Translating ethical and quality principles for the effective, safe and fair development, deployment and use of artificial intelligence technologies in healthcare.

OBJECTIVE: The complexity and rapid pace of development of algorithmic technologies pose challenges for their regulation and oversight in healthcare settings. We sought to improve our institution's approach to evaluation and governance of algorithmic technologies used in clinical care and operations by creating an Implementation Guide that standardizes evaluation criteria so that local oversight is performed in an objective fashion. MATERIALS AND METHODS: Building on a framework that applies key ethical and quality principles (clinical value and safety, fairness and equity, usability and adoption, transparency and accountability, and regulatory compliance), we created concrete guidelines for evaluating algorithmic technologies at our institution. RESULTS: An Implementation Guide articulates evaluation criteria used during review of algorithmic technologies and details what evidence supports the implementation of ethical and quality principles for trustworthy health AI. Application of the processes described in the Implementation Guide can lead to algorithms that are safer as well as more effective, fair, and equitable upon implementation, as illustrated through 4 examples of technologies at different phases of the algorithmic lifecycle that underwent evaluation at our academic medical center. DISCUSSION: By providing clear descriptions/definitions of evaluation criteria and embedding them within standardized processes, we streamlined oversight processes and educated communities using and developing algorithmic technologies within our institution. CONCLUSIONS: We developed a scalable, adaptable framework for translating principles into evaluation criteria and specific requirements that support trustworthy implementation of algorithmic technologies in patient care and healthcare operations.

Curating a knowledge base for individuals with coinfection of HIV and SARS-CoV-2: a study protocol of EHR-based data mining and clinical implementation.

INTRODUCTION: Despite a higher risk of severe COVID-19 disease in individuals with HIV, the interactions between SARS-CoV-2 and HIV infections remain unclear. To delineate these interactions, multicentre Electronic Health Records (EHR) hold existing promise to provide full-spectrum and longitudinal clinical data, demographics and sociobehavioural data at individual level. Presently, a comprehensive EHR-based cohort for the HIV/SARS-CoV-2 coinfection has not been established; EHR integration and data mining methods tailored for studying the coinfection are urgently needed yet remain underdeveloped. METHODS AND ANALYSIS: The overarching goal of this exploratory/developmental study is to establish an EHR-based cohort for individuals with HIV/SARS-CoV-2 coinfection and perform large-scale EHR-based data mining to examine the interactions between HIV and SARS-CoV-2 infections and systematically identify and validate factors contributing to the severe clinical course of the coinfection. We will use a nationwide EHR database in the USA, namely, National COVID Cohort Collaborative (N3C). Ultimately, collected clinical evidence will be implemented and used to pilot test a clinical decision support prototype to assist providers in screening and referral of at-risk patients in real-world clinics. ETHICS AND DISSEMINATION: The study was approved by the institutional review boards at the University of South Carolina (Pro00121828) as non-human subject study. Study findings will be presented at academic conferences and published in peer-reviewed journals. This study will disseminate urgently needed clinical evidence for guiding clinical practice for individuals with the coinfection at Prisma Health, a healthcare system in collaboration.

A framework for the oversight and local deployment of safe and high-quality prediction models.

Artificial intelligence/machine learning models are being rapidly developed and used in clinical practice. However, many models are deployed without a clear understanding of clinical or operational impact and frequently lack monitoring plans that can detect potential safety signals. There is a lack of consensus in establishing governance to deploy, pilot, and monitor algorithms within operational healthcare delivery workflows. Here, we describe a governance framework that combines current regulatory best practices and lifecycle management of predictive models being used for clinical care. Since January 2021, we have successfully added models to our governance portfolio and are currently managing 52 models.

Overcoming barriers to the adoption and implementation of predictive modeling and machine learning in clinical care: what can we learn from US academic medical centers?

