What influences how physicians frame information for patients? An explorative quantitative study on hospital interactions from different specialties

Abstract

Background

Physicians frame medical information for patients in different ways, impacting patient outcomes. What underlies their framing choices has not been investigated. 

Objective

To explore the use and function of information framing practices in medical interactions.

Design

Explorative, quantitative observational study with a mixed-methods design.

Setting and participants

28 videorecorded hospital interactions, ranging from short-term/acute (orthopaedic surgery, gynaecology) to long-term/chronic care (oncology, gastroenterology) and involving 14 physicians and 28 patients.

Outcome measures

Using a previously developed coding system, we analysed physicians’ framing practices. We extracted information sharing functions qualitatively, checking 10% of the analysis with an independent assessment from the physicians. To explore whether variation in physicians’ use of information frames stemmed from individual or clinical specialty differences, we estimated intraclass correlations. To assess how their use of information frames varied at the macro level (across clinical specialties) and micro (the immediate function in the interaction), we estimated linear mixed models adjusted for the number of words.

Results

Variation in framing practices was mainly observed at the level of physicians nested within the same clinical unit (9.5% of the overall variance explained vs 1.3% for the unit level and 0.7% for the individual level). Physicians from different clinical units differed significantly in how they framed information, with the main differences between the gastroenterology and oncology units (mean difference=1.88; 95% CI 0.97; 2.79; p<0.001). The primary function of their information sharing also varied significantly across clinical specialties (all p<0.001 in χ² tests). Physicians from more short-term care units were oriented towards shared understanding, from gastroenterology towards inviting and convincing efforts and from oncology towards personal communication.

Conclusion

Results revealed signature marks of clinical units in terms of information sharing practices. Physicians’ information framing choices were driven both by the macro level (the clinical unit) and micro (the specific function for sharing information at that moment), thus highlighting potential areas for future interventions.

STRENGTHS AND LIMITATIONS OF THIS STUDY

• This is the first study quantifying physicians’ detailed information framing practices as observed in real-life consultations from four hospital units.

• The study employs an innovative mixed-methods design, combining qualitative and quantitative approaches to link micro-level interactional practices with organisational-level patterns across clinical specialties.

• One person performed the majority of the analysis, and no inter-rater reliability was established for the analysis of framing devices.

• By focusing only on physicians’ information sharing practices, the study did not capture how patients influence or shape information sharing.

Introduction

Sharing medical information with patients is an ethical duty and core task for physicians.¹,⁴ As curators of medical knowledge,^{5 6} physicians are expected to share a variety of medical information with patients, including the patient’s medical and psychosocial condition, lifestyle, self-care, therapy and treatment.⁷ Therefore, medical information sharing is crucial, and at times vital, for patients. In literature reviews, medical information sharing has been directly and indirectly (eg, through understanding and health literacy) linked to various patient health outcomes, including better informed medical decisions, commitment to treatment and beneficial behavioural changes.^{8 9} However, the way physicians share information has received too little attention in training and research. A recent, large scoping review concluded that only 39 studies on medical communication (out of >9000) actually tested physicians’ medical information sharing practices.¹⁰ In addition, the experimental studies testing medical information sharing practices were heterogeneous in terms of interventions and outcomes, and most of the tested practices were not based on previous evidence or theory.¹⁰

Outside the medical setting, information sharing is often described as comprising two main elements: content (what is shared, the subject matter) and form (how it is shared, the framing).¹¹ How information is shared has been the focus of many scholars of language use.¹² For example, Clark and Fox Tree revealed that speakers systematically use brief signals to forecast whether a delay while speaking will be shorter (‘uh’) or longer (‘um’).¹³ While these dysfluencies do not contribute specific content, they add information about the state of mind of the speaker and about the flow of the conversation itself.¹⁴ Speakers spontaneously use a large variety of such performance additions (ie, ‘framing devices’) to mark how the listener should take the information, to display the speaker’s relationship to the information and to manage the flow of the conversation.¹⁵

Within healthcare, the study of information framing typically has focused on public health risk messages.¹⁶ While such studies have demonstrated a surprising effect of specific framing practices on health outcomes,¹⁷ they have operationalised framing using only a limited range of behaviours (eg, negative vs positive framing of information).^{16 17} In conversations with patients, physicians may have a wider repertoire. A qualitative study of patient-physician interactions revealed 66 unique information framing devices that physicians used to frame medical information for patients.¹⁵ For example, to display their subjective stance, physicians framed information with their professional reasoning by adding ‘I think that… there are two treatment options’. Or they displayed professional competence by saying ‘we routinely perform… both of these options’ (for a full overview see ¹⁵). The field currently lacks knowledge about how physicians use these framing practices and what may influence which kinds they use. Understanding these influences is essential for identifying where to intervene to improve communication practices, ultimately for the benefit of patients.

