Validation of an EHR Algorithm to Identify Adult and Pediatric Patients with Asthma in West Chicago.
Monday, March 7, 2016
South Exhibit Hall H (Convention Center)
Ashvini Biswas, MD, Byung H. Yu, MD, Christopher D. Codispoti, MD PhD, Sindhura Bandi, MD
Rationale: An electronic health record(EHR) algorithm to identify patients with asthma was developed. This algorithm was tested to identify asthma subjects and healthy controls among pediatric and adult patients.

Methods: The EHR was queried from 1/1/2012 to 11/30/2014 at Rush University Medical Center(RUMC) and Cook County Hospital(CCH). Asthma cases required at least one meaningful clinical encounter with an asthma diagnosis(ICD-9 code 493.xx) and either a second encounter with an asthma diagnosis or current asthma medication prescription. Control patients required two meaningful encounters without an asthma diagnosis or asthma medication. A random sample of 100 children and 100 adults(50 asthma subjects, 50 controls) at each site were manually reviewed by two physicians. Agreement between the EHR algorithm and chart review was determined by kappa score, along with positive predictive and negative predictive values(PPV and NPV).

Results: At RUMC, for the combined group, the agreement between algorithm classification and chart review was fair to good(Κ=0.51) with PPV=75%, NPV=76%. This was similar to that observed in the pediatric(Κ=0.68, PPV=82%, NPV=86%), and better to that observed in the adult(Κ=0.34, PPV=68%, NPV=66%) subgroups. At CCH, for the combined group, the agreement between the algorithm and chart review was excellent(Κ=0.81) with PPV=84%, NPV=97%. This was similar to that observed in the pediatric(Κ=0.80, PPV=82%, NPV=98%) and adult(κ=0.82, PPV=86%, NPV=96%) subgroups.

Conclusions: The EHR algorithm demonstrated good to excellent positive and negative predictive values for identifying subjects with asthma in the combined age groups of the reviewed samples at both private and public hospitals in west Chicago.