Recommendations for the safe, effective use of adaptive CDS in the US healthcare system: an AMIA position paper

AI-Assisted Decision-Making in Long-Term Care: Qualitative Study on Prerequisites for Responsible Innovation

BACKGROUND

Although the use of artificial intelligence (AI)-based technologies, such as AI-based decision support systems (AI-DSSs), can help sustain and improve the quality and efficiency of care, their deployment creates ethical and social challenges. In recent years, a growing prevalence of high-level guidelines and frameworks for responsible AI innovation has been observed. However, few studies have specified the responsible embedding of AI-based technologies, such as AI-DSSs, in specific contexts, such as the nursing process in long-term care (LTC) for older adults.

OBJECTIVE

Prerequisites for responsible AI-assisted decision-making in nursing practice were explored from the perspectives of nurses and other professional stakeholders in LTC.

METHODS

Semistructured interviews were conducted with 24 care professionals in Dutch LTC, including nurses, care coordinators, data specialists, and care centralists. A total of 2 imaginary scenarios about AI-DSSs were developed beforehand and used to enable participants articulate their expectations regarding the opportunities and risks of AI-assisted decision-making. In addition, 6 high-level principles for responsible AI were used as probing themes to evoke further consideration of the risks associated with using AI-DSSs in LTC. Furthermore, the participants were asked to brainstorm possible strategies and actions in the design, implementation, and use of AI-DSSs to address or mitigate these risks. A thematic analysis was performed to identify the opportunities and risks of AI-assisted decision-making in nursing practice and the associated prerequisites for responsible innovation in this area.

RESULTS

The stance of care professionals on the use of AI-DSSs is not a matter of purely positive or negative expectations but rather a nuanced interplay of positive and negative elements that lead to a weighed perception of the prerequisites for responsible AI-assisted decision-making. Both opportunities and risks were identified in relation to the early identification of care needs, guidance in devising care strategies, shared decision-making, and the workload of and work experience of caregivers. To optimally balance the opportunities and risks of AI-assisted decision-making, seven categories of prerequisites for responsible AI-assisted decision-making in nursing practice were identified: (1) regular deliberation on data collection; (2) a balanced proactive nature of AI-DSSs; (3) incremental advancements aligned with trust and experience; (4) customization for all user groups, including clients and caregivers; (5) measures to counteract bias and narrow perspectives; (6) human-centric learning loops; and (7) the routinization of using AI-DSSs.

CONCLUSIONS

The opportunities of AI-assisted decision-making in nursing practice could turn into drawbacks depending on the specific shaping of the design and deployment of AI-DSSs. Therefore, we recommend considering the responsible use of AI-DSSs as a balancing act. Moreover, considering the interrelatedness of the identified prerequisites, we call for various actors, including developers and users of AI-DSSs, to cohesively address the different factors important to the responsible embedding of AI-DSSs in practice.

Resilient Artificial Intelligence in Health: Synthesis and Research Agenda Toward Next-Generation Trustworthy Clinical Decision Support

Artificial intelligence (AI)-based clinical decision support systems are gaining momentum by relying on a greater volume and variety of secondary use data. However, the uncertainty, variability, and biases in real-world data environments still pose significant challenges to the development of health AI, its routine clinical use, and its regulatory frameworks. Health AI should be resilient against real-world environments throughout its lifecycle, including the training and prediction phases and maintenance during production, and health AI regulations should evolve accordingly. Data quality issues, variability over time or across sites, information uncertainty, human-computer interaction, and fundamental rights assurance are among the most relevant challenges. If health AI is not designed resiliently with regard to these real-world data effects, potentially biased data-driven medical decisions can risk the safety and fundamental rights of millions of people. In this viewpoint, we review the challenges, requirements, and methods for resilient AI in health and provide a research framework to improve the trustworthiness of next-generation AI-based clinical decision support.

Considerations for Quality Control Monitoring of Machine Learning Models in Clinical Practice

Integrating machine learning (ML) models into clinical practice presents a challenge of maintaining their efficacy over time. While existing literature offers valuable strategies for detecting declining model performance, there is a need to document the broader challenges and solutions associated with the real-world development and integration of model monitoring solutions. This work details the development and use of a platform for monitoring the performance of a production-level ML model operating in Mayo Clinic. In this paper, we aimed to provide a series of considerations and guidelines necessary for integrating such a platform into a team's technical infrastructure and workflow. We have documented our experiences with this integration process, discussed the broader challenges encountered with real-world implementation and maintenance, and included the source code for the platform. Our monitoring platform was built as an R shiny application, developed and implemented over the course of 6 months. The platform has been used and maintained for 2 years and is still in use as of July 2023. The considerations necessary for the implementation of the monitoring platform center around 4 pillars: feasibility (what resources can be used for platform development?); design (through what statistics or models will the model be monitored, and how will these results be efficiently displayed to the end user?); implementation (how will this platform be built, and where will it exist within the IT ecosystem?); and policy (based on monitoring feedback, when and what actions will be taken to fix problems, and how will these problems be translated to clinical staff?). While much of the literature surrounding ML performance monitoring emphasizes methodological approaches for capturing changes in performance, there remains a battery of other challenges and considerations that must be addressed for successful real-world implementation.

A Machine Learning Application to Classify Patients at Differing Levels of Risk of Opioid Use Disorder: Clinician-Based Validation Study

Background

Despite restrictive opioid management guidelines, opioid use disorder (OUD) remains a major public health concern. Machine learning (ML) offers a promising avenue for identifying and alerting clinicians about OUD, thus supporting better clinical decision-making regarding treatment.

