Retrospective validation study of a machine learning-based software for empirical and organism-targeted antibiotic therapy selection.

Errors in antibiotic prescriptions are frequent, often resulting from the inadequate coverage of the infection-causative microorganism. The efficacy of iAST, a machine-learning-based software offering empirical and organism-targeted antibiotic recommendations, was assessed. The study was conducted in a 12-hospital Spanish institution. After model fine-tuning with 27,531 historical antibiograms, 325 consecutive patients with acute infections were selected for retrospective validation. The primary endpoint was comparing each of the top three of iAST's antibiotic recommendations' success rates (confirmed by antibiogram results) with the antibiotic prescribed by the physicians. Secondary endpoints included examining the same hypothesis within specific study population subgroups and assessing antibiotic stewardship by comparing the percentage of antibiotics recommended that belonged to different World Health Organization AWaRe groups within each arm of the study. All of iAST first three recommendations were non-inferior to doctor prescription in the primary endpoint analysis population as well as the secondary endpoint. The overall success rate of doctors' empirical treatment was 68.93%, while that of the first three iAST options was 91.06% (P < 0.001), 90.63% (P < 0.001), and 91.06% (P < 001), respectively. For organism-targeted therapy, the doctor's overall success rate was 84.16%, and that of the first three ranked iAST options was 97.83% (P < 0.001), 94.09% (P < 0.001), and 91.30% (P < 0.001), respectively. In empirical therapy, compared to physician prescriptions, iAST demonstrated a greater propensity to recommend access antibiotics, fewer watch antibiotics, and higher reserve antibiotics. In organism-targeted therapy, iAST advised a higher utilization of access antibiotics. The present study demonstrates iAST accuracy in predicting antibiotic susceptibility, showcasing its potential to promote effective antibiotic stewardship. CLINICAL TRIALS: This study is registered with ClinicalTrials.gov as NCT06174519.

Getting to 100%: Research Priorities and Unanswered Questions to Inform the US Debate on Universal Health Insurance Coverage.

A majority of Americans favor universal health insurance, but there is uncertainty over how best to achieve this goal. Whatever the insurance design that is implemented, additional details that must be considered include breadth of services covered, restrictions and limits on volumes of services, cost-sharing for individuals, and pricing. In the hopes that research can inform this ongoing debate, we review evidence supporting different models for achieving universal coverage in the US and identify areas where additional research and stakeholder input is needed. Key areas in need of further research include how care should be organized, how costs can be reduced, and what healthcare services universal insurance should cover.

What's in a name? A comparison of attitudes towards artificial intelligence (AI) versus augmented human intelligence (AHI).

BACKGROUND: "Artificial intelligence" (AI) is often referred to as "augmented human intelligence" (AHI). The latter term implies that computers support-rather than replace-human decision-making. It is unclear whether the terminology used affects attitudes and perceptions in practice. METHODS: In the context of a quality improvement project implementing AI/AHI-based decision support in a regional health system, we surveyed staff's attitudes about AI/AHI, randomizing question prompts to refer to either AI or AHI. RESULTS: Ninety-three staff completed surveys. With a power of 0.95 to detect a difference larger than 0.8 points on a 5-point scale, we did not detect a significant difference in responses to six questions regarding attitudes when respondents were alternatively asked about AI versus AHI (mean difference range: 0.04-0.22 points; p > 0.05). CONCLUSION: Although findings may be setting-specific, we observed that use of the terms "AI" and "AHI" in a survey on attitudes of clinical staff elicited similar responses.

Development and validation of an electronic medical record-based alert score for detection of inpatient deterioration outside the ICU.

