More than a feeling? What does compassion in healthcare ‘look like’ to patients?

Abstract Objective Compassion is important to patients and their families, predicts positive patient and practitioner outcomes, and is a professional requirement of physicians around the globe. Yet, despite the value placed on compassion, the empirical study of compassion remains in its infancy and little is known regarding what compassion ‘looks like’ to patients. The current study addresses limitations in prior work by asking patients what physicians do that helps them feel cared for. Methods Topic modelling analysis was employed to identify empirical commonalities in the text responses of 767 patients describing physician behaviours that led to their feeling cared for. Results Descriptively, seven meaningful groupings of physician actions experienced as compassion emerged: listening and paying attention (71% of responses), following‐up and running tests (11%), continuity and holistic care (8%), respecting preferences (4%), genuine understanding (2%), body language and empathy (2%) and counselling and advocacy (1%). Conclusion These findings supplement prior work by identifying concrete actions that are experienced as caring by patients. These early data may provide clinicians with useful information to enhance their ability to customize care, strengthen patient–physician relationships and, ultimately, practice medicine in a way that is experienced as compassionate by patients. Public Contribution This study involves the analysis of data provided by a diverse sample of patients from the general community population of New Zealand.


| INTRODUCTION
Compassion has been defined as an emotion, 1 a motivation 2 and a virtuous response. 3 At the least, it involves both feeling and action components 2,4,5 the awareness of suffering and acting to alleviate it. 6 In medicine, compassion is desired by patients, mandated by medical regulatory bodies and increasingly linked to positive outcomes for patients and families, professionals and healthcare systems. 7 Patients and families rate compassion among the most important healthcare requirements, [8][9][10][11][12] recalling it years later. 13,14 Compassionate care predicts faster recovery, 15 greater autonomy, 16 lower intensive care utilization 17 and more responsible healthcare management. 18 Similarly, compassion-related constructs have been associated with objective benefits, including better disease control 19 and reduced metabolic complications 20 among patients with diabetes. Compassion is thus central to both the practice of effective medicine and essential in the preferences of those receiving professional care.
However, despite the value patients place on compassion and the benefits it may have, data circumscribing patients' experiences of compassion are lacking. 21 Complicating the study of compassion in healthcare is the fact that it has often been confused with other terms, such as empathy, sympathy and concern. 22 However, while sympathy shares some surface similarities with compassion, sympathy can arise in response to a range of feelings while compassion more specifically arises in response to the suffering of another and necessarily includes a motivation to relieve suffering. 3 Similarly, while empathy is periodically conflated with compassion, empathy does not require action and it may be difficult to sustain over long periods of time. 23 More to the point, while compassion is experienced as distinct/ preferable to empathy or sympathy, 24 exactly what compassionate care entails or 'looks like' to patients is unclear. To date, studies suggest effective communication, 25 interpersonal connections, 26,27 understanding, 28 being present, empathizing, 26 taking action and providing individualized care 29 are important to the experience of compassion. Other studies highlight the importance of touch in the experience of compassion, 26 safety, authenticity and connection. 30 One study explored how doctors communicate compassion by developing a taxonomy of compassionate physician behaviours in the realm of oncology. 31 Analyses suggested that the recognition of the patient's suffering, emotional resonance, and movement towards addressing suffering were all important elements in compassion interactions. Of note, these behaviours were experienced as compassion across a conversation rather than in terms of a single event, and silence was associated with emotional resonance.
Importantly, patient data suggest that feeling cared for often takes only a moment, while nonpatient views often imply that compassionate care is time-consuming. 27,32 It is also possible that compassion may be experienced differently between the healthcare provider expressing compassion, the patient receiving it, or others observing the interaction. Nonetheless, evidence to date suggests that patients experience care when practitioners are emotionally present, communicate effectively, enter into their experience and display understanding and kindness. While these factors are clearly important to patients, what physicians might actually do to create the experience of care remains unknown.
More broadly, there are at least three significant limitations to prior studies of the patient experience of care. First, most patient studies have been conducted in nursing contexts 18,26,33 or palliative care samples. 21,24,34 While such contributions are important, findings may be less applicable to general patient samples. For example, caring behaviours are often thought to be 'part and parcel' of nursing, 35 40 Specifically, the risk with such designs is that in creating and refining coding systems, researchers may (involuntarily) impose their own beliefs, knowledge and interests, which may (or may not) reflect patient meaning regarding compassion. While text analysis also has limitations 41 such as not being able to interpret latent context (e.g., humour, irony or polysemes), this approach can reveal unbiased themes as well as themes that researchers might not notice or code for to deepen our understanding of compassionate care.
In contributing to this nascent area of study, the current report presents data from a large sample of community-based patients, identifying the physicians' actions that are seen as characterizing caring behaviour for patients. 42 In shifting the focus from patients' experiences of care to identifying physicians' actions that communicate compassion and using an analytic framework that avoids some forms of researcher bias, the current study addresses several limitations in the prior research of compassionate care and outcomes related to patient-physicians relationships interventions. 43 Findings can thus supplement existing work in helping to identify the physician's actions that matter to patients and thus offer clinicians an initial glimpse at a future compassion tool kit with the potential to enhance their ability to customize care, strengthen the patient-physician relationship and, ultimately, practice in ways that are experienced as caring.

