Background: Compassion, considered a cornerstone of quality healthcare by patients, families, clinicians, and policymakers, is believed to have wide-ranging benefits, including improving clinical outcomes, increasing patient satisfaction with services, and enhancing the quality of information gathered from patients. Compassion-filled interactions may also promote clinician wellbeing, decrease burnout, and improve mental health. Compassion is thought to consist of five domains: 1) recognizing suffering, 2) understanding the universality of suffering in human experience, 3) feeling empathy towards the person suffering, 4) tolerating uncomfortable feelings aroused in response to the suffering person, and 5) being motivated to act or acting to alleviate suffering. Most currently available compassion measurement tools are based on self-perception and thus, may not adequately reflect how patients perceive compassion in their clinicians. Few tools measure clinician compassion using patient perception, and all were designed for specific patient populations limiting their use in the general hospitalized population. Therefore, we designed a tool to assess patients’ perception of clinicians’ compassion during hospitalization.

Methods: We developed a set of questions to adequately represent 3 of the 5 compassion domains; the other two domains (understanding the universality of suffering and tolerating uncomfortable feelings) are not perceived by patients. These questions, answered on a 5-point Likert scale, were then reviewed by a small group of patients with a healthcare background and the final tool consisted of 20 questions. Hospitalized patients older than 18 years and who provided consent were asked to complete this tool and provide some demographic information. We used exploratory factor analysis (EFA) to determine factor loading and confirmatory factor analysis (CFA) using structural modeling for validation of our construct. The goodness of fit statistics were examined for CFA model fit. For each patient mean perceived compassion score and score for each of the three domains was calculated. We used Student’s t-test with unequal variances to determine if demographic characteristics affect perceived compassion. All analyses were performed using Stata 16.1.

Results: Of the 106 participants, 55 were females, 34 were African Americans, 30 were married, 44 had Medicare, 51 were older than 60 years, 32 had hypertension, and 31 had diabetes. The EFA identified three questions with limited communality and five additional questions with loadings on multiple factors – these eight questions were excluded from further analyses. The Cronbach alpha for the tool with the remaining 12 questions was 0.94 suggesting excellent reliability. The CFA found adequate convergent and discriminant validity and showed a fit of the model to the data (Figure). Using this 12-question tool, we calculated the mean scores of each of the three domains and an overall compassion score. The mean (SD) overall compassion score was 4.13 (0.63), the recognition of suffering score was 1.02 (0.84), the empathy score was 4.16 (0.72), and the acting to alleviate suffering score was 4.19 (0.58). There were no differences in the overall compassion score or the three domains by population characteristics (Table).

Conclusions: We present evidence for the reliability and validity of a new instrument to assess hospitalized patients’ perception of clinician compassion. This instrument is short, easy, can be completed in a short time.

IMAGE 1: FIGURE: Results of the Confirmatory Factor Analysis and Fit Indices of the tool to measure the perception of hospitalized patients of their clinician’s compassion. The figure shows standardized results where variances of the latent variables are fixed at 1. Standardized coefficients are along the arrows pointing towards the respective questions from latent variables; all coefficients were statistically significant (all P<0.001). Also displayed are estimates of the variances of the error terms for each question.

IMAGE 2: TABLE: Overall Compassion and three domain distributions (mean with standard deviation) by study population characteristics. No difference between the two categories of a characteristic was significant when examined by t-test.