Validation of International Classification of Diseases criteria to identify severe influenza hospitalizations

Abstract

In this cohort study of hospitalized patients with linked medical record data, we developed International Classification of Diseases (ICD) criteria that accurately identified laboratory-confirmed, severe influenza hospitalizations (positive predictive value [PPV] 80%, 95% confidence interval [CI] 71-87%), which we validated through medical record documentation. These criteria identify patients with clinically important influenza illness outcomes to inform evaluation of preventive and therapeutic interventions and public health policy recommendations.

Validity of ICD‐10‐based algorithms to identify patients with influenza in inpatient and outpatient settings

“…6 In a single-center US study of hospitalized, publicly insured patients, ICD-9/10-based algorithms had a PPV of 85% (80% for severe influenza), similar to our finding for ICD-10-CM codes (any position). 7 In a study of outpatient infections diagnosed within an integrated US health system, a combined ICD-9/10-based algorithm had PPV of 80%-100% for clinical (rather than test-positive) diagnoses of influenza, again similar to our results for ICD-10-CM codes. 8 Prior studies have used different algorithms as proxies for influenza.…”

“…Using the same algorithm that we tested, Hamilton et al found that ICD‐10‐CM codes had a similar performance for identifying influenza in hospitalized patients in Ontario except for a higher PPV (91%); they did not report the PPV of primary influenza discharge diagnoses 6 . In a single‐center US study of hospitalized, publicly insured patients, ICD‐9/10‐based algorithms had a PPV of 85% (80% for severe influenza), similar to our finding for ICD‐10‐CM codes (any position) 7 . In a study of outpatient infections diagnosed within an integrated US health system, a combined ICD‐9/10‐based algorithm had PPV of 80%–100% for clinical (rather than test‐positive) diagnoses of influenza, again similar to our results for ICD‐10‐CM codes 8 …”

“…The focus on commercially insured persons tested in a single healthcare system may have limited generalizability, although our findings are comparable with those from other recent validation studies. [6][7][8] In addition, we only included patients tested for influenza within the linked dataset; untested patients may have had different risks and severity of influenza, and people with influenza in the source claims data were younger and had…”

“…Linkage between claims and EHR data enabled examination of the performance of ICD‐10‐CM‐based algorithms based on a reliable reference standard of influenza testing, limiting misclassification. Inclusion of test‐negative subjects allowed us to evaluate sensitivity, specificity, and NPV in addition to PPV, unlike recent US studies 7,8 . Furthermore, we characterized and compared algorithm performance within both inpatient and outpatient settings and across multiple subgroups, finding few differences based on age or year.…”

“…Inclusion of test-negative subjects allowed us to evaluate sensitivity, specificity, and NPV in addition to PPV, unlike recent US studies. 7,8 Furthermore, we characterized and compared algorithm performance within both inpatient and outpatient settings and across multiple subgroups, finding few differences based on age or year.…”