There is little known about how academic medical centers (AMCs) in the US develop, implement, and maintain predictive modeling and machine learning (PM and ML) models. We conducted semi-structured interviews with leaders from AMCs to assess their use of PM and ML in clinical care, understand associated challenges, and determine recommended best practices. Each transcribed interview was iteratively coded and reconciled by a minimum of 2 investigators to identify key barriers to and facilitators of PM and ML adoption and implementation in clinical care. Interviews were conducted with 33 individuals from 19 AMCs nationally. AMCs varied greatly in the use of PM and ML within clinical care, from some just beginning to explore their utility to others with multiple models integrated into clinical care. Informants identified 5 key barriers to the adoption and implementation of PM and ML in clinical care: (1) culture and personnel, (2) clinical utility of the PM and ML tool, (3) financing, (4) technology, and (5) data. Recommendation to the informatics community to overcome these barriers included: (1) development of robust evaluation methodologies, (2) partnership with vendors, and (3) development and dissemination of best practices. For institutions developing clinical PM and ML applications, they are advised to: (1) develop appropriate governance, (2) strengthen data access, integrity, and provenance, and (3) adhere to the 5 rights of clinical decision support. This article highlights key challenges of implementing PM and ML in clinical care at AMCs and suggests best practices for development, implementation, and maintenance at these institutions.

Implementation and Continuous Monitoring of an Electronic Health Record Embedded Readmissions Clinical Decision Support Tool.

Unplanned hospital readmissions represent a significant health care value problem with high costs and poor quality of care. A significant percentage of readmissions could be prevented if clinical inpatient teams were better able to predict which patients were at higher risk for readmission. Many of the current clinical decision support models that predict readmissions are not configured to integrate closely with the electronic health record or alert providers in real-time prior to discharge about a patient's risk for readmission. We report on the implementation and monitoring of the Epic electronic health record-"Unplanned readmission model version 1"-over 2 years from 1/1/2018-12/31/2019. For patients discharged during this time, the predictive capability to discern high risk discharges was reflected in an AUC/C-statistic at our three hospitals of 0.716-0.760 for all patients and 0.676-0.695 for general medicine patients. The model had a positive predictive value ranging from 0.217-0.248 for all patients. We also present our methods in monitoring the model over time for trend changes, as well as common readmissions reduction strategies triggered by the score.

Telehealth transformation: COVID-19 and the rise of virtual care.

The novel coronavirus disease-19 (COVID-19) pandemic has altered our economy, society, and healthcare system. While this crisis has presented the U.S. healthcare delivery system with unprecedented challenges, the pandemic has catalyzed rapid adoption of telehealth, or the entire spectrum of activities used to deliver care at a distance. Using examples reported by U.S. healthcare organizations, including ours, we describe the role that telehealth has played in transforming healthcare delivery during the 3 phases of the U.S. COVID-19 pandemic: (1) stay-at-home outpatient care, (2) initial COVID-19 hospital surge, and (3) postpandemic recovery. Within each of these 3 phases, we examine how people, process, and technology work together to support a successful telehealth transformation. Whether healthcare enterprises are ready or not, the new reality is that virtual care has arrived.

Identified themes of interactive visualizations overlayed onto EHR data: an example of improving birth center operating room efficiency.

OBJECTIVE: While electronic health record (EHR) systems store copious amounts of patient data, aggregating those data across patients can be challenging. Visual analytic tools that integrate with EHR systems allow clinicians to gain better insight and understanding into clinical care and management. We report on our experience building Tableau-based visualizations and integrating them into our EHR system. MATERIALS AND METHODS: Visual analytic tools were created as part of 12 clinician-initiated quality improvement projects. We built the visual analytic tools in Tableau and linked it within our EPIC environment. We identified 5 visual themes that spanned the various projects. To illustrate these themes, we choose 1 exemplary project which aimed to improve obstetric operating room efficiency. RESULTS: Across our 12 projects, we identified 5 visual themes that are integral to project success: scheduling & optimization (in 11/12 projects); provider assessment (10/12); executive assessment (8/12); patient outcomes (7/12); and control and goal charts (2/12). DISCUSSION: Many visualizations share common themes. Identification of these themes has allowed our internal team to be more efficient and directed in developing visualizations for future projects. CONCLUSION: Organizing visual analytics into themes can allow informatics teams to more efficiently provide visual products to clinical collaborators.

