Validity of self-reported mammography uptake in the Belgian health interview survey: selection and reporting bias.

Finaba Berete,Stefaan Demarest,Olivier Bruyère,Françoise Renard,Rana Charafeddine,Herman Van Oyen,Johan Van Der Heyden,Jean Tafforeau

doi:10.1093/eurpub/ckaa217

Abstract

BackgroundThe validity of self-reported mammography uptake is often questioned. We assessed the related selection and reporting biases among women aged 50–69 years in the Belgian Health Interview Survey (BHIS) using reimbursement data for mammography stemming from the Belgian Compulsory Health Insurance organizations (BCHI).MethodsIndividual BHIS 2013 data (n = 1040) were linked to BCHI data 2010–13 (BHIS–BCHI sample). Being reimbursed for mammography within the last 2-years was used as the gold standard. Selection bias was assessed by comparing BHIS estimates reimbursement rates in BHIS–BCHI with similar estimates from the Echantillon Permanent/Permanente Steekproef (EPS), a random sample of BCHI data, while reporting bias was investigated by comparing self-reported versus reimbursement information in the BHIS–BCHI. Reporting bias was further explored through measures of agreement and logistic regression.ResultsMammography uptake rates based on self-reported information and reimbursement from the BHIS–BCHI were 75.5% and 69.8%, respectively. In the EPS, it was 64.1%. The validity is significantly affected by both selection bias {relative size = 8.93% [95% confidence interval (CI): 3.21–14.64]} and reporting bias [relative size = 8.22% (95% CI: 0.76–15.68)]. Sensitivity was excellent (93.7%), while the specificity was fair (66.4%). The agreement was moderate (kappa = 0.63). Women born in non-EU countries (OR = 2.81, 95% CI: 1.54–5.13), with high household income (OR = 1.27, 95% CI: 1.02–1.60) and those reporting poor perceived health (OR = 1.41, 95% CI: 1.14–1.73) were more likely to inaccurately report their mammography uptake.ConclusionsThe validity of self-reported mammography uptake in women aged 50–69 years is affected by both selection and reporting bias. Both administrative and survey data are complementary when assessing mammography uptake.

Highlights

We investigate the validity of self-reported mammography attendance in the Belgian Health Interview Survey (BHIS), as a proxy of screening uptake by assessing the associated selection and reporting biases
The BHIS collects information on mammography uptake by means of a self-administered questionnaire in women aged 15 years and older: Have you ever had mammograms? ‘Yes/No’ and for those who respond ‘Yes’, the time lapse since her last mammograms: ‘When was the last time you had mammograms?’ the BHIS collects data on a wide range of other health and health-related topics such as demographic information, socioeconomic status (SES) and self-reported health status, life style and health services use
The absolute and relative size of the selection bias is 5.72% points and 8.93%, respectively

Summary

Introduction

Breast cancer is the most common cancer in terms of incidence among women both in developed and developing countries[1,2,3] and the second cause of cancer death among women after lung cancer in most developed countries.[4]Early detection of breast cancer through mammography screening is recognized as being effective in reducing mortality[5,6]; in women aged 50–69 years.[7,8,9] Literature suggests that with a screening attendance reaching 70%, a reduction in breast cancer mortality by about 25% might be expected.[8,10] European guidelines recommend biennial mammography screening for women aged 50–69 years.[11]Valid methods of determining and monitoring breast cancer screening (screening) uptake are important to evaluate screening programs.[6,12,13] Underestimating screening prevalence could lead to waste of resources, while overestimation could lead to missed opportunities for improving screening.[6]. We assessed the related selection and reporting biases among women aged 50–69 years in the Belgian Health Interview Survey (BHIS) using reimbursement data for mammography stemming from the Belgian Compulsory Health Insurance organizations (BCHI). Selection bias was assessed by comparing BHIS estimates reimbursement rates in BHIS–BCHI with similar estimates from the Echantillon Permanent/Permanente Steekproef (EPS), a random sample of BCHI data, while reporting bias was investigated by comparing self-reported versus reimbursement information in the BHIS–BCHI. Conclusions: The validity of self-reported mammography uptake in women aged 50–69 years is affected by both selection and reporting bias. Both administrative and survey data are complementary when assessing mammography uptake

Objectives

Methods

Results

Discussion

Conclusion