Cost-of-illness studies based on massive data: a prevalence-based, top-down regression approach

Area of Science:

• Health economics
• Biostatistics
• Data science

Background:

• Routine data availability for cost-of-illness (COI) studies is increasing.
• Existing analysis methods are not suitable for massive datasets.

Purpose of the Study:

• To present a novel analysis method for COI studies using massive routine data.
• To assess the numerical efficiency gains of the proposed method.

Main Methods:

• A prevalence-based, top-down regression approach was developed.
• The method involves data aggregation, generalized additive model (GAM) fitting, cost prediction, and uncertainty quantification.
• Applied to German sickness funds data for chronic lung disease (over 7.3 million insured).

Main Results:

• The innovative method demonstrated practical computational efficiency (19 minutes).
• Analysis of individual-level data showed disproportionately increased computational time and a high risk of model failure (approx. 80% for 6 million subjects).

Conclusions:

• The proposed COI analysis method offers significant computational efficiency gains.
• This approach may lead to more precise cost estimates in large-scale health economic evaluations.