Objective

This study aimed to assess the clinical validity of an ML application designed to identify and alert clinicians of different levels of OUD risk by comparing it to a structured review of medical records by clinicians.

Methods

The ML application generated OUD risk alerts on outpatient data for 649,504 patients from 2 medical centers between 2010 and 2013. A random sample of 60 patients was selected from 3 OUD risk level categories (n=180). An OUD risk classification scheme and standardized data extraction tool were developed to evaluate the validity of the alerts. Clinicians independently conducted a systematic and structured review of medical records and reached a consensus on a patient's OUD risk level, which was then compared to the ML application's risk assignments.

Results

A total of 78,587 patients without cancer with at least 1 opioid prescription were identified as follows: not high risk (n=50,405, 64.1%), high risk (n=16,636, 21.2%), and suspected OUD or OUD (n=11,546, 14.7%). The sample of 180 patients was representative of the total population in terms of age, sex, and race. The interrater reliability between the ML application and clinicians had a weighted kappa coefficient of 0.62 (95% CI 0.53-0.71), indicating good agreement. Combining the high risk and suspected OUD or OUD categories and using the review of medical records as a gold standard, the ML application had a corrected sensitivity of 56.6% (95% CI 48.7%-64.5%) and a corrected specificity of 94.2% (95% CI 90.3%-98.1%). The positive and negative predictive values were 93.3% (95% CI 88.2%-96.3%) and 60.0% (95% CI 50.4%-68.9%), respectively. Key themes for disagreements between the ML application and clinician reviews were identified.

Conclusions

A systematic comparison was conducted between an ML application and clinicians for identifying OUD risk. The ML application generated clinically valid and useful alerts about patients' different OUD risk levels. ML applications hold promise for identifying patients at differing levels of OUD risk and will likely complement traditional rule-based approaches to generating alerts about opioid safety issues.

Applying human-centered design to the construction of a cirrhosis management clinical decision support system

BACKGROUND

Electronic health record (EHR)-based clinical decision support is a scalable way to help standardize clinical care. Clinical decision support systems have not been extensively investigated in cirrhosis management. Human-centered design (HCD) is an approach that engages with potential users in intervention development. In this study, we applied HCD to design the features and interface for a clinical decision support system for cirrhosis management, called CirrhosisRx.

METHODS

We conducted technical feasibility assessments to construct a visual blueprint that outlines the basic features of the interface. We then convened collaborative-design workshops with generalist and specialist clinicians. We elicited current workflows for cirrhosis management, assessed gaps in existing EHR systems, evaluated potential features, and refined the design prototype for CirrhosisRx. At the conclusion of each workshop, we analyzed recordings and transcripts.

RESULTS

Workshop feedback showed that the aggregation of relevant clinical data into 6 cirrhosis decompensation domains (defined as common inpatient clinical scenarios) was the most important feature. Automatic inference of clinical events from EHR data, such as gastrointestinal bleeding from hemoglobin changes, was not accepted due to accuracy concerns. Visualizations for risk stratification scores were deemed not necessary. Lastly, the HCD co-design workshops allowed us to identify the target user population (generalists).

CONCLUSIONS

This is one of the first applications of HCD to design the features and interface for an electronic intervention for cirrhosis management. The HCD process altered features, modified the design interface, and likely improved CirrhosisRx's overall usability. The finalized design for CirrhosisRx proceeded to development and production and will be tested for effectiveness in a pragmatic randomized controlled trial. This work provides a model for the creation of other EHR-based interventions in hepatology care.

Enabling the clinical application of artificial intelligence in genomics: a perspective of the AMIA Genomics and Translational Bioinformatics Workgroup

OBJECTIVE

Given the importance AI in genomics and its potential impact on human health, the American Medical Informatics Association-Genomics and Translational Biomedical Informatics (GenTBI) Workgroup developed this assessment of factors that can further enable the clinical application of AI in this space.

PROCESS

A list of relevant factors was developed through GenTBI workgroup discussions in multiple in-person and online meetings, along with review of pertinent publications. This list was then summarized and reviewed to achieve consensus among the group members.

CONCLUSIONS

Substantial informatics research and development are needed to fully realize the clinical potential of such technologies. The development of larger datasets is crucial to emulating the success AI is achieving in other domains. It is important that AI methods do not exacerbate existing socio-economic, racial, and ethnic disparities. Genomic data standards are critical to effectively scale such technologies across institutions. With so much uncertainty, complexity and novelty in genomics and medicine, and with an evolving regulatory environment, the current focus should be on using these technologies in an interface with clinicians that emphasizes the value each brings to clinical decision-making.

Applicability Area: A novel utility-based approach for evaluating predictive models, beyond discrimination

Translating prediction models into practice and supporting clinicians' decision-making demand demonstration of clinical value. Existing approaches to evaluating machine learning models emphasize discriminatory power, which is only a part of the medical decision problem. We propose the Applicability Area (ApAr), a decision-analytic utility-based approach to evaluating predictive models that communicate the range of prior probability and test cutoffs for which the model has positive utility; larger ApArs suggest a broader potential use of the model. We assess ApAr with simulated datasets and with three published medical datasets. ApAr adds value beyond the typical area under the receiver operating characteristic curve (AUROC) metric analysis. As an example, in the diabetes dataset, the top model by ApAr was ranked as the 23rd best model by AUROC. Decision makers looking to adopt and implement models can leverage ApArs to assess if the local range of priors and utilities is within the respective ApArs.

