Implementation Research and Practice (Nov 2022)

Improving the feasibility of fidelity measurement for community-based quality assurance: Partial- versus full-session observations of supervisor adherence and competence

  • Jason E. Chapman,
  • Zoe M. Alley,
  • Sonja K. Schoenwald

DOI
https://doi.org/10.1177/26334895221135263
Journal volume & issue
Vol. 3

Abstract

Read online

Background Clinical supervision is a common quality assurance method for supporting the implementation and sustainment of evidence-based interventions (EBIs) in community mental health settings. However, assessing and supporting supervisor fidelity requires efficient and effective measurement methods. This study evaluated two observational coding approaches that are potentially more efficient than coding full sessions: a randomly selected 15-min segment and the first case discussion of the session. Method Data were leveraged from a randomized trial of an Audit and Feedback (A&F) intervention for supervisor Adherence and Competence. Supervisors ( N = 57) recorded and uploaded weekly group supervision sessions for 7 months, with one session observationally coded each month ( N = 374). Of the coded sessions, one was randomly selected for each supervisor, and a random 15-min segment was coded. Additionally, the first case discussion was coded for the full sample of sessions. Results Across all models (and controlling for the proportion of the session covered by the partial observation), Adherence and Competence scores from partial observations were positively and significantly associated with scores from full sessions. In all cases, partial observations were most accurate when the level of Adherence and Competence was moderate. At lower levels, partial observations were underestimates, and at higher levels, they were overestimates. Conclusions The results suggest that efficient observational measurement can be achieved while retaining a general level of measurement effectiveness. Practically, first-case discussions are easier to implement, whereas 15-min segments have fewer potential threats to validity. Evaluation of resource requirements is needed, along with determining whether A&F effects are retained if feedback is based on partial observations. Nevertheless, more efficient observational coding could increase the feasibility of routine fidelity monitoring and quality assurance strategies, including A&F, which ultimately could support the implementation and sustainment of effective supervision practices and EBIs in community practice settings. Plain Language Summary: When delivering evidence-based mental health interventions in community-based practice settings, a common quality assurance method is clinical supervision. To support supervisors, assessment methods are needed, and those methods need to be both efficient and effective. Ideally, supervision sessions would be recorded, and trained coders would rate the supervisor’s use of specific strategies. In most settings, though, this requires too many resources. The present study evaluated a more efficient approach. The data came from an existing randomized trial of an Audit and Feedback intervention for enhancing supervisor Adherence and Competence. This included 57 supervisors and 374 sessions across seven months of monitoring. Instead of rating full supervision sessions, a more efficient approach was to have coders rate partial sessions. Two types of partial observations were considered: a randomly selected 15-minute segment of the session and the first case discussion of the session. The aim was to see if partial observations and full observations led to similar conclusions about Adherence and Competence. In all cases, they did. The scores were most similar for sessions with moderate levels of Adherence and Competence. If Adherence and Competence were low, partial observations were underestimates, but if they were high, partial observations were overestimates. Observing partial sessions is more efficient, but in terms of accuracy, the benefits and limitations should be evaluated in light of how the scores will be used. Additionally, future research should consider whether Audit and Feedback interventions have the same effect if feedback is based on observations of partial sessions.