approximately 90-min sessions across three different reading rooms (body, neuroradiology, general). The frequency of interruptions was documented including characteristics such as timing, severity, method, and length. Two hundred twenty-five interruptions (43 Teams, 47 phone, 89 in-person, 46 other) occurred, averaging 2 min and 5 s with 5.2 interruptions per hour. Microsoft Teams interruptions averaged 1 min 12 s with only 60.5% during cases. In-person interruptions averaged 2 min 12 s with 82% during cases. Phone interruptions averaged 2 min and 48 s with 97.9% during cases. A substantial portion of reading room interruptions occur via predominantly asynchronous communication tools, a new development compared to prior literature. Interruptions via predominantly asynchronous communications tools are shorter and less likely to occur during cases. In our practice, we are developing tools and mechanisms to promote asynchronous communication to harness these benefits.

Patient Safety and Artificial Intelligence in Clinical Care.

This Viewpoint offers 3 recommendations for health care organizations and other stakeholders to consider as part of the Health and Human Services' artificial intelligence safety program.

A text mining approach to categorize patient safety event reports by medication error type.

Patient safety reporting systems give healthcare provider staff the ability to report medication related safety events and errors; however, many of these reports go unanalyzed and safety hazards go undetected. The objective of this study is to examine whether natural language processing can be used to better categorize medication related patient safety event reports. 3,861 medication related patient safety event reports that were previously annotated using a consolidated medication error taxonomy were used to develop three models using the following algorithms: (1) logistic regression, (2) elastic net, and (3) XGBoost. After development, models were tested, and model performance was analyzed. We found the XGBoost model performed best across all medication error categories. 'Wrong Drug', 'Wrong Dosage Form or Technique or Route', and 'Improper Dose/Dose Omission' categories performed best across the three models. In addition, we identified five words most closely associated with each medication error category and which medication error categories were most likely to co-occur. Machine learning techniques offer a semi-automated method for identifying specific medication error types from the free text of patient safety event reports. These algorithms have the potential to improve the categorization of medication related patient safety event reports which may lead to better identification of important medication safety patterns and trends.

Electronic health record "gag clauses" and the prevalence of screenshots in peer-reviewed literature.

OBJECTIVE: To determine whether the Office of the National Coordinator's policy change restricting the use of "gag clauses" in contracts between electronic health record (EHR) vendors and healthcare facilities increased the prevalence of screenshots in peer-reviewed literature. MATERIALS AND METHODS: We reviewed EHR usability and safety-related peer-reviewed journal articles from 2015 to 2023 and quantified the number of articles containing screenshots. For those that did not contain screenshots, we identified whether they would have benefited from screenshots. RESULTS: When gag clauses were permitted 6 of 79 (7.6%) of articles contained screenshots and 8 (10.1%) would have benefited from screenshots. When gag clauses were restricted 3 of 40 (7.5%) contained screenshots and 8 (20%) would have benefited from screenshots. DISCUSSION: The policy change does not appear to have an impact on the prevalence of screenshots in peer-reviewed literature. CONCLUSIONS: Additional steps are necessary to promote the use of screenshots in peer-reviewed literature.

Variations in Physician Telemedicine Provision.

This cross-sectional study assesses variation in the provision of telemedicine services among primary care physicians and quantifies the extent to which this variation may be explained by the individual physician vs temporal, patient, or visit factors.

The impact of COVID-19 on primary care accessibility and the role of telehealth for patients with chronic conditions.

Objectives: The objective of this study is to quantify how long patients took to complete their rescheduled primary care appointment pre-pandemic (2019) and during an initial pandemic period (2020). In doing so, the study evaluates telehealth's role in helping primary care patients - particularly in patients with chronic conditions - withstand COVID's significant disruption in care. Methods: Cancelled and completed primary care appointments for adult patients were extracted from the beginning of the pandemic (March 1 to July 31, 2020) and a similar period pre-pandemic (March 1 to July 31, 2019). Days to the subsequent completed visit after cancellation (through June 30, 2021) and appointment modality (in-person, phone, video) were examined. Statistical testing was done to determine statistical significance, and a linear regression was run to control for effects of other study variables. Results: Pre-pandemic patients with chronic conditions needed 52.3 days on average to reschedule their cancelled in-person appointment. During the early pandemic period, chronic condition patients who saw their provider in-person took on average 78.8 days. During the same pre-pandemic period, patients with chronic conditions had their average wait time decrease to 51.5 days when rescheduling via telehealth. These differences were similar for patients without chronic conditions. Conclusions: This analysis shows that telehealth created return to care timelines comparable to the pre-pandemic period which is especially important for patients with chronic conditions. Public interest summary: Telehealth visits (i.e., talking with a physician via phone or video call) help patients continue to receive the medical care they need - especially during disruptive periods such as the COVID pandemic. Access to telehealth is the strongest predictor in determining how soon a patient will complete their reschedule primary care appointment. Because telehealth is so important, health care providers and systems need to continue to offer patients the ability to talk with their physician via phone or video call.

