When models diverge: Harmonizing pneumococcal vaccination cost-effectiveness analyses to inform PCV20 and PCV21 use...

Policymakers use decision models to assess pneumococcal vaccine cost-effectiveness, but different models often yield inconsistent results. We systematically compared 2 published models evaluating 20- and 21-valent pneumococcal conjugate vaccines (PCV20/PCV21) in adults aged 50+ years.

Markov decision model structures, parameters, and assumptions were iteratively modified, comparing PCV20 use in higher-risk 50-64-year-olds and all 65-year-olds to either PCV20 or PCV21 for all at age 50 or 65. One model was adjusted toward the other by harmonizing strategies, aligning parameter values, performing sensitivity analyses to identify influential parameters, and modifying sensitive parameters to narrow differences. Other analyses compared differing cohort structures and vaccine timing.

Initial results diverged widely, driven by differing disease risk, vaccination likelihood and serotype coverage, and childhood vaccination indirect effects - all uncertain and debated. Results were extremely sensitive to parameter variation due to minuscule incremental differences between strategies. Parameter harmonization produced substantial convergence. Other factors that reduced divergence were vaccine effectiveness assumptions, cohort age, and single-age vs. multi-age cohorts.

Divergent model results were mitigated through harmonization. Cost-effectiveness ratios were very sensitive to alternative model inputs. Model comparisons can identify uncertainty, reveal evidence gaps, enhance transparency, and potentially improve decision making.