Evaluation of EHR Data for Fulfilling Criteria Specified for Public Health Disease Reporting

Monday, June 20, 2016: 2:44 PM
Tubughnenq' 6 / Boardroom, Dena'ina Convention Center
Shan He , Intermountain Healthcare, murray, UT
Catherine Staes , University of Utah, Salt Lake City, UT
Jim Murphy , Intermountain Healthcare, Salt Lake City, UT
Shu McGarvey , Northrop Grumman Information Systems, Atlanta, GA
Darren Mann , Intermountain Healthcare, Salt Lake City, UT
Sidney Thornton , Intermountain Healthcare, Salt Lake City, UT
BACKGROUND: Disease surveillance is dependent on reports from healthcare providers.  Reporting criteria in CSTE Position Statements include clinical, laboratory, pharmacologic, and epidemiologic evidence, but the quality and availability over time of this information in an electronic health record (EHR) is unknown. In 2015, Intermountain Healthcare collaborated with public health authorities, a clinical decision support (CDS) vendor, and the CDC/CSTE Reportable Condition Knowledge Management System (RCKMS) team. Epidemiologists harmonized reporting criteria for Tuberculosis, Chlamydia, Pertussis, and Lead. As a healthcare provider, our objectives were to a) automate use of external CDS services to identify patients with reportable conditions, b) evaluate the effectiveness of individual reporting criterion and c) identify common causes in the instance of low sensitivity.   

METHODS: First, Intermountain implemented data-driven triggers based on 'trigger codes' provided by RCKMS, then sent patient encounter summaries to the CDS service for determination whether the case required reporting. Next, we queried the EHR data from 2011 to 2015 to retrieve ‘cases’ identified by symptomatic, diagnostic, or laboratory criterion to evaluate each set of criteria. For example, for Pertussis, we assessed the sensitivity of 'apnea and cough and age<=2', vs. 'a Pertussis diagnosis', vs. 'any lab test positive for B. pertussis’.  For this preliminary analysis, the reference standard was the union of confirmed cases from the problem list, laboratory results, and billing codes. Finally, to identify the common causes for false negative or positive results for each criterion, we manually reviewed a subset of the patient records, including clinical notes.

RESULTS: Intermountain successfully interfaced with the external service-based CDS.  Among trigger codes provided by RCKMS, few (2%-15%) were actively used. Among all reportable Pertussis cases, 67% had a positive laboratory result, 6% had Pertussis in the problem list, 38% had a pertussis-related billing code, and no record met the symptom & age criteria. Some records (10%) met more than one criterion. Each reporting criterion was evaluated for effectiveness and timeliness. Most laboratory results were recorded in a structured and coded format, whereas diagnoses and symptoms were not always documented in a structured format by physicians. Manual review of patient records also revealed potential duplicate reporting when patients seek care at multiple health systems.

CONCLUSIONS: Healthcare providers may rely on automated central CDS services to identify reportable cases. The effectiveness of the CDS service can be impacted by EHR data quality and availability. Timely documentation of structured and coded information will improve disease reporting.