Not Simply More of the Same: Distinguishing between Patient Heterogeneity and Parameter Uncertainty. (November 2014)

• Record Type: Journal Article
• Authors: Vemer, Pepijn; Goossens, Lucas M. A.; Rutten-van Mölken, Maureen P. M. H.
• Is Part Of: Medical decision making. Volume 34:Number 8(2014)
• Page Start: 1048
• Page End: 1058
• Publication Date: 2014-11
• Subjects: COPD -- heterogeneity -- uncertainty -- Markov models; Medical policy -- Periodicals; Clinical medicine -- Decision making -- Periodicals; Medicine -- Periodicals; Médecine clinique -- Prise de décision -- Périodiques; 362.1
• DOI: 10.1177/0272989X14550499
• Languages: English
• ISSNs: 0272-989X

Abstract

In cost-effectiveness (CE) Markov models, heterogeneity in the patient population is not automatically taken into account. We aimed to compare methods of dealing with heterogeneity on estimates of CE, using a case study in chronic obstructive pulmonary disease (COPD). We first present a probabilistic sensitivity analysis (PSA) in which we sampled only from distributions representing parameter uncertainty. This ignores any heterogeneity. Next, we explored heterogeneity by presenting results for subgroups, using a method that samples parameter uncertainty simultaneously with heterogeneity in a single-loop PSA. Finally, we distinguished parameter uncertainty from heterogeneity in a double-loop PSA by performing a nested simulation within each PSA iteration. Point estimates and uncertainty differed substantially between methods. The incremental CE ratio (ICER) ranged from €4900 to €13, 800. The single-loop PSA led to a substantially different shape of the CE plane and an overestimation of the uncertainty compared with the other 3 methods. The CE plane for the double-loop PSA showed substantially less uncertainty and a stronger negative correlation between the difference in costs and the difference in effects compared with the other methods. This came at the cost of higher calculation times. Not accounting for heterogeneity, subgroup analysis and the double-loop PSA can be viable options, depending on the decision makers' information needs. The single-loop PSA should not be used in CE research. It disregards the fundamental differences between heterogeneity and sampling uncertainty and overestimates uncertainty as a result.