Designing risk prediction models for ambulatory no-shows across different specialties and clinics.

Objective: As available data increases, so does the opportunity to develop risk scores on more refined patient populations. In this paper we assessed the ability to derive a risk score for a patient no-showing to a clinic visit. Methods: Using data from 2 264 235 outpatient appointments we assessed the performance of models built across 14 different specialties and 55 clinics. We used regularized logistic regression models to fit and assess models built on the health system, specialty, and clinic levels. We evaluated fits based on their discrimination and calibration. Results: Overall, the results suggest that a relatively robust risk score for patient no-shows could be derived with an average C-statistic of 0.83 across clinic level models and strong calibration. Moreover, the clinic specific models, even with lower training set sizes, often performed better than the more general models. Examination of the individual models showed that risk factors had different degrees of predictability across the different specialties. Implementation of optimal modeling strategies would lead to capturing an additional 4819 no-shows per-year. Conclusion: Overall, this work highlights both the opportunity for and the importance of leveraging the available electronic health record data to develop more refined risk models.

A mobile app for securely capturing and transferring clinical images to the electronic health record: description and preliminary usability study.

BACKGROUND: Photographs are important tools to record, track, and communicate clinical findings. Mobile devices with high-resolution cameras are now ubiquitous, giving clinicians the opportunity to capture and share images from the bedside. However, secure and efficient ways to manage and share digital images are lacking. OBJECTIVE: The aim of this study is to describe the implementation of a secure application for capturing and storing clinical images in the electronic health record (EHR), and to describe initial user experiences. METHODS: We developed CliniCam, a secure Apple iOS (iPhone, iPad) application that allows for user authentication, patient selection, image capture, image annotation, and storage of images as a Portable Document Format (PDF) file in the EHR. We leveraged our organization's enterprise service-oriented architecture to transmit the image file from CliniCam to our enterprise clinical data repository. There is no permanent storage of protected health information on the mobile device. CliniCam also required connection to our organization's secure WiFi network. Resident physicians from emergency medicine, internal medicine, and dermatology used CliniCam in clinical practice for one month. They were then asked to complete a survey on their experience. We analyzed the survey results using descriptive statistics. RESULTS: Twenty-eight physicians participated and 19/28 (68%) completed the survey. Of the respondents who used CliniCam, 89% found it useful or very useful for clinical practice and easy to use, and wanted to continue using the app. Respondents provided constructive feedback on location of the photos in the EHR, preferring to have photos embedded in (or linked to) clinical notes instead of storing them as separate PDFs within the EHR. Some users experienced difficulty with WiFi connectivity which was addressed by enhancing CliniCam to check for connectivity on launch. CONCLUSIONS: CliniCam was implemented successfully and found to be easy to use and useful for clinical practice. CliniCam is now available to all clinical users in our hospital, providing a secure and efficient way to capture clinical images and to insert them into the EHR. Future clinical image apps should more closely link clinical images and clinical documentation and consider enabling secure transmission over public WiFi or cellular networks.

A safe practice standard for barcode technology.

OBJECTIVE: Safety advocates have identified barcode verification technology as an important tool to improve health-care practices. METHODS: We evaluated the evidence for the role of barcode technology in improving a wide range of medication safety outcomes across a broad range of settings. Important implementation issues were highlighted to guide standards for the safe adoption of barcode technology. RESULTS: Adverse drug events are common, occurring frequently in both inpatient and outpatient settings. Although approximately half of all preventable adverse drug events in inpatients result from medication errors arising from transcription, dispensing, and administration, these errors are far less likely to be caught than in any of the earlier stages of the medication use process and are therefore most amenable to improvement. When integrated with electronic medication administration records, barcode systems are associated with complete elimination of transcription errors. Furthermore, barcode-assisted dispensing systems are associated with 93% to 96% reductions in dispensing errors, and 85% reductions in potential adverse drug events in dispensing. Most studies have reported large and significant reductions in administration errors by up to 80% after implementation of barcode medication administration systems. Although most studies of barcode technology have been conducted in the adult inpatient setting, the limited data available also support their benefit in pediatric and outpatient settings. CONCLUSIONS: There is growing evidence for the efficacy of barcode solutions in improving overall medication safety. Standards for the implementation of barcode technology are proposed.

