Abstract
IntroductionThe ICD-9 codes for acute respiratory illness (ARI) andpneumonia/influenza (P&I) are commonly used in ARI surveillance;however, few studies evaluate the accuracy of these codes or theimportance of ICD-9 position. We reviewed ICD-9 codes reportedamong patients identified through severe acute respiratory infection(SARI) surveillance to compare medical record documentation withmedical coding and evaluated ICD-9 codes assigned to patients withinfluenza detections.MethodsThe Minnesota Department of Health (MDH) conducted SARIsurveillance at three hospitals. All hospitalized patients withsubmission of a physician-ordered upper respiratory specimens(e.g., sputum, throat or nasopharyngeal swabs) were enrolled.A medical chart review was conducted to identify those meetingSARI criteria, defined as patients admitted to an inpatient ward withnew onset of respiratory symptoms or acute exacerbation of chronicrespiratory conditions. Enrolled patients who did not meet the SARIcriteria were categorized as non-SARI. Residual material from theupper respiratory specimens were submitted to MDH for influenzatesting by RT-PCR. Demographic and clinical data, including upto eight ICD-9 codes, were collected through the medical recordreview. Patients with an ICD-9 code indicating ARI (460 to 466)or P&I (480 to 488) were defined as having an ARI/P&I code. Wecompared the frequency of ARI/P&I codes by SARI clinical criteriaand influenza detection and evaluated the position of the reportedARI/P&I code.ResultsFrom May 2013 through August 2015, we enrolled 5,950patients, of which 4,449 (75%) met SARI criteria and 1501 did not(non-SARI). An ARI/P&I code in any position was found in 61%(2705) of SARI vs. 16% (241) of non-SARI patients (odds ratio [OR]8.1, 95% confidence interval [CI] 7.0-9.4); an ARI/P&I code in thefirst position was found in 40% of SARI vs 7% of non-SARI patients(OR=8.6, 95% CI 7.0-10.5). Among SARI patients with at least oneARI/P&I code, 66% had their first or only ARI/P&I code in the 1stposition, 25% in the 2ndposition, and 6% in the 3rdposition. Foridentification of SARI, sensitivity/specificity was 61%/84% for ARI/P&I codes in any position and 40%/93% for ARI/P&I codes in the 1stposition. Among SARI patients, codes for pneumonia (486) and acutebronchiolitis (466.11, 466.19) were commonly reported. The mostfrequent codes among SARI patients without an ARI/P&I code werefever (780.6), acute respiratory failure (518.81), and asthma (493.92)(Table). Influenza was detected among 8% (351) of SARI patients.An ARI/P&I code in any position was more common in influenza-positive vs. influenza-negative SARI patients (77% vs 59%, OR 2.4,95% CI 1.8-3.1). An ARI/P&I code in the 1stposition was slightlymore common in influenza-positive vs -negative patients though notsignificant (44% vs 40%).ConclusionsAmong patients from whom a respiratory specimen was collected,administrative data identified those meeting SARI with moderatesensitivity and high specificity, and with lower sensitivity but greaterspecificity when limited to the 1stICD-9 position. Pneumonia andacute bronchiolitis ICD-9 codes were frequent ARI/P&I codes amongSARI patients. Further investigation is needed to determine the valueof including additional ICD-9 codes, such as respiratory distress andacute asthma exacerbation, in identifying SARI.