Outreach Decay
A patient outreach program can work and still fail.
A cluster RCT published in JAMA Network Open in Oct 2026 examined whether a centralized, annual, mailed FIT outreach program could improve screening rates across 16 FQHC clinics in San Diego County. The intervention involved a mailed educational leaflet, a FIT kit with a prepaid return envelope and automated phone and text reminders at 2 and 3 weeks for those who hadn't responded. The materials were provided in both English and Spanish. The hub organization was responsible for the outreach, so the clinic site staff weren't the bottleneck.
The results at years 1 and 2 were meaningful:
In year 1, 39.0% of patients in the intervention clinics were screened compared to 27.8% in the usual care clinics (a difference of 11.2 pp).
In year 2, the difference was maintained at 37.8% v 27.0% (a difference of 10.8 pp). Both differences were statistically significant. The intervention, as designed, delivered a sustained and clinically meaningful improvement over two consecutive annual cycles.
Year 3 is where it becomes instructive.
The intervention itself continued to operate, but the organization lost its ability to measure it, these are not the same thing.
Year three
At the year-three screening, the rate in the intervention clinics was 39.4% (very similar to years one and two), indicating that the intervention arm was stable. What changed was in the usual care arm. The control clinics, which had been at 27.8% and 27.0% in prior years, rose to 36.5% in year three, closing the gap to 2.9 points and pushing the primary endpoint to non-significance.
Some of the convergence is due to real improvement in usual care (the end of pandemic suppression, quality improvement work within the FQHCs and co-interventions that control sites, which were rolled out organically during the trial period). Some of it, however, is due to a measurement artifact. The authors point out that in year 3, the EHR query used to capture screening status was run two months before the end of the observation window (rather than at or after the end of the window in prior years). That shift in timing meant that patients who had been screened in year 2 were flagged as not yet due for their repeat FIT, which reduced the apparent eligible population and made the comparisons across arms very difficult.
The authors are very open about the difficulties with intervention fidelity in year 3, which, in turn, limits our ability to isolate the intervention's effect. They do not conclude that the intervention failed. They conclude that the trial wasn't set up to measure whether it worked cleanly.
The distinction that matters for healthcare marketers
The study is being cited in the context of the lesson on how patient outreach programs lose their effectiveness over time. That's not what the data actually show. The correct interpretation is that the problem of maintaining the infrastructure necessary to deliver outreach consistently and to measure it accurately is a separate issue from the problem of designing the outreach in the first place.
This is the distinction that really matters for anyone running a preventive care program. A FIT mailing protocol is not difficult to design at all. It's a letter, a kit, a prepaid envelope and a follow-up text. The logic is simple, and the evidence base is solid. What the trial shows is that the real difficulty lies in maintaining that protocol over multiple annual cycles through the inevitable churn of staff, EHR changes, competing demands and budgetary pressure. The hub model was designed to remove the onus of execution from the individual clinics, and it worked for two years. The problem in year three was not a clinic failure at all. It was a timing error in how the hub processed the EHR data.
The program and process erode at different rates and for different reasons.
What this means for program design
The practical implication is that any preventive care outreach program needs two things to be explicitly designed rather than just assumed.
The first is the intervention itself (who gets it, when, via what channel, in what language and with what follow-up).
The second is the fidelity infrastructure (who owns each step, how timing is checked, what data review takes place prior to each cycle going out and how the organization checks that what was meant to go out actually does go out).
Most organizations put their money into the first and fail to invest in the second. The intervention design is the obvious one, the one that you can see, you can document, and you can present to leadership. The fidelity infrastructure is the operational detail: the scheduling discipline, the data query protocols, the ownership structures and the sort of cross-system coordination that doesn't feature on a campaign brief. It's also the thing that gets nibbled away first when there's a change in staffing, when a grant cycle ends, or when another initiative takes attention.
The Martinez et al. trial was conducted with a central hub to reduce the operational burden on individual clinics. Even with that level of centralization, year three was a measurement failure of such magnitude that the authors felt it was appropriate to designate year two as the primary outcome. That is not a reason to abandon the centralized outreach model. It is a reason to treat fidelity review as a core component of program operations rather than as a quality improvement exercise on top of the program.
For any organization running annual or repeat patient outreach (for any form of preventive care service, not just colorectal cancer screening), the year 3 result will inevitably prompt a particular question as you enter the next cycle: is the process that generated last year's outreach still in place?
The answer is usually no. The staff has changed. The EHR has been updated. The vendor relationship has changed. A data field that used to pull from the right place is now pulling from a different table. The mailing list was built on last year's logic, which included a rule that no longer exists. None of these are major failures. Any one of them is quite sufficient to give you a year three result.
The intervention that worked in year one did so because someone had actually made it work in that particular configuration with that particular team. The question for year three is whether that configuration still exists or whether the organization is simply running a version of the program that looks like the original but isn't.
A program that works once proves the point. A program that works for three years in a row proves the organization.
Martinez, M.E., Roesch, S., Largaespada, V., Nodora, J.N., Rabin, B.A., Castaneda, S.F., Covin, J., Ortwine, K., Preciado-Hidalgo, Y., Howard, N., Schultz, J., Stamm, N., Halpern, M.T., & Gupta, S. (2026, October 21). Fecal immunochemical test outreach for colorectal cancer screening in community health centers: A randomized clinical trial. JAMA Network Open, 9(10), e2639013. https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2855574
Note: Cluster-randomized trial across 16 FQHC clinics in San Diego County; 20,250 patients at baseline; June 2021 to May 2024. Funded by the National Cancer Institute. Year 2 designated as secondary outcome; year 3 designated as primary. Authors note a protocol modification in year 3 (EHR query timing) as a fidelity limitation affecting the interpretation of the primary endpoint.
Some ideas are worth discussing in the context of your organization.