Utilizing standardized nursing terminologies in implementing an AI-powered fall-prevention tool to improve patient outcomes: a multihospital study

OBJECTIVES

Standardized nursing terminologies (SNTs) are necessary to ensure consistent knowledge expression and compare the effectiveness of nursing practice across settings. This study investigated whether SNTs can support semantic interoperability and outcoming tracking over time by implementing an AI-powered CDS tool for fall prevention across multiple EMR systems.

MATERIALS AND METHODS

The study involved 3 tertiary academic hospitals and 1 public hospital with different EMR systems and nursing terms, and employed an AI-powered CDS tool that determines the fall risk within the next hour (prediction model) and recommends tailored care plans (CDS functions; represented by SNTs). The prediction model was mapped to local data elements and optimized using local data sets. The local nursing statements in CDS functions were mapped using an ICNP-based inpatient fall-prevention catalog. Four implementation models were compared, and patient outcomes and nursing activities were observed longitudinally at one site.

RESULTS

The postimplementation approach was practical for disseminating the AI-powered CDS tool for nursing. The 4 hospitals successfully implemented prediction models with little performance variation; the AUROCs were 0.8051-0.9581. The nursing process data contributed markedly to fall-risk predictions. The local nursing statements on preventing falls covered 48.0%-86.7% of statements. There was no significant longitudinal decrease in the fall rate (P = .160, 95% CI = -1.21 to 0.21 per 1000 hospital days), but rates of interventions provided by nurses were notably increased.

CONCLUSION

SNTs contributed to achieving semantic interoperability among multiple EMR systems to disseminate AI-powered CDS tools and automatically track nursing and patient outcomes.

Evaluating the predictive ability of natural language processing in identifying tertiary/quaternary cases in prioritization workflows for interhospital transfer

Objectives

Tertiary and quaternary (TQ) care refers to complex cases requiring highly specialized health services. Our study aimed to compare the ability of a natural language processing (NLP) model to an existing human workflow in predictively identifying TQ cases for transfer requests to an academic health center.

Materials and methods

Data on interhospital transfers were queried from the electronic health record for the 6-month period from July 1, 2020 to December 31, 2020. The NLP model was allowed to generate predictions on the same cases as the human predictive workflow during the study period. These predictions were then retrospectively compared to the true TQ outcomes.

Results

There were 1895 transfer cases labeled by both the human predictive workflow and the NLP model, all of which had retrospective confirmation of the true TQ label. The NLP model receiver operating characteristic curve had an area under the curve of 0.91. Using a model probability threshold of ≥0.3 to be considered TQ positive, accuracy was 81.5% for the NLP model versus 80.3% for the human predictions (P = .198) while sensitivity was 83.6% versus 67.7% (P<.001).

Discussion

The NLP model was as accurate as the human workflow but significantly more sensitive. This translated to 15.9% more TQ cases identified by the NLP model.

Conclusion

Integrating an NLP model into existing workflows as automated decision support could translate to more TQ cases identified at the onset of the transfer process.

The Impact of Clinical Decision Support on Health Disparities and the Digital Divide

OBJECTIVES

This literature review summarizes relevant studies from the last three years (2020-2022) related to clinical decision support (CDS) and CDS impact on health disparities and the digital divide. This survey identifies current trends and synthesizes evidence-based recommendations and considerations for future development and implementation of CDS tools.

METHODS

We conducted a search in PubMed for literature published between 2020 and 2022. Our search strategy was constructed as a combination of the MEDLINE®/PubMed® Health Disparities and Minority Health Search Strategy and relevant CDS MeSH terms and phrases. We then extracted relevant data from the studies, including priority population when applicable, domain of influence on the disparity being addressed, and the type of CDS being used. We also made note of when a study discussed the digital divide in some capacity and organized the comments into general themes through group discussion.

RESULTS

Our search yielded 520 studies, with 45 included at the conclusion of screening. The most frequent CDS type in this review was point-of-care alerts/reminders (33.3%). Health Care System was the most frequent domain of influence (71.1%), and Blacks/African Americans were the most frequently included priority population (42.2%). Throughout the literature, we found four general themes related to the technology divide: inaccessibility of technology, access to care, trust of technology, and technology literacy.This survey revealed the diversity of CDS being used to address health disparities and several barriers which may make CDS less effective or potentially harmful to certain populations. Regular examinations of literature that feature CDS and address health disparities can help to reveal new strategies and patterns for improving healthcare.

Artificial Intelligence-enabled Decision Support in Surgery: State-of-the-art and Future Directions

OBJECTIVE

To summarize state-of-the-art artificial intelligence-enabled decision support in surgery and to quantify deficiencies in scientific rigor and reporting.

BACKGROUND

To positively affect surgical care, decision-support models must exceed current reporting guideline requirements by performing external and real-time validation, enrolling adequate sample sizes, reporting model precision, assessing performance across vulnerable populations, and achieving clinical implementation; the degree to which published models meet these criteria is unknown.

METHODS

Embase, PubMed, and MEDLINE databases were searched from their inception to September 21, 2022 for articles describing artificial intelligence-enabled decision support in surgery that uses preoperative or intraoperative data elements to predict complications within 90 days of surgery. Scientific rigor and reporting criteria were assessed and reported according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews guidelines.

