Differences in morbidity measures and risk factor identification using multiple data sources: the case of stroke

Area of Science:

• Epidemiology
• Public Health
• Biostatistics

Background:

• Epidemiologic studies rely on diverse data sources like interviews, medical records, and death certificates.
• The selection of data sources can significantly impact the validity of study findings.
• Understanding how different case-ascertainment methods affect risk estimates is crucial for accurate research.

Purpose of the Study:

• To evaluate the impact of seven distinct case-ascertainment algorithms on stroke incidence and risk factor estimates.
• To determine if variations in data source combinations lead to substantive differences in study outcomes.
• To assess the reliability of different methodologies in defining case status for stroke.

Main Methods:

• Utilized the NHANES I Epidemiologic Follow-up Study (NHEFS) data.
• Developed and applied seven different algorithms for case ascertainment.
• Each algorithm combined various data sources to define stroke case status.
• Analyzed differences in incidence rates and relative risks associated with stroke risk factors.

Main Results:

• Seven case identification models produced significantly different stroke incidence estimates.
• Case characteristics were largely consistent across models, excluding those identified solely by death certificate.
• Relative risk estimates were similar across six of the seven models.
• The death certificate-only model yielded notably higher relative risk estimates.

Conclusions:

• Despite variations in incidence estimates, case characteristics and stroke risk associations remained similar across most methods.
• Reliance on death certificates alone for case identification presented the most significant divergence.
• Analyses utilizing self-reported data demonstrated validity and utility in stroke incidence studies.