Repository hosted by TU Delft Library

Home · Contact · About · Disclaimer ·

The impact of response bias on estimates of health care utilization in a metropolitan area: The use of administrative data

Publication files not online:

Author: Reijneveld, S.A. · Stronks, K.
Institution: TNO Preventie en Gezondheid
Source:International Journal of Epidemiology, 6, 28, 1134-1140
Identifier: 235230
Keywords: Health · Health care utilization · Reproducibility of results · Response bias · Medical geography · Methodology · Metropolitan area · Dental care · Epidemiological data · Ethnology · Health care utilization · Health insurance · Health survey · Hospital care · Interview · Marriage · Medical care · Prescription · Register · Risk · Socioeconomics · Technical aid · Adolescent · Adult · Aged · Bias (Epidemiology) · Consumer Participation · Data Collection · Delivery of Health Care · Female · Health Care Surveys · Humans · Male · Middle Aged · Needs Assessment · Netherlands · Random Allocation · Risk Assessment · Urban Health Services · Utilization Review


Background. Surveys among the general population are an important method for collecting epidemiological data on health and utilization of health care in that population. Selective non-response may affect the validity of these data. This study examines the impact of response bias on estimates of health care utilization and on risk estimates for utilization in relation to demographic and socioeconomic characteristics, using administrative data on use of health care. Methods. Data on registered health care utilization were extracted from health insurance register and linked to respondents (2934; 62.7%) and non-respondents (1744) in a personal health interview survey among adult residents from the lower two-thirds income bracket in Amsterdam, the Netherlands. Results. Estimates of registered healthcare utilization are higher if based on respondents only, than if they are based on the entire target sample. This goes for prescription drugs, specialist medical care, paramedical care, dental care and medical aids, but not for hospital care. Most risk estimates of registered utilization for background characterisitcs (gender, family composition, marital status, year of settlement, affluence of neighbourhood and ethnicity) differ only slightly and without statistical significance. If different, most estimates based on respondents only are somewhat higher. The largest differences are found for age (average overestimation of risks for age groups 35-64 and 65+ years compared to that of 16-34 years: 16% and 17%, respectively). Conclusions. In this study, response bias affects estimates of registered health care utilization but hardly affects risk estimates of utilization by background characteristics.