With this study, we aimed to provide a better understanding of the use and function of information framing practices in medical interactions; therefore, we investigated physicians’ use of a wide range of information framing devices in a variety of physician-patient interactions from four hospital settings. We specifically explored whether physicians’ use of information framing practices varied due to clinical specialties, individual or task differences:

1. Does the way physicians frame information vary based on differences between individuals or clinical specialties?

2. Are different information sharing functions associated with different information framing practices?

Methods

This is an explorative, quantitative observational study with a mixed-methods design, combining qualitative and quantitative analysis of interactions and linking micro-level interactional practices with organisational-level patterns across clinical specialties.

Data collection and sample

Between 2019 and 2023, we videorecorded 28 physician-patient interactions at a university hospital in Norway from four clinical specialties: gynaecology, orthopaedic surgery, gastroenterology and oncology. The selection of the hospital specialties was intended to represent diverse clinical arenas ranging from more short-term/acute care (gynaecology, orthopaedic surgery) to more long-term/chronic care (oncology, gastroenterology). To increase the opportunity to observe nuanced and deliberate practices that might isolate the effects of specialty or task demands, we asked unit leaders to select physicians who were ‘good communicators’; 14 physicians participated. We selected consultations by checking each physician’s planned visits for the next month and included those in which sharing medical information was likely to be an important part of the agenda (eg, diagnostic and decision-making visits were included, but short, follow-up visits were not). We aimed for two consultations from each of the 14 physicians, balanced on patient gender, age, education and self-reported health status. This approach ensured variation across physicians while preserving the hierarchical structure of the data (patients nested within physicians nested within clinical units). Table 1 provides information about the patients’ and physicians’ characteristics.

Table 1. Patient and physician characteristics.

Patients (n=28)
Female, n (%)	12 (43)
Age, median (range)	55 (21–80)
Higher education (university), n (%)	6 (21)
First visit with the physician, n (%)	7 (25)
Health literacy<adequate cut-off 33*, n (%)	12 (43)
Poor or very poor health status, n (%)	4 (14)
Experienced symptoms of depression or anxiety in the last week, n (%)	9 (32)
Physicians (n=14)
Female, n (%)	7 (50)
Age, median (range)	49 (34–58)
Years of experience, median (range)	21 (4–27)
Previous communication skills training, n (%)	9 (64)

^*Health literacy assessed using the Norwegian validation of the short version of the European Health Literacy Survey.¹⁷

For descriptive purposes, patients were asked to fill in a questionnaire prior to the consultation (sociodemographic information, health and emotional status) as well as a health literacy questionnaire.¹⁸

Physicians participated in a video-reflexive interview 1 week after the recordings. Prior to the interviews, JM selected 1–3 min excerpts from each visit, during which the physician was providing medical information. During the interviews, JM showed the excerpts and asked the physician to reflect on their information sharing practices during those excerpts.

Ethics

Both physicians and patients provided written informed consent before collecting data and video recording the consultation. The study was approved by the hospital data protection committee (ID: 2020_174; 20/11994) and deemed exempt from review by the Regional Committee for Research Ethics on medicine and health (ID: 239229).

Patient and public involvement

A user panel including physicians, patients with different cultural backgrounds and policymakers was consulted during the design phase of the project. Their feedback informed the design of the informed consent forms and the way the project was presented to the public and participants.

