Modelling length bias in a longitudinally-linked record system of HIV cases

Area of Science:

• Public Health
• Epidemiology
• Health Services Research

Background:

• HIV/AIDS remains a significant public health concern.
• Intravenous drug use (IVDU) is a known risk factor for HIV infection.
• Accurate data on healthcare utilization for different HIV patient groups is crucial for resource allocation.

Purpose of the Study:

• To compare hospital admission rates between HIV patients with a history of IVDU and other HIV patients.
• To address and correct for length-biased sampling in the analysis of hospital admissions.
• To provide a more accurate understanding of healthcare needs for IVDU HIV patients.

Main Methods:

• Utilized New York's Statewide Planning and Research Cooperative System (SPARCS) hospital discharge data.
• Linked hospital discharge reports to create longitudinal case histories.
• Developed and applied a statistical model to correct for length-biased sampling in identifying IVDU status.

Main Results:

• The study identified a significant difference in mean hospital admissions between IVDU HIV patients and other HIV patients.
• The developed model demonstrated that the difference in admissions was less pronounced after correcting for length-biased sampling.
• Findings suggest that previous estimates may have overestimated admission differences due to identification bias.

Conclusions:

• Hospital admission patterns differ between IVDU HIV patients and other HIV patients.
• Correcting for length-biased sampling provides a more accurate comparison of healthcare utilization.
• This research refines our understanding of the healthcare needs of IVDU HIV patients.