Factors Associated With Variability in the Performance of a Proprietary Sepsis Prediction Model Across 9 Networked Hospitals in the US.

This cohort study uses data from electronic health records to assess variability in a sepsis prediction model across 9 hospitals.

OpenSep: a generalizable open source pipeline for SOFA score calculation and Sepsis-3 classification.

EHR-based sepsis research often uses heterogeneous definitions of sepsis leading to poor generalizability and difficulty in comparing studies to each other. We have developed OpenSep, an open-source pipeline for sepsis phenotyping according to the Sepsis-3 definition, as well as determination of time of sepsis onset and SOFA scores. The Minimal Sepsis Data Model was developed alongside the pipeline to enable the execution of the pipeline to diverse sources of electronic health record data. The pipeline's accuracy was validated by applying it to the MIMIC-IV version 1.0 data and comparing sepsis onset and SOFA scores to those produced by the pipeline developed by the curators of MIMIC. We demonstrated high reliability between both the sepsis onsets and SOFA scores, however the use of the Minimal Sepsis Data model developed for this work allows our pipeline to be applied to more broadly to data sources beyond MIMIC.

Sepsis Prediction for the General Ward Setting.

Objective: To develop and evaluate a sepsis prediction model for the general ward setting and extend the evaluation through a novel pseudo-prospective trial design. Design: Retrospective analysis of data extracted from electronic health records (EHR). Setting: Single, tertiary-care academic medical center in St. Louis, MO, USA. Patients: Adult, non-surgical inpatients admitted between January 1, 2012 and June 1, 2019. Interventions: None. Measurements and Main Results: Of the 70,034 included patient encounters, 3.1% were septic based on the Sepsis-3 criteria. Features were generated from the EHR data and were used to develop a machine learning model to predict sepsis 6-h ahead of onset. The best performing model had an Area Under the Receiver Operating Characteristic curve (AUROC or c-statistic) of 0.862 ± 0.011 and Area Under the Precision-Recall Curve (AUPRC) of 0.294 ± 0.021 compared to that of Logistic Regression (0.857 ± 0.008 and 0.256 ± 0.024) and NEWS 2 (0.699 ± 0.012 and 0.092 ± 0.009). In the pseudo-prospective trial, 388 (69.7%) septic patients were alerted on with a specificity of 81.4%. Within 24 h of crossing the alert threshold, 20.9% had a sepsis-related event occur. Conclusions: A machine learning model capable of predicting sepsis in the general ward setting was developed using the EHR data. The pseudo-prospective trial provided a more realistic estimation of implemented performance and demonstrated a 29.1% Positive Predictive Value (PPV) for sepsis-related intervention or outcome within 48 h.

Respiratory support status from EHR data for adult population: classification, heuristics, and usage in predictive modeling.

OBJECTIVE: Respiratory support status is critical in understanding patient status, but electronic health record data are often scattered, incomplete, and contradictory. Further, there has been limited work on standardizing representations for respiratory support. The objective of this work was to (1) propose a practical terminology system for respiratory support methods; (2) develop (meta-)heuristics for constructing respiratory support episodes; and (3) evaluate the utility of respiratory support information for mortality prediction. MATERIALS AND METHODS: All analyses were performed using electronic health record data of COVID-19-tested, emergency department-admit, adult patients at a large, Midwestern healthcare system between March 1, 2020 and April 1, 2021. Logistic regression and XGBoost models were trained with and without respiratory support information, and performance metrics were compared. Importance of respiratory-support-based features was explored using absolute coefficient values for logistic regression and SHapley Additive exPlanations values for the XGBoost model. RESULTS: The proposed terminology system for respiratory support methods is as follows: Low-Flow Oxygen Therapy (LFOT), High-Flow Oxygen Therapy (HFOT), Non-Invasive Mechanical Ventilation (NIMV), Invasive Mechanical Ventilation (IMV), and ExtraCorporeal Membrane Oxygenation (ECMO). The addition of respiratory support information significantly improved mortality prediction (logistic regression area under receiver operating characteristic curve, median [IQR] from 0.855 [0.852-0.855] to 0.881 [0.876-0.884]; area under precision recall curve from 0.262 [0.245-0.268] to 0.319 [0.313-0.325], both P < 0.01). The proposed generalizable, interpretable, and episodic representation had commensurate performance compared to alternate representations despite loss of granularity. Respiratory support features were among the most important in both models. CONCLUSION: Respiratory support information is critical in understanding patient status and can facilitate downstream analyses.

Comparison of early warning scores for sepsis early identification and prediction in the general ward setting.