Data extraction and analysis

Extraction of physicians’ information framing practices

Analysis focused on identifying and extracting the framing devices physicians used in turns during which they shared medical information. Analysis followed procedures and decisions reported in a previously published study, in which 66 framing devices were qualitatively identified using microanalysis of clinical interaction (MCI),¹⁹ finalised following saturation criteria and expert validation and categorised into nine frames.¹⁵ For the present study, JM watched video recordings and, using a transcript, (1) highlighted physicians’ information sharing turns (according to definitions and criteria previously developed and reported,¹⁵ here summarised in table 2), (2) copied them into a separate Excel worksheet and (3) extracted the framing devices within the turns. In particular, step 3 involved examining the information sharing turns phrase by phrase, first distilling the most basic or generic version of the information in each phrase, then checking whether anything the physician did beyond that basic version corresponded to any of the 66 framing devices and finally assigning a dichotomous score for each framing device (0=not present; 1=present). Once the analysis was completed, framing devices were organised into the nine higher-level frames.¹⁵ Table 2 provides a glossary of terms with examples.

Table 2. Glossary of terms with one worked-through example.

Label	Definition	Example
Physician information sharing turn	A turn during which the physician provides new medical information to the consultation and the patient contributes only with backchannel continuers (ie, not substantial verbalizations/contributions). These turns have four requirements: (1) the timing and topic of the episode are chosen by the physician (ie, not in response to a question from the patient), (2) the patient is in a listening role, (3) the status of the information is treated by the physician as new to the dialogue, and (4) the topic of the information is medical.	‘Let us look a little to your blood tests. Here they are. So what we look at is how much blood percentage, also blood percentage haemoglobin, you have. And this you have it perfectly completely good’
Basic, generic content	A distillation of the most basic content in the information sharing turn	In the above example: ‘The blood tests show that your haemoglobin is good’
Information framing device	Any dialogic mechanism or performance addition to present information beyond the most basic, generic content, which could shape how the patient may perceive and interpret it	In the above example: foreshadowing (‘let us look a little’); minimising (‘a little’); displaying to observe (‘here they are’); specifying (‘blood percentage, also blood percentage haemoglobin); accentuating (‘perfectly, completely’)
Information frame	A higher-level way of organising framing devices reflecting how information is structured to reach a particular purpose	In the above example: Foreshadowing is part of the frame ‘You may need to think’; minimising is part of the frame ‘This is not important now’; displaying to observe is part of the frame ‘This may be tricky to understand’; specifying is part of the frame ‘You may need to think’; accentuating is part of the frame ‘This is something important’
Information sharing function	The primary goal of an information sharing turn based on what the turn appears to be designed to achieve	In the above example, the turn has the function of ‘explaining’. Explaining turns are focused on informing, educating and increasing knowledge or understanding in the patient about the illness and care. These are characterised by consistent content delivered, explanations and details that deliver ‘objective, close to neutral, facts’ and that presuppose a lack of knowledge from the patient

Extraction of the function of the physicians’ information sharing turns

Informed by MCI,¹⁹ physicians’ information sharing turns were qualitatively and inductively analysed to identify their predominant function, that is, what the turn appeared to be designed to do. JM assigned preliminary function categories, collected similar and contrasting examples and worked recursively from examples to categories. JM also noted which observable behaviours were playing a part in decisions (ie, the rationale behind each choice). Using this process, JM developed definitions for each observed category of functions and constructed a decision tree to support systematic analytical decision-making (see figure 1).

Figure 1. Decision tree for analysing the function of the information sharing turns.

Turns in which the function clearly shifted were divided. When the main function was unclear, JM discussed the turn with a group of 8–10 researchers experienced in inductive analysis of clinical interactions. These discussions helped to refine definitions and differentiate among categories. Preliminary categories of functions with operational definitions, examples and the decision tree were discussed with the same group of interaction researchers and with an interdisciplinary advisory board of experts in medical information sharing. The advisory board consisted of eight researchers: three with linguistic, health science and psycholinguistic backgrounds, three who were also communication skills teachers and two who were also practising physicians.

To further check the validity of the analyst’s choices, we used the physicians’ video-reflexive interviews. During these interviews, they were asked to watch selected clips from their own video recordings and reflect on their information sharing practices, which often included discussing their intended function for sharing information at that moment. The clips used in the video-reflexive interviews included 58 physicians’ information sharing turns (10% of the 610 physicians’ information sharing turns identified in the data). The analyst’s decisions about the function of the turns were overall aligned with the physicians’ perspectives, with 79% in complete agreement and 19% in partial agreement with physicians indicating multiple functions, and in one case, the physician expressed a function that was not matching the analyst’s choice. Therefore, the categories were finalised based on saturation criteria, with consensus used to resolve ambiguities and validation of 10% of the coded turns against physicians’ own accounts.

