Quantifying and reducing statistical uncertainty in sample-based health program costing studies in low- and...

Area of Science:

• Health economics
• Public health program evaluation
• Statistical modeling

Background:

• Public health program costs in low- and middle-income countries are often not tracked.
• Estimating these costs using facility-based surveys introduces statistical uncertainty.
• Quantifying this uncertainty is crucial for decision-making but often not reported.

Purpose of the Study:

• To provide an overview of statistical uncertainty in complex costing surveys.
• To emphasize sources of uncertainty in program cost estimation.
• To highlight methods for calculating and reducing uncertainty.

Main Methods:

• Describing formulae and resampling techniques (e.g., bootstrap) for uncertainty computation.
• Overviewing calibration as a method to decrease uncertainty using auxiliary information.
• Demonstrating uncertainty reduction in a national immunization program costing study.

Main Results:

• Statistical uncertainty in program cost estimates can be quantified using standard errors and confidence intervals.
• Calibration, using readily available auxiliary data, significantly reduces uncertainty.
• Uncertainty reduction is applicable to estimating total program costs and intervention effects.

Conclusions:

• Measures of statistical uncertainty are vital for health policy and future study design.
• Technical challenges and software awareness hinder the reporting of uncertainty.
• Modern statistical methods, like calibration, improve cost estimation by reducing uncertainty.