RESULTS

Sample size ranged from 163-2,882,526, with 8/36 articles (22.2%) featuring sample sizes of less than 2000; 7 of these 8 articles (87.5%) had below-average (<0.83) area under the receiver operating characteristic or accuracy. Overall, 29 articles (80.6%) performed internal validation only, 5 (13.8%) performed external validation, and 2 (5.6%) performed real-time validation. Twenty-three articles (63.9%) reported precision. No articles reported performance across sociodemographic categories. Thirteen articles (36.1%) presented a framework that could be used for clinical implementation; none assessed clinical implementation efficacy.

CONCLUSIONS

Artificial intelligence-enabled decision support in surgery is limited by reliance on internal validation, small sample sizes that risk overfitting and sacrifice predictive performance, and failure to report confidence intervals, precision, equity analyses, and clinical implementation. Researchers should strive to improve scientific quality.

Clinical use of artificial intelligence requires AI-capable organizations

Artificial intelligence-based algorithms are being widely implemented in health care, even as evidence is emerging of bias in their design, problems with implementation, and potential harm to patients. To achieve the promise of using of AI-based tools to improve health, healthcare organizations will need to be AI-capable, with internal and external systems functioning in tandem to ensure the safe, ethical, and effective use of AI-based tools. Ideas are starting to emerge about the organizational routines, competencies, resources, and infrastructures that will be required for safe and effective deployment of AI in health care, but there has been little empirical research. Infrastructures that provide legal and regulatory guidance for managers, clinician competencies for the safe and effective use of AI-based tools, and learner-centric resources such as clear AI documentation and local health ecosystem impact reviews can help drive continuous improvement.

Competencies for the Use of Artificial Intelligence-Based Tools by Health Care Professionals

PURPOSE

The expanded use of clinical tools that incorporate artificial intelligence (AI) methods has generated calls for specific competencies for effective and ethical use. This qualitative study used expert interviews to define AI-related clinical competencies for health care professionals.

METHOD

In 2021, a multidisciplinary team interviewed 15 experts in the use of AI-based tools in health care settings about the clinical competencies health care professionals need to work effectively with such tools. Transcripts of the semistructured interviews were coded and thematically analyzed. Draft competency statements were developed and provided to the experts for feedback. The competencies were finalized using a consensus process across the research team.

RESULTS

Six competency domain statements and 25 subcompetencies were formulated from the thematic analysis. The competency domain statements are: (1) basic knowledge of AI: explain what AI is and describe its health care applications; (2) social and ethical implications of AI: explain how social, economic, and political systems influence AI-based tools and how these relationships impact justice, equity, and ethics; (3) AI-enhanced clinical encounters: carry out AI-enhanced clinical encounters that integrate diverse sources of information in creating patient-centered care plans; (4) evidence-based evaluation of AI-based tools: evaluate the quality, accuracy, safety, contextual appropriateness, and biases of AI-based tools and their underlying data sets in providing care to patients and populations; (5) workflow analysis for AI-based tools: analyze and adapt to changes in teams, roles, responsibilities, and workflows resulting from implementation of AI-based tools; and (6) practice-based learning and improvement regarding AI-based tools: participate in continuing professional development and practice-based improvement activities related to use of AI tools in health care.

CONCLUSIONS

The 6 clinical competencies identified can be used to guide future teaching and learning programs to maximize the potential benefits of AI-based tools and diminish potential harms.

Artificial intelligence-based clinical decision support in pediatrics

Machine learning models may be integrated into clinical decision support (CDS) systems to identify children at risk of specific diagnoses or clinical deterioration to provide evidence-based recommendations. This use of artificial intelligence models in clinical decision support (AI-CDS) may have several advantages over traditional "rule-based" CDS models in pediatric care through increased model accuracy, with fewer false alerts and missed patients. AI-CDS tools must be appropriately developed, provide insight into the rationale behind decisions, be seamlessly integrated into care pathways, be intuitive to use, answer clinically relevant questions, respect the content expertise of the healthcare provider, and be scientifically sound. While numerous machine learning models have been reported in pediatric care, their integration into AI-CDS remains incompletely realized to date. Important challenges in the application of AI models in pediatric care include the relatively lower rates of clinically significant outcomes compared to adults, and the lack of sufficiently large datasets available necessary for the development of machine learning models. In this review article, we summarize key concepts related to AI-CDS, its current application to pediatric care, and its potential benefits and risks. IMPACT: The performance of clinical decision support may be enhanced by the utilization of machine learning-based algorithms to improve the predictive performance of underlying models. Artificial intelligence-based clinical decision support (AI-CDS) uses models that are experientially improved through training and are particularly well suited toward high-dimensional data. The application of AI-CDS toward pediatric care remains limited currently but represents an important area of future research.

A Sociotechnical Systems Framework for the Application of Artificial Intelligence in Health Care Delivery

In the coming years, artificial intelligence (AI) will pervade almost every aspect of the health care delivery system. AI has the potential to improve patient safety (e.g. diagnostic accuracy) as well as reduce the burden on clinicians (e.g. documentation-related workload); however, these benefits are yet to be realized. AI is only one element of a larger sociotechnical system that needs to be considered for effective AI application. In this paper, we describe the current challenges of integrating AI into clinical care and propose a sociotechnical systems (STS) approach for AI design and implementation. We demonstrate the importance of an STS approach through a case study on the design and implementation of a clinical decision support (CDS). In order for AI to reach its potential, the entire work system as well as clinical workflow must be systematically considered throughout the design of AI technology.

