A Comparison of Different Approaches for Costing Medication Use in an Economic Evaluation

Estimating individual-level medication costs in an economic evaluation can involve extensive data collection and handling. Implications of detailed versus general approaches are unclear.

To compare costing approaches in a trial-based economic evaluation.

We applied four costing approaches to prescribed medication data from the Tumour necrosis factor inhibitors Against Combination Intensive Therapy randomized controlled trial. A detailed micro-costing approach was used as a base case, against which other approaches were compared: costing medications used by at least 1.5% of patients; costing medications on the basis of only chemical name; applying a generic prescription charge rather than a medication-specific cost. We quantitatively examined resulting estimates of prescribed medication and total care costs, and qualitatively examined trial conclusions.

Medication costs made up 6% of the total health and social care costs. There was good agreement in prescribed medication costs (concordance correlation coefficient [CCC] 0.815, 0.819, and 0.989) and excellent agreement in total costs (CCC 0.990, 0.995, and 0.995) between approaches 1 and 2. Approaches 3 and 4 had poor agreement with approach 1 on prescribed medication costs (CCC 0.246-0.700 and 0.033-0.333, respectively), but agreement on total care costs remained good (CCC 0.778-0.993 and 0.729-0.986, respectively).

Because medication costs comprised only a small proportion of total costs, the less resource-intensive approaches had substantial impacts on medication cost estimates, but had little impact on total care costs and did not significantly impact the trial's cost-effectiveness conclusions. There is room for research efficiencies without detriment to an evaluation in which medication costs are likely to form a small proportion of total costs.