The clinic that worked - until it didn't
A county launches a program for its highest-poverty census tracts: free blood-pressure screening, gym vouchers, nutrition classes, and nurse follow-up calls. At 12 months, mean systolic pressure in the enrolled tracts has dropped 9 mmHg, and the gap versus affluent tracts has narrowed by half. At 36 months, with the program still running and attendance stable, mean systolic pressure in the enrolled tracts has drifted back up and the gap is within 1 mmHg of where it started. Program staff conclude the community 'lost motivation' and propose adding a financial incentive for attendance.
1. Offer the strongest structural explanation for the 36-month pattern, and state what it predicts about the incentive proposal.
Read the worked answer
Fundamental cause theory: socioeconomic position affects health not through any single mechanism but through access to flexible resources - money, knowledge, power, prestige, and beneficial social connections - that can be used to avoid risk regardless of which specific risk is currently salient. When you close one pathway (gyms, classes, screening), advantage reasserts itself through whichever pathway is next available (housing quality, occupational exposures, food access, time, stress physiology).
Prediction: the 12-month narrowing reflects the program occupying the dominant pathway temporarily; drift back is mechanism replacement, not motivational decay. The attendance-incentive proposal targets individual effort, which is the exact variable the theory says is not the binding constraint - so expect small short-term gains and renewed drift, plus a political cost: framing the problem as motivation locates the failure in the community rather than in resource distribution.
Discriminating evidence would be a rise in upstream risk exposure (housing instability, work schedules, food insecurity) over the same 36 months in the enrolled tracts.
2. Staff point out attendance never fell. Why does that fact matter diagnostically?
Read the worked answer
It rules out the individual-behavior explanation the staff are offering. If participation held steady while outcomes worsened, then the behavior the program targeted was maintained and the outcome changed anyway - so something outside the program's behavior chain must have moved.
This is a mediation check: the proposed causal chain is program → attendance/behavior → blood pressure. Attendance is stable (first link intact), blood pressure regressed (final link broken), so either the intermediate behavioral change decayed while attendance held (people attending but no longer exercising - check it), or a competing exposure increased.
That possibility should be tested before any incentive is added: an incentive increases attendance, and attendance was already stable, so the proposal would likely produce no change - a clean, falsifiable prediction that also protects against spending on the wrong lever.
3. The program is refunded with a mandate to 'narrow the gap within 2 years.' What design would you argue for, and what would you tell the funders about the timeline?
Read the worked answer
Argue for interventions aimed at resource distribution rather than individual behavior: stable housing, transportation, paid leave and scheduling flexibility, income supports, and food environment changes - plus retaining the clinical components, which are cheap and do help individuals even if they do not shift the population gap.
On timeline, be explicit that the 2-year mandate is in tension with the mechanism. Programs that occupy one pathway can produce narrowing within a year - the county already did it. But that narrowing is fragile precisely because it is pathway-specific, so a 2-year endpoint will reward exactly the intervention design that produced the 36-month rebound. Recommend measuring both the 12-month marker and upstream exposures (housing, food insecurity, employment precarity) as leading indicators, with the gap as a lagging indicator.
Also name the confound honestly: secular trends in the affluent tracts can widen a gap without anything changing in the poor tracts, so the comparison group's trajectory should be modeled, not assumed flat.