BACKGROUND: Patients in general medical-surgical wards who experience unplanned transfer to the intensive care unit (ICU) show evidence of physiologic derangement 6-24h prior to their deterioration. With increasing availability of electronic medical records (EMRs), automated early warning scores (EWSs) are becoming feasible. OBJECTIVE: To describe the development and performance of an automated EWS based on EMR data. MATERIALS AND METHODS: We used a discrete-time logistic regression model to obtain an hourly risk score to predict unplanned transfer to the ICU within the next 12h. The model was based on hospitalization episodes from all adult patients (18years) admitted to 21 Kaiser Permanente Northern California (KPNC) hospitals from 1/1/2010 to 12/31/2013. Eligible patients met these entry criteria: initial hospitalization occurred at a KPNC hospital; the hospitalization was not for childbirth; and the EMR had been operational at the hospital for at least 3months. We evaluated the performance of this risk score, called Advanced Alert Monitor (AAM) and compared it against two other EWSs (eCART and NEWS) in terms of their sensitivity, specificity, negative predictive value, positive predictive value, and area under the receiver operator characteristic curve (c statistic). RESULTS: A total of 649,418 hospitalization episodes involving 374,838 patients met inclusion criteria, with 19,153 of the episodes experiencing at least one outcome. The analysis data set had 48,723,248 hourly observations. Predictors included physiologic data (laboratory tests and vital signs); neurological status; severity of illness and longitudinal comorbidity indices; care directives; and health services indicators (e.g. elapsed length of stay). AAM showed better performance compared to NEWS and eCART in all the metrics and prediction intervals. The AAM AUC was 0.82 compared to 0.79 and 0.76 for eCART and NEWS, respectively. Using a threshold that generated 1 alert per day in a unit with a patient census of 35, the sensitivity of AAM was 49% (95% CI: 47.6-50.3%) compared to the sensitivities of eCART and NEWS scores of 44% (42.3-45.1) and 40% (38.2-40.9), respectively. For all three scores, about half of alerts occurred within 12h of the event, and almost two thirds within 24h of the event. CONCLUSION: The AAM score is an example of a score that takes advantage of multiple data streams now available in modern EMRs. It highlights the ability to harness complex algorithms to maximize signal extraction. The main challenge in the future is to develop detection approaches for patients in whom data are sparser because their baseline risk is lower.

Why the C-statistic is not informative to evaluate early warning scores and what metrics to use.

Metrics typically used to report the performance of an early warning score (EWS), such as the area under the receiver operator characteristic curve or C-statistic, are not useful for pre-implementation analyses. Because physiological deterioration has an extremely low prevalence of 0.02 per patient-day, these metrics can be misleading. We discuss the statistical reasoning behind this statement and present a novel alternative metric more adequate to operationalize an EWS. We suggest that pre-implementation evaluation of EWSs should include at least two metrics: sensitivity; and either the positive predictive value, number needed to evaluate, or estimated rate of alerts. We also argue the importance of reporting each individual cutoff value.

Transient neonatal zinc deficiency due to a new autosomal dominant mutation in gene SLC30A2 (ZnT-2).

Transient neonatal zinc deficiency (TNZD) has a clinical presentation similar to that of acrodermatitis enteropathica but is caused by a low zinc concentration in maternal breast milk. TNZD becomes clinically evident during breastfeeding and is resolved by weaning and the introduction of complementary nutrition. We present a 4-month-old girl with TNZD due to a new autosomal dominant mutation (663delC) in the maternal SLC30A2 gene not previously described in the literature.

An operational guide to translational clinical machine learning in academic medical centers.

Few published data science tools are ever translated from academia to real-world clinical settings for which they were intended. One dimension of this problem is the software engineering task of turning published academic projects into tools that are usable at the bedside. Given the complexity of the data ecosystem in large health systems, this task often represents a significant barrier to the real-world deployment of data science tools for prospective piloting and evaluation. Many information technology companies have created Machine Learning Operations (MLOps) teams to help with such tasks at scale, but the low penetration of home-grown data science tools in regular clinical practice precludes the formation of such teams in healthcare organizations. Based on experiences deploying data science tools at two large academic medical centers (Beth Israel Deaconess Medical Center, Boston, MA; Mayo Clinic, Rochester, MN), we propose a strategy to facilitate this transition from academic product to operational tool, defining the responsibilities of the principal investigator, data scientist, machine learning engineer, health system IT administrator, and clinician end-user throughout the process. We first enumerate the technical resources and stakeholders needed to prepare for model deployment. We then propose an approach to planning how the final product will work from data extraction and analysis to visualization of model outputs. Finally, we describe how the team should execute on this plan. We hope to guide health systems aiming to deploy minimum viable data science tools and realize their value in clinical practice.

Addressing positive multi-cancer early detection tests in head and neck Surgery: Experience with head and neck work up for high-risk referrals.

