OBJECTIVES: The choice of modeling approach is a crucial question faced by health economists. The choice is based partly on the advantages and limitations of the modeling techniques. Thus, in this study, we evaluated the trade-offs between cohort- and patient-level Markov modeling techniques.

METHODS: We used identical datasets and similar assumptions to construct two models: a Markov cohort model (CH) and a patient-level simulation (PL) model. The relative difference in patient proportion was quantified using the newly introduced metric “health state disparity index.” The absolute difference values between the CH and PL models divided by the patient proportions in the corresponding state of the CH model were weighted by the differences in patient proportions between the models in each cell. PL involved 10,000 iterations for each patient. The sample size for the PL was determined after 100 runs using different numbers of patients, and the sample size chosen was the one balancing the low coefficient of variation and acceptable computational burden.

RESULTS: The base case results of the CH and PL were almost identical. The stability was higher, and parameter testing was easier with CH. Flexibility for structure adjustments and analysis of individual pathways was higher with PL. Sensitivity analyses required significantly more time for PL. Interpreting the results of the PL requires more skilled calibers than those of the CH. Adapting to extensive structural changes is more challenging in CH. Both models require some knowledge and experience with Excel formulas and Visual Basic for Application (VBA).

CONCLUSIONS: Choosing between cohort and patient-level modeling techniques involves a careful assessment of their strengths and limitations, depending on the analytical needs and complexities of the case. This decision underscores the need for a flexible and informed model selection process to optimize outcomes.