These come from syntheses and trials on organizational change generally, not only middle managers, graded with the same rubric as the Atlas studies.
Williams et al. 2025 · cluster randomized trial (the WISDOM trial), 21 youth mental health clinics (452 youths); this paper is in Psychiatric Services, read from its abstract · rated High strength
The implementation-leadership program (LOCI) did not sustain higher measurement-based care completion or better symptom improvement, a clear null result. Among the 81 youths who received the care, fidelity was higher at LOCI sites (24% against 1%), though low in both arms (rated Medium).
The Atlas cases show changes fading once under way; this trial found that a leadership program's advantage did not carry into the sustainment period, in a controlled setting. It is the same trial as Williams et al. 2024 (Implementation Science) and 2026 (Translational Behavioral Medicine), so it counts once; the 2026 paper, counting clinic-level monthly uptake rather than completion and symptoms, found a smaller advantage still there at 35 months, so the papers agree that the gain shrank rather than held.
Becker et al. 2025 · cluster randomized trial, 28 opioid treatment programs (186 providers) · rated High strength
Providers at programs given an enhanced strategy (incentives plus team facilitation) were more likely to reach an excellent competence benchmark, significant in two of three models, but whether programs sustained the practice did not differ (4 vs 6 programs); with 28 programs, power to detect a sustainment difference was limited, and that comparison is rated Medium.
It adds that extra launch support may have lifted provider skill, though not in every model, while whether programs kept the practice going did not differ; the Atlas cases cannot test this, and the trial had little power to detect a difference in sustainment.
Hughes et al. 2020 · meta-analysis of pooled correlations on training sustainment · rated Medium strength
Peer, supervisor and organizational support were all positively related to long-term sustainment of learned skills, with peer and supervisor support showing the strongest relationships.
It is about keeping trained skills in use, not about a whole change, and it is correlational; it fits the Atlas group on what keeps a change moving.
Still untested in the Evidence Library (Loop Part 4): no trial of a pause or initiative freeze, and no test of review cadence. Both trials are in health care.