Using a medical simulation center as an electronic health record usability laboratory.

Usability testing is increasingly being recognized as a way to increase the usability and safety of health information technology (HIT). Medical simulation centers can serve as testing environments for HIT usability studies. We integrated the quality assurance version of our emergency department (ED) electronic health record (EHR) into our medical simulation center and piloted a clinical care scenario in which emergency medicine resident physicians evaluated a simulated ED patient and documented electronically using the ED EHR. Meticulous planning and close collaboration with expert simulation staff was important for designing test scenarios, pilot testing, and running the sessions. Similarly, working with information systems teams was important for integration of the EHR. Electronic tools are needed to facilitate entry of fictitious clinical results while the simulation scenario is unfolding. EHRs can be successfully integrated into existing simulation centers, which may provide realistic environments for usability testing, training, and evaluation of human-computer interactions.

Impact of an automated email notification system for results of tests pending at discharge: a cluster-randomized controlled trial.

BACKGROUND AND OBJECTIVE: Physician awareness of the results of tests pending at discharge (TPADs) is poor. We developed an automated system that notifies responsible physicians of TPAD results via secure, network email. We sought to evaluate the impact of this system on self-reported awareness of TPAD results by responsible physicians, a necessary intermediary step to improve management of TPAD results. METHODS: We conducted a cluster-randomized controlled trial at a major hospital affiliated with an integrated healthcare delivery network in Boston, Massachusetts. Adult patients with TPADs who were discharged from inpatient general medicine and cardiology services were assigned to the intervention or usual care arm if their inpatient attending physician and primary care physician (PCP) were both randomized to the same study arm. Patients of physicians randomized to discordant study arms were excluded. We surveyed these physicians 72 h after all TPAD results were finalized. The primary outcome was awareness of TPAD results by attending physicians. Secondary outcomes included awareness of TPAD results by PCPs, awareness of actionable TPAD results, and provider satisfaction. RESULTS: We analyzed data on 441 patients. We sent 441 surveys to attending physicians and 353 surveys to PCPs and received 275 and 152 responses from 83 different attending physicians and 112 different PCPs, respectively (attending physician survey response rate of 63%). Intervention attending physicians and PCPs were significantly more aware of TPAD results (76% vs 38%, adjusted/clustered OR 6.30 (95% CI 3.02 to 13.16), p<0.001; 57% vs 33%, adjusted/clustered OR 3.08 (95% CI 1.43 to 6.66), p=0.004, respectively). Intervention attending physicians tended to be more aware of actionable TPAD results (59% vs 29%, adjusted/clustered OR 4.25 (0.65, 27.85), p=0.13). One hundred and eighteen (85%) and 43 (63%) intervention attending physician and PCP survey respondents, respectively, were satisfied with this intervention. CONCLUSIONS: Automated email notification represents a promising strategy for managing TPAD results, potentially mitigating an unresolved patient safety concern. CLINICAL TRIAL REGISTRATION: ClinicalTrials.gov (NCT01153451).

Unrealized potential and residual consequences of electronic prescribing on pharmacy workflow in the outpatient pharmacy.

INTRODUCTION: Electronic prescribing systems have often been promoted as a tool for reducing medication errors and adverse drug events. Recent evidence has revealed that adoption of electronic prescribing systems can lead to unintended consequences such as the introduction of new errors. The purpose of this study is to identify and characterize the unrealized potential and residual consequences of electronic prescribing on pharmacy workflow in an outpatient pharmacy. METHODS: A multidisciplinary team conducted direct observations of workflow in an independent pharmacy and semi-structured interviews with pharmacy staff members about their perceptions of the unrealized potential and residual consequences of electronic prescribing systems. We used qualitative methods to iteratively analyze text data using a grounded theory approach, and derive a list of major themes and subthemes related to the unrealized potential and residual consequences of electronic prescribing. RESULTS: We identified the following five themes: Communication, workflow disruption, cost, technology, and opportunity for new errors. These contained 26 unique subthemes representing different facets of our observations and the pharmacy staff's perceptions of the unrealized potential and residual consequences of electronic prescribing. DISCUSSION: We offer targeted solutions to improve electronic prescribing systems by addressing the unrealized potential and residual consequences that we identified. These recommendations may be applied not only to improve staff perceptions of electronic prescribing systems but also to improve the design and/or selection of these systems in order to optimize communication and workflow within pharmacies while minimizing both cost and the potential for the introduction of new errors.

