Retrospective case note review

Discharge letters for 250 patients were examined (54% male; 46% female). Of these, 248 (99.2%) were inpatient discharges and two were day therapy discharges (0.8%); 249 (99.6%) were medical physician discharge letters (typically sent between doctors), and one was a nursing discharge letter (typically sent between nursing staff). Most discharged patients were aged between 50 and 89 years (n = 220, 88%) and of white ethnicity (n = 226, 90.4%), with a range of 80–100% across hospices (see Additional file 4). The total number of words in the discharge letter corpus was 90,598 with 4048 different word types. The 100 most frequent content words (i.e. nouns, verbs…) for the corpus are in Additional file 5 alongside concordance samples; the top 25 content words are in Table 2.

Table 2 The 25 most frequent content words in the discharge letter corpus (lemmatised)

Most patients had a diagnosis of cancer (n = 205, 82.0%); for 199 of these patients, cancer was the primary diagnosis (79.6%) with a range of 70–96% across sites. 45.6% of patients (n = 114) had co-morbidities documented in the discharge letters. The documented reason for admission for most was symptom management (n = 232, 92.8%). Admission length ranged from 1 to 240 days (median 17 days). The main reasons for discharge given were symptoms “managed/resolved” (n = 188, 75.2%), and/or that the “patient wishes to die/for care at home” (n = 93, 37.2%) (see Fig. 1 for reasons for discharge results).

Fig. 1

Documentation of reasons for discharge in discharge letters

Patients’ needs described in discharge letters included those that led to admission, those managed during the admission, and those which were ongoing at the point of discharge. Documented needs covered physical, social, psychological, and spiritual domains (see Fig. 2); this included 84 (33.6%) with one domain, 115 (46.0%) with two, 45 (18.0%) with three, and two (0.8%) with all four.

Fig. 2

Broad domains of need in hospice discharge letters

Four patients (1.6%) had needs separate from the domains, described as functional needs for two, with the other two admitted for respite care. Other patient needs included: “capacity or communication needs” (n = 27, 10.8%), “functional needs” (n = 124, 49.6%), and “changing/dynamic needs” (n = 126, 50.4%) such as change in condition (n = 86). Carer/family needs were coded for 31 letters (12.4%). Additionally, the patient advanced care plan (ACP) was noted as discussed or updated in 222 discharge letters (n = 88.8%). Further data extraction results are in Additional file 6.

Focus groups

Five focus groups took place, one for each participating hospice, in November 2021. Groups took around 1 h; recordings ranged from 52 to 59 minutes. In total, 25 participants took part with 3–6 per site, including doctors (n = 15), nurses (n = 4), and allied healthcare professionals (n = 6). Six themes were identified; these are listed in Table 3.

Table 3 Themes identified during analysisTriangulated findingsStructure, content, and transmission of discharge letters

Discharge letters varied in format from structured or templated letters to free text or narrative-style letters, as well as combinations of both. The mode of sharing discharge letters varied, with references to electronic and hard copies of letters as well as synchronous and asynchronous communication (see Table 4). In many cases (n = 158, 63.2%), the letter was sent to the general practitioner on the day of discharge (n = 113) or < 48 hours (n = 45); there were no cases where this exceeded 3 weeks. However, some participants did identify timing of letters as an area for improvement:

“I think what we can improve on is how quickly we send the letters, ‘cause I think especially for medicines it’s a slow turn around from a [general practitioner]. And to getting it from your pharmacy and things like Medi boxes, we don’t send them straightaway.” [PCS4].

Table 4 Variability in transmission of discharge communication

The content and depth of detail in the letters varied across the sample, both within and between hospices, something that was acknowledged by focus group participants. For example, several participants acknowledged that letters were often idiosyncratic and dependent upon clinician preferences and experience:

“I think every medic has their own style essentially I am very wordy person. I would write probably 3 pages letter and it would probably include psychological distress. Other people may not add that. So I think depends on how experienced you are… It’s a personal style of writing the letters.” [PCS17].

“We come with our own sort of prejudice and bias in what we think is important…if you have a particular interest in sort of the psychology aspects of things then you’re more likely to like add that in...” [PCS7].

One participant outlined the function of the discharge letters and their purpose in a broader sense, “This is meant to be a holistic letter that describes a person so that the receiving [general practitioner] or whoever reads it get a full picture of this persons’ needs” [PCS18].

Despite variation, there were some commonalities in letter content; for example, the patient’s diagnosis was extracted for all letters, and corpus linguistics analysis found the following content elements occurred in over 100 letters: “care” (244 letters); “plan(s)” (178 letters); “follow up” (130 letters); and “management” (105 letters). “Care” frequently co-occurred or collocated [63, 64] with “preferred” (log-likelihood = 4221.6) and “place” (log-likelihood = 3821.7); see concordance line sample in Table 5.

