Hospital versus population referents in two case-referent studies

Area of Science:

• Epidemiology
• Biostatistics

Background:

• Case-referent studies are crucial for investigating disease etiology.
• The selection of appropriate referent groups (hospital-based vs. population-based) is critical and can introduce systematic errors.
• Potential biases include nonresponse, information bias, and selection bias.

Purpose of the Study:

• To compare hospital-based and population-based referents in case-referent studies.
• To evaluate potential sources of systematic errors, specifically selection and information bias, between the two referent types.
• To inform the choice of referent group in epidemiological research.

Main Methods:

• Two case-referent studies (astrocytoma and pancreatic cancer) utilized both hospital and population referents.
• Systematic errors were assessed by comparing reported exposures between the two referent series.
• Analyses included comparisons of 'hard' and 'soft' questionnaire items and validation against medical records.

Main Results:

• Significant differences in reported exposures were observed between hospital and population referents.
• These differences were attributed primarily to selection bias, reflecting differing exposure experiences, rather than information bias.
• Similar exposure discrepancies were noted for both 'hard' and 'soft' questionnaire items, and when comparing reported versus recorded radiographs.
• Nonresponse rates were slightly higher in the population referent group.

Conclusions:

• Hospital referents may be less prone to selection bias compared to population referents in case-referent studies.
• Both referent types exhibit underreporting of exposure, highlighting the need for careful data collection.
• The findings suggest that selection bias is a more significant concern than information bias when choosing between hospital and population referents.