Temporal phenotyping of medically complex children via PARAFAC2 tensor factorization.

OBJECTIVE: Our aim is to extract clinically-meaningful phenotypes from longitudinal electronic health records (EHRs) of medically-complex children. This is a fragile set of patients consuming a disproportionate amount of pediatric care resources but who often end up with sub-optimal clinical outcome. The rise in available electronic health records (EHRs) provide a rich data source that can be used to disentangle their complex clinical conditions into concise, clinically-meaningful groups of characteristics. We aim at identifying those phenotypes and their temporal evolution in a scalable, computational manner, which avoids the time-consuming manual chart review. MATERIALS AND METHODS: We analyze longitudinal EHRs from Children's Healthcare of Atlanta including 1045 medically complex patients with a total of 59,948 encounters over 2 years. We apply a tensor factorization method called PARAFAC2 to extract: (a) clinically-meaningful groups of features (b) concise patient representations indicating the presence of a phenotype for each patient, and (c) temporal signatures indicating the evolution of those phenotypes over time for each patient. RESULTS: We identified four medically complex phenotypes, namely gastrointestinal disorders, oncological conditions, blood-related disorders, and neurological system disorders, which have distinct clinical characterizations among patients. We demonstrate the utility of patient representations produced by PARAFAC2, towards identifying groups of patients with significant survival variations. Finally, we showcase representative examples of the temporal phenotypic trends extracted for different patients. DISCUSSION: Unsupervised temporal phenotyping is an important task since it minimizes the burden on behalf of clinical experts, by relegating their involvement in the output phenotypes' validation. PARAFAC2 enjoys several compelling properties towards temporal computational phenotyping: (a) it is able to handle high-dimensional data and variable numbers of encounters across patients, (b) it has an intuitive interpretation and (c) it is free from ad-hoc parameter choices. Computational phenotypes, such as the ones computed by our approach, have multiple applications; we highlight three of them which are particularly useful for medically complex children: (1) integration into clinical decision support systems, (2) interpretable mortality prediction and 3) clinical trial recruitment. CONCLUSION: PARAFAC2 can be applied to unsupervised temporal phenotyping tasks where precise definitions of different phenotypes are absent, and lengths of patient records are varying.

COPA: Constrained PARAFAC2 for Sparse & Large Datasets.

PARAFAC2 has demonstrated success in modeling irregular tensors, where the tensor dimensions vary across one of the modes. An example scenario is modeling treatments across a set of patients with the varying number of medical encounters over time. Despite recent improvements on unconstrained PARAFAC2, its model factors are usually dense and sensitive to noise which limits their interpretability. As a result, the following open challenges remain: a) various modeling constraints, such as temporal smoothness, sparsity and non-negativity, are needed to be imposed for interpretable temporal modeling and b) a scalable approach is required to support those constraints efficiently for large datasets. To tackle these challenges, we propose a COnstrained PARAFAC2 (COPA) method, which carefully incorporates optimization constraints such as temporal smoothness, sparsity, and non-negativity in the resulting factors. To efficiently support all those constraints, COPA adopts a hybrid optimization framework using alternating optimization and alternating direction method of multiplier (AO-ADMM). As evaluated on large electronic health record (EHR) datasets with hundreds of thousands of patients, COPA achieves significant speedups (up to 36× faster) over prior PARAFAC2 approaches that only attempt to handle a subset of the constraints that COPA enables. Overall, our method outperforms all the baselines attempting to handle a subset of the constraints in terms of speed, while achieving the same level of accuracy. Through a case study on temporal phenotyping of medically complex children, we demonstrate how the constraints imposed by COPA reveal concise phenotypes and meaningful temporal profiles of patients. The clinical interpretation of both the phenotypes and the temporal profiles was confirmed by a medical expert.

Predicting Changes in Pediatric Medical Complexity using Large Longitudinal Health Records.

Medically complex patients consume a disproportionate amount of care resources in hospitals but still often end up with sub-optimal clinical outcomes. Predicting dynamics of complexity in such patients can potentially help improve the quality of care and reduce utilization of hospital resources. In this work, we model the change prediction of medical complexity using a large dataset of 226K pediatric patients over 5 years from Children's Healthcare of Atlanta (CHOA). We compare different classification methods including logistic regression, random forest, gradient boosting trees, and multilayer perceptron in predicting whether patients will change their complexity status in the last year based on the data from previous years. We achieved an area under the ROC curve (AUC) of 88% for predicting noncomplex patients becoming complex and 74% for predicting complex patients staying complex. We also identify the factors associated with the change in complexity of patients.

