Association between obesity and statin use on mortality and hospital encounters in atrial fibrillation.

Background: Obesity increases risk of atrial fibrillation (AF) at least in part due to pro-inflammatory effects, but has been paradoxically associated with improved mortality. Although statins have pleiotropic anti-inflammatory properties, their interaction with obesity and clinical outcomes in AF is unknown. We explored the relationship between BMI, statin use, and all-cause mortality and AF/congestive heart failure (CHF)-related encounters, hypothesizing that statin exposure may be differentially associated with improved outcomes in overweight/obesity. Methods: This was a single center retrospective cohort study of adults with AF diagnosed between 2011-2018. Patients were grouped by body mass index (BMI) and statin use at time of AF diagnosis. Outcomes included all-cause mortality and ED or inpatient encounters for AF or CHF. Results and Conclusions: A total of 2503 subjects were included (median age 66 years, 43.4 % female, median BMI 29.8 kg/m2, 54.6 % on baseline statin therapy). Increasing BMI was associated with decreased mortality hazard but not associated with AF/CHF encounter risk. Adjusting for statin-BMI interaction, demographics, and cardiovascular comorbidities, overweight non-statin users experienced improved mortality (adjusted hazard ratio [aHR] 0.55, 95 % CI 0.35-0.84) compared to statin users (aHR 0.98, 95 % CI 0.69-1.40; interaction P-value = 0.013). Mortality hazard was consistently lower in obese non-statin users than in statin users, however interaction was insignificant. No significant BMI-statin interactions were observed in AF/CHF encounter risk. In summary, statin use was not differentially associated with improved mortality or hospitalization risk in overweight/obese groups. These findings do not support statins for secondary prevention of adverse outcomes based on overweight/obesity status alone.

Opportunistic detection of type 2 diabetes using deep learning from frontal chest radiographs.

Deep learning (DL) models can harness electronic health records (EHRs) to predict diseases and extract radiologic findings for diagnosis. With ambulatory chest radiographs (CXRs) frequently ordered, we investigated detecting type 2 diabetes (T2D) by combining radiographic and EHR data using a DL model. Our model, developed from 271,065 CXRs and 160,244 patients, was tested on a prospective dataset of 9,943 CXRs. Here we show the model effectively detected T2D with a ROC AUC of 0.84 and a 16% prevalence. The algorithm flagged 1,381 cases (14%) as suspicious for T2D. External validation at a distinct institution yielded a ROC AUC of 0.77, with 5% of patients subsequently diagnosed with T2D. Explainable AI techniques revealed correlations between specific adiposity measures and high predictivity, suggesting CXRs' potential for enhanced T2D screening.

Lessons learned: Development of COVID-19 clinical staging models at a large urban research institution.

Background/Objective: The University of Illinois at Chicago (UIC), along with many academic institutions worldwide, made significant efforts to address the many challenges presented during the COVID-19 pandemic by developing clinical staging and predictive models. Data from patients with a clinical encounter at UIC from July 1, 2019 to March 30, 2022 were abstracted from the electronic health record and stored in the UIC Center for Clinical and Translational Science Clinical Research Data Warehouse, prior to data analysis. While we saw some success, there were many failures along the way. For this paper, we wanted to discuss some of these obstacles and many of the lessons learned from the journey. Methods: Principle investigators, research staff, and other project team members were invited to complete an anonymous Qualtrics survey to reflect on the project. The survey included open-ended questions centering on participants' opinions about the project, including whether project goals were met, project successes, project failures, and areas that could have been improved. We then identified themes among the results. Results: Nine project team members (out of 30 members contacted) completed the survey. The responders were anonymous. The survey responses were grouped into four key themes: Collaboration, Infrastructure, Data Acquisition/Validation, and Model Building. Conclusion: Through our COVID-19 research efforts, the team learned about our strengths and deficiencies. We continue to work to improve our research and data translation capabilities.

Pharmacist-guided pharmacogenetic service lowered warfarin-related hospitalizations.