OBJECTIVES: Blood-based multi-cancer early detection (MCED) tests are now commercially available. However, there are currently no consensus guidelines available for head and neck cancer (HNC) providers to direct work up or surveillance for patients with a positive MCED test. We seek to describe cases of patients with positive MCED tests suggesting HNC and provide insights for their evaluation. METHODS: Retrospective chart review of patients referred to Otolaryngology with an MCED result suggesting HNC. Patients enrolled in prospective MCED clinical trials were excluded. Cancer diagnoses were confirmed via frozen-section pathology. RESULTS: Five patients were included (mean age: 69.2 years, range 50-87; 4 male) with MCED-identified-high-risk for HNC or lymphoma. Only patient was symptomatic. After physical exam and follow-up head and neck imaging, circulating tumor HPV DNA testing, two patients were diagnosed with p16 + oropharyngeal squamous cell carcinomas and underwent appropriate therapy. A third patient had no evidence of head and neck cancer but was diagnosed with sarcoma of the thigh. The remaining two patients had no evidence of malignancy after in-depth workup. CONCLUSIONS: In this retrospective study, 2 of 5 patients referred to Otolaryngology with a positive MCED result were diagnosed with HPV + oropharyngeal squamous cell carcinoma. We recommend that positive HNC MCED work up include thorough head and neck examination with flexible laryngoscopy and focused CT or MRI imaging. Given the potential for inaccurate MCED tissue of origin classification, PET/CT may be useful in specific situations. For a patient with no cancer identified, development of clear guidelines is warranted.

Development of an Automated Triage System for Longstanding Dizzy Patients Using Artificial Intelligence.

Objective: To report the first steps of a project to automate and optimize scheduling of multidisciplinary consultations for patients with longstanding dizziness utilizing artificial intelligence. Study Design: Retrospective case review. Setting: Quaternary referral center. Methods: A previsit self-report questionnaire was developed to query patients about their complaints of longstanding dizziness. We convened an expert panel of clinicians to review diagnostic outcomes for 98 patients and used a consensus approach to retrospectively determine what would have been the ideal appointments based on the patient's final diagnoses. These results were then compared retrospectively to the actual patient schedules. From these data, a machine learning algorithm was trained and validated to automate the triage process. Results: Compared with the ideal itineraries determined retrospectively with our expert panel, visits scheduled by the triage clinicians showed a mean concordance of 70%, and our machine learning algorithm triage showed a mean concordance of 79%. Conclusion: Manual triage by clinicians for dizzy patients is a time-consuming and costly process. The formulated first-generation automated triage algorithm achieved similar results to clinicians when triaging dizzy patients using data obtained directly from an online previsit questionnaire.

Blockchain for Electronic Vaccine Certificates: More Cons Than Pros?

Electronic vaccine certificates (EVC) for COVID-19 vaccination are likely to become widespread. Blockchain (BC) is an electronic immutable distributed ledger and is one of the more common proposed EVC platform options. However, the principles of blockchain are not widely understood by public health and medical professionals. We attempt to describe, in an accessible style, how BC works and the potential benefits and drawbacks in its use for EVCs. Our assessment is BC technology is not well suited to be used for EVCs. Overall, blockchain technology is based on two key principles: the use of cryptography, and a distributed immutable ledger in the format of blockchains. While the use of cryptography can provide ease of sharing vaccination records while maintaining privacy, EVCs require some amount of contribution from a centralized authority to confirm vaccine status; this is partly because these authorities are responsible for the distribution and often the administration of the vaccine. Having the data distributed makes the role of a centralized authority less effective. We concluded there are alternative ways to use cryptography outside of a BC that allow a centralized authority to better participate, which seems necessary for an EVC platform to be of practical use.

Assessing the Generalizability of a Clinical Machine Learning Model Across Multiple Emergency Departments.

Objective: To assess the generalizability of a clinical machine learning algorithm across multiple emergency departments (EDs). Patients and Methods: We obtained data on all ED visits at our health care system's largest ED from May 5, 2018, to December 31, 2019. We also obtained data from 3 satellite EDs and 1 distant-hub ED from May 1, 2018, to December 31, 2018. A gradient-boosted machine model was trained on pooled data from the included EDs. To prevent the effect of differing training set sizes, the data were randomly downsampled to match those of our smallest ED. A second model was trained on this downsampled, pooled data. The model's performance was compared using area under the receiver operating characteristic (AUC). Finally, site-specific models were trained and tested across all the sites, and the importance of features was examined to understand the reasons for differing generalizability. Results: The training data sets contained 1918-64,161 ED visits. The AUC for the pooled model ranged from 0.84 to 0.94 across the sites; the performance decreased slightly when Ns were downsampled to match those of our smallest ED site. When site-specific models were trained and tested across all the sites, the AUCs ranged more widely from 0.71 to 0.93. Within a single ED site, the performance of the 5 site-specific models was most variable for our largest and smallest EDs. Finally, when the importance of features was examined, several features were common to all site-specific models; however, the weight of these features differed. Conclusion: A machine learning model for predicting hospital admission from the ED will generalize fairly well within the health care system but will still have significant differences in AUC performance across sites because of site-specific factors.

