Identifying community pharmacists’ preferences for attributes of public health interventions in Kenya: a discrete choice experiment

Community pharmacies are increasingly recognized as access points for public health interventions (PHIs) such as vaccination, family planning services, and disease screening. In Kenya, evidence suggests the feasibility of pharmacy-delivered PHIs; however, the uptake remains inconsistent. This is partly attributed to poor programme design without taking pharmacy providers preferences into consideration. We employed a discrete choice experiment (DCE) to investigate community pharmacists' preferences for attributes of PHIs delivered in community pharmacies in Kenya. We constructed a Bayesian efficient design and conducted a DCE survey among 663 community pharmacy providers in Makueni, Nairobi, and Kisumu counties in Kenya from January 2025 to March 2025. Panel multinomial mixed logit, generalized multinomial logit, and latent class models were used in the analysis. We also estimated willingness to pay (WTP) and willingness to accept (WTA) estimates using cost and profit margins as the monetary estimates, respectively. We found that community pharmacists were willing to offer PHIs with a low preference for opting out (β = -3.5723, P < 0.01). Preferences for PHIs significantly increased with higher profit margins (β = 0.028, P < 0.01) and decreased with higher cost of equipment (β = -0.00023, P < 0.01). There were higher preferences for PHIs that require moderate training (β = 0.266, P < 0.01) and extensive training (β = 0.141, P < 0.05) compared to no additional training and lower preferences for PHIs with complex interventions compared to simple interventions (β = -0.323, P < 0.01). The WTP estimates showed that providers were willing to pay Khs. 11 738 (USD 90) for moderate training and Kshs. 7327 (USD 56) for extensive training. Moreover, the WTA estimates showed that providers were willing to accept a 10.9% increase in profit margin in order to deliver complex interventions. In addition to this, a three-class latent class model revealed preference heterogeneity among the respondents. These findings can be used to inform the design of PHIs to enhance uptake and acceptability among providers.