Search

Abstract Background The healthcare sector is rapidly integrating artificial intelligence-derived predictive models (AIDPM) to enhance clinical decision support, operational efficiency, and patient experiences. However, research on management strategies for AIDPM acquisition, deployment, and governance remains limited. This study examines changes in AIDPM integration and governance since 2021, with a particular focus on large language models and health equity considerations. Results Our survey of health system leaders achieved a 49% response rate (32/65). While 84% of institutions reported using AIDPM in clinical practice, only 53% had established dedicated teams for these models. Compared to 2021, there was a significant increase in representation from experts in clinical informatics, operations, and quality improvement on AIDPM teams. Most organizations (41%) primarily purchased AIDPM from external vendors. Support for integrating large language models into healthcare practices was unanimous among respondents. The principal obstacles to AIDPM adoption included regulatory concerns, data security, workflow integration, and clinician acceptance. A large majority (72%) of respondents supported government regulation of AI in healthcare. While 76% of organizations reported having a team member dedicated to health equity, ethicists and diversity leaders were underrepresented on AIDPM teams (18%). Organizations reported various efforts to promote health equity, but involvement of frontline clinicians in AIDPM development and its impact on health equity was significantly less common. Conclusions Clinical adoption of AIDPM faces challenges due to the absence of established best practices. Health system leaders strongly support federal regulations for AI in healthcare. These regulations could provide quality and safety standards. The study highlights the need for developing evaluation guidelines, especially for large language models. It also reveals a lack of uniform involvement of frontline clinicians and equity experts in AIDPM governance. Their involvement could increase adoption and trust of these new AI tools. Future research should assess healthcare systems' adherence to emerging regulations and best practice frameworks. This research should emphasize patient safety and health equity. These findings underscore the urgent need for a comprehensive roadmap. This roadmap would guide the responsible implementation of AIDPM in healthcare settings.

Abstract Background Malaria control in Panama is problematic due to the high diversity of morphologically similar Anopheles mosquito species, which makes identification of vectors of human Plasmodium challenging. Strategies by Panamanian health authorities to bring malaria under control targeting Anopheles vectors could be ineffective if they tackle a misidentified species. Methods A rapid mass spectrometry identification procedure was developed to accurately and timely sort out field-collected Neotropical Anopheles mosquitoes into vector and non-vector species. Matrix-assisted laser desorption/ionization (MALDI) mass spectra of highly-abundant proteins were generated from laboratory-reared mosquitoes using different extraction protocols, body parts, and sexes to minimize the amount of material from specimen vouchers needed and optimize the protocol for taxonomic identification. Subsequently, the mass spectra of field-collected Neotropical Anopheles mosquito species were classified using a combination of custom-made unsupervised (i.e., Principal component analysis—PCA) and supervised (i.e., Linear discriminant analysis—LDA) classification algorithms. Results Regardless of the protocol used or the mosquito species and sex, the legs contained the least intra-specific variability with enough well-preserved proteins to differentiate among distinct biological species, consistent with previous literature. After minimizing the amount of material needed from the voucher, one leg was enough to produce reliable spectra between specimens. Further, both PCA and LDA were able to classify up to 12 mosquito species, from different subgenera and seven geographically spread localities across Panama using mass spectra from one leg pair. LDA demonstrated high discriminatory power and consistency, with validation and cross-validation positive identification rates above 93% at the species level. Conclusion The selected sample processing procedure can be used to identify field-collected Anopheles species, including vectors of Plasmodium, in a short period of time, with a minimal amount of tissue and without the need of an expert mosquito taxonomist. This strategy to analyse protein spectra overcomes the drawbacks of working without a reference library to classify unknown samples. Finally, this MALDI approach can aid ongoing malaria eradication efforts in Panama and other countries with large number of mosquito’s species by improving vector surveillance in epidemic-prone sites such as indigenous Comarcas.

BackgroundRegular medical care is important for people living with HIV. A no-show predictive model among people with HIV could improve clinical care by allowing providers to proactively engage patients at high risk of missing appointments. Epic, a major provider of electronic medical record systems, created a model that predicts a patient’s probability of being a no-show for an outpatient health care appointment; however, this model has not been externally validated in people with HIV. ObjectiveWe examined the performance of Epic’s no-show model among people with HIV at an academic medical center and assessed whether the performance was impacted by the addition of demographic and HIV clinical information. MethodsWe obtained encounter data from all in-person appointments among people with HIV from January 21 to March 30, 2022, at the University of Chicago Medicine. We compared the predicted no-show probability at the time of the encounter to the actual outcome of these appointments. We also examined the performance of the Epic model among people with HIV for only HIV care appointments in the infectious diseases department. We further compared the no-show model among people with HIV for HIV care appointments to an alternate random forest model we created using a subset of seven readily accessible features used in the Epic model and four additional features related to HIV clinical care or demographics. ResultsWe identified 674 people with HIV who contributed 1406 total scheduled in-person appointments during the study period. Of those, we identified 331 people with HIV who contributed 440 HIV care appointments. The performance of the Epic model among people with HIV for all appointments in any outpatient clinic had an area under the receiver operating characteristic curve (AUC) of 0.65 (95% CI 0.63-0.66) and for only HIV care appointments had an AUC of 0.63 (95% CI 0.59-0.67). The alternate model we created for people with HIV attending HIV care appointments had an AUC of 0.78 (95% CI 0.75-0.82), a significant improvement over the Epic model restricted to HIV care appointments (P200 copies per mL, lower CD4 T cell counts (both 50 to

