Designing and evaluating contextualized drug-drug interaction algorithms.

OBJECTIVE: Alert fatigue is a common issue with off-the-shelf clinical decision support. Most warnings for drug-drug interactions (DDIs) are overridden or ignored, likely because they lack relevance to the patient's clinical situation. Existing alerting systems for DDIs are often simplistic in nature or do not take the specific patient context into consideration, leading to overly sensitive alerts. The objective of this study is to develop, validate, and test DDI alert algorithms that take advantage of patient context available in electronic health records (EHRs) data. METHODS: Data on the rate at which DDI alerts were triggered but for which no action was taken over a 3-month period (override rates) from a single tertiary care facility were used to identify DDIs that were considered a high-priority for contextualized alerting. A panel of DDI experts developed algorithms that incorporate drug and patient characteristics that affect the relevance of such warnings. The algorithms were then implemented as computable artifacts, validated using a synthetic health records data, and tested over retrospective data from a single urban hospital. RESULTS: Algorithms and computable knowledge artifacts were developed and validated for a total of 8 high priority DDIs. Testing on retrospective real-world data showed the potential for the algorithms to reduce alerts that interrupt clinician workflow by more than 50%. Two algorithms (citalopram/QT interval prolonging agents, and fluconazole/opioid) showed potential to filter nearly all interruptive alerts for these combinations. CONCLUSION: The 8 DDI algorithms are a step toward addressing a critical need for DDI alerts that are more specific to patient context than current commercial alerting systems. Data commonly available in EHRs can improve DDI alert specificity.

Decompensation in Critical Care: Early Prediction of Acute Heart Failure Onset.

BACKGROUND: Heart failure is a leading cause of mortality and morbidity worldwide. Acute heart failure, broadly defined as rapid onset of new or worsening signs and symptoms of heart failure, often requires hospitalization and admission to the intensive care unit (ICU). This acute condition is highly heterogeneous and less well-understood as compared to chronic heart failure. The ICU, through detailed and continuously monitored patient data, provides an opportunity to retrospectively analyze decompensation and heart failure to evaluate physiological states and patient outcomes. OBJECTIVE: The goal of this study is to examine the prevalence of cardiovascular risk factors among those admitted to ICUs and to evaluate combinations of clinical features that are predictive of decompensation events, such as the onset of acute heart failure, using machine learning techniques. To accomplish this objective, we leveraged tele-ICU data from over 200 hospitals across the United States. METHODS: We evaluated the feasibility of predicting decompensation soon after ICU admission for 26,534 patients admitted without a history of heart failure with specific heart failure risk factors (ie, coronary artery disease, hypertension, and myocardial infarction) and 96,350 patients admitted without risk factors using remotely monitored laboratory, vital signs, and discrete physiological measurements. Multivariate logistic regression and random forest models were applied to predict decompensation and highlight important features from combinations of model inputs from dissimilar data. RESULTS: The most prevalent risk factor in our data set was hypertension, although most patients diagnosed with heart failure were admitted to the ICU without a risk factor. The highest heart failure prediction accuracy was 0.951, and the highest area under the receiver operating characteristic curve was 0.9503 with random forest and combined vital signs, laboratory values, and discrete physiological measurements. Random forest feature importance also highlighted combinations of several discrete physiological features and laboratory measures as most indicative of decompensation. Timeline analysis of aggregate vital signs revealed a point of diminishing returns where additional vital signs data did not continue to improve results. CONCLUSIONS: Heart failure risk factors are common in tele-ICU data, although most patients that are diagnosed with heart failure later in an ICU stay presented without risk factors making a prediction of decompensation critical. Decompensation was predicted with reasonable accuracy using tele-ICU data, and optimal data extraction for time series vital signs data was identified near a 200-minute window size. Overall, results suggest combinations of laboratory measurements and vital signs are viable for early and continuous prediction of patient decompensation.

The Connected Intensive Care Unit Patient: Exploratory Analyses and Cohort Discovery From a Critical Care Telemedicine Database.

