Patient-Treatment Match Value
An important source of physician learning arises from heterogeneity in patient responses to treatment. Even when average treatment effects are well understood, physicians often face uncertainty about how particular patients will respond, creating scope for learning about patient–treatment match quality. This form of learning is central to personalized medicine and provides a dynamic explanation for variation in treatment patterns across physicians and over time.
The literature on learning about match value emphasizes that physicians update beliefs not only about treatments in general, but about how treatments perform for different types of patients. Learning therefore depends on diagnostic signals, observed outcomes, and the physician’s willingness to experiment across patient populations. Early structural work in pharmaceutical markets formalizes this idea by modeling how learning about heterogeneous treatment effects shapes prescribing behavior following new drug entry (Coscelli and Shum (2004); Crawford and Shum (2005)). These models show that uncertainty about match quality can generate gradual diffusion, persistence in prescribing, and patient-level sorting across treatments.
More recent work brings these ideas directly into clinical settings, highlighting the role of diagnostic skill and information in guiding experimentation. Currie and MacLeod (2020) develops a model in which physicians learn about both treatment effectiveness and patient-specific match quality, showing how higher diagnostic skill leads to better targeting of therapies and improved patient outcomes. Together, this literature frames learning about match value as a key mechanism linking uncertainty, experimentation, and heterogeneity in care, distinct from both learning-by-doing and social learning.
Potential papers for presentation today include:
- Coscelli and Shum (2004) — learning about heterogeneous treatment effects with patient spillovers
- Crawford and Shum (2005) — dynamic learning and matching in pharmaceutical demand
- Currie and MacLeod (2020) — physician learning, diagnostic skill, and treatment matching