Clinician sentiments related to implementation of evidence-based treatment for hoarding in older adults

Introduction The effects of hoarding disorder can often be debilitating, with accumulated possessions interfering with everyday use of home spaces and the ability to engage in activities of daily living safely. For older adults, these challenges are magnified, with hoarding in older age also being a predictor for increased levels of disability (Ayers et al., 2014). Given the insubstantial effects of solely cognitive-behavioral interventions for hoarding disorder (Ayers et al., 2011), cognitive rehabilitation consisting of compensatory cognitive training combined with exposure and sorting therapy (CREST trial; Ayers et al., 2018) was investigated as a potential avenue for effectively reducing hoarding symptoms in older adults. When compared to case management alone, the CREST intervention has been proven to be an efficacious treatment. In addition to measures showing objective symptom reduction, participants’ subjective evaluation of the intervention was that it was “helpful” (Davidson et al., 2021). Patient experiences of helpfulness are crucial for treatment adherence and retention. However, clinician experiences are also critical, as they can influence the rate at which new evidence-based interventions are adopted into clinical practice. Thus, this study aimed to evaluate the trajectory of clinicians’ sentiments surrounding patient engagement, comparing the CREST intervention with case management for the treatment of older adults with hoarding disorder (Ayers et al., 2018). Methods The Valence Aware Dictionary and sEntiment Reasoner (VADER) package for R (Hutto & Gilbert, 2014) was used to conduct a sentiment analysis of clinician self-reflections embedded in the case notes for 15 older adults enrolled in the CREST trial (9 receiving CREST; 6 receiving case management) over 60 sessions. Of note, not all cases had reported self-reflections for each session. The VADER package calculates compound valence scores using sums of positive/negative valence scores of words within strings of text, with a range of -1 to 1 for each score (most negative to most positive). Thus, higher scores indicate a clinician used more positive language to describe a particular visit (e.g., good, hopeful, inspiring), lower scores indicate a clinician used more negative language (e.g., bad, disappointing, unable), and scores near zero indicate a clinician used neutral language (e.g., refrigerator). Results Over 60 sessions, clinician ratings of their experiences treating each participant ranged from -.91 to .90 (Median = .33) for both conditions combined, with most reflections centered on how clinicians perceived their success with participants or participants’ likelihood for improvement. Overall, clinician sentiments shifted insignificantly over time (B = -0.009, p > .37). Clinician ratings for the CREST group were similar (Range -.91 to .90, Median = .37), with no significant shift over time (B = -.01, p = .38), whereas clinician sentiments for the case management group had a narrower range of scores (-.67 to .80), with similarly insignificant changes in sentiment over time (B = -.008, p = .68). Conclusions Clinician sentiments recorded after each session tended to be on the lower end of “positive,” but had a wide range. Notably, these sentiments did not change in a significant way across treatment. The specific domain of these notes analyzed for this presentation involved primarily self-reflective statements of clinician views of success on their own part and progress for participants. Given that the CREST trial was successful in reducing hoarding symptoms, clinicians’ reported perceptions for otherwise difficult tasks may have been tempered by visibly positive changes in participants. Alternatively, clinicians’ self-efficacy could account for the stability of sentiment over time. Further investigation into other domains of these sessions, such as participant goals, treatment barriers, and attempted components of the CREST intervention may clarify other factors about this trial that could have influenced clinician self-reflections. Additionally, further investigation is warranted into the clinical use of sentiment analysis tools to analyze participant and clinician comments as an indicator of clinical success. This research was funded by VA Career Development Award (CSRD-068-10S). The effects of hoarding disorder can often be debilitating, with accumulated possessions interfering with everyday use of home spaces and the ability to engage in activities of daily living safely. For older adults, these challenges are magnified, with hoarding in older age also being a predictor for increased levels of disability (Ayers et al., 2014). Given the insubstantial effects of solely cognitive-behavioral interventions for hoarding disorder (Ayers et al., 2011), cognitive rehabilitation consisting of compensatory cognitive training combined with exposure and sorting therapy (CREST trial; Ayers et al., 2018) was investigated as a potential avenue for effectively reducing hoarding symptoms in older adults. When compared to case management alone, the CREST intervention has been proven to be an efficacious treatment. In addition to measures showing objective symptom reduction, participants’ subjective evaluation of the intervention was that it was “helpful” (Davidson et al., 2021). Patient experiences of helpfulness are crucial for treatment adherence and retention. However, clinician experiences are also critical, as they can influence the rate at which new evidence-based interventions are adopted into clinical practice. Thus, this study aimed to evaluate the trajectory of clinicians’ sentiments surrounding patient engagement, comparing the CREST intervention with case management for the treatment of older adults with hoarding disorder (Ayers et al., 2018). The Valence Aware Dictionary and sEntiment Reasoner (VADER) package for R (Hutto & Gilbert, 2014) was used to conduct a sentiment analysis of clinician self-reflections embedded in the case notes for 15 older adults enrolled in the CREST trial (9 receiving CREST; 6 receiving case management) over 60 sessions. Of note, not all cases had reported self-reflections for each session. The VADER package calculates compound valence scores using sums of positive/negative valence scores of words within strings of text, with a range of -1 to 1 for each score (most negative to most positive). Thus, higher scores indicate a clinician used more positive language to describe a particular visit (e.g., good, hopeful, inspiring), lower scores indicate a clinician used more negative language (e.g., bad, disappointing, unable), and scores near zero indicate a clinician used neutral language (e.g., refrigerator). Over 60 sessions, clinician ratings of their experiences treating each participant ranged from -.91 to .90 (Median = .33) for both conditions combined, with most reflections centered on how clinicians perceived their success with participants or participants’ likelihood for improvement. Overall, clinician sentiments shifted insignificantly over time (B = -0.009, p > .37). Clinician ratings for the CREST group were similar (Range -.91 to .90, Median = .37), with no significant shift over time (B = -.01, p = .38), whereas clinician sentiments for the case management group had a narrower range of scores (-.67 to .80), with similarly insignificant changes in sentiment over time (B = -.008, p = .68). Clinician sentiments recorded after each session tended to be on the lower end of “positive,” but had a wide range. Notably, these sentiments did not change in a significant way across treatment. The specific domain of these notes analyzed for this presentation involved primarily self-reflective statements of clinician views of success on their own part and progress for participants. Given that the CREST trial was successful in reducing hoarding symptoms, clinicians’ reported perceptions for otherwise difficult tasks may have been tempered by visibly positive changes in participants. Alternatively, clinicians’ self-efficacy could account for the stability of sentiment over time. Further investigation into other domains of these sessions, such as participant goals, treatment barriers, and attempted components of the CREST intervention may clarify other factors about this trial that could have influenced clinician self-reflections. Additionally, further investigation is warranted into the clinical use of sentiment analysis tools to analyze participant and clinician comments as an indicator of clinical success.