The choice of health care modeling approaches is driven by trade-offs between various
modeling techniques. This study evaluates cohort (CH) versus patient-level (PL) Markov
modeling techniques within a cost-effectiveness analysis framework to understand the
practical decisions analysts face. Both the CH and PL models were constructed using
identical datasets and similar assumptions. Each model included eight health states
to capture disease severity and symptom types and allowed switching from first-line
to second-line treatment. We assessed model outcomes and performance using various
quantitative and qualitative techniques. The CH and PL models yielded very similar
base case results; only minor differences in functionality and outcome consistency
were detected. The CH model offered greater stability and easier parameter testing,
while the PL model provided superior flexibility for structural adjustments and detailed
patient pathway and subgroup analysis. However, the PL model required substantially
more computational time for sensitivity analyses and more technical skills to understand
and interpret patient pathways and model results. CH modeling faced more challenges
when extensive structural changes were initiated. Choosing between CH and PL modeling
techniques involves the careful assessment of trade-offs between the need for a flexible
and informed model and the optimization of human and computational resources.
