Heterogeneity of Treatment Effect in a Randomized Trial of a Communication Intervention

Interventions to promote serious illness conversations have shown promise in promoting high-quality care. However, in randomized trials, some participants may benefit more from the intervention than others.

To examine heterogeneity of treatment effect and identify subgroups of patients with serious illness who might benefit most from interventions to enhance communication about goals of care.

We used data from a multi-center cluster-randomized trial evaluating a communication intervention to increase goals-of-care discussions in the outpatient setting. Patients (n = 249 intervention, n = 288 usual care) had serious illness with an expected median survival of two years. Using model-based recursive partitioning, we tested heterogeneity of the intervention's effect on the occurrence of patient-reported goals-of-care discussions, electronic health record documentation of goals-of-care discussions, patients' ratings of quality of communication, and patients' symptoms of psychological distress at three and six months.

We found two significant interactions. For patients' overall rating of clinician communication (n = 251), the intervention effect was positive for patients with higher household income, but not those with lower income (P < 0.001). For patients' symptoms of depression at six months (n = 288), the intervention was associated with fewer symptoms of depression among those whose self-assessed health was poor, but not among those with fair to excellent health (P < 0.001).

Identifying heterogeneity of treatment effect can be a valuable exercise following completion of a randomized trial. Interactions between the intervention and patient income and self-assessed health suggest these factors could be used to design more effective interventions to enhance communication about goals of care.