Open questions and research gaps for monitoring and updating AI-enabled tools in clinical settings

As the implementation of artificial intelligence (AI)-enabled tools is realized across diverse clinical environments, there is a growing understanding of the need for ongoing monitoring and updating of prediction models. Dataset shift-temporal changes in clinical practice, patient populations, and information systems-is now well-documented as a source of deteriorating model accuracy and a challenge to the sustainability of AI-enabled tools in clinical care. While best practices are well-established for training and validating new models, there has been limited work developing best practices for prospective validation and model maintenance. In this paper, we highlight the need for updating clinical prediction models and discuss open questions regarding this critical aspect of the AI modeling lifecycle in three focus areas: model maintenance policies, performance monitoring perspectives, and model updating strategies. With the increasing adoption of AI-enabled tools, the need for such best practices must be addressed and incorporated into new and existing implementations. This commentary aims to encourage conversation and motivate additional research across clinical and data science stakeholders.

Computerized Clinical Decision Support Systems for the Early Detection of Sepsis Among Pediatric, Neonatal, and Maternal Inpatients: Scoping Review

BACKGROUND

Sepsis is a severe condition associated with extensive morbidity and mortality worldwide. Pediatric, neonatal, and maternal patients represent a considerable proportion of the sepsis burden. Identifying sepsis cases as early as possible is a key pillar of sepsis management and has prompted the development of sepsis identification rules and algorithms that are embedded in computerized clinical decision support (CCDS) systems.

OBJECTIVE

This scoping review aimed to systematically describe studies reporting on the use and evaluation of CCDS systems for the early detection of pediatric, neonatal, and maternal inpatients at risk of sepsis.

METHODS

MEDLINE, Embase, CINAHL, Cochrane, Latin American and Caribbean Health Sciences Literature (LILACS), Scopus, Web of Science, OpenGrey, ClinicalTrials.gov, and ProQuest Dissertations and Theses Global (PQDT) were searched by using a search strategy that incorporated terms for sepsis, clinical decision support, and early detection. Title, abstract, and full-text screening was performed by 2 independent reviewers, who consulted a third reviewer as needed. One reviewer performed data charting with a sample of data. This was checked by a second reviewer and via discussions with the review team, as necessary.

RESULTS

A total of 33 studies were included in this review-13 (39%) pediatric studies, 18 (55%) neonatal studies, and 2 (6%) maternal studies. All studies were published after 2011, and 27 (82%) were published from 2017 onward. The most common outcome investigated in pediatric studies was the accuracy of sepsis identification (9/13, 69%). Pediatric CCDS systems used different combinations of 18 diverse clinical criteria to detect sepsis across the 13 identified studies. In neonatal studies, 78% (14/18) of the studies investigated the Kaiser Permanente early-onset sepsis risk calculator. All studies investigated sepsis treatment and management outcomes, with 83% (15/18) reporting on antibiotics-related outcomes. Usability and cost-related outcomes were each reported in only 2 (6%) of the 31 pediatric or neonatal studies. Both studies on maternal populations were short abstracts.

CONCLUSIONS

This review found limited research investigating CCDS systems to support the early detection of sepsis among pediatric, neonatal, and maternal patients, despite the high burden of sepsis in these vulnerable populations. We have highlighted the need for a consensus definition for pediatric and neonatal sepsis and the study of usability and cost-related outcomes as critical areas for future research.

INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID)

RR2-10.2196/24899.

Defining AMIA's artificial intelligence principles

Recent advances in the science and technology of artificial intelligence (AI) and growing numbers of deployed AI systems in healthcare and other services have called attention to the need for ethical principles and governance. We define and provide a rationale for principles that should guide the commission, creation, implementation, maintenance, and retirement of AI systems as a foundation for governance throughout the lifecycle. Some principles are derived from the familiar requirements of practice and research in medicine and healthcare: beneficence, nonmaleficence, autonomy, and justice come first. A set of principles follow from the creation and engineering of AI systems: explainability of the technology in plain terms; interpretability, that is, plausible reasoning for decisions; fairness and absence of bias; dependability, including "safe failure"; provision of an audit trail for decisions; and active management of the knowledge base to remain up to date and sensitive to any changes in the environment. In organizational terms, the principles require benevolence-aiming to do good through the use of AI; transparency, ensuring that all assumptions and potential conflicts of interest are declared; and accountability, including active oversight of AI systems and management of any risks that may arise. Particular attention is drawn to the case of vulnerable populations, where extreme care must be exercised. Finally, the principles emphasize the need for user education at all levels of engagement with AI and for continuing research into AI and its biomedical and healthcare applications.

Computerized Clinical Decision Support Systems for the Early Detection of Sepsis Among Adult Inpatients: Scoping Review

Background

Sepsis is a significant cause of morbidity and mortality worldwide. Early detection of sepsis followed promptly by treatment initiation improves patient outcomes and saves lives. Hospitals are increasingly using computerized clinical decision support (CCDS) systems for the rapid identification of adult patients with sepsis.

Objective

This scoping review aims to systematically describe studies reporting on the use and evaluation of CCDS systems for the early detection of adult inpatients with sepsis.