Assessing socioeconomic bias in machine learning algorithms in health care: a case study of the HOUSES index.

OBJECTIVE: Artificial intelligence (AI) models may propagate harmful biases in performance and hence negatively affect the underserved. We aimed to assess the degree to which data quality of electronic health records (EHRs) affected by inequities related to low socioeconomic status (SES), results in differential performance of AI models across SES. MATERIALS AND METHODS: This study utilized existing machine learning models for predicting asthma exacerbation in children with asthma. We compared balanced error rate (BER) against different SES levels measured by HOUsing-based SocioEconomic Status measure (HOUSES) index. As a possible mechanism for differential performance, we also compared incompleteness of EHR information relevant to asthma care by SES. RESULTS: Asthmatic children with lower SES had larger BER than those with higher SES (eg, ratio = 1.35 for HOUSES Q1 vs Q2-Q4) and had a higher proportion of missing information relevant to asthma care (eg, 41% vs 24% for missing asthma severity and 12% vs 9.8% for undiagnosed asthma despite meeting asthma criteria). DISCUSSION: Our study suggests that lower SES is associated with worse predictive model performance. It also highlights the potential role of incomplete EHR data in this differential performance and suggests a way to mitigate this bias. CONCLUSION: The HOUSES index allows AI researchers to assess bias in predictive model performance by SES. Although our case study was based on a small sample size and a single-site study, the study results highlight a potential strategy for identifying bias by using an innovative SES measure.

Non-linear machine learning models incorporating SNPs and PRS improve polygenic prediction in diverse human populations.

Polygenic risk scores (PRS) are commonly used to quantify the inherited susceptibility for a trait, yet they fail to account for non-linear and interaction effects between single nucleotide polymorphisms (SNPs). We address this via a machine learning approach, validated in nine complex phenotypes in a multi-ancestry population. We use an ensemble method of SNP selection followed by gradient boosted trees (XGBoost) to allow for non-linearities and interaction effects. We compare our results to the standard, linear PRS model developed using PRSice, LDpred2, and lassosum2. Combining a PRS as a feature in an XGBoost model results in a relative increase in the percentage variance explained compared to the standard linear PRS model by 22% for height, 27% for HDL cholesterol, 43% for body mass index, 50% for sleep duration, 58% for systolic blood pressure, 64% for total cholesterol, 66% for triglycerides, 77% for LDL cholesterol, and 100% for diastolic blood pressure. Multi-ancestry trained models perform similarly to specific racial/ethnic group trained models and are consistently superior to the standard linear PRS models. This work demonstrates an effective method to account for non-linearities and interaction effects in genetics-based prediction models.

Using nursing notes to improve clinical outcome prediction in intensive care patients: A retrospective cohort study.

OBJECTIVE: Electronic health record documentation by intensive care unit (ICU) clinicians may predict patient outcomes. However, it is unclear whether physician and nursing notes differ in their ability to predict short-term ICU prognosis. We aimed to investigate and compare the ability of physician and nursing notes, written in the first 48 hours of admission, to predict ICU length of stay and mortality using 3 analytical methods. MATERIALS AND METHODS: This was a retrospective cohort study with split sampling for model training and testing. We included patients ≥18 years of age admitted to the ICU at Beth Israel Deaconess Medical Center in Boston, Massachusetts, from 2008 to 2012. Physician or nursing notes generated within the first 48 hours of admission were used with standard machine learning methods to predict outcomes. RESULTS: For the primary outcome of composite score of ICU length of stay ≥7 days or in-hospital mortality, the gradient boosting model had better performance than the logistic regression and random forest models. Nursing and physician notes achieved area under the curves (AUCs) of 0.826 and 0.796, respectively, with even better predictive power when combined (AUC, 0.839). DISCUSSION: Models using only nursing notes more accurately predicted short-term prognosis than did models using only physician notes, but in combination, the models achieved the greatest accuracy in prediction. CONCLUSIONS: Our findings demonstrate that statistical models derived from text analysis in the first 48 hours of ICU admission can predict patient outcomes. Physicians' and nurses' notes are both uniquely important in mortality prediction and combining these notes can produce a better predictive model.