Statistical analyses

We employed both descriptive and inferential statistics. In this data set, we have a hierarchical structure with patients being nested within the physicians and the physicians nested within the clinical unit. To explore whether physicians’ use of information frames varies across individual physicians or clinical units, we estimated intraclass correlations by empty linear mixed models. To assess how the use of information frames varies across clinical specialties, we estimated linear mixed models with random effects for physicians, one for each of the nine information frames. To assess how the use of information frames varies across information sharing functions, we estimated similar models but with random effects for physicians nested within clinical unit. The models included fixed effects for either clinical specialties or functions as categorical variables. To control for differences in turns’ length, the models were adjusted for the number of words, a continuous variable. The results are presented as mean differences between categories with the corresponding 95% CIs and p values. Standard diagnostic tests were performed to assess the model assumptions, and no major deviations were found. The differences between clinical units in the main function of their information sharing were assessed by χ² test. Although this is an exploratory study, to adjust for multiple testing, we set a significance level at or below 0.001.

Results

In total, 610 physicians’ information sharing turns were identified in the 28 consultations (mean length=23 min; SD=13; min, max=9–63). These turns included 5714 framing devices, which were mostly within the frames of either displaying a professional stance towards the information (ie, ‘This comes from me as a physician’) or orienting the patient to information that was complex and possibly difficult to understand (ie, ‘This may be tricky to understand’) (18% and 16% of the total devices, respectively).

Does the way physicians frame information vary based on differences between individuals or clinical specialties?

When exploring whether physicians’ use of information framing practices was related to individual or clinical specialties’ differences, we found that physicians within the same clinical unit tended to use information frames more similarly to each other than to physicians in other units (9.5% of the overall variance in the use of information frames explained by the level of physicians nested within the same clinical unit, compared with 1.3% for the unit level and 0.7% for the individual level). Especially for the frame ‘Do we agree that we share this knowledge?’, 11.5% of the variance could be explained by the level of the physicians nested within clinical units. For specific information frames, we also observed clustering within the same clinical units (6.2% of explained variance in frame ‘Do we agree that we share this knowledge?’; 3.4% in ‘This is not important now’; 2.3% in ‘This is something important’) and within the same physicians (4.1% in frame ‘This may be tricky to understand’), even though the number of observations per physician was limited.

When exploring how the clinical units specifically differed in terms of physicians’ use of information framing practices, the linear mixed model revealed significant differences especially between the gastroenterology and the oncology unit for the overall use of information framing practices (online supplemental table 1; table 3 for the descriptive statistics).

Table 3. Descriptive statistics (mean, SD) of number of framing devices within the information frames in the 610 physicians’ information sharing turns, by clinical unit.

Information frames	Total mean (SD)	Gynaecology (n=147)	Gastroenterology (n=181)	Orthopaedic surgery (n=175)	Oncology (n=107)
Frames that are about ‘shared understanding’
1. Do we agree that we share this knowledge?	0.78 (1.06)	1.24 (1.39)	0.62 (0.84)	0.69 (0.98)	0.52 (0.77)
2. I don't like where I am/you are going with this	0.56 (0.90)	0.63 (1.08)	0.37 (0.74)	0.72 (0.91)	0.54 (0.83)
Frames that orient to ‘how to take the information’
3. This may be tricky to understand	1.53 (1.99)	1.63 (2.01)	1.53 (2.33)	1.58 (1.92)	1.30 (1.29)
4. You may need to think	0.85 (1.12)	0.92 (1.31)	0.88 (1.08)	0.87 (1.15)	0.68 (0.78)
5. This is something important	0.82 (1.41)	1.24 (2.19)	0.66 (0.98)	0.69 (0.96)	0.69 (1.14)
6. This is not important now	0.84 (1.25)	1.25 (1.69)	0.82 (1.08)	0.63 (1.00)	0.64 (1.05)
Frames that display the ‘relationship to the information’
7. This comes from me as a physician	1.68 (1.97)	1.97 (2.33)	1.93 (2.21)	1.26 (1.51)	1.58 (1.54)
8. This comes from me as a person	1.15 (1.64)	1.26 (1.88)	1.14 (1.58)	1.07 (1.57)	1.16 (1.54)
9. This is directed to you as a unique person	1.16 (1.49)	1.19 (1.58)	1.13 (1.52)	1.05 (1.41)	1.35 (1.46)
Total	9.37 (8.26)	11.33 (10.85)	9.09 (8.11)	8.56 (6.86)	8.46 (5.75)

