A Computationally Efficient Method for Probabilistic Parameter Threshold Analysis for Health Economic Evaluations.

Area of Science:

• Health economics
• Decision analysis
• Biostatistics

Background:

• Threshold analysis determines cost-effectiveness by identifying critical input parameter values.
• Traditional deterministic threshold analysis assumes linearity and independence of inputs, leading to inaccuracies in complex models.
• Nonlinearity and parameter correlation in cost-effectiveness models necessitate advanced analytical approaches.

Purpose of the Study:

• To introduce and evaluate a probabilistic method for threshold analysis.
• To account for joint uncertainty in all input parameters without assuming model linearity.
• To compare the proposed method against deterministic and gold-standard probabilistic approaches.

Main Methods:

• Deterministic Threshold Analysis (DTA): Varies one input at a time, fixing others at their mean.
• 2-level Monte Carlo simulation: Considered the gold standard for probabilistic sensitivity analysis.
• Generalized Additive Model (GAM): A regression-based method utilizing probabilistic sensitivity analysis samples to directly estimate thresholds.

Main Results:

• DTA failed to identify cost-effective thresholds for typhoid vaccination strategies in Uganda.
• Probabilistic methods (2-level Monte Carlo and GAM) provided accurate threshold estimates for vaccination strategies.
• GAM demonstrated comparable accuracy to the 2-level Monte Carlo approach but was significantly more computationally efficient (430x faster).

Conclusions:

• For nonlinear cost-effectiveness models, the GAM approach offers a computationally efficient and accurate alternative to traditional methods.
• DTA is unreliable and should be avoided due to its tendency to produce incorrect threshold values.
• Probabilistic threshold analysis is crucial for accurate health economic evaluations, especially when dealing with model nonlinearity and parameter uncertainty.