Background: The authors aimed to assess outcomes with a pharmacogenetic (PGx)-informed, pharmacist-guided, personalized consult service for warfarin dosing. Methods: This retrospective cohort study included patients admitted with thromboembolic events. Eligible subjects received either PGx-informed (n = 389) or historical non-PGx pharmacist-guided warfarin dosing (Hx; n = 308) before hospital discharge. The composite of admission with bleeding or thromboembolic events over 90 days after the discharge was compared between the PGx and Hx groups. Results: The rate ratio (95% CI) of the composite of bleeding or thromboembolic admissions for PGx versus Hx was 0.32 (0.12-0.82). The estimated hazard ratio was 0.43 (0.16-1.12). Conclusion: A PGx-informed warfarin dosing service was associated with decreased bleeding and thromboembolic encounters.

Effect of Electronic Health Record Clinical Decision Support on Contextualization of Care: A Randomized Clinical Trial.

Importance: Contextualizing care is a process of incorporating information about the life circumstances and behavior of individual patients, termed contextual factors, into their plan of care. In 4 steps, clinicians recognize clues (termed contextual red flags), clinicians ask about them (probe for context), patients disclose contextual factors, and clinicians adapt care accordingly. The process is associated with a desired outcome resolution of the presenting contextual red flag. Objective: To determine whether contextualized clinical decision support (CDS) tools in the electronic health record (EHR) improve clinician contextual probing, attention to contextual factors in care planning, and the presentation of contextual red flags. Design, Setting, and Participants: This randomized clinical trial was performed at the primary care clinics of 2 academic medical centers with different EHR systems. Participants were adults 18 years or older consenting to audio record their visits and their physicians between September 6, 2018, and March 4, 2021. Patients were randomized to an intervention or a control group. Analyses were performed on an intention-to-treat basis. Interventions: Patients completed a previsit questionnaire that elicited contextual red flags and factors and appeared in the clinician's note template in a contextual care box. The EHR also culled red flags from the medical record, included them in the contextual care box, used passive and interruptive alerts, and proposed relevant orders. Main Outcomes and Measures: Proportion of contextual red flags noted at the index visit that resolved 6 months later (primary outcome), proportion of red flags probed (secondary outcome), and proportion of contextual factors addressed in the care plan by clinicians (secondary outcome), adjusted for study site and for multiple red flags and factors within a visit. Results: Four hundred fifty-two patients (291 women [65.1%]; mean [SD] age, 55.6 [15.1] years) completed encounters with 39 clinicians (23 women [59.0%]). Contextual red flags were not more likely to resolve in the intervention vs control group (adjusted odds ratio [aOR], 0.96 [95% CI, 0.57-1.63]). However, the intervention increased both contextual probing (aOR, 2.12 [95% CI, 1.14-3.93]) and contextualization of the care plan (aOR, 2.67 [95% CI, 1.32-5.41]), controlling for whether a factor was identified by probing or otherwise. Across study groups, contextualized care plans were more likely than noncontextualized plans to result in improvement in the presenting red flag (aOR, 2.13 [95% CI, 1.38-3.28]). Conclusions and Relevance: This randomized clinical trial found that contextualized CDS did not improve patients' outcomes but did increase contextualization of their care, suggesting that use of this technology could ultimately help improve outcomes. Trial Registration: ClinicalTrials.gov Identifier: NCT03244033.

Changing from mandatory to optional genotyping results in higher acceptance of pharmacist-guided warfarin dosing.

Aim: We evaluated the clinical acceptance and feasibility of a pharmacist-guided personalized consult service following its transition from a mandatory (mPGx) to optional (oPGx) CYP2C9/VKORC1/CYP4F2 genotyping for warfarin. Methods: A total of 1105 patients were included. Clinical acceptance and feasibility outcomes were analyzed using bivariate and multivariable analyses. Results: After transitioning to optional genotyping, genotype testing was still ordered in a large segment of the eligible population (52.1%). Physician acceptance of pharmacist-recommended doses improved from 83.9% (mPGx) to 86.6% (oPGx; OR: 1.3; 95% CI: 1.1-1.5; p = 0.01) with a shorter median genotype result turnaround time (oPGX: 23.6 h vs mPGX: 25.1 h; p < 0.01). Conclusion: Ordering of genotype testing and provider acceptance of dosing recommendations remained high after transitioning to optional genotyping.

