Feedback mechanisms of change: How problem alerts reported by youth clients and their caregivers impact clinician-reported session content

Abstract

Objective: This study explored how clinician-reported content addressed in treatment sessions was predicted by clinician feedback group and multi-informant cumulative problem alerts that appeared in computerized feedback reports for 299 clients aged 11-18 years receiving home-based community mental health treatment.

Method: Measures included a clinician report of content addressed in sessions and additional measures of treatment progress and process (e.g., therapeutic alliance) completed by clinicians, clients, and their caregivers. Item responses in the top 25th percentile in severity from these measures appeared as "problem alerts" on corresponding computerized feedback reports. Clinicians randomized to the feedback group received feedback weekly while the control group did not. Analyses were conducted using the Cox proportional hazards regression for recurrent events.

Results: For all content domains, the results of the survival analyses indicated a robust effect of the feedback group on addressing specific content in sessions, with feedback associated with shorter duration to first occurrence and increased likelihood of addressing or focusing on a topic compared to the non-feedback group.

Conclusion: There appears to be an important relationship between feedback and cumulative problem alerts reported by multiple informants as they influence session content.

Keywords: Session Report Form; child psychotherapy; measurement feedback system; mental health services research; process research; technology in psychotherapy research and training.

Figure 1. Kaplan-Meier Survival Plot for the SRF Topic Addressing Behavioral Issues by Randomized Group (feedback = 1, non-feedback = 0)

Note: The feedback group was coded as 1, with the non-feedback group coded as 0.

Figure 2

Forest plot summarizing results of Cox regression for recurrent events where addressing content in sessions was predicted by feedback group, clinician- or client/caregiver-reported dynamic problem-specific severity, and their interaction.* * This table only includes the simplest model chosen for interpretation for each topic by reporter (Cl = Clinician, CY = Client and/or caregiver). The results for all models including statistics are presented in the appendix (Table 1 for clinician-reported measures and Table 2 for client and/or caregiver-reported measures). All reported models were significant at an adjusted p-value of p < 0.01. S = Simple model (likelihood of addressing/ focusing on content was higher for clients whose clinicians were in the feedback condition compared to the non-feedback condition); A = Adjusted model (relative likelihood that feedback increased addressing/focusing on content after adjusting for cumulative problem alerts); I = Interaction model (the relative likelihood for the feedback condition of addressing/focusing on content where cumulative problem alerts were present). Where the interaction model was significant, the hazard ratio of feedback was calculated by summing the hazard ratios for feedback plus the interaction term where the cumulative problem alert = 1.