The objective of this study was to directly compare the ability of commonly used early warning scores (EWS) for early identification and prediction of sepsis in the general ward setting. For general ward patients at a large, academic medical center between early-2012 and mid-2018, common EWS and patient acuity scoring systems were calculated from electronic health records (EHR) data for patients that both met and did not meet Sepsis-3 criteria. For identification of sepsis at index time, National Early Warning Score 2 (NEWS 2) had the highest performance (area under the receiver operating characteristic curve: 0.803 [95% confidence interval [CI]: 0.795-0.811], area under the precision recall curves: 0.130 [95% CI: 0.121-0.140]) followed NEWS, Modified Early Warning Score, and quick Sequential Organ Failure Assessment (qSOFA). Using validated thresholds, NEWS 2 also had the highest recall (0.758 [95% CI: 0.736-0.778]) but qSOFA had the highest specificity (0.950 [95% CI: 0.948-0.952]), positive predictive value (0.184 [95% CI: 0.169-0.198]), and F1 score (0.236 [95% CI: 0.220-0.253]). While NEWS 2 outperformed all other compared EWS and patient acuity scores, due to the low prevalence of sepsis, all scoring systems were prone to false positives (low positive predictive value without drastic sacrifices in sensitivity), thus leaving room for more computationally advanced approaches.

A Pragmatic Machine Learning Model To Predict Carbapenem Resistance.

Infection caused by carbapenem-resistant (CR) organisms is a rising problem in the United States. While the risk factors for antibiotic resistance are well known, there remains a large need for the early identification of antibiotic-resistant infections. Using machine learning (ML), we sought to develop a prediction model for carbapenem resistance. All patients >18 years of age admitted to a tertiary-care academic medical center between 1 January 2012 and 10 October 2017 with ≥1 bacterial culture were eligible for inclusion. All demographic, medication, vital sign, procedure, laboratory, and culture/sensitivity data were extracted from the electronic health record. Organisms were considered CR if a single isolate was reported as intermediate or resistant. Patients with CR and non-CR organisms were temporally matched to maintain the positive/negative case ratio. Extreme gradient boosting was used for model development. In total, 68,472 patients met inclusion criteria, with 1,088 patients identified as having CR organisms. Sixty-seven features were used for predictive modeling. The most important features were number of prior antibiotic days, recent central venous catheter placement, and inpatient surgery. After model training, the area under the receiver operating characteristic curve was 0.846. The sensitivity of the model was 30%, with a positive predictive value (PPV) of 30% and a negative predictive value of 99%. Using readily available clinical data, we were able to create a ML model capable of predicting CR infections at the time of culture collection with a high PPV.

Comparison of Sepsis Definitions as Automated Criteria.

OBJECTIVES: Assess the impact of heterogeneity among established sepsis criteria (Sepsis-1, Sepsis-3, Centers for Disease Control and Prevention Adult Sepsis Event, and Centers for Medicare and Medicaid severe sepsis core measure 1) through the comparison of corresponding sepsis cohorts. DESIGN: Retrospective analysis of data extracted from electronic health record. SETTING: Single, tertiary-care center in St. Louis, MO. PATIENTS: Adult, nonsurgical inpatients admitted between January 1, 2012, and January 6, 2018. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: In the electronic health record data, 286,759 encounters met inclusion criteria across the study period. Application of established sepsis criteria yielded cohorts varying in prevalence: Centers for Disease Control and Prevention Adult Sepsis Event (4.4%), Centers for Medicare and Medicaid severe sepsis core measure 1 (4.8%), International Classification of Disease code (7.2%), Sepsis-3 (7.5%), and Sepsis-1 (11.3%). Between the two modern established criteria, Sepsis-3 (n = 21,550) and Centers for Disease Control and Prevention Adult Sepsis Event (n = 12,494), the size of the overlap was 7,763. The sepsis cohorts also varied in time from admission to sepsis onset (hr): Sepsis-1 (2.9), Sepsis-3 (4.1), Centers for Disease Control and Prevention Adult Sepsis Event (4.6), and Centers for Medicare and Medicaid severe sepsis core measure 1 (7.6); sepsis discharge International Classification of Disease code rate: Sepsis-1 (37.4%), Sepsis-3 (40.1%), Centers for Medicare and Medicaid severe sepsis core measure 1 (48.5%), and Centers for Disease Control and Prevention Adult Sepsis Event (54.5%); and inhospital mortality rate: Sepsis-1 (13.6%), Sepsis-3 (18.8%), International Classification of Disease code (20.4%), Centers for Medicare and Medicaid severe sepsis core measure 1 (22.5%), and Centers for Disease Control and Prevention Adult Sepsis Event (24.1%). CONCLUSIONS: The application of commonly used sepsis definitions on a single population produced sepsis cohorts with low agreement, significantly different baseline demographics, and clinical outcomes.