Latent Class Analysis of Prescribing Behavior of Primary Care Physicians in the Veterans Health Administration.

BACKGROUND: Benzodiazepines, opioids, proton-pump inhibitors (PPIs), and antibiotics are frequently prescribed inappropriately by primary care physicians (PCPs), without sufficient consideration of alternative options or adverse effects. We hypothesized that distinct groups of PCPs could be identified based on their propensity to prescribe these medications. OBJECTIVE: To identify PCP groups based on their propensity to prescribe benzodiazepines, opioids, PPIs, and antibiotics, and patient and PCP characteristics associated with identified prescribing patterns. DESIGN: Retrospective cohort study using VA data and latent class regression analyses to identify prescribing patterns among PCPs and examine the association of patient and PCP characteristics with class membership. PARTICIPANTS: A total of 2524 full-time PCPs and their patient panels (n = 2,939,636 patients), from January 1, 2017, to December 31, 2018. MAIN MEASURES: We categorized PCPs based on prescribing volume quartiles for the four drug classes, based on total days' supply dispensed of each medication by the PCP to their patients (expressed as days' supply per 1000 panel patient-days). We used latent class analysis to group PCPs based on prescribing and used multinomial logistic regression to examine patient and PCP characteristics associated with latent class membership. KEY RESULTS: PCPs were categorized into four groups (latent classes): low intensity (23% of cohort), medium-intensity overall/high-intensity PPI (36%), medium-intensity overall/high-intensity opioid (20%), and high intensity (21%). PCPs in the high-intensity group were predominantly in the highest quartile of prescribers for all four drugs (68% in the highest quartile for benzodiazepine, 86% opioids, 64% PPIs, 62% antibiotics). High-intensity PCPs (vs. low intensity) were substantially less likely to be female (OR: 0.30, 95% CI: 0.21-0.42) or practice in the northeast versus other census regions (OR: 0.10, 95% CI: 0.06-0.17). CONCLUSIONS: VA PCPs can be classified into four clearly differentiated groups based on their prescribing of benzodiazepines, opioids, PPIs, and antibiotics, suggesting an underlying typology of prescribing. High-intensity PCPs were more likely to be male.

Detecting Racial/Ethnic Health Disparities Using Deep Learning From Frontal Chest Radiography.

