Patient-physician discussions about costs: definitions and impact on cost conversation incidence estimates

Background: Nearly one in three Americans are financially burdened by their medical expenses. To mitigate financial distress, experts recommend routine physician-patient cost conversations. However, the content and incidence of these conversations are unclear, and rigorous definitions are lacking. We sought to develop a novel set of cost conversation definitions, and determine the impact of definitional variation on cost conversation incidence in three clinical settings.

Methods: Retrospective, mixed-methods analysis of transcribed dialogue from 1,755 outpatient encounters for routine clinical management of breast cancer, rheumatoid arthritis, and depression, occurring between 2010-2014. We developed cost conversation definitions using summative content analysis. Transcripts were evaluated independently by at least two members of our multi-disciplinary team to determine cost conversation incidence using each definition. Incidence estimates were compared using Pearson's Chi-Square Tests.

Results: Three cost conversation definitions emerged from our analysis: (a) Out-of-Pocket (OoP) Cost--discussion of the patient's OoP costs for a healthcare service; (b) Cost/Coverage--discussion of the patient's OoP costs or insurance coverage; (c) Cost of Illness- discussion of financial costs or insurance coverage related to health or healthcare. These definitions were hierarchical; OoP Cost was a subset of Cost/Coverage, which was a subset of Cost of Illness. In each clinical setting, we observed significant variation in the incidence of cost conversations when using different definitions; breast oncology: 16, 22, 24% of clinic visits contained cost conversation (OOP Cost, Cost/Coverage, Cost of Illness, respectively; P < 0.001); depression: 30, 38, 43%, (P < 0.001); and rheumatoid arthritis, 26, 33, 35%, (P < 0.001).

Conclusions: The estimated incidence of physician-patient cost conversation varied significantly depending on the definition used. Our findings and proposed definitions may assist in retrospective interpretation and prospective design of investigations on this topic.