In detail, physicians used the ‘shared understanding’ types of framing devices (frames 1 and 2) more often in acute care units (gynaecology, orthopaedic surgery) than in chronic care units (oncology, gastroenterology), with the most striking difference between the gynaecology unit and the gastroenterology unit (mean difference for frame 1=−0.56; 95% CI −0.85; –0.27; mean difference for frame 2=−0.22; 95% CI −0.30; −0.15).

Clinical units also differed in physicians’ use of framing devices that displayed their relationship to the information (frames 7–9). This difference was especially observable between the oncology unit, where physicians used more personal frames (‘this comes from me as a person’ and ‘this is directed to you as a unique person’) than the gynaecology unit (mean difference for frame 8=0.42; 95% CI 0.23; 0.61; mean difference for frame 9=0.65; 95% CI 0.28; 1.02).

Finally, clinical units were more similar in the use of framing devices that provided orientation to the patient on how to take the information (frames 3–6). The main differences were observed in the gastroenterology unit, where physicians used more ‘this is something important’ devices compared with the acute care units (mean difference with gynaecology=0.47; 95% CI 0.67; 0.27; mean difference with orthopaedic surgery=0.26; 95% CI 0.19; 0.33).

Are different information sharing functions associated with different information frames?

We identified five main categories of functions of the information sharing turns: ‘explaining’, ‘inviting’, ‘convincing’, ‘reassuring’ and ‘meta-communicating’ (a document with definitions and examples for each function is available on request from the corresponding author). We assessed 327 information sharing turns as having the function of ‘explaining’, 171 as ‘inviting’, 96 as ‘meta-communicating’, 41 as ‘reassuring’ and 15 as ‘convincing’. The function of the information sharing turns differed significantly between clinical units (all p<0.001 in χ² tests): the orthopaedic surgery unit was featured by information sharing turns with the function of ‘explaining’, the gastroenterology unit was featured by ‘inviting’ and ‘convincing’, the gynaecology unit by ‘meta-communicating’ and the oncology unit by information sharing turns with the function of ‘reassuring’ (see table 4).

Table 4. Cross-table between clinical unit and function of information sharing turns, numbers are frequencies (%), n=610.

Clinical unit	Convincing	Explaining	Inviting	Meta-communicating	Reassuring	Total
Gynaecology Gastroenterology Orthopaedic surg. Oncology	2 (1.4) 8 (4.4) 2 (1.1) 3 (2.8)	60 (40.8) 80 (44.2) 130 (74.3) 57 (53.3)	36 (24.5) 55 (30.4) 24 (13.7) 16 (15.0)	45 (30.6) 29 (16.0) 6 (3.4) 16 (15.0)	4 (2.7) 9 (5.0) 13 (7.4) 15 (14.0)	147 181 175 107
Total	15	327	131	96	41	610

In bold font are displayed the highest frequencies, all p<0.001 in χ2 tests

We then asked how the framing practices mapped onto the different functions of the information sharing turns (see table 5 for the descriptive statistics). The results of the linear mixed model (online supplemental table 2) showed some differences in framing practices based on the different functions of the information sharing turns. In particular, turns with the function of ‘explaining’ and ‘meta-communicating’ were more featured by framing efforts in the direction of reaching ‘shared understanding’ (frames 1 and 2), especially compared with turns with the function of ‘reassuring’.

Table 5. Descriptive statistics (mean, SD) of the number of framing devices within the information frames in the 610 physicians’ information sharing turns, by information sharing goals.