PURPOSE: The aim of this study was to assess racial/ethnic and socioeconomic disparities in the difference between atherosclerotic vascular disease prevalence measured by a multitask convolutional neural network (CNN) deep learning model using frontal chest radiographs (CXRs) and the prevalence reflected by administrative hierarchical condition category codes in two cohorts of patients with coronavirus disease 2019 (COVID-19). METHODS: A CNN model, previously published, was trained to predict atherosclerotic disease from ambulatory frontal CXRs. The model was then validated on two cohorts of patients with COVID-19: 814 ambulatory patients from a suburban location (presenting from March 14, 2020, to October 24, 2020, the internal ambulatory cohort) and 485 hospitalized patients from an inner-city location (hospitalized from March 14, 2020, to August 12, 2020, the external hospitalized cohort). The CNN model predictions were validated against electronic health record administrative codes in both cohorts and assessed using the area under the receiver operating characteristic curve (AUC). The CXRs from the ambulatory cohort were also reviewed by two board-certified radiologists and compared with the CNN-predicted values for the same cohort to produce a receiver operating characteristic curve and the AUC. The atherosclerosis diagnosis discrepancy, Δvasc, referring to the difference between the predicted value and presence or absence of the vascular disease HCC categorical code, was calculated. Linear regression was performed to determine the association of Δvasc with the covariates of age, sex, race/ethnicity, language preference, and social deprivation index. Logistic regression was used to look for an association between the presence of any hierarchical condition category codes with Δvasc and other covariates. RESULTS: The CNN prediction for vascular disease from frontal CXRs in the ambulatory cohort had an AUC of 0.85 (95% confidence interval, 0.82-0.89) and in the hospitalized cohort had an AUC of 0.69 (95% confidence interval, 0.64-0.75) against the electronic health record data. In the ambulatory cohort, the consensus radiologists' reading had an AUC of 0.89 (95% confidence interval, 0.86-0.92) relative to the CNN. Multivariate linear regression of Δvasc in the ambulatory cohort demonstrated significant negative associations with non-English-language preference (ß = -0.083, P < .05) and Black or Hispanic race/ethnicity (ß = -0.048, P < .05) and positive associations with age (ß = 0.005, P < .001) and sex (ß = 0.044, P < .05). For the hospitalized cohort, age was also significant (ß = 0.003, P < .01), as was social deprivation index (ß = 0.002, P < .05). The Δvasc variable (odds ratio [OR], 0.34), Black or Hispanic race/ethnicity (OR, 1.58), non-English-language preference (OR, 1.74), and site (OR, 0.22) were independent predictors of having one or more hierarchical condition category codes (P < .01 for all) in the combined patient cohort. CONCLUSIONS: A CNN model was predictive of aortic atherosclerosis in two cohorts (one ambulatory and one hospitalized) with COVID-19. The discrepancy between the CNN model and the administrative code, Δvasc, was associated with language preference in the ambulatory cohort; in the hospitalized cohort, this discrepancy was associated with social deprivation index. The absence of administrative code(s) was associated with Δvasc in the combined cohorts, suggesting that Δvasc is an independent predictor of health disparities. This may suggest that biomarkers extracted from routine imaging studies and compared with electronic health record data could play a role in enhancing value-based health care for traditionally underserved or disadvantaged patients for whom barriers to care exist.

Telemedicine Readiness Across Medical Conditions in a US National Representative Sample of Older Adults.

Telemedicine has provided older adults the ability to seek care remotely during the coronavirus disease (COVID-19) pandemic. However, it is unclear how diverse medical conditions play a role in telemedicine uptake. A total of 3379 participants (≥65 years) were interviewed in 2018 as part of the National Health and Aging Trends Study. We assessed telemedicine readiness across multiple medical conditions. Most chronic medical conditions and mood symptoms were significantly associated with telemedicine unreadiness, for physical or technical reasons or both, while cancer, hypertension, and arthritis were significantly associated with telemedicine readiness. Our findings suggest that multiple medical conditions play a substantial role in telemedicine uptake among older adults in the US. Therefore, comorbidities should be taken into consideration when promoting and adopting telemedicine technologies among older adults.

Indication alerts to improve problem list documentation.

BACKGROUND: Problem lists represent an integral component of high-quality care. However, they are often inaccurate and incomplete. We studied the effects of alerts integrated into the inpatient and outpatient computerized provider order entry systems to assist in adding problems to the problem list when ordering medications that lacked a corresponding indication. METHODS: We analyzed medication orders from 2 healthcare systems that used an innovative indication alert. We collected data at site 1 between December 2018 and January 2020, and at site 2 between May and June 2021. We reviewed random samples of 100 charts from each site that had problems added in response to the alert. Outcomes were: (1) alert yield, the proportion of triggered alerts that led to a problem added and (2) problem accuracy, the proportion of problems placed that were accurate by chart review. RESULTS: Alerts were triggered 131 134, and 6178 times at sites 1 and 2, respectively, resulting in a yield of 109 055 (83.2%) and 2874 (46.5%), P< .001. Orders were abandoned, for example, not completed, in 11.1% and 9.6% of orders, respectively, P<.001. Of the 100 sample problems, reviewers deemed 88% ± 3% and 91% ± 3% to be accurate, respectively, P = .65, with a mean of 90% ± 2%. CONCLUSIONS: Indication alerts triggered by medication orders initiated in the absence of a justifying diagnosis were useful for populating problem lists, with yields of 83.2% and 46.5% at 2 healthcare systems. Problems were placed with a reasonable level of accuracy, with 90% ± 2% of problems deemed accurate based on chart review.

