Referral bias in diabetes research

Area of Science:

• Endocrinology
• Epidemiology
• Public Health

Background:

• Diabetes mellitus is a significant public health concern with a complex presentation in the general population.
• Understanding the true prevalence and clinical spectrum of diabetes requires examining community-based data.
• Medical records and clinical settings may not capture the full scope of the disease.

Purpose of the Study:

• To assess the proportion of known diabetic individuals identified through medical record review in Rochester, Minnesota.
• To evaluate the extent to which hospitalized or clinic-attended diabetic patients represent the broader diabetic population.
• To identify potential biases in the observed clinical spectrum of diabetes within different healthcare settings.

Main Methods:

• Retrospective review of medical records for known diabetic individuals in Rochester, Minnesota over a one-year period.
• Analysis of hospitalization and diabetes clinic attendance rates among the identified diabetic population.
• Comparison of clinical features and disease severity between community-based and healthcare-accessed diabetic cohorts.

Main Results:

• Only approximately 50% of known diabetic residents were identified via a one-year medical record review.
• Fewer than 20% of diabetic residents were hospitalized or attended a diabetes clinic during the study year.
• Patients seen in hospitals or diabetes clinics showed an overrepresentation of insulin-dependent diabetes and vascular complications.

Conclusions:

• Medical record reviews and clinical data significantly underestimate the prevalence of diabetes in the community.
• Referral bias distorts the perceived clinical spectrum of diabetes, overemphasizing severe cases and complications.
• Epidemiological studies of diabetes must account for referral bias when comparing data from different medical care settings.