Go to content

Identifying obstructive sleep apnea in administrative data: a study of diagnostic accuracy

Share

Background — Health administrative (HA) databases are increasingly used to identify surgical patients with obstructive sleep apnea (OSA) for research purposes, primarily using diagnostic codes. Such means to identify patients with OSA are not validated. The authors determined the accuracy of case-ascertainment algorithms for identifying patients with OSA with the use of HA data.

Methods — Clinical data derived from an academic health sciences network within a universal health insurance plan were used as the reference standard. The authors linked patients to HA data and retrieved all claims in the 2 yr before surgery to determine the presence of any diagnostic codes, diagnostic procedures, or therapeutic interventions consistent with OSA.

Results — The authors identified 4,965 patients (2003 to 2012) who underwent preoperative polysomnogram. Of these, 4,353 patients were linked to HA data; 2,427 of these (56%) had OSA based on diagnosis by a sleep physician or the apnea hypopnea index. A claim for a polysomnogram and receipt of a positive airway pressure device had a sensitivity, specificity, and positive likelihood ratio (+LR) for OSA of 19, 98, and 10.9%, respectively. An International Classification of Diseases, Tenth Revision, code for sleep apnea in hospitalization abstracts was 9% sensitive and 98% specific (+LR, 4.5). A physician billing claim for OSA (International Classification of Diseases, Ninth Revision, 780.5) was 58% sensitive and 38% specific (+LR, 0.9). A polysomnogram and a positive airway pressure device or any code for OSA was 70% sensitive and 36% specific (+LR, 1.1).

Conclusions — No code or combination of codes provided a +LR high enough to adequately identify patients with OSA. Existing studies using administrative codes to identify OSA should be interpreted with caution.

Information

Citation

McIsaac DI, Gershon A, Wijeysundera D, Bryson GL, Badner N, van Walraven C. Anesthesiology. 2015; 123(2):253-63. Epub 2015 May 1.

View Source