Validation of ICD-10 diagnostic coding for influenza in the Danish National Patient Registry

Background

The accuracy of recorded diagnosis codes for hospital admissions due to influenza in the Danish national registries is uncertain. We evaluated positive predictive value (PPV) and sensitivity of ICD-10 codes for influenza by comparing to the reference standard of influenza test results.

Methods

Hospital admissions were assessed in the Danish National Patient Registry (DNPR), and influenza test results in the Danish Microbiology Database (MiBa). First, we report the proportion of positive influenza virus tests within seven days of admission among hospital admissions with a discharge influenza ICD-10 code (PPV). Second, we report the proportion with ICD-10 codes for influenza among patients with an admission registered with seven days of a positive influenza virus test (sensitivity).

Results

From January 2012 – November 2022 a total of 18,761 admissions were registered with one of the 22 influenza ICD-10 codes in DNPR. Overall, there was a positive influenza test in 16,754 of the admissions (87.9 % = overall PPV, 95 % CI: 87.4–88.3). The PPV was highest for older patient groups (93.7 % in patients >80 years vs. 78.0 % in patients < 11 years), and for admissions that occurred in recent years (95.8 % in 2022 vs. 52.4 % in 2012). Among 33,834 hospitals admissions with a positive influenza test, less than half (n = 16,421, 48.5 % = sensitivity (95 % CI: 48.0 – 49.1 %)) were registered with an influenza ICD-10 code.

Conclusions

ICD-10 diagnoses codes have relatively high positive predictive value, but the sensitivity is low. Furthermore, the PPV depend on age and calendar year.

What is new

• •
  Danish national registries have reasonable positive predictive value for influenza ICD-10 codes.

• •
  Positive predictive value varies with time of hospital admission and age of the patient.

• •
  Studies based on ICD-10 codes alone underestimates the number of patients with influenza due to low sensitivity.