Information frames	Convincing (n=15)	Explaining (n=327)	Inviting (n=131)	Meta-communicating (n=96)	Reassuring (n=41)
Frames that are about ‘shared understanding’
1. Do we agree that we share this knowledge?	0.60 (0.83)	0.75 (1.06)	0.85 (1.01)	0.95 (1.26)	0.37 (0.62)
2. I don't like where I am/you are going with this	0.87 (1.13)	0.67 (0.94)	0.53 (0.93)	0.26 (0.60)	0.37 (0.70)
Frames that orient to ‘how to take the information’
3. This may be tricky to understand	2.47 (2.95)	1.86 (2.20)	1.37 (1.79)	0.67 (0.97)	1.02 (1.19)
4. You may need to think	1.53 (1.77)	0.94 (1.21)	0.79 (0.97)	0.57 (0.83)	0.71 (0.87)
5. This is something important	1.27 (1.22)	0.91 (1.54)	0.67 (1.34)	0.43 (0.82)	1.27 (1.48)
6. This is not important now	1.13 (2.07)	0.82 (1.23)	0.77 (1.21)	1.08 (1.35)	0.49 (0.81)
Frames that display the ‘relationship to the information’
7. This comes from me as a physician	2.73 (2.19)	1.80 (2.19)	2.13 (1.81)	0.83 (0.99)	0.90 (1.39)
8. This comes from me as a person	2.13 (2.61)	1.26 (1.80)	0.95 (1.29)	0.89 (1.20)	1.22 (1.62)
9. This is directed to you as a unique person	1.27 (2.37)	1.21 (1.56)	1.20 (1.45)	0.66 (1.08)	1.80 (1.27)
Total	14.00 (12.93)	10.23 (9.09)	9.28 (7.39)	6.33 (5.24)	8.15 (4.97)

Other observed differences pertained to ‘meta-communicating’ turns, where physicians used on average fewer framing practices that displayed their professional stance (ie, the frame ‘This comes from me as a physician’). Also, ‘reassuring’ turns had on average more framing efforts directing to the patient as a unique person.

Discussion

With this study, we shed light on factors that may systematically influence physicians’ real-life use of framing practices when sharing medical information with patients.

First, the findings suggested that physicians have framing styles that are consistent but also likely influenced by the unique clinical units’ features, demands and/or internal cultures. This finding highlights the potential influence of local, group-level communication practices (ie, shared by small teams or departments within units) on shaping communication practices with patients. Patient-physician communication does not exist in a vacuum but rather is contextualised in an environment that may help or hinder it.²⁰ According to social cognitive theories, behaviours indeed result from the dynamic interplay of personal, behavioural and environmental factors.²¹,²³ The Theory of Planned Behaviour,²² the Theory of Reasoned Action²³ and the Social Cognitive Theory of Bandura²¹ have been widely applied in healthcare settings and shown to predict clinicians’ behaviours, including those of physicians.²⁴,²⁷ According to these theories, social influences contribute to physicians’ behaviours, such as through peer modeling, social norms or perceived resources and barriers. However, compared with individual-level factors, such as physicians’ beliefs about capabilities, the impact of these social influences has mixed evidence.²⁴,²⁷ In general, such theories aim to explain causal relations between environmental inputs and behavioural outputs, focusing on internal mental constructs. In contrast, our findings emphasise the role of social environmental factors over individual cognitive factors. Accordingly, efforts to improve physicians’ information framing practices in talks with patients should focus on localised workplace cultures: Communication is situated in the clinic, and shared communication practices are continuously shaped and normalised by employees.²⁸,³⁰ Indeed, interventions aimed at benefiting patient outcomes (eg, for patient safety³¹) have specifically highlighted the level of the wards and the group of clinicians within the wards. This insight could help identify a potentially powerful target for future interventions—the group level, where shared practices may be shaped and reinforced more directly than at the individual or whole-unit level. Future studies may test interventions at the group level by using constructs from social cognitive or sociocultural theories (eg, observational learning) to train mastery of information framing.