Methods

The protocol for this scoping review was previously published. A total of 10 electronic databases (MEDLINE, Embase, CINAHL, the Cochrane database, LILACS [Latin American and Caribbean Health Sciences Literature], Scopus, Web of Science, OpenGrey, ClinicalTrials.gov, and PQDT [ProQuest Dissertations and Theses]) were comprehensively searched using terms for sepsis, CCDS, and detection to identify relevant studies. Title, abstract, and full-text screening were performed by 2 independent reviewers using predefined eligibility criteria. Data charting was performed by 1 reviewer with a second reviewer checking a random sample of studies. Any disagreements were discussed with input from a third reviewer. In this review, we present the results for adult inpatients, including studies that do not specify patient age.

Results

A search of the electronic databases retrieved 12,139 studies following duplicate removal. We identified 124 studies for inclusion after title, abstract, full-text screening, and hand searching were complete. Nearly all studies (121/124, 97.6%) were published after 2009. Half of the studies were journal articles (65/124, 52.4%), and the remainder were conference abstracts (54/124, 43.5%) and theses (5/124, 4%). Most studies used a single cohort (54/124, 43.5%) or before-after (42/124, 33.9%) approach. Across all 124 included studies, patient outcomes were the most frequently reported outcomes (107/124, 86.3%), followed by sepsis treatment and management (75/124, 60.5%), CCDS usability (14/124, 11.3%), and cost outcomes (9/124, 7.3%). For sepsis identification, the systemic inflammatory response syndrome criteria were the most commonly used, alone (50/124, 40.3%), combined with organ dysfunction (28/124, 22.6%), or combined with other criteria (23/124, 18.5%). Over half of the CCDS systems (68/124, 54.8%) were implemented alongside other sepsis-related interventions.

Conclusions

The current body of literature investigating the implementation of CCDS systems for the early detection of adult inpatients with sepsis is extremely diverse. There is substantial variability in study design, CCDS criteria and characteristics, and outcomes measured across the identified literature. Future research on CCDS system usability, cost, and impact on sepsis morbidity is needed.

International Registered Report Identifier (IRRID)

RR2-10.2196/24899

Peeking into a black box, the fairness and generalizability of a MIMIC-III benchmarking model

As artificial intelligence (AI) makes continuous progress to improve quality of care for some patients by leveraging ever increasing amounts of digital health data, others are left behind. Empirical evaluation studies are required to keep biased AI models from reinforcing systemic health disparities faced by minority populations through dangerous feedback loops. The aim of this study is to raise broad awareness of the pervasive challenges around bias and fairness in risk prediction models. We performed a case study on a MIMIC-trained benchmarking model using a broadly applicable fairness and generalizability assessment framework. While open-science benchmarks are crucial to overcome many study limitations today, this case study revealed a strong class imbalance problem as well as fairness concerns for Black and publicly insured ICU patients. Therefore, we advocate for the widespread use of comprehensive fairness and performance assessment frameworks to effectively monitor and validate benchmark pipelines built on open data resources.

Development of a computer-aided text message platform for user engagement with a digital Diabetes Prevention Program: a case study

Digital Diabetes Prevention Programs (dDPP) are novel mHealth applications that leverage digital features such as tracking and messaging to support behavior change for diabetes prevention. Despite their clinical effectiveness, long-term engagement to these programs remains a challenge, creating barriers to adherence and meaningful health outcomes. We partnered with a dDPP vendor to develop a personalized automatic message system (PAMS) to promote user engagement to the dDPP platform by sending messages on behalf of their primary care provider. PAMS innovates by integrating into clinical workflows. User-centered design (UCD) methodologies in the form of iterative cycles of focus groups, user interviews, design workshops, and other core UCD activities were utilized to defined PAMS requirements. PAMS uses computational tools to deliver theory-based, automated, tailored messages, and content to support patient use of dDPP. In this article, we discuss the design and development of our system, including key requirements and features, the technical architecture and build, and preliminary user testing.

Predicting the Travel Distance of Patients to Access Healthcare Using Deep Neural Networks

OBJECTIVE

Improving geographical access remains a key issue in determining the sufficiency of regional medical resources during health policy design. However, patient choices can be the result of the complex interactivity of various factors. The aim of this study is to propose a deep neural network approach to model the complex decision of patient choice in travel distance to access care, which is an important indicator for policymaking in allocating resources.

METHOD

We used the 4-year nationwide insurance data of Taiwan and accumulated the possible features discussed in earlier literature. This study proposes the use of a convolutional neural network (CNN)-based framework to make predictions. The model performance was tested against other machine learning methods. The proposed framework was further interpreted using Integrated Gradients (IG) to analyze the feature weights.

RESULTS

We successfully demonstrated the effectiveness of using a CNN-based framework to predict the travel distance of patients, achieving an accuracy of 0.968, AUC of 0.969, sensitivity of 0.960, and specificity of 0.989. The CNN-based framework outperformed all other methods. In this research, the IG weights are potentially explainable; however, the relationship does not correspond to known indicators in public health.

CONCLUSIONS

Our results demonstrate the feasibility of the deep learning-based travel distance prediction model. It has the potential to guide policymaking in resource allocation. Clinical and Translational Impact Statement- Deep learning technology is feasible in investigating the distance that patients would travel while accessing care. It is a tool that integrates complex interactive variables with highly imbalanced data distributions.