Validation of a deep learning, value-based care model to predict mortality and comorbidities from chest radiographs in COVID-19.

We validate a deep learning model predicting comorbidities from frontal chest radiographs (CXRs) in patients with coronavirus disease 2019 (COVID-19) and compare the model's performance with hierarchical condition category (HCC) and mortality outcomes in COVID-19. The model was trained and tested on 14,121 ambulatory frontal CXRs from 2010 to 2019 at a single institution, modeling select comorbidities using the value-based Medicare Advantage HCC Risk Adjustment Model. Sex, age, HCC codes, and risk adjustment factor (RAF) score were used. The model was validated on frontal CXRs from 413 ambulatory patients with COVID-19 (internal cohort) and on initial frontal CXRs from 487 COVID-19 hospitalized patients (external cohort). The discriminatory ability of the model was assessed using receiver operating characteristic (ROC) curves compared to the HCC data from electronic health records, and predicted age and RAF score were compared using correlation coefficient and absolute mean error. The model predictions were used as covariables in logistic regression models to evaluate the prediction of mortality in the external cohort. Predicted comorbidities from frontal CXRs, including diabetes with chronic complications, obesity, congestive heart failure, arrhythmias, vascular disease, and chronic obstructive pulmonary disease, had a total area under ROC curve (AUC) of 0.85 (95% CI: 0.85-0.86). The ROC AUC of predicted mortality for the model was 0.84 (95% CI,0.79-0.88) for the combined cohorts. This model using only frontal CXRs predicted select comorbidities and RAF score in both internal ambulatory and external hospitalized COVID-19 cohorts and was discriminatory of mortality, supporting its potential use in clinical decision making.

Improving the In-Hospital Mortality Prediction of Diabetes ICU Patients Using a Process Mining/Deep Learning Architecture.

Diabetes intensive care unit (ICU) patients are at increased risk of complications leading to in-hospital mortality. Assessing the likelihood of death is a challenging and time-consuming task due to a large number of influencing factors. Healthcare providers are interested in the detection of ICU patients at higher risk, such that risk factors can possibly be mitigated. While such severity scoring methods exist, they are commonly based on a snapshot of the health conditions of a patient during the ICU stay and do not specifically consider a patient's prior medical history. In this paper, a process mining/deep learning architecture is proposed to improve established severity scoring methods by incorporating the medical history of diabetes patients. First, health records of past hospital encounters are converted to event logs suitable for process mining. The event logs are then used to discover a process model that describes the past hospital encounters of patients. An adaptation of Decay Replay Mining is proposed to combine medical and demographic information with established severity scores to predict the in-hospital mortality of diabetes ICU patients. Significant performance improvements are demonstrated compared to established risk severity scoring methods and machine learning approaches using the Medical Information Mart for Intensive Care III dataset.

Home Health Care for Patients Without Shelter.

Home health care (HHC) is a well-established model of caring for patients in their homes, which has not been robustly applied to benefit patients without regular access to shelter. This article describes Chicago Street Medicine, an organization that implements HHC to improve health outcomes and care continuity for patients experiencing homelessness.

Predicting clinical outcomes among hospitalized COVID-19 patients using both local and published models.

BACKGROUND: Many models are published which predict outcomes in hospitalized COVID-19 patients. The generalizability of many is unknown. We evaluated the performance of selected models from the literature and our own models to predict outcomes in patients at our institution. METHODS: We searched the literature for models predicting outcomes in inpatients with COVID-19. We produced models of mortality or criticality (mortality or ICU admission) in a development cohort. We tested external models which provided sufficient information and our models using a test cohort of our most recent patients. The performance of models was compared using the area under the receiver operator curve (AUC). RESULTS: Our literature review yielded 41 papers. Of those, 8 were found to have sufficient documentation and concordance with features available in our cohort to implement in our test cohort. All models were from Chinese patients. One model predicted criticality and seven mortality. Tested against the test cohort, internal models had an AUC of 0.84 (0.74-0.94) for mortality and 0.83 (0.76-0.90) for criticality. The best external model had an AUC of 0.89 (0.82-0.96) using three variables, another an AUC of 0.84 (0.78-0.91) using ten variables. AUC's ranged from 0.68 to 0.89. On average, models tested were unable to produce predictions in 27% of patients due to missing lab data. CONCLUSION: Despite differences in pandemic timeline, race, and socio-cultural healthcare context some models derived in China performed well. For healthcare organizations considering implementation of an external model, concordance between the features used in the model and features available in their own patients may be important. Analysis of both local and external models should be done to help decide on what prediction method is used to provide clinical decision support to clinicians treating COVID-19 patients as well as what lab tests should be included in order sets.