BACKGROUND: Many intensive care units (ICUs) utilize telemedicine in response to an expanding critical care patient population, off-hours coverage, and intensivist shortages, particularly in rural facilities. Advances in digital health technologies, among other reasons, have led to the integration of active, well-networked critical care telemedicine (tele-ICU) systems across the United States, which in turn, provide the ability to generate large-scale remote monitoring data from critically ill patients. OBJECTIVE: The objective of this study was to explore opportunities and challenges of utilizing multisite, multimodal data acquired through critical care telemedicine. Using a publicly available tele-ICU, or electronic ICU (eICU), database, we illustrated the quality and potential uses of remote monitoring data, including cohort discovery for secondary research. METHODS: Exploratory analyses were performed on the eICU Collaborative Research Database that includes deidentified clinical data collected from adult patients admitted to ICUs between 2014 and 2015. Patient and ICU characteristics, top admission diagnoses, and predictions from clinical scoring systems were extracted and analyzed. Additionally, a case study on respiratory failure patients was conducted to demonstrate research prospects using tele-ICU data. RESULTS: The eICU database spans more than 200 hospitals and over 139,000 ICU patients across the United States with wide-ranging clinical data and diagnoses. Although mixed medical-surgical ICU was the most common critical care setting, patients with cardiovascular conditions accounted for more than 20% of ICU stays, and those with neurological or respiratory illness accounted for nearly 15% of ICU unit stays. The case study on respiratory failure patients showed that cohort discovery using the eICU database can be highly specific, albeit potentially limiting in terms of data provenance and sparsity for certain types of clinical questions. CONCLUSIONS: Large-scale remote monitoring data sources, such as the eICU database, have a strong potential to advance the role of critical care telemedicine by serving as a testbed for secondary research as well as for developing and testing tools, including predictive and prescriptive analytical solutions and decision support systems. The resulting tools will also inform coordination of care for critically ill patients, intensivist coverage, and the overall process of critical care telemedicine.

Epidemiology of Pediatric Traumatic Brain Injury and Hypothalamic-Pituitary Disorders in Arizona.

Traumatic brain injury (TBI) in children can result in long-lasting social, cognitive, and neurological impairments. In adults, TBI can lead to endocrinopathies (endocrine system disorders), but this is infrequently reported in children. Untreated endocrinopathies can elevate risks of subsequent health issues, such that early detection in pediatric TBI survivors can initiate clinical interventions. To understand the risk of endocrinopathies following pediatric TBI, we identified patients who had experienced a TBI and subsequently developed a new-onset hypothalamic regulated endocrinopathy (n = 498). We hypothesized that pediatric patients who were diagnosed with a TBI were at higher risk of being diagnosed with a central endocrinopathy than those without a prior diagnosis of TBI. In our epidemiological assessment, we identified pediatric patients enrolled in the Arizona Health Care Cost Containment System (AHCCCS) from 2008 to 2014 who were diagnosed with one of 330 TBI International Classification of Diseases (ICD)-9 codes and subsequently diagnosed with one of 14 central endocrinopathy ICD-9 codes. Additionally, the ICD-9 code data from over 600,000 Arizona pediatric patients afforded an estimate of the incidence, prevalence, relative risk, odds ratio, and number needed to harm, regarding the development of a central endocrinopathy after sustaining a TBI in Arizona Medicaid pediatric patients. Children with a TBI diagnosis had 3.22 times the risk of a subsequent central endocrine diagnosis compared with the general population (±0.28). Pediatric AHCCCS patients with a central endocrine diagnosis had 3.2-fold higher odds of a history of a TBI diagnosis than those without an endocrine diagnosis (±0.29). Furthermore, the number of patients with a TBI diagnosis for one patient to receive a diagnosis of a central endocrine diagnosis was 151.2 (±6.12). Female subjects were more likely to present with a central endocrine diagnosis after a TBI diagnosis compared to male subjects (64.1 vs. 35.9%). These results are the first state-wide epidemiological study conducted to determine the risk of developing a hypothalamic-pituitary disorder after a TBI in the pediatric population. Our results contribute to a body of knowledge demonstrating a TBI etiology for idiopathic endocrine disorders, and thus advise physicians with regard to TBI follow-up care that includes preventive screening for endocrine disorders.

Evaluating ICU Clinical Severity Scoring Systems and Machine Learning Applications: APACHE IV/IVa Case Study.

Clinical scoring systems have been developed for many specific applications, yet they remain underutilized for common reasons such as model inaccuracy and difficulty of use. For intensive care units specifically, the Acute Physiology and Chronic Health Evaluation (APACHE) score is used as a decision-making tool and hospital efficacy measure. In an attempt to alleviate the general underlying limitations of scoring instruments and demonstrate the utility of readily available medical databases, machine learning techniques were used to evaluate APACHE IV and IVa prediction measures in an open-source, teleICU research database. The teleICU database allowed for large-scale evaluation of APACHE IV and IVa predictions by comparing predicted values to the actual, recorded patient outcomes along with preliminary exploration of new predictive models for patient mortality and length of stay in both the hospital and the ICU. An increase in performance was observed in the newly developed models trained on the APACHE input variables highlighting avenues of future research and illustrating the utility of teleICU databases for model development and evaluation.