The hazard of software updates to clinical workstations: a natural experiment.

Emergency department (ED) electronic tracking boards provide a snapshot view of patient status and a quick link to other clinical applications, such as a web-based image viewer client to view current and previous radiology images from the picture archiving and communication systems (PACS). We describe a case where an update to Microsoft Internet Explorer severed the link between the ED tracking board and web-based image viewer. The loss of this link resulted in decreased web-based image viewer access rates for ED patients during the 10 days of the incident (2.8 views/study) compared with image review rates for a similar 10-day period preceding this event (3.8 views/study, p<0.001). Single-click user interfaces that transfer user and patient contexts are efficient mechanisms to link disparate clinical systems. Maintaining hazard analyses and rigorously testing all software updates to clinical workstations, including seemingly minor web-browser updates, are important to minimize the risk of unintended consequences.

Adverse drug events caused by serious medication administration errors.

OBJECTIVE: To determine how often serious or life-threatening medication administration errors with the potential to cause harm (potential adverse drug events) result in actual harm (adverse drug events (ADEs)) in the hospital setting. DESIGN: Retrospective chart review of clinical events following observed medication administration errors. BACKGROUND: Medication errors are common at the medication administration stage for inpatients. While many errors can cause harm, it is unclear exactly how often. METHODS: In a previous study where 14 041 medication administrations were directly observed, 1271 medication administration errors were discovered, of which 133 had the potential to cause serious or life-threatening harm and were considered serious or life-threatening potential adverse drug events. As a follow-up, clinical reviewers conducted detailed chart review of serious or life-threatening potential ADEs to determine if they caused an ADE. Reviewers assessed severity of the ADE and attribution to the error. RESULTS: Ten (7.5% (95% CI 6.98 to 8.01)) actual ADEs resulted from the 133 serious and life-threatening potential ADEs, of which 6 resulted in significant, three in serious, and one life threatening injury. Therefore 4 (3% (95% CI 2.12 to 3.6)) of serious or life threatening potential ADEs led to serious or life threatening ADEs. Half of the ADEs were caused by dosage or monitoring errors for anti-hypertensives. CONCLUSIONS: Unintercepted potential ADEs at the medication administration stage can cause serious patient harm. At hospitals where 6 million doses are administered per year, about 4000 preventable ADEs would be attributable to medication administration errors annually.

Cognitive errors and logistical breakdowns contributing to missed and delayed diagnoses of breast and colorectal cancers: a process analysis of closed malpractice claims.

PURPOSE: To erform a process analysis of missed and delayed diagnoses of breast and colorectal cancers to identify: (1) the cognitive and logistical factors that lead to these diagnostic errors, and (2) prevention strategies. METHODS: Using 56 cases (43 breast, 13 colon) of missed and delayed diagnosis, we performed structured analyses to identify specific points in the diagnostic process in which errors occurred. Each error was classified as either a cognitive error or logistical breakdown. Finally, two physician-investigators identified strategies to prevent the errors in each case. RESULTS: Virtually all cases involved one or more cognitive errors (53/56, 95 %) and approximately half (31/56, 55 %) involved logistical breakdowns. The clinical activity most prone to cognitive error was the selection of the diagnostic strategy, both during the office visit (25/56, 45 %) and during interpretation of test results (22/50, 44 %). Arrangement of follow-up visits with a primary care physician (8/29, 28 %) or specialist physician (7/29, 26 %) were especially prone to logistical breakdowns. Adherence to current clinical guidelines could have prevented at least one error in 66 % of cases and assistance from a patient advocate could have prevented at least one error in 48 % of cases. CONCLUSIONS: Cognitive errors and logistical breakdowns are common among missed and delayed diagnoses of breast and colorectal cancers. Prevention strategies should focus on ensuring improving the effectiveness and use of clinical guidelines in the selection of diagnostic strategy, both during office visits and when interpreting test results. Tools to facilitate communication and to ensure that follow-up visits occur should also be considered.

