Development of case definition criteria for querying insurance databases in heart failure epidemiological investigations: rationale and design of the HF-EPI-HUN study
R Sepp, B Mazzag, R Frigyesy, K Mihalyi, S Z Toth, H Takacs, A Kormanyos, L Szabo, M Lainscak, T HabonAbstract
Background and aims
Prevalence estimates for heart failure (HF) demonstrate significant global variability, ranging from approximately 0.37% to 6%. This wide variation is likely due, at least in part, to the lack of standardized query methods for data sources, such as clinical or administrative databases. In this work, we aimed to establish data-proven query strategies to obtain more reliable epidemiological data on HF in Hungary.
Methods and results
The administrative database of the Hungarian National Health Insurance Fund (NHIF) was queried for the period spanning January 1, 2009, to December 31, 2023. Mirroring the most general criteria currently used for querying HF occurrence data, we queried for HF codes (ICD-10 codes I11.0, I13.0, I13.2, I42, I50, and J81) as a single entry in the database, both inpatient and outpatient entries, either as a primary or secondary diagnosis. This initial query identified 1,099,068 unique patients, corresponding to 2,001,658 distinct HF code entries. The analysis of the distribution of these patients and code entries by diagnosis type revealed that the majority of HF code entries were recorded as a secondary diagnosis (e.g., complication, comorbidity, or death-related), totaling 1,376,932 entries (69%); and only 624,726 entries (31%) were coded as a primary diagnosis. Altogether, 186,333 patients (17% of all patients) received death-related secondary codes as a single diagnosis during the 15-year-long observational period, and only 54-57% of HF patients and 47-51% of HF cases received secondary HF diagnosis in conjunction with a cardiac disease-related primary diagnosis ("I" category ICD code), making these two scenarios unlikely representing "true" HF cases. As for outpatients, a total of 476,885 patients were identified, however, only 172,833 patients had at least two episodes of outpatient care, out of which 119,648 had the two outpatient encounters within two years. Through this process we established four case definition criteria for querying administrative databases for HF: i) inpatients with an HF code as a primary diagnosis; ii) inpatients with an HF code as a secondary diagnosis (categorized as "complication" or "comorbidity", but not as "cause of death") AND who have a primary diagnosis from the 'I' code category; iii) outpatients with an HF code who had a minimum of two outpatient care encounters within two years, iv) outpatients with an HF code who died within 60 days of their initial HF diagnosis, based on the assumption that they received a late diagnosis for their condition. Applying these four criteria to identify highly probable "true" HF cases reduced the total number of identified patients to 755,517 patients, 31% less compared to the initial criteria.
Conclusions
The development of these stringent query criteria, also reflecting current clinical practice, are expected to facilitate the standardization required for obtaining more realistic data in heart failure (HF) epidemiology.