Status of AI-Enabled Clinical Decision Support Systems Implementations in China

BACKGROUND

AI-enabled Clinical Decision Support Systems (AI + CDSSs) were heralded to contribute greatly to the advancement of health care services. There is an increased availability of monetary funds and technical expertise invested in projects and proposals targeting the building and implementation of such systems. Therefore, understanding the actual system implementation status in clinical practice is imperative.

OBJECTIVES

The aim of the study is to understand (1) the current situation of AI + CDSSs clinical implementations in Chinese hospitals and (2) concerns regarding AI + CDSSs current and future implementations.

METHODS

We investigated 160 tertiary hospitals from six provinces and province-level cities. Descriptive analysis, two-sided Fisher exact test, and Mann-Whitney U-test were utilized for analysis.

RESULTS

Thirty-eight of the surveyed hospitals (23.75%) had implemented AI + CDSSs. There were statistical differences on grade, scales, and medical volume between the two groups of hospitals (implemented vs. not-implemented AI + CDSSs, p <0.05). On the 5-point Likert scale, 81.58% (31/38) of respondents rated their overall satisfaction with the systems as "just neutral" to "satisfied." The three most common concerns were system functions improvement and integration into the clinical process, data quality and availability, and methodological bias.

CONCLUSION

While AI + CDSSs were not yet widespread in Chinese clinical settings, professionals recognize the potential benefits and challenges regarding in-hospital AI + CDSSs.

Algorithm Change Protocols in the Regulation of Adaptive Machine Learning-Based Medical Devices

One of the greatest strengths of artificial intelligence (AI) and machine learning (ML) approaches in health care is that their performance can be continually improved based on updates from automated learning from data. However, health care ML models are currently essentially regulated under provisions that were developed for an earlier age of slowly updated medical devices—requiring major documentation reshape and revalidation with every major update of the model generated by the ML algorithm. This creates minor problems for models that will be retrained and updated only occasionally, but major problems for models that will learn from data in real time or near real time. Regulators have announced action plans for fundamental changes in regulatory approaches. In this Viewpoint, we examine the current regulatory frameworks and developments in this domain. The status quo and recent developments are reviewed, and we argue that these innovative approaches to health care need matching innovative approaches to regulation and that these approaches will bring benefits for patients. International perspectives from the World Health Organization, and the Food and Drug Administration’s proposed approach, based around oversight of tool developers’ quality management systems and defined algorithm change protocols, offer a much-needed paradigm shift, and strive for a balanced approach to enabling rapid improvements in health care through AI innovation while simultaneously ensuring patient safety. The draft European Union (EU) regulatory framework indicates similar approaches, but no detail has yet been provided on how algorithm change protocols will be implemented in the EU. We argue that detail must be provided, and we describe how this could be done in a manner that would allow the full benefits of AI/ML-based innovation for EU patients and health care systems to be realized.

Timing of antibiotic therapy in the ICU

Severe or life threatening infections are common among patients in the intensive care unit (ICU). Most infections in the ICU are bacterial or fungal in origin and require antimicrobial therapy for clinical resolution. Antibiotics are the cornerstone of therapy for infected critically ill patients. However, antibiotics are often not optimally administered resulting in less favorable patient outcomes including greater mortality. The timing of antibiotics in patients with life threatening infections including sepsis and septic shock is now recognized as one of the most important determinants of survival for this population. Individuals who have a delay in the administration of antibiotic therapy for serious infections can have a doubling or more in their mortality. Additionally, the timing of an appropriate antibiotic regimen, one that is active against the offending pathogens based on in vitro susceptibility, also influences survival. Thus not only is early empiric antibiotic administration important but the selection of those agents is crucial as well. The duration of antibiotic infusions, especially for β-lactams, can also influence antibiotic efficacy by increasing antimicrobial drug exposure for the offending pathogen. However, due to mounting antibiotic resistance, aggressive antimicrobial de-escalation based on microbiology results is necessary to counterbalance the pressures of early broad-spectrum antibiotic therapy. In this review, we examine time related variables impacting antibiotic optimization as it relates to the treatment of life threatening infections in the ICU. In addition to highlighting the importance of antibiotic timing in the ICU we hope to provide an approach to antimicrobials that also minimizes the unnecessary use of these agents. Such approaches will increasingly be linked to advances in molecular microbiology testing and artificial intelligence/machine learning. Such advances should help identify patients needing empiric antibiotic therapy at an earlier time point as well as the specific antibiotics required in order to avoid unnecessary administration of broad-spectrum antibiotics.

Automation in nursing decision support systems: A systematic review of effects on decision making, care delivery, and patient outcomes

OBJECTIVE

The study sought to summarize research literature on nursing decision support systems (DSSs ); understand which steps of the nursing care process (NCP) are supported by DSSs, and analyze effects of automated information processing on decision making, care delivery, and patient outcomes.

MATERIALS AND METHODS

We conducted a systematic review in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement. PubMed, CINAHL, Cochrane, Embase, Scopus, and Web of Science were searched from January 2014 to April 2020 for studies focusing on DSSs used exclusively by nurses and their effects. Information about the stages of automation (information acquisition, information analysis, decision and action selection, and action implementation), NCP, and effects was assessed.

RESULTS

Of 1019 articles retrieved, 28 met the inclusion criteria, each studying a unique DSS. Most DSSs were concerned with two NCP steps: assessment (82%) and intervention (86%). In terms of automation, all included DSSs automated information analysis and decision selection. Five DSSs automated information acquisition and only one automated action implementation. Effects on decision making, care delivery, and patient outcome were mixed. DSSs improved compliance with recommendations and reduced decision time, but impacts were not always sustainable. There were improvements in evidence-based practice, but impact on patient outcomes was mixed.