Acid Resistance of Glass Ionomer Cement Restorative Materials.

In view of the need for aesthetics, restorations of teeth will typically be completed using tooth colored restorative materials. With the advent of biomimetic restorative materials, such as glass ionomer cements (GIC), much greater emphasis is now being placed on how well such materials can resist the challenge of acids that are present in foods and drinks, or gastric contents that are regurgitated. This laboratory study compared the dissolution and behavior of five GIC materials (GC Fuji® VII, GC Fuji® Bulk, GC Fuji® IX Fast, Fuji® IX Extra and GC Equia® Forte Fil) when exposed to three acids (citric acid, phosphoric acid and lactic acid), versus ultrapure deionized water, which was used as a control. Discs of each material GIC were submerged in solutions and percentage weight changes over time determined. Subsequently, the GIC materials were also placed as a part of standardized Class II sandwich restorations in bovine teeth (n = 20), and submerged in the solutions, and the extent of GIC dissolution and protection of the adjacent tooth was scored. Weight loss increased with time and with acid concentration. Overall, the most soluble material was GC Fuji® IX Extra, while GC Fuji® IX Fast and GC Fuji® Bulk were less soluble, and the least soluble material was GC Equia® Forte Fil. The most destructive solution for both the discs and for GIC restorations in teeth was 10% citric acid, while the least destructive acid was 0.1% lactic acid. The more recent GIC materials GC Fuji® Bulk and GC Equia® Forte Fil showed increased acid resistance over the older GIC materials, and this further justifies their use in open sandwich Class II restorations in more hostile environments.

Spot the difference: comparing results of analyses from real patient data and synthetic derivatives.

BACKGROUND: Synthetic data may provide a solution to researchers who wish to generate and share data in support of precision healthcare. Recent advances in data synthesis enable the creation and analysis of synthetic derivatives as if they were the original data; this process has significant advantages over data deidentification. OBJECTIVES: To assess a big-data platform with data-synthesizing capabilities (MDClone Ltd., Beer Sheva, Israel) for its ability to produce data that can be used for research purposes while obviating privacy and confidentiality concerns. METHODS: We explored three use cases and tested the robustness of synthetic data by comparing the results of analyses using synthetic derivatives to analyses using the original data using traditional statistics, machine learning approaches, and spatial representations of the data. We designed these use cases with the purpose of conducting analyses at the observation level (Use Case 1), patient cohorts (Use Case 2), and population-level data (Use Case 3). RESULTS: For each use case, the results of the analyses were sufficiently statistically similar (P > 0.05) between the synthetic derivative and the real data to draw the same conclusions. DISCUSSION AND CONCLUSION: This article presents the results of each use case and outlines key considerations for the use of synthetic data, examining their role in clinical research for faster insights and improved data sharing in support of precision healthcare.

Novel Visualization of Clostridium difficile Infections in Intensive Care Units.

BACKGROUND: Accurate and timely surveillance and diagnosis of healthcare-facility onset Clostridium difficile infection (HO-CDI) is vital to controlling infections within the hospital, but there are limited tools to assist with timely outbreak investigations. OBJECTIVES: To integrate spatiotemporal factors with HO-CDI cases and develop a map-based dashboard to support infection preventionists (IPs) in performing surveillance and outbreak investigations for HO-CDI. METHODS: Clinical laboratory results and Admit-Transfer-Discharge data for admitted patients over two years were extracted from the Information Warehouse of a large academic medical center and processed according to Center for Disease Control (CDC) National Healthcare Safety Network (NHSN) definitions to classify Clostridium difficile infection (CDI) cases by onset date. Results were validated against the internal infection surveillance database maintained by IPs in Clinical Epidemiology of this Academic Medical Center (AMC). Hospital floor plans were combined with HO-CDI case data, to create a dashboard of intensive care units. Usability testing was performed with a think-aloud session and a survey. RESULTS: The simple classification algorithm identified all 265 HO-CDI cases from 1/1/15-11/30/15 with a positive predictive value (PPV) of 96.3%. When applied to data from 2014, the PPV was 94.6% All users "strongly agreed" that the dashboard would be a positive addition to Clinical Epidemiology and would enable them to present Hospital Acquired Infection (HAI) information to others more efficiently. CONCLUSIONS: The CDI dashboard demonstrates the feasibility of mapping clinical data to hospital patient care units for more efficient surveillance and potential outbreak investigations.