Midkiff K, Johannes C, Tennis P, Calingaert B, McQuay L, Harris DH, Andrews E. Identifying causes of death in a large cohort of asthma patients. Poster presented at the 29th International Conference on Pharmacoepidemiology & Therapeutic Risk Management; August 2013. Montreal, Canada. [abstract] Pharmacoepidemiol Drug Saf. 2013 Aug; 22(Suppl 1):300.

BACKGROUND: A large multisite collaboration examined asthma mortality using data from 10 large health insurers/data partners. A cohort of 994,627 patients aged ≥ 4 years fulfilling a claims definition of persistent asthma were followed for fact and cause of death via linkage to the United States (US) National Death Index (NDI). NDI linkage results may include multiple matches per patient, with the most likely match usually selected by manual review. An automated algorithm (AA) was used to process the NDI results, due to the large number of patients and need for a standard method that could be applied by 10 separate data sites to select the most likely match. The AA, based on one used widely by US cancer registries, uses varying combinations and completeness of patient identifiers.

OBJECTIVES: To determine the level of agreement between the AA and the NDI most likely matches, and to characterize the causes of death in this cohort of asthma patients.

METHODS: Each data partner submitted patient identifiers to the NDI for linkage to determine fact and cause of death using a common selection procedure that screened for up to 2 years after the last claim. The level of agreement between the AA most likely match and the NDI most likely match was derived. Causes of death were tabulated.

RESULTS: Deaths totaled 31,931 in the 618,870 persons submitted to the NDI. The most common underlying cause of death (UCOD) was chronic obstructive pulmonary disease (COPD), then lung cancer and heart disease. Of 18,553 deaths identified during follow-up, asthma was listed among causes of death on the death certificate in 5.1% of deaths and was the UCOD in 1.5% of deaths. Over 30% of patients submitted to the NDI had at least one possible match returned. The AA-defined most likely match agreed with the NDI best match more than 99% of the time, regardless of whether the patient vital status was known at the time of NDI linkage.

CONCLSUIONS: The application of a standard but more liberal AA provided an efficient way to assess many potential NDI matches but did not yield a meaningful increase in the number of additional deaths identified over the NDI’s best match. Asthma as the UCOD was rare in this cohort of asthma patients.

Share on: