Improving best practice for patients receiving hospital discharge letters: a realist review

Objective To understand how different outcomes are achieved from adult patients receiving hospital discharge letters from inpatient and outpatient settings. Design Realist review conducted in six main steps: (1) development of initial theory, (2) searching, (3) screening and selection, (4) data extraction and analysis, (5) data synthesis and (6) programme theory (PT) refinement. Eligibility criteria Documents reporting evidence that met criteria for relevance to the PT. Documents relating solely to mental health or children aged <18 years were excluded. Analysis Data were extracted and analysed using a realist logic of analysis. Texts were coded for concepts relating to context, mechanism, outcome configurations (CMOCs) for the intervention of patients receiving discharge letters. All outcomes were considered. Based on evidence and our judgement, CMOCs were labelled ‘positive’ or ‘negative’ in order to clearly distinguish between contexts where the intervention does and does not work. Results 3113 documents were screened and 103 were included. Stakeholders contributed to refining the PT in step 6. The final PT included 48 CMOCs for how outcomes are affected by patients receiving discharge letters. ‘Patient choice’ emerged as a key influencer to the success (or not) of the intervention. Important contexts were identified for both ‘positive’ CMOCs (eg, no new information in letter) and ‘negative’ CMOCs (eg, letter sent without verifying patient contact details). Two key findings were that patient understanding is possibly greater than clinicians perceive, and that patients tend to express strong preference for receiving letters. Clinician concerns emerged as a barrier to wider sharing of discharge letters with patients, which may need to be addressed through organisational policies and direction. Conclusions This review forms a starting point for explaining outcomes associated with whether or not patients receive discharge letters. It suggests several ways in which current processes might be modified to support improved practice and patient experience.

letters.  The engagement of patients, GPs and policy makers in refining the programme theory increased relevance and rigour of the theory.  The programme theory is likely to be applicable and relevant to multiple healthcare settings.  The exclusion criteria imposed restrictions on the programme theory such that evidence relating to children, those with mental health problems or lacking capacity is not considered.  Only sources written in the English Language were included. Discharge communication may follow an inpatient or outpatient discharge; it typically comprises written discharge information in the form of a discharge letter or summary. It is a well-established practice that the physician who is to follow up patient care, typically the GP or equivalent (1) , should receive written discharge communication from the discharging physician; this practice supports continuity of care between specialist services and primary care. Patients are sometimes included in this communication, and while within the UK this is considered to be 'good practice' (2) , is not standardised.
The Department of Health in the UK describes patient copies of letters as a "right" and recommend patients should be copied in where appropriate as a "rule", unless there is risk of harm (2,3) . This practice was intended to support patient understanding and wellbeing, increase patient safety and the quality of information sent, and improve doctor-patient relationships (2,3) . More recently, the Academy of Medical Royal Colleges (AMRC) released the "please write to me" (4) initiative. The initiative's purpose is to encourage doctors to write to directly patients in simple plain English to increase understanding. Despite these initiatives and guidelines, studies within and outside the UK report both benefits (5)(6)(7)(8)(9)(10)(11)(12)(13) (e.g. high patient satisfaction), and drawbacks (9,10,(13)(14)(15)(16) (e.g. patient confusion) of patients receiving their letters. Hence, many patients do not receive copies of such letters, but the reasons for this and the subsequent consequences remain unclear (2,3,17) .
We recently published a protocol (18) for this review which fully details the background and methods we used. As summarised in the protocol paper: "Whether or not it is beneficial for patients to receive written discharge communication, and, if so, for whom, when, how, why, and whether this should be a direct copy or personalised letter remains equivocal. We could find no review specific to this question; we only found reviews of copying letters in general (19,20) ." (18) Thus, the evidence on patients receiving discharge letters is unclear and it was concluded that consolidation of the evidence through a realist review is required.

METHODS
A realist review is a, 'theory-driven, interpretative approach to the synthesis of evidence' (21) . Synthesising evidence involves interrogating data sources to develop, refine and test context, mechanism, and outcome configurations (CMOCs). "Context may be conceptualised as external factors that influence mechanisms (22) . "Mechanisms" are hidden, context causal forces that produce "outcomes". (22) Following Pawson (22)(23)(24)(25) , CMOCs should be configured and consolidated to build and develop a realist programme theory or theorised explanation of how an intervention works or not. The intervention under scrutiny 'patients receiving discharge letters' was defined by the review team as 'the patient being given or sent any form of written (paper or digital) hospital discharge communication; this could be a direct copy, patient-directed letter, or a combination.' The protocol (18) considers realist review methodology in depth and argues a realist approach is apt and useful approach for this review. Briefly, we argued that: a realist review has the potential to identify how positive outcomes may be reproduced and has capacity to account for complexity. The intervention is complex in that the form of discharge communication can vary and the quality of communication is highly context-dependent.
The review design (Figure 1) was informed by previous literature, consists of six steps (22,26,27) and is further described in the protocol paper (18) .

Figure 1 Review design
The aim of the review is to understand and explain how the different outcomes are produced from adult patients receiving written discharge letters. Outcomes may be simplified into desired/beneficial or 'positive' (e.g. satisfaction) and undesired/detrimental or 'negative' (e.g. anxiety). The objectives are to conduct a realist review of the intervention (patients receiving copies of discharge communication); to develop a programme theory (PT); and to make best practice recommendations for the intervention.
The research questions (RQs) are as follows: RQ1: What positive and negative outcomes have been reported on patients receiving written discharge communication?
RQ2: What are the important contexts which are associated with whether the mechanisms produce the different outcomes, and why?

Programme theory development (step 1)
The task of locating existing theories to develop an initial rough PT was achieved through a scoping search. Theories were sought which contributed toward understanding how patients receiving discharge letters works or not. Search terms were based on the intervention (e.g. patient cop(y)ies). Published resources and healthcare websites were searched to ascertain a range of evidence (see Supplementary file 1). Sources were selected based on their "relevance" (22)(23)(24) to the PT; where relevance concerns 'does the [source] address the theory under test?' (23) . Crucially, the whole source did not need to inform the PT but we considered the relevance and contribution of sections of the document (7) . During this phase, research team judgement was needed to decide the stopping point for programme theory development as was the need to balance the degree of comprehensiveness and practicalities (18) .
Twenty seven documents were selected from the scoping search (see Supplementary file 2). All documents were then interrogated and coded for any CMOCs, concepts, or theories which could inform development of a PT. These were consolidated to form Figure 2, the initial PT.

Search strategy (step 2)
The electronic searching was purposive and guided by the initial PT. A search strategy was developed which was piloted and adapted for MEDLINE until a diverse and relevant range of search results were yielded (target 500-3,000). Thus, there was no strict search "threshold" and the most important criterion for search results to meet was relevance (22,23) . In line with a realist approach, searching was iterative, and the strategy was refined for each database (see Supplementary file 3). Sources included electronic databases, healthcare sites, and grey literature.
The search strategy was not intended to be exhaustive, but provided a large enough overview to be meaningful for PT development (23) . Evidence was searched up until September 2017; publications were monitored thereafter but no new evidence affected the PT. In total, 3113 documents were selected for screening.

Selection and appraisal of documents (step 3)
Inclusion or exclusion of source evidence for the review were according to the following criteria:

Inclusion criteria:
• Full text or section of source had relevance (22,23) to informing the PT • Relate to inpatients/outpatients discharged from hospital to GP (or equivalent) • Relate to discharge where discharge letter is sent to GP (may also be copied to patient) • Source written or published in English Exclusion criteria: • Specific to discharge to units/physicians other than GPs (or equivalent), e.g. another hospital • Specific to discharge of patients who lack cognitive capacity, e.g. dementia, or where there may be higher risk of harm, e.g. mental health discharge • Lack of written communication having taken place, e.g. telephone only • Specifically relate to patients <18 years • Source not written or published in English The criteria were developed to include evidence that encompassed a variety of patients and be relevant across different healthcare settings. The exclusion criteria posed limitations on the review; children under 18 (where the parent would often be the letter recipient), patients with particularly specialised communicative needs (e.g. patients without capacity) or where the intervention may have a higher potential risk of causing harm (e.g. psychiatric discharge documents) were excluded. The communication needs of some of these patients may be more complex and variable within and between groups and therefore was not possible within review scope. The first exclusion criterion states patient discharge communication to those other than GPs or family or community physicians was excluded. This is because the review specifically focussed on discharge communication to GPs and patients rather than referrals or care-handovers. Furthermore, the review aimed to develop a theory for patients receiving discharge communication and inclusion of hospital-hospital discharge may have reduced clarity and produced a less focussed theory.
Once KW had screened the documents by title and abstract, second reviewer EM screened a random 10% test selection; as recommended by Wong et al. (26) . Inter-reviewer agreement was set at kappa measure K≥.8 (28) . A result K<.8 would require all documents to be second screened. Inter-reviewer agreement was calculated as sufficient (K=0.82). In the first screening phase, 611 duplicates were removed and 2,341 documents excluded; this left 161 documents. The full texts of these 161 documents were then screened, primarily for relevance (22,23) by KW, with EM screening a random 10% sample. Inter-reviewer agreement was again high (K=0.92). Eighty eight documents were excluded at this stage leaving 73 for inclusion.
In addition, hand-searching of bibliographies, 'cited by' searching, and contacting experts was undertaken. This identified a further 30 relevant documents, creating a total of 103 documents. Supplementary file 4 provides the final document list. The selection process is summarised in Figure 3.  (29) diagram (document selection process)

Data extraction and analysis (step 4)
A "hybrid" (26,30) approach to data extraction was undertaken. This allowed extraction of both descriptive document characteristics and annotation of CMOC ideas for synthesis and integration into the PT (22,23) . The Excel data extraction form (see Supplementary file 5) was designed iteratively to record pertinent document details. Final columns included: author(s), year, geographical information, healthcare system, design aim, no. of participants intervention, clinical speciality, inclusion and exclusion criteria, findings/conclusions, rigour/quality assessment (22,23) , topic focus, form of discharge communication e.g. discharge summary, participant mix, staff mix, and relevance score (22)(23)(24) .
Documents were also annotated in NVivo for CMOCs and PT ideas. Annotations were guided by the initial PT devised in step 1.

Data synthesis (step 5)
During step 5, data and annotations of PT ideas and CMOs were consolidated. A realist analytic approach, following the work of Pawson (22)(23)(24)31) , was used to interrogate the theory during data synthesis. Pawson (22)(23)(24)31) presents several different frameworks for synthesising data evidence. We selected the framework (23) entitled "synthesis to consider the same theory in comparative settings", which involves five analytical strategy steps. This framework was chosen as it assumes theories sometimes "work" and "do not work" according to the particular setting; Pawson et al. (23) describe this as 'aim [ing] to make sense of the patterns of winners and losers'. Hence, this framework is suitable for the research questions which focus on cause and context of positive outcomes "winners" and negative outcomes "losers". Thus, data synthesis was grounded on the assumption that the intervention outcomes of the intervention may differ according to context.
The following realist analytical strategy steps (23) were undertaken simultaneously 1. Juxtaposition of data sources -align sources to build upon/clarify each other 2. Reconciliation of data discrepancies -explore reasons for data disparities 3. Adjudication of data -data quality consideration of trustworthiness/relevance 4. Consolidation of data -inference of Mechanisms for outcomes 5. Situation of evidence -consideration of intervention settings 'Juxtaposition of data sources' was achieved using NVivo 'nodes'. Annotations were labelled and coded as nodes. The nodes were named according to ideas or concepts around the PT and often formed groupings of mechanisms, outcomes and contexts. NVivo node coding resulted in 19 nodes seen in Table 1.  Clinician context (views) 23 57 Confidentiality 12 15 Context (when it does not work) 29 46 Context (when it does work) 54 107 Cost/resources 20 33 Dictate in front of patient 3 5 Doctor patient relationship 5 7 GP preference 4 8 NHS policy or contextual standards (international) 30 51 Outcomes (positive) 58 128 Outcomes (negative) 22 28 Patient as delivery method 2 2 Patient harm 24 33 Patient letters 18 34 Patient preference 37 94 Patient recall 11 12 Queries and contact 10 12 Understanding 46 88 During, 'reconciliation of data discrepancies' (22,23,26) and 'adjudication of data' (22,23,26) , NVivo was used for scanning and comparing data to identify disparities. Adjudicating and situating evidence was important to reconcile discrepancies (22,23,26) . Following node coding, a CMOC table was constructed (see Supplementary file 6) for consolidation of data and annotations. During table completion, we identified CMOCs where the intervention does and does not work.

Programme theory refinement (step 6)
Review step 6 was to consider stakeholder perspectives to test and refine the PT in light of the synthesised data (23) . Stakeholder views are for "checking" that the PT aligns with real-life experiences (7) . We consulted three groups: local policy makers and health service commissioners, GPs and service-users/patients. These groups were selected due to accessibility and their differing roles. Groups were contacted through University links.

Document characteristics
The 103 evidence sources were from 16 countries across various continents with most emanating from England (54%), the US (17%), and Australia (7%). Healthcare settings were split between insurance style systems (23%) and publically funded systems (77%), such as the NHS. The date range of the sources was from 1979-2017 and the total number of participants detailed across the research studies was 16,383; this included staff and patient participants but there was not enough detail across all of the studies to quantify the participant type proportions. Most had been published in the 10 years prior to the search: [1970][1971][1972][1973][1974][1975][1976][1977][1978][1979]  The evidence covered a wide range of specialties. Most specified inclusion of adult patients only (over 18 years) but often did not detail the exact patient ages in the write up; a few studies focussed on elderly patients. Information relating to patient demographics e.g. gender, was often not found in the sources and hence these were not summarised. Many sources instead focused on the speciality under consideration in the document and clinical presentations of interest to that speciality e.g. abnormal ECGs (32) . Participants who were staff included medical students, doctors of all training grades, nurses, GPs, non-specified hospital staff, and non-clinical staff. However, the majority of documents (66%) either did not provide staff participant details or they were irrelevant e.g. guideline document, no participants. The type of discharge communication that the evidence related to was varied: direct copies (48%), discharge instructions (13%), pictures (1%), personal discharge packs (1%), personalised letters (13%), information booklets (9%), multiple types of discharge communication (7%), and other (11%). Where the sources came from showed some variation such as Department of Health archive (3%) and conference listing (5%) but the greatest number of sources were from journals (68%).

Quality and document rigour
The findings of this review in the following sections should be considered in light of the quality and rigour appraisal results of included documents to avoid over-interpretation of the findings. During data extraction in step four, documents were quality appraised for rigour and evaluated for relevance (22,23) . The concept of rigour is defined as 'whether the methods used to generate the relevant data are credible and trustworthy' (21) . Relevance and rigour were scored on a scale from very low to very high and factors such as document type e.g. opinion piece or scientific trial paper, were considered. It is acknowledged that the appraisal process was subjective. Documents were not excluded solely based on rigour as extracts of documents with a lower quality score may still have valid contributions (9) . The full quality appraisal results are in the data extraction table (see Supplementary file 5).
The quality of evidence varied, with 53% of sources graded as medium or above for relevance and 80% for rigour. The remainder were graded medium/low (relevance=9%, rigour=18%), low (relevance=24%, rigour=1%) or very low (relevance=14%, rigour=1%). Information relating to setting and context was not always found or was insufficient. The source type was mixed: discussion pieces (20%), survey-based study (19%), guideline documents (13%), conference abstract (7%), review (5%), interview-based study (5%), experimental study (5%), pilot study (5%), randomised controlled trial or randomised intervention study (5%), non-randomised intervention study (3%), report document (3%), cohort study (2%), mixed methods (not covered above) (1%) and other e.g. PhD thesis (8%). Notably, the most common type was "discussion piece"; these were often solely based on the opinion of one individual and so did not always provide strong evidence. In some areas, evidence relating to the PT were thin, these included: negative outcomes, intervention cost, recent clinician views, doctor-patient relationships, personalised patient letters and nil intervention. Thus, some CMOCs were constrained by source data. Evidence was also thin in relation to data disparities, for example the reasons why conflicting attitudes or results occurred were often not described. Although, context and outcome information was generally well supported, mechanisms were frequently omitted. Hence, where possible, the expertise of the research team was drawn on to infer reasons for disparities and what the likely mechanism(s) were within any CMOC. Hence, the CMOCs and PT presented in the following sections are based on source data where possible but have also been supported by stakeholder feedback and research team inferences.

Context-mechanism-outcome configurations
The following section provides an overview of theories in the form of a narrative of how patients receiving discharge letters does or does not work, as informed by the included sources. The sub-headed themes emerged during data interrogation and consolidation. Sections contain references to CMOCs, quotations from data texts, and references. Quotations have been chosen which illustrate the described theories and highlight key elements of CMOCs. The full table of 48 CMOCs is found in Supplementary file 6.

Patient preference/choice
Allowing patients to make their own choice for receiving letters may: reduce unnecessary resource strain [CMOC14], only take minimal time (12) , make patients feel more involved in their care (2,12,(33)(34)(35)(36)(37) [CMOC2], increase satisfaction (34,(38)(39)(40)(41)(42) [CMOC14, CMOC41, CMOC47] and aid information acceptance (43) : ''I wanted to know as much as possible about what was going on with my body" (43) (p. 73) "Sometimes for whatever reason you don't fully take on board what the doctor has told you. I found the letter useful to read over and digest properly what was written" (2) (p. 3) Many patients report that receiving letters is useful (2,7,36,37,39,(44)(45)(46) . Patients may show friends/family to help them better understand their condition/treatment (43,(47)(48)(49)(50)(51) . Patients may use letters as a reference/reminder for the consultation (42,43) : "My mind went blank after seeing the doctor and the letter reminded me of what had been said" (36) (p. 83) Across a range of specialties and settings, the reported patient preference for receiving copies of their discharge letter is generally high (79%-97%) (7, 11, 34, 35, 42-44, 46, 52-55) . However, not all patients may find letters helpful (40) or necessary (7,34,49) , and some may not want to be reminded of their diagnosis (34) , which could decrease satisfaction, and generate queries (56) if these patients were sent letters without a choice [CMOC40]. Hence, several studies argue in favour of respecting patient choice and suggest the patient's right to "opt out" needs to be addressed (7,38,39,43,46,47,53,(57)(58)(59) [CMOC41]. In situations where the patient is not offered a choice, such as third party information or risk of harm (2) , the clinician should be able to justify this decision (60) . In relation to sensitive information or social diseases, patients generally do not object to this being included in the letter as long as it has 'some relevance' (60) .
A couple of sources suggested training doctors on writing patient letters can mean they produce letters more meaningful to patients (13,72) [CMOC13]. However, the evidence around training in relation to the intervention was limited and needs further research. "It is not considered good practice to send the discharge summary home with the patient as there is no guarantee that the information will be passed on to the general practitioner" (89)

Dictating letters in front of patients
Evidence for this concept was somewhat thin. One study suggested that dictating letters in front of patients can make patients feel less in need of a copy of the letter (11) . Another paper suggested this practice may also provide a context that triggers patients to challenge inaccuracies, improving letter quality (83) [CMOC22, CMOC30]: "The content of letters to GPs is sometimes incorrect and this may be remedied by dictating the letter in front of the patient." (83)

Confidentiality
There are concerns and legal implications surrounding potential confidentiality breaches associated with patients receiving letters, particularly when they are sent out in the post (15,35,42,49,52,54,58,90) . One recent paper (52) (2013), which looked at confidentiality, continued to stress risks around postal communication and the importance of secure information transfer: "There is a substantial risk of breaching patient confidentiality when distributing correspondence by post. A well-designed security arrangement is therefore required to ensure the safety of confidential information." (52) Some documents (52,57,58,60,91) suggested ways to reduce potential risk of confidentiality breach through communication platforms and the processes involved in sending letters e.g. verifying patient contact details before sending letter (60) [CMOC3, CMOC27, CMOC28].

Patient harm
Patient anxiety or "harm" in general are often cited as reasons for clinicians not wanting to copy letters, particularly in "bad news" settings (17,36,41,43,46,47,49,56,61,92) [CMOC6]. Letter inaccuracies can cause concern leaving patients feeling confused or anxious (45) [CMOC19]. Nevertheless, the letter can reassure the patients their problems are being handled (54) and initial anxiety can settle or be nullified by the usefulness of the letter (44,46,49,60,93) [CMOC37]. Moreover, one study published in the Lancet in 1991 suggested patient letters in "bad news" settings may be more useful than "good news": "Patients who had received bad news found the letter significantly more useful in helping them to understand and remember what they had been told during the consultation than did patients receiving good news… almost half the patients receiving bad news found their letter distressing to some extent; however, with 1 exception, all patients were pleased to have received it." (39) (Pp.924-925) Although the above paper was published in 1991, there has been no recent evidence or system changes to dispute the notion that "bad news" letters may be of particular use to the patient. Hence, despite risk of initial "harm", "bad news" letters should perhaps not be avoided.

Clinician views
Confusion around 'letter comprehensibility' and lack of 'patient understanding' were the commonest clinician reservations relating to the intervention (15,33,35,36,43,47,49,50,54,88,98,99) . However, as covered in the comprehension section, patients are often reported as understanding their letters (11,33,44,57,60) and furthermore they tend to express strong preference for receiving such letters (7, 11, 34, 35, 42-44, 46, 52-55) . Thus, it may be inferred from the evidence that patient understanding of letters is higher than clinicians' perceive. The following from a recent (2016) abstract concisely summarises an example of patient and clinician view disparity: "While some oncologists assess the copy letters as inappropriate for supplemental patient-oncologist-communication, breast cancer patients regard this tool as predominantly gainful. Oncologists appear to stick to their traditional perspective which perceives the copy letter mainly as a communication tool from doctor to doctor." (50) Notably, most of the evidence reporting clinician views was published from 2002-2008 and current evidence on clinician perspectives remains limited. Moreover, although sources often referred to conflicting clinician views, information on why attitudes differ was extremely thin. Overall, better understanding of current clinician views on copying discharge letters to patients is required. Further research should address reasons behind different viewpoints to include patients and practitioners.

Cost and resources
The estimated costs associated with the intervention varied (47) but this must be considered in the context that included documents spanned a wide time range and thus factors such as inflation need to be considered. In addition, robust health economic analyses were not found in the included sources. Documents (15,33,42,54,83) referred to "cost" or financial implications [CMOC25] of sending letters in different ways such as use of consumables (33,45,54,55,101) [CMOC10], and secretarial (33,44,45,54) [CMOC10] and clinician time required (45) . A few sources (34,36,44,46,55,60,102) , including guideline documents and research papers, suggested that benefits were such that associated costs were minimal, or  even reduced by patients being more informed from receiving discharging communication  [CMOC7, CMOC25, CMOC42]. However, as these views were based on personal comment or studies with weak methodologies, the true cost consequences remain unknown.

Autonomy
One source suggested that when patients are not given letters, they may feel less involved in their care, resulting in reduced patient autonomy (103) [CMOC1, CMOC6]: "To refuse to provide such information if this is the patient's wish is to deny their autonomy." (103) (p.388) Conversely, some evidence was found that providing patients with written discharge letters is their "right" (3,58) , may create a sense of involvement, and increase patient autonomy and satisfaction (103,104) [CMOC2, CMOC4, CMOC5, CMOC8, CMOC14].

Doctor-patient relationships
Very few documents were found which considered the intervention in terms of the doctor-patient relationship. However, the limited evidence that was found indicated that patients receiving letters has the potential to improve communication, trust and the doctorpatient relationship [CMOC9] (33,43,47) .

Stakeholder perspectives
As detailed in step 6, the final review step was to refine the programme theory through stakeholder perspectives. Three groups were consulted: local commissioners, GPs and service-users/patients. Groups were relatively small and not representative of the general population; due to feasibility it was not possible to achieve diverse and representative group samples.
All group discussions were centred on the programme theory; members were encouraged to critique and feedback on the PT diagram. Between all the groups, concepts not covered or explored in adequate detail in the PT diagram were raised in regards to: the importance of comprehensible language and terminology, difficulty and problems retaining verbal information only e.g. where anaesthesia used, patient choice of receiving letters, illegibility of handwritten discharge communication, critical context of prior patient communication of a high quality to increase likelihood of understanding discharge letters, issues around personalised patient letters considering NHS resource availability, and concerns around writing a letter which meets the needs of both GP and patient. The commissioner and GP representatives also emphasised the importance of patient safety and that this should be central to best practice recommendations. In addition, the patient group reported reading a letter about themselves written in third person was peculiar. The patient group felt patient letters were very important for patients taking responsibility for their health in line with the NHS promotion of patient-centred and patient-led care.
Several different members across the various groups commented that in practice, patients do not always receive their letters, despite this process being best practice. Recommendations were suggested to rectify this by the commissioner members to include: clinicians should assume when writing letters that they could be made available to the patient, early clinician and student training in good letter writing and record keeping, and that hospitals should support the initiative e.g. audits.

Cycling of review steps
As a realist review is an iterative process, following the initial six step cycle, steps may be repeated in order to attain "theoretical saturation" (22,24) . There was substantial overlap between documents which is demonstrated by the fact there are 103 documents in the review but only 48 CMOCs. Furthermore, as described in step two, new publications were followed and consulted for evidence and provided no new or conflicting programme theory knowledge. Thus, it was deemed that "theoretical saturation" in accordance with Pawson's realist review methodology (22)(23)(24)(25) was attained and no further searching or step cycling was required.

Resultant programme theory
The PT was systematically updated and finalised following review steps 1-6 ( Figure  4). Thus, several changes were made to the initial theory ( Figure 2) to produce the resultant theory below ( Figure 4) and these are summarised here.
The feasibility of providing a personal patient letter with current NHS setting of resource provision was removed from the PT. Therefore, the personalised letter section of the PT instead focuses on alternative healthcare settings e.g. insurance-based, which is the setting where the majority of research on this type of intervention was conducted. In addition, the box of contextual influences was deleted, and the points integrated into the overall diagram. Patient outcomes throughout the PT were simplified and clarified, for example, the outcomes such as 'empowered patients' and 'reduce patient anxiety' were simplified to the outcome 'improved patient well-being'. The final PT has more refined and distinct pathways for when the patient does and does not receive letters and the subsequent respective outcomes. Furthermore, contexts for when the patient does receive their letter(s) were condensed into an aligned grouping of five key contexts for when the intervention may be theorised to work and four key contexts for when the intervention may be theorised not to work. In the resultant PT, CMOCs have been "grouped" where overlap was apparent, for example, all resources are labelled simply as "resources" as most of the data concurrently referred to financial, administrative and clinician time resources.

Statement of principal findings
This review of 103 sources summarises and expands upon existing evidence by moving beyond "benefits" and "drawbacks" of patients receiving letters alone, and considering contexts of when as well as how the intervention works. Although the review focuses on the UK health system, our use of realist review has enabled identification of findings that may be transferable to other healthcare settings. RQ1 asked about positive and negative outcomes of the intervention. Positive outcomes include: increased patient satisfaction (34,(38)(39)(40)(41)(42) , improved doctor-patient relationship and trust (33,43,47) , heightened patient knowledge (32,75) , improved letter and record quality (2,13,33,45) , and reduced anxiety (17) . Negative outcomes include patient queries (47) , confusion (33,49,54,61) , and anxiety (47,49) . RQ2 enquired after the important contexts for triggering these outcomes. Important contexts for positive outcomes include: letters written in plain English with minimal abbreviations (60) , lay explanations or simplified terms in brackets for medical jargon e.g. myocardial infarction (heart attack) (47,60) , written information provided alongside verbal explanation, no new information in letter (2) or value judgements (36) , letter translation (38,76,81) where relevant, training clinicians on letter writing practice (13,72) , use of pictures and glossaries where relevant (75,84,85) , letters only given to patients who choose to have them (47,60) , and where there is no identified risk of harm (2) or confidentiality breach (60) . This review has produced two key findings, which are important but not wholly surprising. The first is that the reviewed evidence indicates that patient "understanding" of their discharge letters is greater than clinicians perceive (7, 11, 15, 33-36, 43-47, 49, 50, 56, 60-64, 88, 98, 99) . However, reasons behind patient and clinician perceived comprehension discrepancies were unclear. It is important to situate the first finding in terms of the study exclusion criteria and participant diversity across the evidence reviewed, for example, it is likely that patients who participate in research on this topic have a greater level of interest and literacy than those who did not participate. One or a number of demographic groups not involved in the studies, either by choice not to participate or by exclusion, may have accounted for a portion of those who clinicians perceive to have low understanding. Thus, evidence for low patient understanding was lacking, and this requires further research. The second key finding is that in a number of contexts, patients expressed preference for receiving correspondence (7, 11, 34, 35, 42-44, 46, 52-55) . Patients can continue to use the letter(s) to refer to beyond discharge (34,93) as a medication list reminder, and to share with friends/relatives as desired (34,81) . Nevertheless, patient choice should still be acknowledged as the review did find evidence that not all patients want their letters; a practical way of addressing this would be to check with the patient that they want a letter in the first instance (7,34,40,47,49,60) .
Systems for monitoring patient letters, e.g. the Newcastle Trust Policy for auditing and sharing letters with patients (60) seems prudent moving forward (17) . This is of particular relevance in the NHS given that guidelines for copying letters have been widely available since 2003 (2) and yet in practice, many patients do not receive letters (42,47,52,79,105) . Given the wider context of a drive for patient-led care and patient-centred communication and decision-making (103) , this review is timely and relevant. The review findings have the potential to influence policy and improve practice. The results demonstrate how care can be improved through patient choice and good quality letter provision. However, current clinician views (33,35,36,44,49,50,65,88) and hierarchical mimicking of practices of seniors (100) pose a barrier (40) to implementation and need addressing.

Review limitations
For this review we followed the RAMESES quality and publication standards for realist reviews (106,107) . Quality assessment and analysis is to a degree dependent on F o r p e e r r e v i e w o n l y reviewer skills and reflexivity (108,109) . Furthermore, analysis and inferences were 'subjective and interpretative' (110,111) . However, because the steps we have taken for this review are transparent, other review teams can see and make judgements on result plausibility.
The resultant PT is limited by the quality and content of evidence reviewed. Some of the evidence found in sources was markedly thin, particularly in relation to costing information, recent clinician viewpoints, personalised letter copies, and influence on the doctor-patient relationship. Furthermore, there were a greater number of CMOCs relating to positive outcomes than negative outcomes, that is, when the intervention does work than doesn't. This may be rationalised by publication bias towards positive findings. Similarly, CMOCs for patients not receiving letters (nil intervention) were thin. Consequently, these evidence limitations constrained the detail available in the resultant PT in these areas.
The review is also limited by the inclusion and exclusion criteria; not all patient groups could be considered. Furthermore, the PT is limited by the representativeness of patient groups within sources. The usefulness of the programme theory to a general health population is affected by the participation bias; certain groups are more and less likely to participate and ethnic minorities, illiterate patients and other hard to reach groups are likely under-represented. It is essential to take this into account when considering the weight of the review recommendations for informing best practice of discharge communication.
It is hoped the PT may be used as a starting point for future research and be useful and practicable for informing policy and guidelines. It is acknowledged that the review is not exhaustive. However, this is not the intention of a realist review (112) .

Suggestions for future research
Further research is needed to define the cost benefits of copying patients into discharge letters. Potential barriers such as clinician views and the current limited available clinician training on letter writing should be addressed. Since patient and clinician views were sometimes conflicting, a study which parallels both views alongside the same patient cases to understand reasons for any discrepancies would be useful and may provide valuable insights. Furthermore, the PT would benefit from further refinement given the identified limitations and in particular from more data from alternative settings.

CONCLUSION
The resultant PT forms a basis for explaining how, when, why and for whom this intervention does and does not work. The resultant PT makes suggestions for how best practice of patients receiving discharge letters may be improved, although it should be considered that evidence for some aspects of the PT was lacking.
The review's key findings are that patient understanding is possibly greater than clinicians' perceive, patient choice is instrumental to increasing likelihood of desired patient outcomes, and clinician views of the intervention is a barrier to practice implementation. This barrier could be addressed through clinician training and organisational initiatives which guide, mandate, and monitor the intervention. Without hospital organisational support, it is unlikely this practice will be consistently adopted into practice given the barriers identified in the review. Given the drive for patient-centred care, involving patients in their healthcare communication is not something that should be applied inconsistently.
In conclusion, this review outlines how the intervention of patients receiving their discharge letters may work to increase the likelihood of positive effects. Author contributions: KW was the lead reviewer. EM completed the second reviewer tasks. All authors contributed toward the programme theory through discussions. KW is responsible for the design and drafting of the initial manuscript. GW, ES, SS and JD critically reviewed and edited the final manuscript. All authors read and approved the final manuscript.
Acknowledgements: Samantha Johnson, our information specialist, for critically appraising the search strategy.    1. personalised patient discharge letter AND "patient discharge" AND written AND patient information AND copy AND personalised AND copy letter AND doctor letter AND discharge communication copy OR letter "discharge from hospital"Google Scholar search: (many thousands of results when using above terms) 1. (personalised patient discharge letter) AND (discharge from hospital) AND "patient discharge" AND written AND patient information AND copy AND personalised AND copy letter AND doctor letter AND discharge communication AND written AND patient copy Broad searches for X, Y, Z (websites and sources without indexing or electronic searching) e.g.

Department of health Used for: HSCIC and EPPI-CENTRE And Open Grey
The below search terms were entered into GMC yielding several hundred results. As results from GMC must be exported singularly these were screened on the webpage. The searches found no relevant results.   Patients who did not speak English (14) or who were considered too ill or disabled to be interviewed (7)  none specified inpatients none specified none specified breast unit patients none specified n/a n/a n/a adults all patients none specified F o r p e e r r e v i e w o n l y none specified inpatients none specified "over 18 years of age and discharged on analgesics with verbal instructions only (e.g. paracetamol, NSAIDs and/or tramadol, as reported in patient's file) were eligible to take part. Exclusion criteria were: patients unable to read and speak Dutch, patients with missing data (particularly telephone number), patients who could not be contacted by telephone within 3 days of their ED visit, patients without access to e-mail and patients who refused to take part." ED patients who met criteria none specified none specified asthma patients, none specified none specified chronic lower back pain patients hospital nurses and clinicians none specified none specified none specified n/a n/a n/a iving patients an information booklet at discharge from hospital appreciably increased the accuracy and thoroughness of their recall of important medical details concerning their illness and its treatment. The booklet was shown to be feasible, helpful in the outpatient clinic, and preferred by most general practitioners.' medium low Giving an ipformation card to all patients at discharge was feasible and favoured by most patients and their general practitioners. Having made minor changes in design, we think that we have produced an information card that is a convenient size and will improve communication between patients, their general practitioners, and hospital doctors. We now issue this card routinely to all patients discharged from our ward and hope that it might be widely adopted.' medium very low received no complaints from this practice. low/medium low "In our study, almost half the patients receiving bad news found their letter distressing to some extent; however, with 1 exception, all patients were pleased to have received it." "The study shows that letters to "patients are a useful method of improving patient satisfaction and recall in clinical consultations. They also provide a permanent record of the consultation, which can be kept for future reference, and encourage greater patient involvement in their care." publication, rigour -med/high medium reports and documents should be shared with patients low/medium very low "The recent introduction of new information technology to the NHS has not only improved the ability to carry out surgical audit, it has allowed the production of computerised discharge information on patients. This shorter, more structured form of discharge summary is clearly more acceptable to general practitioners and potentially may be used by all hospital specialties to provide a speedier, more efficient communication service to the general practitioner." medium medium "On the basis of our findings we suggest that genetic counsellors send patients a letter summarising their consultations as this may result not only in an increase in the patient's understanding, but may prevent the miscommunication of genetic information within the family." Opinion based but published -low/med low "Given that the majority of doctors supported the provision to some or all patients of summary letters, and the majority of oncologists supported the provision of a consultation audiotape in at least some cases, it may be useful to establish a forum where physicians and patients can develop guidelines for the use of these information aids. Such guidelines might suggest appropriate criteria for determining when to offer such aids, as well as a suggested format." publication, rigour -med/high medium "For many people with eye disease, particularly if vision is threatened, fear of the unknown is a major cause of stress and anxiety. Carefully structured correspondence, written in a format that can be read easily and understood, can significantly enhance the management of many of our patients." low/medium low/medium It is recommended that structured, pre-formatted instruction sheets be provided to all patients discharged to home, that emergency departments establish uniform policies to promote best practice in communication, and that the use of discharge instructions be considered as an emergency department performance indicator.' publication, rigour -med/high low a competency based approach to test whether individuals have heard and understood their instructions may be better medium low/medium F o r p e e r r e v i e w o n l y "The idea that patients should receive copies of letters and summaries is not new, and research has shown that it meets with high satisfaction from the patients." Letter -low/medium low-med "Despite our wish to see this policy implemented nationally as soon as possible, we believe that it would be counterproductive to do this without first addressing the considerable concerns of both clinicians and certain patient groups." Opinion based but published -low/med low-med "It should now be the exception to write letters that we would not wish patients to read." low low "Patients receive a discharge information summary at the time of discharge." medium/high medium "We agree with Chantler and Johnson that most patients (even those with cancer) wish to receive copy letters and most patients find them useful. However, it is essential to offer patients a choice in this matter because some patients do not wish to receive copy letters." Opinion based but published -low/med medium Sending patients a copy of correspondence to their GP is one means of aiding communication and improving overall satisfaction.' medium medium/high "letters between clinicians about an individual patient's care will be copied to the patient as of right" medium/high medium/high "These guidelines show that the issues to be tackled are mainly those already required throughout the health service and in good professional practice. They include good communication, obtaining consent, record keeping and handling (including IT support systems), and provisions to meet legal requirements to prevent discrimination on the grounds of disability or race. In particular, issues involve provisions around confidentiality and data protection." medium/high medium/high "Patients found the intervention usable and reported increased confidence to ask questions. A definitive randomised controlled trial is required to establish the effect of the education package on patient outcomes." medium low New training is needed in order to implement the practice of copying letters to patients -doctors are not yet ready publication -high medium  Consultants who participated in the exercise did not perceive any additional difficulties in implementing this practice. This small study found that 100% of patients receiving a copy letter found it useful. The fact that around one-fifth of patients did not receive such copy letters within 2 weeks as intended is worrying, and requires further investigation. Sending a copy letter involves a relatively trivial cost for a practice which patients view as a valuable resource.' publication, rigour -med/high medium "If the government is serious about providing more information to patients, then copying letters may be a useful way of facilitating that process of culture change, as well as addressing some of the objections to the principles of more information sharing." Publication, NHS based. Rigour -high low-med "This study demonstrated that sending copies of the consultant oncologist letter to patients, proved to be a useful and valued method of communication with minimal addition to the workload." publication, rigour -med/high medium "Our survey confirms that patients want to receive copy lettersand find it very useful. The beneficial effects outweigh thedrawbacks, which can easily be overcome. We suggest that the benefits of copying letters should be recognized and the process welcomed voluntarily rather than eventually responding to an imposed compulsory directive." publication, rigour -med/high medium  There is diversity in medical and patient opinion about consultants writing directly to patients. These findings reflect fundamental differences in expectations about the nature and quality of communications between doctors and their patients.' medium/high medium "The research demonstrated a high percentage of patients wanting a copy letter, which has huge resource implications in terms of secretarial time, additional stationery and postage costs. Change is difficult and an emotionally charged issue however, using examples of good practice and taking a multi-faceted approach to the change process the initiative has been successfully implemented." medium/high medium/high 40% of patients wanted to see a copy of their clinic letter. Patients should be given a choice. publication, rigour -med/high low-med "Patients appreciate copies of the letter being sent to their GP but comprehension is less good than with a shorter letter written especially to the patient. More attention needs to be paid to making letters to GPs simpler to read without losing the structure and detail liked by GPs. A compromise might be to dictate the letter in front of the patient and to provide a specialityspecific glossary to accompany each letter." publication, rigour -med/high medium F o r p e e r r e v i e w o n l y "The GPs generally found the structure and lack of specific clinical detail in the letters to patients unacceptable.
What we now need to do is to combine results from all of these reports and if only one letter is going to be written, determine the optimal format in terms of structure, content and comprehensibility to serve the needs of both referring doctors and patients." high high "The findings suggest that although personal letters do not substantially improve recall of the clinical encounter, they are feasible, highly valued by patients and acceptable to referring clinicians." high medium Patient receives copy promoting consumer "awareness and understanding" medium medium "preparations are undoubtedly required to meet this new standard of care: after all, it seems to be strongly supported by patients in the NHS." Publication, East Sussex Hospital. Rigour -high med-high "We have therefore demonstrated that the glossary may enhance a patient's understanding of the letter to the GP and the positive, spontaneous comments would certainly suggest that it enhances patient satisfaction with the whole consultation process." Publication, NHS based. Rigour -high low F o r p e e r r e v i e w o n l y "It appears that in general surgery in Wrexham these concerns may be more in keeping with patients' wishes than in other specialities where it is the professionals but not the patients who are concerned about more communication. Dissemination of this policy is a problem in that many junior doctors are not even aware of it, which is a training issue." publication, rigour -med/high medium "We feel that the practice of writing letters directly to patients has numerous benefits. This study demonstrated how writing such letters allows the patients to have their own source of information about their management. We hope to encourage more clinicians, not just those within palliative medicine, to adopt this practice to further the goal of improving patient care." publication, rigour -med/high medium "We suggest that rather than sending patients' copies of all their correspondence" as a routine, there are more secure and cost effective ways to inform patients and allow free access to all information recorded in their medical notes.
publication, rigour -med/high medium The results of this study will help inform how a DDS can be automatically generated from the electronic patient health record after each clinic visit.' low/medium very low  "Copying patients into their GP letters is a practice supported by the majority of patients. It appears to have little impact on the running of the department and is likely to add significantly to the understanding that patients have of their consultation. Hospital departments should consider implementing this useful practice." medium/high medium/high not in favour of patient copies of clinic letters "In the era of target driven medicine, doctors compulsorily copying patients into correspondence could easily become a surrogate Opinion based but published -low/med medium "Before rushing to provide the service that Richards suggests, we must think it through, plan it properly to maximise its benefits to all patients, and decide whether the potential overall benefit is worth the cost." Opinion based but published -low/med medium "This review suggests that researchers should move from examining the benefits and concerns around copying letters to patients, and instead focus on exploring the quality of correspondence and the optimum process of implementing the practice. As patients can ''opt out'' of receiving copy correspondence, audit of service delivery may be better assessed by whether patients have been offered a letter, rather than the current measure of whether one has been received. Practice implications: Copying letters to patients may have a number of important benefits and should be routine practice where patients wish to receive correspondence. Further discussion regarding the style and content of letters would be beneficial, together with attention paid to the mechanisms for recording patient preference. There is also a need for studies in non-medical professions." high medium/high "Many patients unnecessarily avoided activities of daily living after cataract surgery. Providing an additional written sheet did not significantly improve this, whereas a photograph sheet did. Better awareness of the safety and rapid rehabilitation after modern cataract surgery is needed in hospitals and primary care centers." medium/high medium patient letters with test results following discharge was beneficial by not inconviencing patients with a hospital visit and also ensuring test results were communicated quickly medium/high low/medium "It is not considered to be good practice to send the discharge summary home with the patient as there is no guarantee that the information will be passed on to the general practitioner." medium very low Giving patients a written discharge plan medium low "Including information in a discharge summary in native languages significantly improved patients' knowledge of illness and medication. This could be a simple and cost-effective method to improve health communication and health knowledge. This should be replicated in other parts of Sri Lanka and in other countries with similar contextual factors and further evaluated." medium very low "Structured teaching of communication with the patient brings family medicine back to what has actually always been its main part -communication and doctor-patient relationship. Our future aim is to develop students' letters to patients as a new tool in the family medicine course examination. Moreover, we will investigate how they can be used in everyday practice of family medicine." medium low  The majority of patients want to receive a copy of their clinical letters. "There is a substantial risk of breaching patient confidentiality when distributing correspondence by post. A well-designed security arrangement is therefore required to ensure the safety of confidential information. Despite the security concern, mail is still the preferred mode of delivery." publication, rigour -med/high medium "There is considerable evidence and experience to suggest that patients receive good quality letters very positively and with the outcome of improved satisfaction and reduction of anxiety. However, the most pressing implication of the policy relates to those sending the letters if they do not prepare properly and patients receive inappropriate or unsuitable letters, which might cause unnecessary distress or concern." low/medium low "A key limitation of the present study is that only a small number of experts were used to drive the iterative refinement; this cannot be expected to capture all the problems that health consumers might encounter and does not provide a quantitative measure of performance." low/medium very low "standard headings for the clinical information that should be recorded in the discharge record and included in the discharge summary communication from hospital to GP and patient." Information and advice given to patient should be provided in written discharge communication medium/high low "Writing of a letter to their first patients may be a useful tool for students to personally experience the practice of medicine and establish better partnership with patients in health care." medium low "Internationally, evidence evaluating critical care discharge information is scant. Many complex and interwoven factors can affect physical and psychosocial health outcomes after discharge from critical care, making it difficult to extrapolate the effects of information giving alone. Our understanding of service users' perspectives in this important area also is quite limited. Findings of this review should inform the further development of information for this population and the design of future studies investigating this neglected area of critical care practice. CCN" high medium The law requires written and oral information to be given to patients (France). The written form is often not used. medium medium "Advice given to patients should be in language understandable to the individual6,31,33 and cover aspects of the treatment plan including self-management that encourages a shared-care model of health." publication, rigour -med/high very low  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46 F o r p e e r r e v i e w o n l y The comprehensive discharge bundle was not effective in reducing the rate of readmission and increasing patient satisfaction, but medical discharge summaries were sent faster to the general practitioner and a trend to a longer time to readmission was present.' publication, rigour -med/high very low The current methods of delivering ED discharge instructions are not optimal for patient learning. Proper health education for patients after an ED visit has a potential to improve patient understanding of health information, decrease unnecessary return visit to the ED, and positively affect health' medium low HF patients' comprehension of discharge instructions is inadequate. Patients with limited education and those that do not speak English as a primary language are more likely to have poorer discharge understanding and higher rates of 30-day readmissions.' medium/high low  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46 "A simple patient-directed letter delivered during a brief discussion improves patient understanding of their hospitalisation and postdischarge recommendations, which is otherwise limited. Further evaluation of this brief and wellreceived intervention is indicated, with the goal of improving patient understanding, satisfaction and clinical outcomes." high high "Patient discharge summaries are likely to be a useful adjunct to existing discharge information, but further work is required to determine when and how they should be provided. With appropriate training and support, it is feasible for nurses to write dischargesummaries in a busy critical care environment." high high "patients have a right to these hospital reports, and it is the responsibility of physicians to provide them." medium medium Ask patient how they prefer to receive information. "Use discharge instructions that meet health literacy needs. Materials should be written at a 5th grade or lower reading level. Consider revising written materials to address the health literacy needs of all patients. Use readability tests, divide complex information into bullet points, and modify document font, layout, and design to revise written materials to improve readability." medium low  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46 F o r p e e r r e v i e w o n l y Patient discharge summaries are not routinely produced despite the reported benefits for patients on receiving this information medium medium Therefore, nursing and hospital leaders must develop and have a formal discharge planning process in place that includes interprofessional collaboration to effectively meet the treatment needs of patients and achieve adequate healthcare results.' medium very low Consider the patient's learning style19,20 medium low "Patient understanding of discharge instructions is an important aspect of ED provider-patient communication and adequate care. Our investigation reveals possible areas of communication as well as specific patient populations to target to ensure proper and safe understanding." medium/high medium  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46 F o r p e e r r e v i e w o n l y "Findings from this pilot RCT provide important preliminary data regarding the circumstances under which an intervention based on the principles of UCCDIP could be effective, and the sample size required to demonstrate this." high medium "The form, designed to be completed by junior doctors as part of discharge processes, includes a text box entitled "clinical narrative" which asks the discharging clinician to tell the story of the admission, encouraging them to do so in a way that might be easily understood. Patients themselves receive a printed copy at discharge, aiming to reinforce the importance of making the narrative readable. The documentation also includes the list of medications on which a patient has been discharged as well as specific boxes to document any medications that have been discontinued and any changes made to dosages, flagging up those factors most important for a GP to have quick sight of." medium medium  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46 Participants across literacy groups and settings identified multiple actionable areas for improvement in the ED discharge process. These included the use of simplified/lay language, increased visual learning and demonstration, and the desire for complete information. Individuals with limited literacy may particularly benefit from increased attention to consistency.' medium low "We believe that if we are able to provide clear and understandable instructions at discharge we might partially reduce inappropriate usage of the ED due to chronic disease, reduce costs and satisfy the patients' needs." medium/high low patient summaries improve patient understanding and decrease queries received by hospital followng patient discharge. medium/high medium "Involving patients at discharge has been shown to be valuable in reducing medication-related readmissions and post-discharge service utilisation; for improving patient outcomes and for supporting patients in understanding how, when, and where to seek help should they need it." medium/high medium  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46 F o r p e e r r e v i e w o n l y "Patient discharge notes are written at too advanced of an educational level. To ensure patient comprehension, dismissal notes should be rewritten to a 6th-grade level." medium/high low "Patients and professionals rated the PPDL positively. Key success factors for implementation were: education of interns, residents and staff, standardization of the content of the PPDL, integrating the PPDL into the electronic medical record and hospitalwide policy." medium/high high "While some oncologists assess the copy letters as inappropriate for supplemental patient-oncologist-communication, breast cancer patients regard this tool as predominantly gainful. Oncologists appear to stick to their traditional perspective which perceives the copy letter mainly as a communication tool from doctor to doctor. Due to their individual experience, patients seem to develop an emotional relationship to the copy letter containing information about their disease. Especially for patients dealing actively with their treatment process, copy letters could be a reasonable instrument." medium medium  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  ." medium/high medium/high "When asked how often patients are discharged home with a copy of their discharge plan to give to the community nurse, 48.2% (n=242) of community respondents said that patients 'sometimes' had a copy, 26.5% (n=133) said patients 'almost always' had a copy and just 1% (n=5) responded that patients are always given a copy of their plan." medium/high medium  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46 "Patients wish to receive copies of their correspondence and feel it improves their understanding of their medical condition. Although we were unable to demonstrate a measurable reduction in anxiety, increase in understanding or satisfaction, we recommend that patients be offered the choice of receiving copies of their clinic correspondence and endoscopy reports." publication, rigour -med/high low-med "Of course there are situations where writing to patients may be inappropriate; breaking bad news is always better face to face. But patients do not want any of the consultation withheld from letters." low/med medium "It is acknowledged that outpatient letters are increasingly written to patients and copied to GPs. In these cases the letter can act as a contract between the patient and their healthcare professional. It is expected that the standards will be applicable to these letters also." medium low  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45   Discharge communication may follow an inpatient or outpatient discharge; it typically comprises written discharge information in the form of a discharge letter or summary. It is a well-established practice that the physician who is to follow up patient care, typically the General Practitioner (GP) or equivalent (1) , should receive written discharge communication from the discharging physician; this practice supports continuity of care between specialist services and primary care. Patients are sometimes included in this communication, and while within the UK this is considered to be 'good practice' (2) , is not standardised.
The Department of Health in the UK describes patient copies of letters as a "right" (3) and recommend patients should be copied in where appropriate as a "rule", unless there is risk of harm (2,3) . This is intended to support patient understanding and wellbeing, increase patient safety and the quality of information sent, and improve doctor-patient relationships (2)(3)(4) . More recently, the Academy of Medical Royal Colleges (AMRC) released the "please write to me" (5) initiative. The initiative encourages doctors to write directly to patients in simple plain English to increase understanding. Despite these initiatives and guidelines, evidence within and outside the UK reports both benefits (6-10) (e.g. patient satisfaction), and drawbacks (11)(12)(13)(14)(15) (e.g. patient confusion) of patients receiving their letters. While patients receive copies of discharge letters inconsistently (16,17) , the reasons for this and the subsequent consequences remain unclear.
Hence, the objectives of the current study were to conduct a realist review of patients receiving discharge communication (the intervention); to develop a programme theory (PT); and to make best practice recommendations. The research questions (RQs) were: RQ1: What positive and negative outcomes have been reported on patients receiving written discharge communication?
RQ2: What are the important contexts which are associated with whether the mechanisms produce the different outcomes, and why?

METHODS
A realist review is a, 'theory-driven, interpretative approach to the synthesis of evidence' (18) . Synthesising evidence involves interrogating data sources to develop, refine and test context, mechanism, and outcome configurations (CMOCs). "Context" may be conceptualised as external factors that influence mechanisms (19) . "Mechanisms" are hidden, context sensitive causal forces that produce "outcomes" (19) . Following Pawson (19)(20)(21)(22) , CMOCs should be configured and consolidated to build and develop a realist programme theory or theorised explanation of how an intervention works or not. The intervention under scrutiny 'patients receiving discharge letters' was defined by the review team as 'the patient being given or sent any form of written (paper or digital) hospital discharge communication; this could be a direct copy, patient-directed letter, or a combination.' The aim of the review is to understand and explain how the different outcomes are produced for adult patients receiving written discharge letters. Outcomes may be simplified into desired/beneficial or 'positive' (e.g. increased satisfaction) and undesired/detrimental or 'negative' (e.g. increased anxiety).
We have previously published the full protocol for this review (23) which justifies the rationale for a realist approach and considers each of the methodological steps in detail. The overall review design was informed by previous literature, driven by the research questions, consists of six steps (19,24,25) and is further described in the protocol paper (23) . This design is summarised in figure 1.

Programme theory development (step 1)
The task of locating existing theories to develop an initial rough PT was achieved through a scoping search. Theories and evidence were sought which aided explanation of how and why patients receiving discharge communication results in different positive effects (e.g. drug adherence) and negative effects (e.g. preventable hospital readmissions). Sources were selected based on their "relevance" (19)(20)(21) to the PT; where relevance concerns 'does the [source] address the theory under test?' (20) . Crucially, the whole source did not need to inform the PT but we considered the relevance and contribution of sections of the document (20) .
Search terms were based on the intervention (e.g. patient cop(y)ies). Published resources and healthcare websites were searched to ascertain a range of evidence (see Supplementary file 1). During this phase, research team judgement was needed to decide the stopping point for programme theory development as was the need to balance the degree of comprehensiveness and practicalities (23) . As the purpose was to locate existing theories and initial concepts, the search was not intended to be comprehensive and the decision was made to screen no more than 30 documents. During the scoping search, search strategies within articles and article indexing were noted in order to inform a more thorough subsequent search in step 2.
Twenty seven documents were selected from the scoping search (see Supplementary file 2). All documents were then interrogated and coded for any CMOCs, concepts, or theories which could inform development of a PT. These were consolidated to form Figure 2, the initial PT.
The initial PT shows two main channels for discharge communication; patient copied into (or not) the hospital to GP letter and patient received a personalised letter. Limited evidence was available for the option of 'patient does not receive copy' as evident in Figure  2. Patients being copied into discharge letters, whether by choice or otherwise, is associated with a large range of mechanisms and outcomes. Contexts such as 'patient literacy level' are likely to influence generation of mechanisms (patient does/does not find letter comprehensible) and outcomes (increase/no increase in patient knowledge), but this was unclear from the evidence reviewed in the scoping search. The scoping search revealed a range of "contextual influences" (e.g. 'time constraints of writer' may affect whether a patient is given a choice about receiving a letter and also the overall letter quality). It was unclear where some CMOCs began and ended (e.g. "is patient given choice about receiving copy of discharge communication" falls between contextual influences and context labels). In Figure 2, f/u stands for "follow up" and the small circled "c" icon refers to an outcome which could also thereafter take the form of a context. In summary, there were clear "gaps" and information missing from the initial PT, confirming that the scoping search in isolation was insufficient for realist theory generation; further evidence and searching was needed to clarify details and simplify CMOCs (steps 2-6).

Search strategy (step 2)
The electronic searching was purposive and guided by the initial PT, results and indexing from step 1. A search strategy was developed which was piloted and adapted for MEDLINE until a diverse and relevant range of search results were yielded (target 500-3,000). In line with a realist approach, searching was iterative, and the strategy was refined for each database (see Supplementary file 3). Sources included electronic databases, healthcare sites, and grey literature.
The search strategy was not intended to be exhaustive, but provided a large enough overview to be meaningful for PT development (20) . Evidence was searched up until September 2017; publications were monitored thereafter but no new evidence affected the PT. In total, 3113 documents were selected for screening.

Selection and appraisal of documents (step 3)
Inclusion or exclusion of source evidence for the review were according to the following criteria:

Inclusion criteria:
• Full text or section of source had relevance (19,20) to informing the PT • Relate to inpatients/outpatients discharged from general hospital setting to GP (or equivalent) • Relate to discharge where 'written discharge communication' is sent to GP or referring physician (may also be copied to patient) • Source written or published in English Exclusion criteria: • Specific to discharge to units/physicians other than GPs (or equivalent), e.g. another hospital • Specific to discharge of patients who lack cognitive capacity, e.g. dementia, or where there may be higher risk of harm, e.g. mental health discharge • Lack of written communication having taken place, e.g. telephone only • Specifically relate to patients <18 years • Source not written or published in English The exclusion criteria posed limitations on the review; children under 18 (where the parent would often be the letter recipient), patients with particularly specialised communicative needs (e.g. patients without capacity) or where the intervention may have a higher potential risk of causing harm (e.g. psychiatric discharge documents) were excluded. The communication needs of some of these patients may be more complex and variable within and between groups and therefore was not possible within review scope. The first exclusion criterion states patient discharge communication to those other than GPs or equivalent (e.g. family or community physicians) was excluded. This is because the review specifically focussed on discharge communication to GPs and patients rather than referrals or care-handovers. Furthermore, the review aimed to develop a theory for patients receiving discharge communication and inclusion of hospital-hospital discharge may have reduced clarity and produced a less focussed theory.
Once KW had screened the documents by title and abstract, second reviewer EM screened a random 10% test selection; this proportion was selected following Wong et al. (24) . Inter-reviewer agreement was set at kappa measure K≥.8 (26) . A result K<.8 would require all documents to be second screened. Inter-reviewer agreement was calculated as sufficient (K=0.82). In the first screening phase, 611 duplicates were removed and 2,341 documents excluded; this left 161 documents.
The full texts of these 161 documents were then screened, primarily for relevance (19,20) by KW, with EM screening a random 10% sample. Inter-reviewer agreement was again sufficient (K=0.92). Eighty eight documents were excluded at this stage leaving 73 for inclusion.
In addition, hand-searching of bibliographies, 'cited by' searching, and contacting experts was undertaken. This identified a further 30 relevant documents, creating a total of 103 documents. Supplementary file 4 provides the final document list. The selection process is summarised in Figure 3.

Data extraction and analysis (step 4)
A hybrid approach to data extraction was undertaken (24,28,29) . This allowed extraction of both descriptive document characteristics and annotation of CMOC ideas for synthesis and integration into the PT (19,20) . A data extraction form was designed iteratively to record pertinent document details. Final columns included: author(s), year, geographical information, healthcare system, design aim, no. of participants intervention, clinical speciality, inclusion and exclusion criteria, findings/conclusions, rigour/quality assessment (19,20) , topic focus, form of discharge communication e.g. discharge summary, participant mix, staff mix, and relevance score (19)(20)(21) .
Documents were also annotated in NVivo for CMOCs and PT ideas. Annotations were guided by the initial PT devised in step 1.

Data synthesis (step 5)
During step 5, data and annotations of PT ideas and CMOs were consolidated. A realist analytic approach, following the work of Pawson (19)(20)(21)30) , was used to interrogate the theory during data synthesis. Pawson (19)(20)(21)30) presents several different frameworks for synthesising data evidence. We selected the framework (20) entitled "synthesis to consider the same theory in comparative settings", which involves five analytical strategy steps. This framework was chosen as it assumes theories sometimes "work" and "do not work" according to the particular setting; Pawson et al. (20) describe this as 'aim[ing] to make sense of the patterns of winners and losers'. Hence, this framework is suitable for the research questions which focus on cause and context of positive outcomes "winners" and negative outcomes "losers". Thus, data synthesis was grounded on the assumption that the outcomes of the intervention may differ according to context.
The following realist analytical strategy steps (20,28,31) were undertaken simultaneously: 1. Juxtaposition of data sources -align sources to build upon/clarify each other 2. Reconciliation of data discrepancies -explore reasons for data disparities 3. Adjudication of data -data quality consideration of trustworthiness/relevance 4. Consolidation of data -inference of Mechanisms for outcomes 5. Situation of evidence -consideration of intervention settings Data synthesis using the analytic strategy 'juxtaposition of data sources' was achieved through utilisation of NVivo 'nodes'. Sections of text were annotated, and coded as nodes. The nodes were named according to ideas or concepts around the programme theory and contained sections of text that were used to build CMOCs. NVivo node coding resulted in 19 nodes seen in Table 1. During, 'reconciliation of data discrepancies' (19,20,24) and 'adjudication of data' (19,20,24) , the data coded within NVivo was used for scanning and comparing data to identify disparities. Adjudicating and situating evidence was important to reconcile discrepancies (19,20,24) . We interpreted the data coded within each node and judgements were formed as to which sections of text might be functioning as contexts, mechanisms or outcomes. We then made assessments about what the CMOC might plausibly be for each CMO based on the data within each node. Where relevant, we also drew on data contained within other nodes to build CMOCs. Following this process, a CMOC table was constructed (see Supplementary file 5) for consolidation of data.
After table completion, following Pawson's framework (20) , it was important to make sense of the "winners" and "losers". CMOCs were primarily labelled according to how evidence was reported in the included documents, such as whether the outcomes were described as desirable or beneficial. Where evidence was limited or outcomes were not clearly described or evaluated, the research team interpreted what data were available and formed judgements about these CMOCs based on content expertise in order to generate "positive" and "negative" labels. CMOCs were not limited to one per document or one per patient experience. Thus, multiple outcomes and CMOCs could be annotated for a single experience; this exemplifies the complexity of the intervention under scrutiny.
Notably, there were a greater number of CMOCs relating to positive outcomes than when the intervention does not work.

Programme theory refinement (step 6): Patient and public involvement
Review step 6 was to consider stakeholder perspectives to test and refine the PT in light of the synthesised data (20) and to assess whether the PT aligns with real-life experiences (18) . We invited comment from local policy makers and health service commissioners, GPs and a patient and public participation group. Groups were selected according to convenience and accessibility through University links. They were invited to suggest refinements to the PT in an entirely voluntary format, and all comments were anonymised. Formal ethical approval was not required (23) but informed involvement was sought.
The evidence covered a wide range of specialties. Most specified inclusion of adult patients only (over 18 years) but often did not detail the exact patient ages in the write up; a few studies focussed on elderly patients. Information relating to patient demographics e.g. gender, was often not found in the sources and hence these were not summarised. Many sources instead focused on the speciality under consideration in the document and clinical presentations of interest to that speciality e.g. ECGs (32) . Participants who were staff included medical students, doctors of all training grades, nurses, GPs, non-specified hospital staff, and non-clinical staff. However, the majority of documents (66%) either did not provide staff participant details or they were irrelevant e.g. guideline document, no participants. The type of discharge communication varied: direct copies (48%), discharge instructions (13%), pictures (1%), personal discharge packs (1%), personalised letters (13%), information booklets (9%), multiple types of discharge communication (7%), and other (11%). Where the sources came from showed some variation such as Department of Health archive (3%) and conference listing (5%) but the greatest number of sources were from journals (68%).

Quality and document rigour
The findings were considered in light of the quality of included documents. During data extraction, documents were quality appraised for rigour and evaluated for relevance (19,20) . The concept of rigour is defined as 'whether the methods used to generate the relevant data are credible and trustworthy' (18) . Relevance and rigour were scored on a scale from very low to very high and factors such as document type (e.g. opinion piece or scientific trial paper) were considered. Documents were not excluded solely based on rigour as extracts of documents with a lower quality score may still have valid contributions (20) . The quality of evidence varied, with 53% of sources graded as medium or above for relevance and 80% for rigour. Information relating to setting and context was often limited.

Context-mechanism-outcome configurations
The following section provides an overview of theories in the form of a narrative of how patients receiving discharge letters does or does not work, as informed by the evidence reviewed. The sub-headed themes emerged during data interrogation and consolidation although many acted as "nodes" in earlier annotation and coding (see Table 1). Sections contain references to CMOCs, quotations from data texts, and references. Quotations have been chosen which illustrate the described theories and highlight key elements of CMOCs. The full table of 48 CMOCs is found in Supplementary file 5.
Evidence relating to some aspects of the PT was limited, particularly in relation to negative outcomes, intervention costs, current clinician views, impact on doctor-patient relationships, personalised patient letters and patients not receiving any intervention. Evidence was also thin in relation to data disparities. Although, context and outcome information was generally well supported, mechanisms were frequently omitted. Where possible, based on the evidence and research team expertise, we inferred reasons for disparities and what the likely mechanism(s) were within any CMOC.
Two sources with the same lead author suggested training medical students on writing patient letters can help produce letters that are more meaningful to patients (66,86) [CMOC13]. However, the evidence around training in relation to the intervention was limited and needs further research.

Personalised or patient-directed discharge letters
Producing a letter which is comprehensible and useful to both GPs and patients has been recognised to be an issue (17,48,78,87) . Patient-directed or personalised patient letters have been proposed [CMOC24, CMOC36]. Patients often rate these letters positively (6,43,88) and this may heighten satisfaction (43) , and improve understanding (6,89) : "Simplifying written communication has also been shown to improve patient comprehension." (6) (p.855) However, personalised letters have the potential to lead to resource consumption (45) [CMOC25], staff time depletion (33,45,89) and patient anxiety that they have been given different information to their GP (89) [CMOC26]. For these reasons, further research which weighs the benefits of personalised patient letters against the drawbacks and costs is needed.

Patient to deliver letter
The context of patients delivering letters seems to have few reported positive outcomes. Posting and electronic transferral of letters may be preferable as: "It is not considered good practice to send the discharge summary home with the patient as there is no guarantee that the information will be passed on to the general practitioner" (90)

Dictating letters in front of patients
Evidence for this concept was somewhat thin. One study suggested that dictating letters in front of patients can make patients feel less in need of a copy of the letter (8) . Another paper suggested this practice may also provide a context that triggers patients to challenge inaccuracies, improving letter quality (91) [CMOC22, CMOC30]: "The content of letters to GPs is sometimes incorrect and this may be remedied by dictating the letter in front of the patient." (91)

Confidentiality
There are concerns and legal implications surrounding potential confidentiality breaches associated with patients receiving letters, particularly when they are sent out in the post (36,41,49,51,54,87,92) . One recent paper (49) (2013), which looked at confidentiality, continued to stress risks around postal communication and the importance of secure information transfer: "There is a substantial risk of breaching patient confidentiality when distributing correspondence by post. A well-designed security arrangement is therefore required to ensure the safety of confidential information." (49)  Some documents (2,36,49,54,56) suggested ways to reduce potential risk of confidentiality breach through communication platforms and the processes involved in sending letters e.g. verifying patient contact details before sending letter (36,59) [CMOC3, CMOC27, CMOC28].

Patient harm
Patient anxiety or "harm" in general are often cited as reasons for clinicians not wanting to copy letters, particularly in "bad news" settings (14-17, 37, 45, 48, 57, 93) [CMOC6]. Letter inaccuracies can cause concern leaving patients feeling confused or anxious (33) [CMOC19]. Nevertheless, the letter can reassure the patients their problems are being handled (51) [CMOC46] and initial anxiety can settle or be nullified by the usefulness of the letter (7, 42-45, 56, 86, 92) [CMOC37]. Moreover, one study (40) published in the Lancet in 1991 suggested patient letters in "bad news" settings may be more useful than "good news": "Patients who had received bad news found the letter significantly more useful in helping them to understand and remember what they had been told during the consultation than did patients receiving good news… almost half the patients receiving bad news found their letter distressing to some extent; however, with 1 exception, all patients were pleased to have received it." (40) (Pp.924-925) Although the above paper was published in 1991, we found no recent evidence or system changes to dispute the notion that "bad news" letters may be of particular use to the patient. Hence, despite risk of initial "harm", "bad news" letters should perhaps not be avoided.
Practical and feasible suggestions were found in some documents for minimising harm or anxiety: not copying letters with information not previously disclosed to the patient (2,3,14,56) [CMOC38], abstain from use of value judgements e.g. pleasant lady (37) [CMOC12], potentially avoid or carefully consider copying letters where there are 'problems of privacy at home' and/or where the patient lacks capacity (2) [CMOC20], and checking the patient consents to a letter (56) [CMOC41].
Confusion around 'letter comprehensibility' and lack of 'patient understanding' were the commonest clinician reservations relating to the intervention (11,16,34,36,37,46,48,51,87,89,98,99) . However, as covered in the comprehension section, patients are often reported as understanding their letters (7, 8, 14, 15, 33-37, 44-46, 56, 57) and furthermore they tend to express strong preference for receiving such letters (7,8,14,35,36,41,44,45,(49)(50)(51)(52) . Thus, it may be inferred from the evidence that patient understanding of letters is possibly higher than clinicians' perceive (34,45,57) . The following from a recent (2016) abstract (46) concisely summarises an example of patient and clinician view disparity: "While some oncologists assess the copy letters as inappropriate for supplemental patient-oncologist-communication, breast cancer patients regard this tool as predominantly gainful. Oncologists appear to stick to their traditional perspective which perceives the copy letter mainly as a communication tool from doctor to doctor." (46) Notably, much of the evidence reporting clinician views was published from 2002-2008 and current evidence on clinician perspectives remains limited. Moreover, although sources occasionally referred to conflicting clinician views, information on why attitudes differ was thin. Overall, better understanding of current clinician views on copying discharge letters to patients is required. Further research should address reasons behind different viewpoints to include patients and practitioners.

Cost and resources
The estimated costs associated with the intervention varied (16) but this must be considered in the context that included documents spanned a wide time range and thus factors such as inflation need to be considered. In addition, robust health economic analyses were not found in the included sources. Documents (16, 17, 33-35, 37, 41, 44, 50-52, 54-56, 87, 91, 101) referred to "cost" or financial implications [CMOC25] of sending letters in different ways such as use of consumables (17, 33, 34, 50-52, 55, 101) [CMOC10], and secretarial (16,17,33,34,37,44,51,55,56,101) [CMOC10] and clinician time required (17,33) . A few sources (2,17,35,37,45,56,102) , including guideline documents and research papers, suggested that benefits were such that associated costs were minimal, or even reduced by patients being more informed from receiving discharging communication [CMOC7, CMOC25, CMOC42]. However, as many of these views were based on personal comment or studies with weak methodologies, the true cost consequences remain unknown.

Stakeholder perspectives
As detailed in step 6, the final review step was to refine the programme theory through stakeholder perspectives. Three groups were consulted: local commissioners, GPs and service-users/patients. Stakeholder involvement took the form of group discussions and email correspondence. As the PT was continually being developed throughout the review process, stakeholders commented on the most recently developed PT at the time of their involvement. Groups were relatively small; due to feasibility it was not possible to achieve diverse and representative group samples.
Group discussions were centred on the programme theory; members were encouraged to critique and feedback on the PT diagram. This included concepts not covered or explored in detail in the PT diagram such as: the importance of comprehensible language and terminology, difficulty and problems retaining verbal information only e.g. following use of anaesthesia, patient choice of receiving letters, illegibility of handwritten discharge communication, critical context of prior patient communication of a high quality to increase likelihood of understanding discharge letters, issues around personalised patient letters considering NHS resource availability, and concerns around writing a letter which meets the needs of both GP and patient. The commissioner and GP representatives emphasised the importance of patient safety and that this should be central to best practice recommendations. In addition, the patient group reported reading a letter about themselves written in third person was peculiar. The patient group felt patient letters were very important for patients taking responsibility for their health in line with the NHS promotion of patientcentred and patient-led care.
Several different members across the various groups commented that in practice, patients do not always receive their letters, despite this process being recognised as best practice. Recommendations were suggested to rectify this by the commissioner members to include: clinicians should assume when writing letters that they could be made available to the patient, early clinician and student training in good letter writing and record keeping, and that hospitals should support the initiative e.g. quality improvement activities and audits.

Cycling of review steps
As a realist review is an iterative process, steps may be repeated. As described in step two, new publications were followed and consulted for evidence but provided no new or conflicting PT knowledge. Thus, it was deemed that "theoretical saturation" (19,21) in accordance with Pawson's realist review methodology (19)(20)(21)(22) was attained and no further searching or step cycling was required.

Resultant programme theory
The PT was systematically updated to produce a resultant PT following review steps 1-6 ( Figure 4). This still shows two main channels for CMOCs: patient copies of letters and patient personalised letters. There remained limited CMOCs for where patients do not receive letters, due to the paucity of evidence available. Contexts for when the patient does receive their letter(s) were condensed into an aligned grouping of five key contexts for when  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y the intervention may be theorised to work and four key contexts for when the intervention may be theorised not to work. The feasibility of providing a personal patient letter was updated on the PT; findings from both stakeholder feedback and data synthesis suggested personalised letters may currently be more feasible in private or insurance-based healthcare settings than in the NHS. In addition, the box of contextual influences was deleted, and the points integrated into the overall diagram. Patient outcomes throughout the PT were simplified and clarified (e.g. the outcomes such as 'empowered patients' and 'reduce patient anxiety' were simplified to the outcome 'improved patient well-being').

Statement of principal findings
This review of 103 sources summarises and expands upon existing evidence by moving beyond "benefits" and "drawbacks" of patients receiving letters alone, and considering contexts of when as well as how the intervention works. Although the review focuses on the UK health system, our use of realist review has enabled identification of findings that may be transferable to other healthcare settings. RQ1 asked about positive and negative outcomes of the intervention. Positive outcomes include: increased patient satisfaction (10,16,35,(39)(40)(41)(42)(43) , improved doctor-patient relationship and trust (14,16) , heightened patient knowledge (43) , improved letter and record quality (2,56,91) , and reduced anxiety (65) . Negative outcomes include patient queries (55) , confusion (48,51) , and anxiety (15,43) .
RQ2 enquired after the important contexts for triggering these outcomes. Important contexts for positive outcomes include: letters written in plain English with minimal abbreviations (56) , lay explanations or simplified terms in brackets for medical jargon (16,56,83) e.g. myocardial infarction (heart attack), written information provided alongside verbal explanation (67) , no new information in letter (2,14) or value judgements (37) , letter translation (39,76,77) where relevant, training clinicians on letter writing practice (2,66,84,86) , use of pictures and glossaries where relevant (16,69,(80)(81)(82) , letters only given to patients who choose to have them (16,33,56) , and where there is no identified risk of harm (2) or confidentiality breach (56) . Important contexts for negative outcomes include: patient choice not acknowledged (56) , only verbal information provided (44) , letter involving terms and a style that is too advanced for patient to comprehend (78) , and letter sent without verifying patient contact details (36,54,59) . This review has produced two key findings, which are important but not wholly surprising. The first is that the reviewed evidence indicates that patients value their discharge letter and their understanding of them is possibly greater than clinicians perceive (34,45,57) . However, reasons behind patient and clinician perceived comprehension discrepancies were unclear. It is important to situate the first finding in terms of the study exclusion criteria and participant diversity across the evidence reviewed, for example, it is likely that patients who participate in research on this topic have a greater level of interest and literacy than those who did not participate. One or a number of demographic groups not involved in the studies, either by choice not to participate or by exclusion, may have accounted for a portion of those who clinicians perceive to have low understanding. Thus, evidence for low patient understanding was limited, and this requires further research. The second key finding is that in a number of contexts, patients expressed preference for receiving correspondence (7,8,14,35,36,41,44,45,(49)(50)(51)(52) . Patients can continue to use the letter(s) to refer to beyond discharge (16,35,37) , as a medication list reminder, and to share with friends/relatives as desired (15,35,37,89) . Nevertheless, patient choice should still be acknowledged as the review did find evidence that not all patients want their letters; a practical way of addressing this would be to check with the patient that they want a letter in the first instance (7,16,54,56) .

Review limitations
For this review we followed the RAMESES quality and publication standards for realist reviews (105,106) . Quality assessment and analysis is to a degree dependent on reviewer skills and reflexivity (107,108) . Furthermore, analysis and inferences were 'subjective and interpretative' (109,110) . However, because the steps we have taken for this review are transparent, other review teams can see and make judgements on result plausibility.
Due to lack of time, it was not possible to involve all stakeholder groups who may be connected with discharge communication. Nonetheless, the review had a specific focus on "receiving" discharge communication and thus stakeholders were targeted who were closely associated or involved in policy of discharge letter receipt.
The resultant PT is limited by the quality and content of evidence reviewed. Some of the evidence found in sources was markedly thin, particularly in relation to costing information, recent clinician viewpoints, personalised letter copies, and influence on the doctor-patient relationship. Furthermore, there were a greater number of CMOCs relating to positive outcomes than negative outcomes, that is, when the intervention does work than doesn't. This may be rationalised by publication bias towards positive findings. Additionally, the binary distinctions between positive and negative outcomes, that is, when the intervention does and does not work, may have imposed oversimplified CMOC labels. "Positive" and "negative" labels were based on evidence presented in the documents reviewed; at times a degree of subjectivity was involved in this process. Although these binary labels (positive/negative) may have oversimplified some CMOCs, we felt the usefulness of clear distinctions between when the intervention was interpreted to "work" (and not) outweighed the drawbacks of this method.
CMOCs for patients not receiving letters (nil intervention) were thin. Consequently, these evidence limitations constrained the detail available in the resultant PT in these areas. Additionally, not all mechanisms could be inferred from the data resulting in some visible mechanism "gaps" within the CMOC table (supplementary file 5).
The review and PT are not exhaustive but this is not the intention of a realist review (111) . Given time and resource constraints, the review was limited to adult patients who had been discharged from general hospital settings, and other patient groups were excluded. Furthermore, the PT is limited by the representativeness and diversity of the patient groups within the sources reviewed. Much of the evidence was drawn from small scale studies conducted in single settings, and even within these there is likely to have been participation bias which will have resulted in the views of ethnic minorities, patients lacking literacy and other hard to reach groups being under-represented.

Suggestions for future research
The PT offers a useful starting point for future research and should be useful and practicable for informing policy and guidelines. Further research is needed to explore the relevance of the PT to groups, such as children and adults being discharged from mental health services, who were excluded from this review and to those, such as hard-to-reach groups, who may have been under-represented in the evidence included in the review. There is also a need for research to define the cost benefits of copying patients into discharge letters in order that the importance of this topic and the consequences of poor practice are recognised by policymakers, managers and professional bodies. Potential barriers such as clinician views and the current limited available clinician training on letter writing should be addressed; research and evaluation is needed to inform how this can be effectively achieved. Since patient and clinician views were sometimes conflicting, a study which parallels both views alongside the same patient cases to understand reasons for any  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y discrepancies would be useful and may provide valuable insights. This is the topic of a PhD that is currently being undertaken by the lead author, and will be reported in due course.

CONCLUSION
The resultant PT forms a basis for explaining how, when, why and for whom this intervention does and does not work. The resultant PT makes suggestions for how best practice of patients receiving discharge letters may be improved to enhance the provision of patient-centred care. Evidence for some aspects of the PT was rather limited, indicating a need for more research.
The key findings are that the value patients place on discharge letters and their understanding of the letters' content is possibly greater than clinicians' perceive, patient choice is instrumental to increasing the likelihood of desired outcomes, and that clinician views may act as a barrier to wider practice implementation. This could be addressed through clinician training and organisational initiatives which guide, mandate, and monitor the intervention. Without such organisational support, it is unlikely that current processes will be consistently improved given the barriers identified in the review.
In conclusion, this review describes how the intervention of patients receiving their discharge letters may work to increase the likelihood of positive effects and reduce potential negative effects.