Personal Formularies of Primary Care Physicians Across 4 Health Care Systems.

Importance: More conservative prescribing has the potential to reduce adverse drug events and patient harm and cost; however, no method exists defining the extent to which individual clinicians prescribe conservatively. One potential domain is prescribing a more limited number of drugs. Personal formularies-defined as the number and mix of unique, newly initiated drugs prescribed by a physician-may enable comparisons among clinicians, practices, and institutions. Objectives: To develop a method of defining primary care physicians' personal formularies and examine how they differ among primary care physicians at 4 institutions; evaluate associations between personal formularies and patient, physician, and practice site characteristics; and empirically derive and examine the variability of the top 200 core drugs prescribed at the 4 sites. Design, Setting, and Participants: This retrospective cohort study was conducted at 4 US health care systems among 4655 internal and family medicine physicians and 4 930 707 patients who had at least 1 visit to these physicians between January 1, 2017, and December 31, 2018. Exposures: Personal formulary size was defined as the number of unique, newly initiated drugs. Main Outcomes and Measures: Personal formulary size and drugs used, physician and patient characteristics, core drugs, and analysis of selected drug classes. Results: The study population included 4655 primary care physicians (2274 women [48.9%]; mean [SD] age, 48.5 [4.4] years) and 4 930 707 patients (16.5% women; mean [SD] age, 51.9 [8.3] years). There were 41 378 903 outpatient prescriptions written, of which 9 496 766 (23.0%) were new starts. Institution median personal formulary size ranged from 150 (interquartile range, 82.0-212.0) to 296 (interquartile range, 230.0-347.0) drugs. In multivariable modeling, personal formulary size was significantly associated with panel size (total number of unique patients with face-to-face encounters during the study period; 1.2 medications per 100 patients), physician's total number of encounters (5.7 drugs per 10% increase), and physician's sex (-6.2 drugs per 100 patients for female physicians). There were 1527 unique, newly prescribed drugs across the 4 sites. Fewer than half the drugs (626 [41.0%]) were used at every site. Physicians' prescribing of drugs from a pooled core list varied from 0% to 100% of their prescriptions. Conclusions and Relevance: Personal formularies, measured at the level of individual physicians and institutions, reveal variability in size and mix of drugs. Similarly, defining a list of commonly prescribed core drugs in primary care revealed interphysician and interinstitutional differences. Personal formularies and core medication lists enable comparisons and may identify outliers and opportunities for safer and more appropriate prescribing.

Predicting self-intercepted medication ordering errors using machine learning.

Current approaches to understanding medication ordering errors rely on relatively small manually captured error samples. These approaches are resource-intensive, do not scale for computerized provider order entry (CPOE) systems, and are likely to miss important risk factors associated with medication ordering errors. Previously, we described a dataset of CPOE-based medication voiding accompanied by univariable and multivariable regression analyses. However, these traditional techniques require expert guidance and may perform poorly compared to newer approaches. In this paper, we update that analysis using machine learning (ML) models to predict erroneous medication orders and identify its contributing factors. We retrieved patient demographics (race/ethnicity, sex, age), clinician characteristics, type of medication order (inpatient, prescription, home medication by history), and order content. We compared logistic regression, random forest, boosted decision trees, and artificial neural network models. Model performance was evaluated using area under the receiver operating characteristic curve (AUROC) and the area under the precision-recall curve (AUPRC). The dataset included 5,804,192 medication orders, of which 28,695 (0.5%) were voided. ML correctly classified voids at reasonable accuracy; with a positive predictive value of 10%, ~20% of errors were included. Gradient boosted decision trees achieved the highest AUROC (0.7968) and AUPRC (0.0647) among all models. Logistic regression had the poorest performance. Models identified predictive factors with high face validity (e.g., student orders), and a decision tree revealed interacting contexts with high rates of errors not identified by previous regression models. Prediction models using order-entry information offers promise for error surveillance, patient safety improvements, and targeted clinical review. The improved performance of models with complex interactions points to the importance of contextual medication ordering information for understanding contributors to medication errors.