| Procedure
The study was advertised via social media postings, email lists and word-of-mouth. Given the potential sensitivity of patient data, community participants were anonymous and data submission was taken as consent. A link directed prospective participants to an information sheet and consent form. Following consent, demographic, healthcare utilization and health information were gathered before specific questions about the relationship and experience of compassion with their physician were delivered.

| Measurement
Compassionate care. In line with the primary research question, patients were asked to describe their experience of compassionate care with their physician. Patients were provided with a brief definition of compassion (Compassion in medicine is the ability to recognize and understand a patient's suffering, coupled with the desire to relieve it) before being asked a single yes/no item to the question: 'do you feel your physician cares for you and wants to help?' if patients selected yes, they were asked to describe specifically what their physician does that made them feel cared for.

| Analyses
First, the text was cleaned in the Python (3.10) programming language 44 by removing stop words (e.g., and, or, that). Bi-grams and tri-grams were created to account for phrases. Second, data were analysed using Latent Dirichlet Allocation (LDA) topic modelling (TM). 45 This inductive quantitative technique searches for latent structures by clustering words with a higher probability of co-occurring in texts than expected to happen by chance. 46 Since LDA modelling arranges these latent structures (or vocabularies) proportionally, 47 we can identify which physician caring actions are referred to more frequently and are hence of greater importance to patients. As an alternative to more traditional qualitative thematic analysis, TM is more robust in application to larger data sets and helps avoid the (involuntary) imposition of researcher bias. The TM analysis was conducted via one of the most widely used tools-Machine Learning for Language Toolkit (MALLET) for Mac, 48 which is considered best in class due to precision in sampling methods. 49 The number of topics was determined based on the coherence score. Coherence scores can be defined as the ease with which topics can be interpreted by taking a median of pairwise word-similarity scores within a given topic for a group of topics. 50  | 1693 chronic health concerns that the participants self-identified (e.g., asthma, arthritis, chronic pain, diabetes, eczema etc.) (39%). Nearly nineout of ten participants made ratings regarding a primary care physician, with others rating either a specialist (2%) or other types of nonspecialized practitioners (e.g., nurse, psychologist, counsellor etc.) (2%) (see Table 1).
Testing the group of patients who reported feeling cared for and provided a text description from the other groups showed some differences. They did not differ from other groups in terms of ethnicity or gender (p > .05). However, people who did not answer the care question were younger than the group of primary interest, t (842) = 2.075, p < .05, although the effect size was small, d = 0.23.
They were also less likely to be diagnosed with any serious chronic illness (χ 2 [1] = 159.35, p < .01, odds ratio [OR]: 24.68) and consistently did not report their doctor's specialization. In comparison to the people who did not find their doctor caring, patients who reported feeling cared for and described how their doctor cared were also more likely to be diagnosed with cancer (χ 2 [1] = 8.61, p < .05, OR: 10.84); they were more likely to make ratings regarding a primary care physician (χ 2 [1] = 3.95, p < .05, OR: 2.17) and less likely to rate another type of nonspecialized practitioner (nurse, psychologist, counsellor) (χ 2 [1] = 4.45, p < .05, OR: 0.33). People who reported feeling cared for but did not provide a written reflection were less likely to be diagnosed with any serious chronic illness, although with a very small OR (χ 2 [1] = 6.42, p < .05, OR: 0.46).
The primary TM analysis revealed eight topics within the texts describing physician behaviours leading to patients feeling cared for. Seven topics were coherent and could be labelled, the eighth could not. In order of their commonality, topics were: listening and paying attention to the patient (71% of texts), following-up and running tests (11%), continuity and holistic care (8%), respecting preferences (4%), genuine understanding (2%), body language and empathy (2%) and counselling and advocacy (2%) (see Table 2).

T A B L E 1 Analyses
Variables Did not answer the care question (N = 108)  Further work is necessary to explore these factors in more depth.
In extending compassion research more broadly, this study provides three core contributions. First, patient perspectives of compassion in a large, diverse and general population were examined using an analytic approach that supplements prior methodologies. 45 As noted, most studies of patient perspectives have been conducted in modestly sized samples and restricted to nursing and palliative care, 21  In addition to doctors' following up, patient commentary suggested that continuity of care, openness to holistic practices, respecting preferences, and expressing genuine concern mattered. Each of these actions has been alluded to in prior studies, 66

ACKNOWLEDGEMENT
This study received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

CONFLICTS OF INTEREST
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflicts of interest.

DATA AVAILABILITY STATEMENT
The datasets presented in this article are not readily available because participants consented to participate with the understanding that access to data would be restricted to the named researchers.
Requests to access the datasets should be directed to Nathan S.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by the Human Participants Ethics Committee, University of Auckland. The patients/participants provided their written informed consent to participate in this study.