Finaba Berete, Stefaan Demarest, Rana Charafeddine, Olivier Bruyère & Johan Van der Heyden
Archives of Public Health, volume 78, Article number: 120 (2020)
Abstract
Background
Health administrative data were increasingly used for chronic diseases (CDs) surveillance purposes. This cross sectional study explored the agreement between Belgian compulsory health insurance (BCHI) data and Belgian health interview survey (BHIS) data for asserting CDs.
Methods
Individual BHIS 2013 data were linked with BCHI data using the unique national register number. The study population included all participants of the BHIS 2013 aged 15 years and older. Linkage was possible for 93% of BHIS-participants, resulting in a study sample of 8474 individuals. For seven CDs disease status was available both through self-reported information from the BHIS and algorithms based on ATC-codes of disease-specific medication, developed on demand of the National Institute for Health and Disability Insurance (NIHDI). CD prevalence rates from both data sources were compared. Agreement was measured using sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) assuming BHIS data as gold standard. Kappa statistic was also calculated. Participants’ sociodemographic and health status characteristics associated with agreement were tested using logistic regression for each CD.
Results
Prevalence from BCHI data was significantly higher for CVDs but significantly lower for COPD and asthma. No significant difference was found between the two data sources for the remaining CDs. Sensitivity was 83% for CVDs, 78% for diabetes and ranged from 27 to 67% for the other CDs. Specificity was excellent for all CDs (above 98%) except for CVDs. The highest PPV was found for Parkinson’s disease (83%) and ranged from 41 to 75% for the remaining CDs. Irrespective of the CDs, the NPV was excellent. Kappa statistic was good for diabetes, CVDs, Parkinson’s disease and thyroid disorders, moderate for epilepsy and fair for COPD and asthma. Agreement between BHIS and BCHI data is affected by individual sociodemographic characteristics and health status, although these effects varied across CDs.
Conclusions
NHIDI’s CDs case definitions are an acceptable alternative to identify cases of diabetes, CVDs, Parkinson’s disease and thyroid disorders but yield in a significant underestimated number of patients suffering from asthma and COPD. Further research is needed to refine the definitions of CDs from administrative data.