Identifying Electronic Health Record Contributions to Diagnostic Error in Ambulatory Settings Through Legal Claims Analysis.

This qualitative study analyzes closed legal claims data to determine whether problems with electronic health records are associated with diagnostic errors, in which part of the diagnostic process errors occur, and the specific types of errors that occur.

"Are we there yet?" Ten persistent hazards and inefficiencies with the use of medication administration technology from the perspective of practicing nurses.

OBJECTIVES: (1) Characterize persistent hazards and inefficiencies in inpatient medication administration; (2) Explore cognitive attributes of medication administration tasks; and (3) Discuss strategies to reduce medication administration technology-related hazards. MATERIALS AND METHODS: Interviews were conducted with 32 nurses practicing at 2 urban, eastern and western US health systems. Qualitative analysis using inductive and deductive coding included consensus discussion, iterative review, and coding structure revision. We abstracted hazards and inefficiencies through the lens of risks to patient safety and the cognitive perception-action cycle (PAC). RESULTS: Persistent safety hazards and inefficiencies related to MAT organized around the PAC cycle included: (1) Compatibility constraints create information silos; (2) Missing action cues; (3) Intermittent communication flow between safety monitoring systems and nurses; (4) Occlusion of important alerts by other, less helpful alerts; (5) Dispersed information: Information required for tasks is not collocated; (6) Inconsistent data organization: Mismatch of the display and the user's mental model; (7) Hidden medication administration technologies (MAT) limitations: Inaccurate beliefs about MAT functionality contribute to overreliance on the technology; (8) Software rigidity causes workarounds; (9) Cumbersome dependencies between technology and the physical environment; and (10) Technology breakdowns require adaptive actions. DISCUSSION: Errors might persist in medication administration despite successful Bar Code Medication Administration and Electronic Medication Administration Record deployment for reducing errors. Opportunities to improve MAT require a deeper understanding of high-level reasoning in medication administration, including control over the information space, collaboration tools, and decision support. CONCLUSION: Future medication administration technology should consider a deeper understanding of nursing knowledge work for medication administration.

Pilot assessment of an on-demand telehealth 'left without being seen' follow-up programme.

INTRODUCTION: On-demand telehealth can have a high rate of patients requesting visits and dropping off without being seen by a provider, especially during the COVID-19 pandemic. METHODS: On-demand telehealth requests made to a large healthcare system in the USA between 15 March 2020 and 31 May 2020 were included for analysis with a focus on patients who were defined as left without being seen (LWBS). As part of a pilot program a registered nurse attempted to call LWBS patients within 24 hours of their telehealth request and asked if they were ok, if they sought care for their original visit reason, what that care was, or if they still needed guidance. This information and patient demographics were analyzed. RESULTS: During the study period there were 21,610 completed on-demand telehealth visits and 1852 patients for whom there were LWBS attempted follow-ups. Most patients LWBS for a reason that originated from the patient and not associated with the provider or telehealth platform. The mean wait time for LWBS patients was 12.4 min compared to patients waiting 15.1 min before engaging with a provider to complete a visit. Of the 1852 total LWBS patients in the follow-up programme, 819 (44.2%) were successfully contacted with a follow-up phone call. Most of these patients (63.2%) already completed or planned to complete a telehealth visit, 13.6% indicated they no longer needed to see a provider, and 12.8% planned or already completed an in-person visit. Only 2.2% went to an emergency department. DISCUSSION: Results suggest patients can effectively self-manage their care needs.

Electronic Health Record Use Issues and Diagnostic Error: A Scoping Review and Framework.

BACKGROUND: Diagnostic errors are a major source of patient harm, most of which are caused by cognitive errors and biases. Despite research showing the relationship between software systems and cognitive processes, the impact of the electronic health record (EHR) on diagnostic error remains unknown. METHODS: We conducted a scoping review of the scientific literature to (1) survey the association between aspects of the EHR and diagnostic error, and (2) through a human-systems integration lens, identify the types of EHR issues and their impact on the stages of the diagnostic process. RESULTS: We analyzed 11 research articles for the relationship between EHR use and diagnostic error. These articles highlight specific technical, usability, and workflow issues with the EHR that pose risks for diagnostic error at every stage of the diagnostic process. DISCUSSION: Although technical problems such as EHR interoperability and data integrity pose critical issues for the diagnostic process, usability and workflow issues such as poor display design, and inability to track test results also hamper clinicians' ability to track, process, and act in the diagnostic process. Current research methods have limited coverage over clinical settings, are not standardized, and rarely include measures of patient harm. CONCLUSIONS: The available evidence shows that EHRs pose risks for diagnostic error throughout the diagnostic process, with most issues involving their incompatibility with providers' cognitive processing. A structured and systematic model of collecting and reporting on these errors is needed to understand how the EHR shapes the diagnostic process and improve diagnostic accuracy.

Electronic Health Record Legal Settlements in the US Since the 2009 Health Information Technology for Economic and Clinical Health Act.

This cross-sectional study describes allegations of misconduct by electronic health record vendors.