Second, we also described each specialty’s distinctive information sharing functions and framing practices. Physicians from short-term care units were oriented towards shared understanding efforts: those in orthopaedic surgery towards explaining and those in gynaecology towards meta-communicating. In contrast, physicians in gastroenterology were oriented towards inviting and convincing efforts, specifically using ‘highlighting’ frames, and physicians in oncology oriented to personal communication, using frames that revealed the physicians' personal stance and tailored to the individual patient. Few previous studies have detailed communication differences between hospital specialties.³²,³⁴ In general, studies suggest that surgical specialties are less patient-centred and receive lower communication ratings than specialties like oncology.^{32 33} This difference may be due to the professional competencies that the different specialties prioritise, with short-term units giving more importance to clinical, technical competencies and long-term care units to relational competencies.³⁴ Though this study took a different approach by contrasting hospital units in terms of actual communication practices (rather than assessed or perceived), findings reflect a similar orientation of the chronic/longitudinal care specialties versus more acute specialties. To disentangle the role played by workplace cultures compared with domains of practice, further studies may compare similar units. This finding also has implications for whether studies located in one clinical setting can be generalised to other settings, reminding about the risk of overgeneralisation when studying medical communication. Similarly, it suggests that interventions may need to be tailored to a clinic’s characteristics and culture.

Third, our results highlighted that the overall task of ‘information sharing’ can incorporate different functions that are not purely explaining and can involve relationship-building and motivational efforts. The observed functions for the information sharing turns reflect some of the communication functions identified in previous studies,^{35 36} for example, ‘reassuring’ reflecting fostering the relationship and responding to emotions, ‘convincing’ and ‘inviting’ reflecting decision-making and enabling self-management. These findings have implications for considering ‘information provision’ as a separate, distinct action, as it may entail a wider variety of communicative functions.

In addition, we found that physicians use different framing practices depending on the function their information sharing is meant to serve—such as reassuring, convincing and inviting, which can change from unit to unit. For instance, when an information sharing turn is designed to reassure the patient, it often includes framing devices that attend to the uniqueness of the other person, while explaining medical information can involve more efforts to repair and build shared understanding. These findings may have implications for teaching clinical communication in the direction of aligning framing practices with information sharing goals. Findings also support models of clinical communication as action-response cycles driven by clinical goals and subgoals,³⁷ where the success depends on how coherent and aligned actions and goals are.^{9 10} Additionally, our study provides basic knowledge about the details of these practices, highlighting how implicit communication—particularly the framing of messages shaped by communicative goals—operates across various settings.

A major limitation of this study is that one person performed most of the analysis. This choice reflected feasibility. We mitigated its effect with the following actions: for framing devices, the analyst followed formalised definitions; physicians’ declared goals were later used to check a subset of the analysis; the analyst engaged in regular interactions and discussions with coauthors and other researchers about analytical choices.

Another limitation relates to the focus on physicians’ information sharing practices, without reference to patients’ responses or actions. This focus minimises insight into how information sharing is tailored to individual patient needs and how patients may understand different information framing practices. However, knowing what influences how physicians share medical information may be a prerequisite for understanding where to intervene to improve communication practice for the benefit of the patient.

Also, the study included physicians that their unit leaders deemed ‘good communicators’, thus potentially reducing generalisability. In addition, although the number of information sharing turns analysed and framing practices identified from physicians was substantial, the study included relatively few physicians, particularly from each clinical unit, raising concerns about the representativeness and generalisability of the findings.

Finally, video recording the consultations may have influenced physicians in the direction of more desirable behaviours.³⁸ However, the camera used for this study was deliberately small to avoid intrusiveness, participants reported not being affected by the camera and previous research showed a limited influence of the Hawthorne effect on most doctor-patient visits.³⁹

With its exploratory purpose and the recognised limitations, the study represents a first step in identifying factors that drive physicians’ information framing practices. Future studies are needed to further check the reliability of the analysis, extend to other groups of participants and include the patients’ actions.

In conclusion, an important feature of medical communication is physicians’ way of sharing information with patients. In one of the first explorative investigations of its kind, this study quantified 18 physicians’ framing practices across different hospital settings and found signature marks of clinical units. This study can inform experimental studies designed to test the effect of different framing practices on patient outcomes. For example, future studies may test different frames when ‘describing treatment options’ on shared decision-making outcomes and on patients’ understanding. Furthermore, it provides a point of departure for exploring the role patient contributions serve in driving physicians’ framing choices. Also, since content and form are inextricably linked, further studies may explore if and how specific topics or contents may modulate how physicians frame information for the patients they are talking with.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.