A Community Health Worker Program and the Question of Evidence

She should distinguish between 'evidence-based' as a registry listing and evidence of effectiveness generated in context. The pilot is a single-group pre-post design: it cannot rule out regression to the mean, secular trends, or selection (participants who enroll may be more motivated). The 0.6% HbA1c drop is modest and within the range attributable to attention and structured contact; the 35% non-attendance rate is a threat to both internal and external validity and suggests the dose actually received varies widely. She should not overclaim. At the same time, the funder's registry criterion is a proxy, and culturally adapted, promotora-delivered programs have a reasonable evidence base in the literature; she can cite the general model while being clear that her local adaptation is not yet locally evaluated. A defensible response: present the pilot as promising feasibility data, propose a modest evaluation design within the expansion (e.g., stepped-wedge rollout across sites, which uses the natural staggered implementation and provides a comparison), and commit to reporting outcomes regardless of direction. A stepped-wedge design is feasible without research staff if data collection is built into routine care, though she should consult an IRB and a biostatistician collaborator, possibly through an academic partnership.

Non-attendance is not noise; it is data about fit. She should characterize who missed sessions and why - work schedules, transportation, caregiving, language variation (e.g., indigenous language speakers within the Spanish-speaking population), distrust, or session content that did not feel relevant. If the program is expanded unchanged, the same attendance pattern will likely reproduce at scale, and the average effect will be diluted. Options include: reducing to a shorter core with booster sessions, offering make-up or telehealth sessions, addressing transportation and childcare more aggressively, and using motivational or peer-navigation strategies to re-engage. She should also reconsider the outcome metric: per-protocol effects among completers may look better than intention-to-treat effects, and the funder and board should see both. Expansion is not automatically warranted by a promising pilot; a staged rollout with pre-specified attendance and outcome targets allows her to learn while scaling. She should present the attendance problem to the board as a design question, not a community-engagement failure, and avoid blaming participants.

She should recommend a staged expansion - e.g., two additional sites in year one with the stepped-wedge evaluation, not all five - contingent on addressing attendance and on a data infrastructure that captures reach, attendance, HbA1c, and patient-reported outcomes. She should frame equity explicitly: the program exists because standard diabetes education has under-served this population; requiring a registry listing that the program cannot yet meet would effectively penalize a culturally tailored intervention for being new. The promotora model's cultural congruence (language, shared experience, trust) is a plausible mechanism, and the evaluation should measure it rather than assume it. She should also be honest about the limits: satisfaction is not effectiveness, and high satisfaction can coexist with modest clinical change. Finally, she should address the sustainability question the board is really asking - one year of funding and no research staff means the evaluation must be lean; she should propose specific, low-burden measures and a named data owner. She should document the decision rationale for the funder, including what would constitute evidence of failure and trigger discontinuation, which is both scientifically and ethically stronger than an open-ended commitment.