PhD Candidate, NYU School of Medicine
Medical Informatics: applying state-of-the-art methodologies from computer science (and statistics, machine learning, data science, data engineering etc.) to real clinical problems, often exploiting electronic medical record data. It is a inherently collaborative area in which we work closely with clinicians, statisticians, data engineers and IT professionals to ensure data-driven results are clinically meaningful with real utility in daily clinical care while being feasible and reliable at scale.
Abstract: The United States has invested substantially in technologies that enable health information exchange (HIE), which in turn can be deployed to reduce avoidable hospital readmission rates in many communities. With avoidable hospital readmissions as the primary focus, this study profiles successful hospital readmission rate reduction initiatives that integrate HIE as a strategy. We hypothesized that the use of HIE is associated with decreased hospital readmissions beyond other observed population health benefits. Results of this systematic review are used to describe and profile successful readmission reduction programs that integrate HIE as a tool.A systematic review of literature provided an understanding of the use of HIE as a strategy to reduce hospital readmission rates. We conducted a review of 4,862 citations written in English about readmission reduction strategies from January 2006 to September 2016 in the MEDLINE-PubMed database. Of these, 106 studies reported 30-day readmission rates as an outcome and only 13 articles reported using HIE.Only a very small number (12%) of hospitals incorporated HIE as a primary tool for evidence-based readmission reduction initiatives. Information exchange between providers has been suggested to play a key role in reducing avoidable readmission rates, yet there is not currently evidence supporting current HIE-enabled readmission initiatives. Most successful readmission reduction programs demonstrate collaboration with primary care providers to augment transitions of care to existing care management functions without additional staff while using effective information exchange capabilities.This research confirms there is very little integration of HIE into health systems readmissions initiatives. There is a great opportunity to achieve population health targets using the HIE infrastructure. Hospitals should consider partnering with primary care clinics to implement multifaceted transitions of care programs to significantly reduce 30-day readmission rates.
Pub.: 11 Jun '17, Pinned: 28 Jun '17
Abstract: Approximately 10% of admissions to acute-care hospitals are associated with an adverse event. Analysis of incident reports helps to understand how and why incidents occur and can inform policy and practice for safer care. Unfortunately our capacity to monitor and respond to incident reports in a timely manner is limited by the sheer volumes of data collected. In this study, we aim to evaluate the feasibility of using multiclass classification to automate the identification of patient safety incidents in hospitals.Text based classifiers were applied to identify 10 incident types and 4 severity levels. Using the one-versus-one (OvsO) and one-versus-all (OvsA) ensemble strategies, we evaluated regularized logistic regression, linear support vector machine (SVM) and SVM with a radial-basis function (RBF) kernel. Classifiers were trained and tested with "balanced" datasets (n_ Type = 2860, n_ SeverityLevel = 1160) from a state-wide incident reporting system. Testing was also undertaken with imbalanced "stratified" datasets (n_ Type = 6000, n_ SeverityLevel =5950) from the state-wide system and an independent hospital reporting system. Classifier performance was evaluated using a confusion matrix, as well as F-score, precision and recall.The most effective combination was a OvsO ensemble of binary SVM RBF classifiers with binary count feature extraction. For incident type, classifiers performed well on balanced and stratified datasets (F-score: 78.3, 73.9%), but were worse on independent datasets (68.5%). Reports about falls, medications, pressure injury, aggression and blood products were identified with high recall and precision. "Documentation" was the hardest type to identify. For severity level, F-score for severity assessment code (SAC) 1 (extreme risk) was 87.3 and 64% for SAC4 (low risk) on balanced data. With stratified data, high recall was achieved for SAC1 (82.8-84%) but precision was poor (6.8-11.2%). High risk incidents (SAC2) were confused with medium risk incidents (SAC3).Binary classifier ensembles appear to be a feasible method for identifying incidents by type and severity level. Automated identification should enable safety problems to be detected and addressed in a more timely manner. Multi-label classifiers may be necessary for reports that relate to more than one incident type.
Pub.: 14 Jun '17, Pinned: 28 Jun '17
Abstract: Physical abuse is a leading cause of pediatric morbidity and mortality. Physicians do not consistently screen for abuse, even in high-risk situations. Alerts in the electronic medical record may help improve screening rates, resulting in early identification and improved outcomes.Triggers to identify children < 2 years old at risk for physical abuse were coded into the electronic medical record at a freestanding pediatric hospital with a level 1 trauma center. The system was run in "silent mode"; physicians were unaware of the system, but study personnel received data on children who triggered the alert system. Sensitivity, specificity, and negative and positive predictive values of the child abuse alert system for identifying physical abuse were calculated.Thirty age-specific triggers were embedded into the electronic medical record. From October 21, 2014, through April 6, 2015, the system was in silent mode. All 226 children who triggered the alert system were considered subjects. Mean (SD) age was 9.1 (6.5) months. All triggers were activated at least once. Sensitivity was 96.8% (95% CI, 92.4-100.0%), specificity was 98.5% (95% CI, 98.3.5-98.7), and positive and negative predictive values were 26.5% (95% CI, 21.2-32.8%) and 99.9% (95% CI, 99.9-100.0%), respectively, for identifying children < 2 years old with possible, probable, or definite physical abuse.Triggers embedded into the electronic medical record can identify young children with who need to be evaluated for physical abuse with high sensitivity and specificity.
Pub.: 24 Jun '17, Pinned: 28 Jun '17
Join Sparrho today to stay on top of science
Discover, organise and share research that matters to you