Development of an Electronic Pediatric All-Cause Harm Measurement Tool Using a Modified Delphi Method.

OBJECTIVES: To have impact on reducing harm in pediatric inpatients, an efficient and reliable process for harm detection is needed. This work describes the first step toward the development of a pediatric all-cause harm measurement tool by recognized experts in the field. METHODS: An international group of leaders in pediatric patient safety and informatics were charged with developing a comprehensive pediatric inpatient all-cause harm measurement tool using a modified Delphi technique. The process was conducted in 5 distinct steps: (1) literature review of triggers (elements from a medical record that assist in identifying patient harm) for inclusion; (2) translation of triggers to likely associated harm, improving the ability for expert prioritization; (3) 2 applications of a modified Delphi selection approach with consensus criteria using severity and frequency of harm as well as detectability of the associated trigger as criteria to rate each trigger and associated harm; (4) developing specific trigger logic and relevant values when applicable; and (5) final vetting of the entire trigger list for pilot testing. RESULTS: Literature and expert panel review identified 108 triggers and associated harms suitable for consideration (steps 1 and 2). This list was pared to 64 triggers and their associated harms after the first of the 2 independent expert reviews. The second independent expert review led to further refinement of the trigger package, resulting in 46 items for inclusion (step 3). Adding in specific trigger logic expanded the list. Final review and voting resulted in a list of 51 triggers (steps 4 and 5). CONCLUSIONS: Application of a modified Delphi method on an expert-constructed list of 108 triggers, focusing on severity and frequency of harms as well as detectability of triggers in an electronic medical record, resulted in a final list of 51 pediatric triggers. Pilot testing this list of pediatric triggers to identify all-cause harm for pediatric inpatients is the next step to establish the appropriateness of each trigger for inclusion in a global pediatric safety measurement tool.

Cloud-based Predictive Modeling System and its Application to Asthma Readmission Prediction.

The predictive modeling process is time consuming and requires clinical researchers to handle complex electronic health record (EHR) data in restricted computational environments. To address this problem, we implemented a cloud-based predictive modeling system via a hybrid setup combining a secure private server with the Amazon Web Services (AWS) Elastic MapReduce platform. EHR data is preprocessed on a private server and the resulting de-identified event sequences are hosted on AWS. Based on user-specified modeling configurations, an on-demand web service launches a cluster of Elastic Compute 2 (EC2) instances on AWS to perform feature selection and classification algorithms in a distributed fashion. Afterwards, the secure private server aggregates results and displays them via interactive visualization. We tested the system on a pediatric asthma readmission task on a de-identified EHR dataset of 2,967 patients. We conduct a larger scale experiment on the CMS Linkable 2008-2010 Medicare Data Entrepreneurs' Synthetic Public Use File dataset of 2 million patients, which achieves over 25-fold speedup compared to sequential execution.

A trigger tool to detect harm in pediatric inpatient settings.

OBJECTIVES: An efficient and reliable process for measuring harm due to medical care is needed to advance pediatric patient safety. Several pediatric studies have assessed the use of trigger tools in varying inpatient environments. Using the Institute for Healthcare Improvement's adult-focused Global Trigger Tool as a model, we developed and pilot tested a trigger tool that would identify the most common causes of harm in pediatric inpatient environments. METHODS: After formal training, 6 academic children's hospitals used this novel pediatric trigger tool to review 100 randomly selected inpatient records per site from patients discharged during the month of February 2012. RESULTS: From the 600 patient charts evaluated, 240 harmful events ("harms") were identified, resulting in a rate of 40 harms per 100 patients admitted and 54.9 harms per 1000 patient days across the 6 hospitals. At least 1 harm was identified in 146 patients (24.3% of patients). Of the 240 total events, 108 (45.0%) were assessed to have been potentially or definitely preventable. The most common patient harms were intravenous catheter infiltrations/burns, respiratory distress, constipation, pain, and surgical complications. CONCLUSIONS: Consistent with earlier rates of all-cause harm in adult hospitals, harm occurs at high rates in hospitalized children. Availability and use of an all-cause harm identification tool will establish the epidemiology of harm and will provide a consistent approach to assessing the effect of interventions on harms in hospitalized children.