Table 5 Sample of 10 random concordance lines for “prefer”

A recurring pattern was “preferred place of care/death” for the patient. Similar constructions to “death” were found for “die/dying” with references to preferred location of death and patient/family knowledge of terminal nature of disease (“she knows she will die from this”) in addition to any issues or barriers to advanced care planning e.g. fear (“fear of dying/being a burden and very anxious which makes all advance care planning discussions difficult”).

The reasons for discharge were not always clear in the letters, and sometimes required interpretation by the clinician data extractors. As one extractor said,

“the reason for discharges was really commonly not specified and…there are quite a few times that I was left thinking I don’t know whether their symptoms have really been sort of managed um, but they’ve been discharged.” [PCS23].

Ways to improve content and quality of discharge letters included, for example, enhanced training for junior doctors and checking of letters by senior clinicians, “I think junior doctors need guidance on how to write discharge letters and what the discharge letter is… there’s not a formal process saying that a senior doctor needs to check it…” [PCS25].

Several focus group participants summarised what discharge letters elements they thought were important or necessary for a “good discharge letter”, for example:

“So diagnosis, reason for admission, what we’ve done to them while they have been here, what their future holds so advanced care planning, preferred choices, what their status is and any social concerns maybe” [PCS1].

However, some participants felt letters are generally comprehensive and content improvement should focus on how to highlight pertinent and urgent information to general practitioners, such as use of bold font and/or headlining key actions and information:

“I’ll do like the meat bulk plan and then action, yeah for [general practitioner] and I put that in bold so if it’s describing something so that they may not read all of that, but at least they’ll have a look.” [PCS3].

Overall, we found that there was a range of ways used to communicate discharge information, and that there did not appear to be a consensus on the structure and content of letters that is most helpful for primary care teams.

Communicating patient (complex) palliative care needs

Specialist palliative care seeks to provide holistic care to patients. Nevertheless, there was a marked focus on physical needs (n = 246, 98.4%) in the discharge letters (see Fig. 2), with pain the most frequent (n = 194), followed by constipation (n = 87), breathlessness/secretions (n = 79), fatigue (n = 61), and nausea/vomiting (n = 54). The second most documented domain of need was psychological/emotional needs (n = 116, 46.4%) such as anxiety (n = 62), depression (n = 26), grief (n = 8), and anger (n = 7). Social needs were the third most documented (n = 89, 35.6%), for example, housing (n = 20) and social isolation (n = 11).

Spiritual needs were the least documented domain of need; this was indicated both in data extraction results (n = 6, 2.4%) and corpus results – “spiritual” had only one hit and the following terms did not occur in any letters: “pastoral”, “religion”, “meaning of life/death”, “existential”. Reasons for low recording of spiritual needs in letters were described by focus group participants as patient and clinician preferences, it was not thought relevant, and perceptions that spiritual needs were too personal to include in the letter. Issues with confidence and understanding were also raised: “It doesn’t surprise me that we don’t reference spiritual needs that that much…I think it’s an area we recognize we’re not great at addressing in person, let alone writing.” [PCS7].

Reasons for discharge included recognition that some discharges occur because referral has taken place too early before complex palliative care needs had developed that necessitated intervention:

“I think the commonest reason for discharge there is, so discharging them from the whole service, is because the referral has come in very early in the patients illness and actually they don’t have any specialist palliative care needs…there’s been a misunderstanding by the referrer… as to what our role is and so that the kind of traditionally held role that you know we are…[going to] provide emotional support to the patient. Actually now we discharge people that just need that because that’s not considered a specialist palliative care need.” [PCS25].

Several participants suggested that all hospice patients should have an element of complexity in their palliative care needs, and yet participants found “complexity” was “…really hard to define… it might be a number of variable factors that make it complex…” [PCS8]. In order to directly acknowledge when “complex(ity)” applies, some participants outlined that letters should use more explicit and clear language:

“They probably did have complex pain, and it’s maybe our annotation of that. It makes it look as though their pain wasn’t a complex symptom, and I believe our discharge letters are freestyle. It’s a narrative, as opposed to ticking complex pain or whatever, so it’s a learning point for us to not underestimate or not describe things as less than they are.” [PCS18].

Variability in use of “complexity” was reflected in the discharge letters. The corpus linguistic analysis found that there were only 16 hits of “complex(ity)” across 14 letters (see Table 6). As seen in Table 6, there were differing contexts for the word “complex(ity)” in the letters. These included references to complex needs and complex medication as well as symptoms.

