Rapid antibiotic administration is known to improve sepsis outcomes, however early diagnosis remains challenging due to complex presentation. Our objective was to develop a model using readily available electronic health record (EHR) data capable of recognizing infant sepsis at least 4 hours prior to clinical recognition. We performed a retrospective case control study of infants hospitalized ≥48 hours in the Neonatal Intensive Care Unit (NICU) at the Children’s Hospital of Philadelphia between September 2014 and November 2017 who received at least one sepsis evaluation before 12 months of age. We considered two evaluation outcomes as cases: culture positive–positive blood culture for a known pathogen (110 evaluations); and clinically positive–negative cultures but antibiotics administered for ≥120 hours (265 evaluations). Case data was taken from the 44-hour window ending 4 hours prior to evaluation. We randomly sampled 1,100 44-hour windows of control data from all times ≥10 days removed from any evaluation. Model inputs consisted of up to 36 features derived from routine EHR data. Using 10-fold nested cross-validation, 8 machine learning models were trained to classify inputs as sepsis positive or negative. When tasked with discriminating culture positive cases from controls, 6 models achieved a mean area under the receiver operating characteristic (AUC) between 0.80–0.82 with no significant differences between them. Including both culture and clinically positive cases, the same 6 models achieved an AUC between 0.85–0.87, again with no significant differences. Machine learning models can identify infants with sepsis in the NICU hours prior to clinical recognition. Learning curves indicate model improvement may be achieved with additional training examples. Additional input features may also improve performance. Further research is warranted to assess potential performance improvements and clinical efficacy in a prospective trial.

Journal of Medical Internet Research - International Scientific Journal for Medical Research, Information and Communication on the Internet

Develop and evaluate an interactive information visualization embedded within the electronic health record (EHR) by following human-centered design (HCD) processes and leveraging modern health information exchange standards. We applied an HCD process to develop a Fast Healthcare Interoperability Resources (FHIR) application that displays a patient’s asthma history to clinicians in a pediatric emergency department. We performed a preimplementation comparative system evaluation to measure time on task, number of screens, information retrieval accuracy, cognitive load, user satisfaction, and perceived utility and usefulness. Application usage and system functionality were assessed using application logs and a postimplementation survey of end users. Usability testing of the Asthma Timeline Application demonstrated a statistically significant reduction in time on task (P < .001), number of screens (P < .001), and cognitive load (P < .001) for clinicians when compared to base EHR functionality. Postimplementation evaluation demonstrated reliable functionality and high user satisfaction. Following HCD processes to develop an application in the context of clinical operations/quality improvement is feasible. Our work also highlights the potential benefits and challenges associated with using internationally recognized data exchange standards as currently implemented. Compared to standard EHR functionality, our visualization increased clinician efficiency when reviewing the charts of pediatric asthma patients. Application development efforts in an operational context should leverage existing health information exchange standards, such as FHIR, and evidence-based mixed methods approaches.

We compared results of a modified version of the Asthma Control Test using parent proxy report (PP-ACT) with results reported by children and parents using the validated Childhood–Asthma Control Test (C-ACT). 104 parent-child dyads with a child aged 6 to 12 years with asthma were randomized to complete PP-ACT followed by C-ACT or C-ACT followed by PP-ACT. Scores ≤19 indicated uncontrolled asthma. We calculated sensitivity, specificity, positive predictive value, and negative predictive value for the PP-ACT in comparison with the C-ACT, and calculated concordance between the 2 scales. The PP-ACT had sensitivity of 86% and negative predictive value of 88% for detecting uncontrolled asthma. More than 75% of surveys were concordant (κ = 0.54, moderate agreement). Our results suggest that while the PP-ACT missed few children with uncontrolled asthma and may simplify reporting of asthma control in circumstances when child report is not feasible or creates barriers to survey receipt, limitations of proxy reporting should be considered.

Objective. To compare practice patterns regarding the diagnosis and management of streptococcal pharyngitis across pediatric primary care practices.Design. Retr...