Effects of an online personal health record on medication accuracy and safety: a cluster-randomized trial.

OBJECTIVE: To determine the effects of a personal health record (PHR)-linked medications module on medication accuracy and safety. DESIGN: From September 2005 to March 2007, we conducted an on-treatment sub-study within a cluster-randomized trial involving 11 primary care practices that used the same PHR. Intervention practices received access to a medications module prompting patients to review their documented medications and identify discrepancies, generating 'eJournals' that enabled rapid updating of medication lists during subsequent clinical visits. MEASUREMENTS: A sample of 267 patients who submitted medications eJournals was contacted by phone 3 weeks after an eligible visit and compared with a matched sample of 274 patients in control practices that received a different PHR-linked intervention. Two blinded physician adjudicators determined unexplained discrepancies between documented and patient-reported medication regimens. The primary outcome was proportion of medications per patient with unexplained discrepancies. RESULTS: Among 121,046 patients in eligible practices, 3979 participated in the main trial and 541 participated in the sub-study. The proportion of medications per patient with unexplained discrepancies was 42% in the intervention arm and 51% in the control arm (adjusted OR 0.71, 95% CI 0.54 to 0.94, p=0.01). The number of unexplained discrepancies per patient with potential for severe harm was 0.03 in the intervention arm and 0.08 in the control arm (adjusted RR 0.31, 95% CI 0.10 to 0.92, p=0.04). CONCLUSIONS: When used, concordance between documented and patient-reported medication regimens and reduction in potentially harmful medication discrepancies can be improved with a PHR medication review tool linked to the provider's medical record. TRIAL REGISTRATION NUMBER: This study was registered at ClinicalTrials.gov (NCT00251875).

Impact of automated alerts on follow-up of post-discharge microbiology results: a cluster randomized controlled trial.

BACKGROUND: Failure to follow up microbiology results pending at the time of hospital discharge can delay diagnosis and treatment of important infections, harm patients, and increase the risk of litigation. Current systems to track pending tests are often inadequate. OBJECTIVE: To design, implement, and evaluate an automated system to improve follow-up of microbiology results that return after hospitalized patients are discharged. DESIGN: Cluster randomized controlled trial. SUBJECTS: Inpatient and outpatient physicians caring for adult patients hospitalized at a large academic hospital from February 2009 to June 2010 with positive and untreated or undertreated blood, urine, sputum, or cerebral spinal fluid cultures returning post-discharge. INTERVENTION: An automated e-mail-based system alerting inpatient and outpatient physicians to positive post-discharge culture results not adequately treated with an antibiotic at the time of discharge. MAIN MEASURES: Our primary outcome was documented follow-up of results within 3 days. Secondary outcomes included physician awareness and assessment of result urgency, impact on clinical assessments and plans, and preferred alerting scenarios. KEY RESULTS: We evaluated the follow-up of 157 post-discharge microbiology results from patients of 121 physicians. We found documented follow-up in 27/97 (28%) results in the intervention group and 8/60 (13%) in the control group [aOR 3.2, (95% CI 1.3-8.4); p=0.01]. Of all inpatient physician respondents, 32/82 (39%) were previously aware of the results, 45/77 (58%) felt the results changed their assessments and plans, 43/77 (56%) felt the results required urgent action, and 67/70 (96%) preferred alerts for current or broader scenarios. CONCLUSION: Our alerting system improved the proportion of important post-discharge microbiology results with documented follow-up, though the proportion remained low. The alerts were well received and may be expanded in the future.