Table 6 All concordance lines for “complex(ity) in hospice discharge letter corpus

The most frequent content collocate for “complex(ity)” was “pain”. In data extraction forms, documentation of complex pain varied; pain (n = 194) was coded by data extractors in 77.6% of discharge letters with a range of 74–86% across hospices. “Complex pain” was coded for 74 patients (29.6%) with a range of 18–66%. For 73 of these, one or more of the complex pain markers (e.g. drug sensitivities; see Additional file 1) had been coded, although combinations of markers varied. This lack of agreement on what constitutes complex pain was reiterated through the differing viewpoints expressed in focus groups, with some participants suggesting it is a “total pain” or involves a psychological element, and others stating that complex pain is that which cannot be managed in the community and therefore is “whatever pain brings you to the hospice.” Notably, the latter view was not reflected in data extraction results as “complex pain” coding was lower than that of “pain” for all hospices. A few hospice staff members suggested that “complex pain” is context-dependent, “because…of the situation” [PCS12], or influenced by the perception of the treating clinician:

“…the person came in with pain and we threw everything at them and we still couldn’t manage it. That I think that to us would be complex pain, whereas to a [general practitioner] it’s maybe they’re on four or five medications. I mean, all the patients we admit with pain are on that, so I suppose it’s your context of complex pain.” [PCS24].

Overall, the results suggested a lack of consensus on what constitutes “complexity”.

Responsibility for future and ongoing care

A sample of lines for the word “discharge” from the sample of discharge letters is found in Table 7 below.

Table 7 Sample of 10 random concordance lines for “discharge”

Differing perspectives were expressed in the focus groups in relation to what “discharge” is, and this was associated with ambiguity in what future care in the community entailed. Terms surrounding community care provision were not frequent in the corpus (e.g. nil hits of “primary care” and “[general practitioner]/GP” occurred in 28 letters); the exception was the word “community” itself (n = 242 in 154 letters). Instances of “community” were typically either statements that the patient had been discharged to the community, or relating to management plans and instructions for the community team, e.g. “This will need to be monitored in the community.” Such language does not necessarily make it clear whether follow-up was to be undertaken by the hospice community palliative team, the primary care team, or both. In the focus groups, some participants posited that primary care undertake this, e.g. “…patients now will be receiving support from [general practitioners]” [PCS25] and others suggested differently, “although they’re being discharged, they’re not…I’m imagining it’s vanishingly rare that we ever discharge to the [general practitioner]” [PCS7]. One participant summarised: “I just feel like there may be that disconnection with within the community, but it’s hard to know because we are not out there.” [PCS20].

In summary, there was a diversity of views and a degree of ambiguity on who was responsible for the patient upon discharge.

Sharing discharge letters with patients and carers

Data extractors could only code whether the discharge letters had been shared with the patient or carer if this was explicitly noted in the letter itself, or recorded in the patient’s notes. Documentation review found that verbal information had been provided to the patient/carer in 130 cases (52%), and that discharge letters had been provided to patients in 52 cases (20.8%) with one instance of the carer receiving the letter (0.4%). Of these 53 instances of patient/carer letter receipt, there were only 3 letters which were identified as explicitly noting this. Corpus searching found only one hit of copy(ies/d), which supports the finding that most letters do not record that a copy was provided to the patient/carer. In all cases where patients/carers were identified as having received the letter, this took place on the day of discharge.

Several focus group participants (for three hospices) noted that letters are not usually sent to patients, e.g., “none of the 50 did [get sent to the patient] and I spoke to our admin about that, and they said that they would only send one if we asked for one.” [PCS2]. Another focus group participant explained that patients receiving discharge letters is not reliably undertaken, “I’m sure a while ago we were told to offer them. But then I think that fell by the wayside” [PCS24]. Additionally, one focus group raised consistency issues with recording of sending letters to patients, “we’re not necessarily good at capturing that” [PCS7].

There appears to be inconsistency within and across hospices, both in providing patients with discharge letters and in documenting what has been provided. During focus groups we explored why this might be. Some participants expressed concerns or outlined difficulties in writing a letter that “covers everybody’s’ needs”, particularly relating to aspects that could be perceived as sensitive or distressing for patients, such as future care planning:

“…say for example, patients haven’t wanted to talk about some end of life issues and then you’ve got to say you have got to sort of communicate that to the community team. But how do you communicate that then without upsetting the patient, if they’ve not wanted to talk about it when they go home?” [PCS24].

Several participants commented on the benefits of patients and carers receiving letters, as well as suggesting that they support this practice:

“…you can speak to people about certain things and it’s like they’ve forgotten or they’ve you know they’ve not really taken it on board and maybe if it was in writing and they had it in the letter then at least they can read over it and see you know what happened” [PCS2].

“I will say the value of a discharge letter for a patient is that they have something, they have a physical copy in their hands such that where communication may be delayed or say - heaven forbid – they get discharged from the hospice and end up in hospital that night. They’ve got an active, up-to-date record and actually a lot of patients feel very empowered with those letters.” [PCS8].

For 48 patients, they received a personalised discharge letter instead of a general practitioner copy; this was due to convention at one of the hospices. This was discussed at the focus groups with suggestions that the language of such letters has to be different or that the content should be abridged, “…not like a detailed letter, that we would send to primary care” [PCS11]. There were mixed views on the concept of patient personalised letters with some participants not supporting this practice within the context of hospice care, “[Ours are] not written like that and I think it’s right that we don’t write them like that” [PCS10].

We found that there was no consensus on whether discharge letters should be shared with patients or carers, nor – if they were shared – what format this might take.