Practical development and operationalization of a 12-hour hospital census prediction algorithm.

Hospital census prediction has well-described implications for efficient hospital resource utilization, and recent issues with hospital crowding due to CoVID-19 have emphasized the importance of this task. Our team has been leading an institutional effort to develop machine-learning models that can predict hospital census 12 hours into the future. We describe our efforts at developing accurate empirical models for this task. Ultimately, with limited resources and time, we were able to develop simple yet useful models for 12-hour census prediction and design a dashboard application to display this output to our hospital's decision-makers. Specifically, we found that linear models with ElasticNet regularization performed well for this task with relative 95% error of +/- 3.4% and that this work could be completed in approximately 7 months.

Using machine learning to improve the accuracy of patient deterioration predictions: Mayo Clinic Early Warning Score (MC-EWS).

OBJECTIVE: We aimed to develop a model for accurate prediction of general care inpatient deterioration. MATERIALS AND METHODS: Training and internal validation datasets were built using 2-year data from a quaternary hospital in the Midwest. Model training used gradient boosting and feature engineering (clinically relevant interactions, time-series information) to predict general care inpatient deterioration (resuscitation call, intensive care unit transfer, or rapid response team call) in 24 hours. Data from a tertiary care hospital in the Southwest were used for external validation. C-statistic, sensitivity, positive predictive value, and alert rate were calculated for different cutoffs and compared with the National Early Warning Score. Sensitivity analysis evaluated prediction of intensive care unit transfer or resuscitation call. RESULTS: Training, internal validation, and external validation datasets included 24 500, 25 784 and 53 956 hospitalizations, respectively. The Mayo Clinic Early Warning Score (MC-EWS) demonstrated excellent discrimination in both the internal and external validation datasets (C-statistic = 0.913, 0.937, respectively), and results were consistent in the sensitivity analysis (C-statistic = 0.932 in external validation). At a sensitivity of 73%, MC-EWS would generate 0.7 alerts per day per 10 patients, 45% less than the National Early Warning Score. DISCUSSION: Low alert rates are important for implementation of an alert system. Other early warning scores developed for the general care ward have achieved lower discrimination overall compared with MC-EWS, likely because MC-EWS includes both nursing assessments and extensive feature engineering. CONCLUSIONS: MC-EWS achieved superior prediction of general care inpatient deterioration using sophisticated feature engineering and a machine learning approach, reducing alert rate.

Forecasting the COVID-19 Pandemic: Lessons learned and future directions.

I.The Coronavirus Disease 2019 (COVID-19) has demonstrated that accurate forecasts of infection and mortality rates are essential for informing healthcare resource allocation, designing countermeasures, implementing public health policies, and increasing public awareness. However, there exist a multitude of modeling methodologies, and their relative performances in accurately forecasting pandemic dynamics are not currently comprehensively understood. In this paper, we introduce the non-mechanistic MIT-LCP forecasting model, and assess and compare its performance to various mechanistic and non-mechanistic models that have been proposed for forecasting COVID-19 dynamics. We performed a comprehensive experimental evaluation which covered the time period of November 2020 to April 2021, in order to determine the relative performances of MIT-LCP and seven other forecasting models from the United States' Centers for Disease Control and Prevention (CDC) Forecast Hub. Our results show that there exist forecasting scenarios well-suited to both mechanistic and non-mechanistic models, with mechanistic models being particularly performant for forecasts that are further in the future when recent data may not be as informative, and non-mechanistic models being more effective with shorter prediction horizons when recent representative data is available. Improving our understanding of which forecasting approaches are more reliable, and in which forecasting scenarios, can assist effective pandemic preparation and management.

Public health impact of delaying second dose of BNT162b2 or mRNA-1273 covid-19 vaccine: simulation agent based modeling study.