BackgroundThe intensive care unit (ICU)–ward transfer poses a particularly high-risk period for patients. The period after transfer has been associated with adverse events and additional work for care teams related to miscommunication or omission of information. Standardized handoff processes have been found to reduce communication errors and adverse patient events in other clinical environments but are understudied at the ICU-ward interface. We previously developed an electronic ICU-ward transfer tool, ICU-PAUSE, which embeds the key elements and diagnostic reasoning to facilitate a safe transfer of care at ICU discharge. ObjectiveThe aim of this study is to evaluate the implementation process of the ICU-PAUSE handoff tool across 10 academic medical centers, including the rate of adoption and acceptability, as perceived by clinical care teams. MethodsICU-PAUSE will be implemented in the medical ICU across 10 academic hospitals, with each site customizing the tool to their institution’s needs. Our mixed methods study will include a combination of a chart review, quantitative surveys, and qualitative interviews. After a 90-day implementation period, we will conduct a retrospective chart review to evaluate the rate of uptake of ICU-PAUSE. We will also conduct postimplementation surveys of providers to assess perceptions of the tool and its impact on the frequency of communication errors and adverse events during ICU-ward transfers. Lastly, we will conduct semistructured interviews of faculty stakeholders with subsequent thematic analysis with the goal of identifying benefits and barriers in implementing and using ICU-PAUSE. ResultsICU-PAUSE was piloted in the medical ICU at Barnes-Jewish Hospital, the teaching hospital of Washington University School of Medicine in St. Louis, in 2019. As of July 2022, implementation of ICU-PAUSE is ongoing at 6 of 10 participating sites. Our results will be published in 2023. ConclusionsOur process of ICU-PAUSE implementation embeds each step of template design, uptake, and customization in the needs of users and key stakeholders. Here, we introduce our approach to evaluate its acceptability, usability, and impact on communication errors according to the tenets of sociotechnical theory. We anticipate that ICU-PAUSE will offer an effective handoff tool for the ICU-ward transition that can be generalized to other institutions. International Registered Report Identifier (IRRID)DERR1-10.2196/40918

BackgroundAutomated and data-driven methods for screening using natural language processing (NLP) and machine learning may replace resource-intensive manual approaches in the usual care of patients hospitalized with conditions related to unhealthy substance use. The rigorous evaluation of tools that use artificial intelligence (AI) is necessary to demonstrate effectiveness before system-wide implementation. An NLP tool to use routinely collected data in the electronic health record was previously validated for diagnostic accuracy in a retrospective study for screening unhealthy substance use. Our next step is a noninferiority design incorporated into a research protocol for clinical implementation with prospective evaluation of clinical effectiveness in a large health system. ObjectiveThis study aims to provide a study protocol to evaluate health outcomes and the costs and benefits of an AI-driven automated screener compared to manual human screening for unhealthy substance use. MethodsA pre-post design is proposed to evaluate 12 months of manual screening followed by 12 months of automated screening across surgical and medical wards at a single medical center. The preintervention period consists of usual care with manual screening by nurses and social workers and referrals to a multidisciplinary Substance Use Intervention Team (SUIT). Facilitated by a NLP pipeline in the postintervention period, clinical notes from the first 24 hours of hospitalization will be processed and scored by a machine learning model, and the SUIT will be similarly alerted to patients who flagged positive for substance misuse. Flowsheets within the electronic health record have been updated to capture rates of interventions for the primary outcome (brief intervention/motivational interviewing, medication-assisted treatment, naloxone dispensing, and referral to outpatient care). Effectiveness in terms of patient outcomes will be determined by noninferior rates of interventions (primary outcome), as well as rates of readmission within 6 months, average time to consult, and discharge rates against medical advice (secondary outcomes) in the postintervention period by a SUIT compared to the preintervention period. A separate analysis will be performed to assess the costs and benefits to the health system by using automated screening. Changes from the pre- to postintervention period will be assessed in covariate-adjusted generalized linear mixed-effects models. ResultsThe study will begin in September 2022. Monthly data monitoring and Data Safety Monitoring Board reporting are scheduled every 6 months throughout the study period. We anticipate reporting final results by June 2025. ConclusionsThe use of augmented intelligence for clinical decision support is growing with an increasing number of AI tools. We provide a research protocol for prospective evaluation of an automated NLP system for screening unhealthy substance use using a noninferiority design to demonstrate comprehensive screening that may be as effective as manual screening but less costly via automated solutions. Trial RegistrationClinicalTrials.gov NCT03833804; https://clinicaltrials.gov/ct2/show/NCT03833804 International Registered Report Identifier (IRRID)DERR1-10.2196/42971