Predicting Prolonged Hospitalization and Supplemental Oxygenation in Patients with COVID-19 Infection from Ambulatory Chest Radiographs using Deep Learning.

RATIONALE AND OBJECTIVES: The clinical prognosis of outpatients with coronavirus disease 2019 (COVID-19) remains difficult to predict, with outcomes including asymptomatic, hospitalization, intubation, and death. Here we determined the prognostic value of an outpatient chest radiograph, together with an ensemble of deep learning algorithms predicting comorbidities and airspace disease to identify patients at a higher risk of hospitalization from COVID-19 infection. MATERIALS AND METHODS: This retrospective study included outpatients with COVID-19 confirmed by reverse transcription-polymerase chain reaction testing who received an ambulatory chest radiography between March 17, 2020 and October 24, 2020. In this study, full admission was defined as hospitalization within 14 days of the COVID-19 test for > 2 days with supplemental oxygen. Univariate analysis and machine learning algorithms were used to evaluate the relationship between the deep learning model predictions and hospitalization for > 2 days. RESULTS: The study included 413 patients, 222 men (54%), with a median age of 51 years (interquartile range, 39-62 years). Fifty-one patients (12.3%) required full admission. A boosted decision tree model produced the best prediction. Variables included patient age, frontal chest radiograph predictions of morbid obesity, congestive heart failure and cardiac arrhythmias, and radiographic opacity, with an internally validated area under the curve (AUC) of 0.837 (95% CI: 0.791-0.883) on a test cohort. CONCLUSION: Deep learning analysis of single frontal chest radiographs was used to generate combined comorbidity and pneumonia scores that predict the need for supplemental oxygen and hospitalization for > 2 days in patients with COVID-19 infection with an AUC of 0.837 (95% confidence interval: 0.791-0.883). Comorbidity scoring may prove useful in other clinical scenarios.

Risk factors associated with medication ordering errors.

OBJECTIVE: We utilized a computerized order entry system-integrated function referred to as "void" to identify erroneous orders (ie, a "void" order). Using voided orders, we aimed to (1) identify the nature and characteristics of medication ordering errors, (2) investigate the risk factors associated with medication ordering errors, and (3) explore potential strategies to mitigate these risk factors. MATERIALS AND METHODS: We collected data on voided orders using clinician interviews and surveys within 24 hours of the voided order and using chart reviews. Interviews were informed by the human factors-based SEIPS (Systems Engineering Initiative for Patient Safety) model to characterize the work systems-based risk factors contributing to ordering errors; chart reviews were used to establish whether a voided order was a true medication ordering error and ascertain its impact on patient safety. RESULTS: During the 16-month study period (August 25, 2017, to December 31, 2018), 1074 medication orders were voided; 842 voided orders were true medication errors (positive predictive value = 78.3 ± 1.2%). A total of 22% (n = 190) of the medication ordering errors reached the patient, with at least a single administration, without causing patient harm. Interviews were conducted on 355 voided orders (33% response). Errors were not uniquely associated with a single risk factor, but the causal contributors of medication ordering errors were multifactorial, arising from a combination of technological-, cognitive-, environmental-, social-, and organizational-level factors. CONCLUSIONS: The void function offers a practical, standardized method to create a rich database of medication ordering errors. We highlight implications for utilizing the void function for future research, practice and learning opportunities.