Evaluation of stability of directly standardized rates for sparse data using simulation methods.

Area of Science:

• Epidemiology
• Biostatistics

Background:

• Directly standardized rates (DSRs) are crucial for comparing disease rates across populations with varying age structures.
• Concerns exist regarding the validity of DSRs when based on a small number of events.

Purpose of the Study:

• To determine the minimum number of events required for a valid Directly Standardized Rate (DSR) in real-world data analysis.
• To evaluate methods for calculating confidence intervals for DSRs.

Main Methods:

• Monte Carlo simulations were employed with Poisson distributions for 19 age groups.
• Varied total events, age-specific risks, and population sizes across 10,000 simulations.
• Calculated DSRs and confidence intervals using normal approximation, Dobson, and Tiwari methods.

Main Results:

• Normal approximation was unsuitable for fewer than 100 events.
• Dobson and Tiwari methods provided suitable confidence intervals with 10 or more events.
• Confidence interval accuracy was unaffected by event distribution across age groups.

Conclusions:

• Directly standardized rates (DSRs) should not be reported with fewer than 10 observed events.
• The Dobson method is preferred for confidence interval calculation due to simpler formulas and slightly better accuracy.