BACKGROUND: Measles, a highly contagious disease, requires rapid detection and prevention by public health agencies. It is difficult, however, to use near real-time syndromic surveillance to find suspect cases because measles symptoms are non-specific (i.e., rash, fever, cough, runny nose, and conjunctivitis). During a measles outbreak in Maricopa and surrounding counties, we aimed to create a procedure for using the National Syndromic Surveillance Program BioSense Platform to identify and prioritize hospital records with measles-like symptoms.
METHODS: We began by exploring the utility of BioSense’s built-in syndrome queries for measles, rash, and fever. The measles query, which searched for the keyword “measles” in the chief complaint and ICD-CM codes, was too specific. Conversely, the rash and fever queries were too sensitive, producing long line-lists with suspect chicken pox, shingles, and herpes cases. To improve our case-finding, we used five queries to search Maricopa County emergency department and inpatient hospital records with phpMyAdmin. We modified the three built-in queries to include common misspellings of measles and eliminate cases that mentioned other conditions. We built a new syndrome query to identify patients with symptoms of cough, runny nose, or conjunctivitis, and a new situational query to find keywords related to exposure sites in our outbreak (e.g., “Clinic A”). An algorithm programmed in SAS produced a priority score (0, lowest – 5, highest) for each record based on their chief complaint and diagnosis. Patients who mentioned an outbreak site and measles received higher scores than those with non-specific rash and fever.
RESULTS: During the enhanced case-finding period (5/26/16 – 6/30/16), the syndromic surveillance queries identified 2,292 records among 88,506 Maricopa County patient visits. SAS produced a line-list that sorted records by descending priority level. On average, 2 records were categorized in levels 5-3 daily, while 62 received scores of 2-0 daily. Records scored 3 and above were reviewed by communicable disease investigators. Among 61 reocords reviewed, 13 patients had been previously reported to MCDPH through alternate channels (e.g., reported by physician). However, BioSense identified 48 new suspect cases that required follow-up. The system was near real-time, but there was at least a 24 hour delay between patient visit and data availability through BioSense.
CONCLUSIONS: This novel procedure helped prioritize our case-finding efforts. Investigators reported that the process was easy to learn and not burdensome, but could be improved by decreasing the delay in data transmission. This use case highlights the importance of timely data transmission to BioSense.