CONCLUSIONS

Current nursing DSSs do not adequately support the NCP and have limited automation. There remain many opportunities to enhance automation, especially at the stage of information acquisition. Further research is needed to understand how automation within the NCP can improve nurses' decision making, care delivery, and patient outcomes.

A survey of extant organizational and computational setups for deploying predictive models in health systems

OBJECTIVE

Artificial intelligence (AI) and machine learning (ML) enabled healthcare is now feasible for many health systems, yet little is known about effective strategies of system architecture and governance mechanisms for implementation. Our objective was to identify the different computational and organizational setups that early-adopter health systems have utilized to integrate AI/ML clinical decision support (AI-CDS) and scrutinize their trade-offs.

MATERIALS AND METHODS

We conducted structured interviews with health systems with AI deployment experience about their organizational and computational setups for deploying AI-CDS at point of care.

RESULTS

We contacted 34 health systems and interviewed 20 healthcare sites (58% response rate). Twelve (60%) sites used the native electronic health record vendor configuration for model development and deployment, making it the most common shared infrastructure. Nine (45%) sites used alternative computational configurations which varied significantly. Organizational configurations for managing AI-CDS were distinguished by how they identified model needs, built and implemented models, and were separable into 3 major types: Decentralized translation (n = 10, 50%), IT Department led (n = 2, 10%), and AI in Healthcare (AIHC) Team (n = 8, 40%).

DISCUSSION

No singular computational configuration enables all current use cases for AI-CDS. Health systems need to consider their desired applications for AI-CDS and whether investment in extending the off-the-shelf infrastructure is needed. Each organizational setup confers trade-offs for health systems planning strategies to implement AI-CDS.

CONCLUSION

Health systems will be able to use this framework to understand strengths and weaknesses of alternative organizational and computational setups when designing their strategy for artificial intelligence.

Artificial intelligence sepsis prediction algorithm learns to say "I don't know"

Sepsis is a leading cause of morbidity and mortality worldwide. Early identification of sepsis is important as it allows timely administration of potentially life-saving resuscitation and antimicrobial therapy. We present COMPOSER (COnformal Multidimensional Prediction Of SEpsis Risk), a deep learning model for the early prediction of sepsis, specifically designed to reduce false alarms by detecting unfamiliar patients/situations arising from erroneous data, missingness, distributional shift and data drifts. COMPOSER flags these unfamiliar cases as indeterminate rather than making spurious predictions. Six patient cohorts (515,720 patients) curated from two healthcare systems in the United States across intensive care units (ICU) and emergency departments (ED) were used to train and externally and temporally validate this model. In a sequential prediction setting, COMPOSER achieved a consistently high area under the curve (AUC) (ICU: 0.925-0.953; ED: 0.938-0.945). Out of over 6 million prediction windows roughly 20% and 8% were identified as indeterminate amongst non-septic and septic patients, respectively. COMPOSER provided early warning within a clinically actionable timeframe (ICU: 12.2 [3.2 22.8] and ED: 2.1 [0.8 4.5] hours prior to first antibiotics order) across all six cohorts, thus allowing for identification and prioritization of patients at high risk for sepsis.

Conflicting information from the Food and Drug Administration: Missed opportunity to lead standards for safe and effective medical artificial intelligence solutions

The Food & Drug Administration (FDA) is considering the permanent exemption of premarket notification requirements for several Class I and II medical device products, including several artificial Intelligence (AI)-driven devices. The exemption is based on the need to rapidly more quickly disseminate devices to the public, estimated cost-savings, a lack of documented adverse events reported to the FDA's database. However, this ignores emerging issues related to AI-based devices, including utility, reproducibility and bias that may not only affect an individual but entire populations. We urge the FDA to reinforce the messaging on safety and effectiveness regulations of AI-based Software as a Medical Device products to better promote fair AI-driven clinical decision tools and for preventing harm to the patients we serve.

How machine learning is embedded to support clinician decision making: an analysis of FDA-approved medical devices

Objective

To examine how and to what extent medical devices using machine learning (ML) support clinician decision making.

Methods

We searched for medical devices that were (1) approved by the US Food and Drug Administration (FDA) up till February 2020; (2) intended for use by clinicians; (3) in clinical tasks or decisions and (4) used ML. Descriptive information about the clinical task, device task, device input and output, and ML method were extracted. The stage of human information processing automated by ML-based devices and level of autonomy were assessed.

Results

Of 137 candidates, 59 FDA approvals for 49 unique devices were included. Most approvals (n=51) were since 2018. Devices commonly assisted with diagnostic (n=35) and triage (n=10) tasks. Twenty-three devices were assistive, providing decision support but left clinicians to make important decisions including diagnosis. Twelve automated the provision of information (autonomous information), such as quantification of heart ejection fraction, while 14 automatically provided task decisions like triaging the reading of scans according to suspected findings of stroke (autonomous decisions). Stages of human information processing most automated by devices were information analysis, (n=14) providing information as an input into clinician decision making, and decision selection (n=29), where devices provide a decision.

Conclusion

Leveraging the benefits of ML algorithms to support clinicians while mitigating risks, requires a solid relationship between clinician and ML-based devices. Such relationships must be carefully designed, considering how algorithms are embedded in devices, the tasks supported, information provided and clinicians’ interactions with them.