Entity

Time filter

Source Type

Indianapolis, IN, United States

Klann J.G.,Massachusetts General Hospital | Klann J.G.,Harvard University | Klann J.G.,The Regenstrief Institute for Health Care | Szolovits P.,Massachusetts Institute of Technology | And 4 more authors.
Journal of Biomedical Informatics | Year: 2014

Objective: Reducing care variability through guidelines has significantly benefited patients. Nonetheless, guideline-based Clinical Decision Support (CDS) systems are not widely implemented or used, are frequently out-of-date, and cannot address complex care for which guidelines do not exist. Here, we develop and evaluate a complementary approach - using Bayesian Network (BN) learning to generate adaptive, context-specific treatment menus based on local order-entry data. These menus can be used as a draft for expert review, in order to minimize development time for local decision support content. This is in keeping with the vision outlined in the US Health Information Technology Strategic Plan, which describes a healthcare system that learns from itself. Materials and methods: We used the Greedy Equivalence Search algorithm to learn four 50-node domain-specific BNs from 11,344 encounters: abdominal pain in the emergency department, inpatient pregnancy, hypertension in the Urgent Visit Clinic, and altered mental state in the intensive care unit. We developed a system to produce situation-specific, rank-ordered treatment menus from these networks. We evaluated this system with a hospital-simulation methodology and computed Area Under the Receiver-Operator Curve (AUC) and average menu position at time of selection. We also compared this system with a similar association-rule-mining approach. Results: A short order menu on average contained the next order (weighted average length 3.91-5.83 items). Overall predictive ability was good: average AUC above 0.9 for 25% of order types and overall average AUC 714-844 (depending on domain). However, AUC had high variance (50-99). Higher AUC correlated with tighter clusters and more connections in the graphs, indicating importance of appropriate contextual data. Comparison with an Association Rule Mining approach showed similar performance for only the most common orders with dramatic divergence as orders are less frequent. Discussion and conclusion: This study demonstrates that local clinical knowledge can be extracted from treatment data for decision support. This approach is appealing because: it reflects local standards; it uses data already being captured; and it produces human-readable treatment-diagnosis networks that could be curated by a human expert to reduce workload in developing localized CDS content. The BN methodology captured transitive associations and co-varying relationships, which existing approaches do not. It also performs better as orders become less frequent and require more context. This system is a step forward in harnessing local, empirical data to enhance decision support. © 2013 Elsevier Inc. Source


Klann J.G.,Massachusetts General Hospital | Klann J.G.,Harvard University | Klann J.G.,Partners HealthCare System Inc. | Anand V.,Indiana University | And 3 more authors.
Journal of the American Medical Informatics Association | Year: 2013

Objective Over 8 years, we have developed an innovative computer decision support system that improves appropriate delivery of pediatric screening and care. This system employs a guidelines evaluation engine using data from the electronic health record (EHR) and input from patients and caregivers. Because guideline recommendations typically exceed the scope of one visit, the engine uses a static prioritization scheme to select recommendations. Here we extend an earlier idea to create patient-tailored prioritization. Materials and methods We used Bayesian structure learning to build networks of association among previously collected data from our decision support system. Using area under the receiver-operating characteristic curve (AUC) as a measure of discriminability (a sine qua non for expected value calculations needed for prioritization), we performed a structural analysis of variables with high AUC on a test set. Our source data included 177 variables for 29 402 patients. Results The method produced a network model containing 78 screening questions and anticipatory guidance (107 variables total). Average AUC was 0.65, which is sufficient for prioritization depending on factors such as population prevalence. Structure analysis of seven highly predictive variables reveals both face-validity (related nodes are connected) and non-intuitive relationships. Discussion We demonstrate the ability of a Bayesian structure learning method to 'phenotype the population' seen in our primary care pediatric clinics. The resulting network can be used to produce patient-tailored posterior probabilities that can be used to prioritize content based on the patient's current circumstances. Conclusions This study demonstrates the feasibility of EHR-driven population phenotyping for patient-tailored prioritization of pediatric preventive care services. Source


Carroll A.E.,Indiana University | Carroll A.E.,The Regenstrief Institute for Health Care | Anand V.,Indiana University | Anand V.,The Regenstrief Institute for Health Care | And 2 more authors.
Applied Clinical Informatics | Year: 2012

Introduction: The identification of key factors influencing responses to prompts and reminders within a computer decision support system (CDSS) has not been widely studied. The aim of this study was to evaluate why clinicians routinely answer certain prompts while others are ignored. Methods: We utilized data collected from a CDSS developed by our research group - the Child Health Improvement through Computer Automation (CHICA) system. The main outcome of interest was whether a clinician responded to a prompt. Results: This study found that, as expected, some clinics and physicians were more likely to address prompts than others. However, we also found clinicians are more likely to address prompts for younger patients and when the prompts address more serious issues. The most striking finding was that the position of a prompt was a significant predictor of the likelihood of the prompt being addressed, even after controlling for other factors. Prompts at the top of the page were significantly more likely to be answered than the ones on the bottom. Conclusions: This study detailed a number of factors that are associated with physicians following clinical decision support prompts. This information could be instrumental in designing better interventions and more successful clinical decision support systems in the future. © Schattauer 2012. Source


Carroll A.E.,Indiana University | Carroll A.E.,The Regenstrief Institute for Health Care | Biondich P.,Indiana University | Biondich P.,The Regenstrief Institute for Health Care | And 5 more authors.
Journal of the American Medical Informatics Association | Year: 2013

Objective To determine if automated screening and just in time delivery of testing and referral materials at the point of care promotes universal screening referral rates for maternal depression. Methods The Child Health Improvement through Computer Automation (CHICA) system is a decision support and electronic medical record system used in our pediatric clinics. All families of patients up to 15 months of age seen between October 2007 and July 2009 were randomized to one of three groups: (1) screening questions printed on prescreener forms (PSF) completed by mothers in the waiting room with physician alerts for positive screens, (2) everything in (1) plus 'just in time' (JIT) printed materials to aid physicians, and (3) a control group where physicians were simply reminded to screen on printed physician worksheets. Results The main outcome of interest was whether physicians suspected a diagnosis of maternal depression and referred a mother for assistance. This occurred significantly more often in both the PSF (2.4%) and JIT groups (2.4%) than in the control group (1.2%) (OR 2.06, 95% CI 1.08 to 3.93). Compared to the control group, more mothers were noted to have depressed mood in the PSF (OR 7.93, 95% CI 4.51 to 13.96) and JIT groups (OR 8.10, 95% CI 4.61 to 14.25). Similarly, compared to the control group, more mothers had signs of anhedonia in the PSF (OR 12.58, 95% CI 5.03 to 31.46) and JIT groups (OR 13.03, 95% CI 5.21 to 32.54). Conclusions Clinical decision support systems like CHICA can improve the screening of maternal depression. Source

Discover hidden collaborations