OBJECTIVE: To estimate population health outcomes with delayed second dose versus standard schedule of SARS-CoV-2 mRNA vaccination. DESIGN: Simulation agent based modeling study. SETTING: Simulated population based on real world US county. PARTICIPANTS: The simulation included 100 000 agents, with a representative distribution of demographics and occupations. Networks of contacts were established to simulate potentially infectious interactions though occupation, household, and random interactions. INTERVENTIONS: Simulation of standard covid-19 vaccination versus delayed second dose vaccination prioritizing the first dose. The simulation runs were replicated 10 times. Sensitivity analyses included first dose vaccine efficacy of 50%, 60%, 70%, 80%, and 90% after day 12 post-vaccination; vaccination rate of 0.1%, 0.3%, and 1% of population per day; assuming the vaccine prevents only symptoms but not asymptomatic spread (that is, non-sterilizing vaccine); and an alternative vaccination strategy that implements delayed second dose for people under 65 years of age, but not until all those above this age have been vaccinated. MAIN OUTCOME MEASURES: Cumulative covid-19 mortality, cumulative SARS-CoV-2 infections, and cumulative hospital admissions due to covid-19 over 180 days. RESULTS: Over all simulation replications, the median cumulative mortality per 100 000 for standard dosing versus delayed second dose was 226 v 179, 233 v 207, and 235 v 236 for 90%, 80%, and 70% first dose efficacy, respectively. The delayed second dose strategy was optimal for vaccine efficacies at or above 80% and vaccination rates at or below 0.3% of the population per day, under both sterilizing and non-sterilizing vaccine assumptions, resulting in absolute cumulative mortality reductions between 26 and 47 per 100 000. The delayed second dose strategy for people under 65 performed consistently well under all vaccination rates tested. CONCLUSIONS: A delayed second dose vaccination strategy, at least for people aged under 65, could result in reduced cumulative mortality under certain conditions.

Emotion Regulation Strategies, Workload Conditions, and Burnout in Healthcare Residents.

BACKGROUND: Burnout syndrome is very prevalent among healthcare residents. Initiatives addressing workload conditions have had limited impact on burnout. The present study aims to explore the contribution of two emotion regulation strategies, namely emotion suppression and cognitive reevaluation, to residents' burnout, while accounting for workload factors. METHODS: Participants were 105 residents (68.6% women; mean age = 27.5, SD = 3.0). They completed measures of workload, burnout, and emotion regulation. The study was cross-sectional. RESULTS: Emotional suppression was associated with higher burnout (depersonalization scale; ß = 0.20, p < 0.05, CI 0.15-2.48) and cognitive revaluation was linked to lower burnout (higher personal accomplishment; ß = 0.35, p < 0.01, CI 0.16-2.56), even after controlling for demographic and workload factors. We found interaction effects between workload variables (supervisor support and number of patient hours) and emotion regulation (p < 0.05). CONCLUSIONS: The relationship between workload, emotion regulation, and burnout seems to be complex. That is, similar work conditions might generate different levels of burnout depending on the resident's emotional regulation strategies. This might partly explain why existing initiatives based on workload changes have had a modest impact on burnout. Results also support including emotion regulation training in prevention and treatment programs targeting burnout during residency.

Implementation of Artificial Intelligence-Based Clinical Decision Support to Reduce Hospital Readmissions at a Regional Hospital.

BACKGROUND: Hospital readmissions are a key quality metric, which has been tied to reimbursement. One strategy to reduce readmissions is to direct resources to patients at the highest risk of readmission. This strategy necessitates a robust predictive model coupled with effective, patient-centered interventions. OBJECTIVE: The aim of this study was to reduce unplanned hospital readmissions through the use of artificial intelligence-based clinical decision support. METHODS: A commercially vended artificial intelligence tool was implemented at a regional hospital in La Crosse, Wisconsin between November 2018 and April 2019. The tool assessed all patients admitted to general care units for risk of readmission and generated recommendations for interventions intended to decrease readmission risk. Similar hospitals were used as controls. Change in readmission rate was assessed by comparing the 6-month intervention period to the same months of the previous calendar year in exposure and control hospitals. RESULTS: Among 2,460 hospitalizations assessed using the tool, 611 were designated by the tool as high risk. Sensitivity and specificity for risk assignment were 65% and 89%, respectively. Over 6 months following implementation, readmission rates decreased from 11.4% during the comparison period to 8.1% (p < 0.001). After accounting for the 0.5% decrease in readmission rates (from 9.3 to 8.8%) at control hospitals, the relative reduction in readmission rate was 25% (p < 0.001). Among patients designated as high risk, the number needed to treat to avoid one readmission was 11. CONCLUSION: We observed a decrease in hospital readmission after implementing artificial intelligence-based clinical decision support. Our experience suggests that use of artificial intelligence to identify patients at the highest risk for readmission can reduce quality gaps when coupled with patient-centered interventions.