A Systematic Review and Bayesian Network Meta-Analysis Investigating the Effectiveness of Psychological Short-Term Interventions in Inpatient Palliative Care Settings

This paper reviews and summarises the evidence of short-term psychosocial interventions (up to 12 sessions delivered within less than eight weeks) on anxiety, depression, and emotional distress in palliative patients in inpatient settings. We screened publications from the following five databases, Embase, PubMed, PsycINFO, Web of Science, and CINAHL, from their inception to 10 September 2021. The eligible studies included controls receiving standard palliative care, actively treated controls, and wait-list controls. Nine studies met the eligibility criteria and reported the effects of five psychosocial interventions in a total of N = 543 patients. We followed PRISMA-guidelines for outcome reporting and the Cochrane Risk of Bias Assessment Tool for assessing study quality. This paper used the network meta-analysis to compare multiple treatments by providing greater statistical power and the cross-validation of observed treatment effects, using the R package BUGSnet. Compared to control groups, the following psychosocial interventions in inpatient settings showed to be superior: life review interventions were the best ranked treatment for improving anxiety and distress, while the top ranked treatment for reducing depression was outlook intervention. The short-term psychosocial interventions investigated in this meta-analysis, especially life review intervention, are feasible and can potentially improve anxiety, depression, and distress in palliative inpatients and should therefore be offered in inpatient settings.


Background
The need for psychological support is tremendous in patients nearing their end of life. Patients in palliative situations often suffer from depression and anxiety or feelings of hopelessness, demoralisation, loss of control, and loss of quality of life [1,2]. Palliative inpatients (PIP) usually stay in hospital for a short period of time and the condition of PIP often rapidly deteriorates within this short timeframe. Long-term psychological interventions over multiple sessions are therefore difficult to perform, which is why PIP could greatly benefit from brief psychotherapies tailored to their specific needs [3,4]. Contrary to long-term psychotherapies, short-term psychotherapies are typically more goal-oriented and tend to focus on specific challenges that are causing patients the greatest amount of adversity in the present. To date there is no uniform definition or consensus on the length or extent of short-term therapies. According to Hazlett-Stevens, short-term psychotherapies lasted up to 10-20 sessions and were delivered within a timespan of approximately two to four months [5]. Stern

Eligibility Criteria
Studies were selected in accordance with the participants, interventions, comparators, outcomes, and study designs (PICOS) framework [19]. The inclusion criteria were as follows: (1) Participants: PIP of 18 years and older; at least 50% of the study population must be inpatients. (2) Intervention: psychological short-term interventions (e.g., mindfulness-based interventions, LRI, CBT, and related approaches) administered within eight weeks from baseline to follow-up assessment; number of sessions may range from one to twelve. (3) Comparison groups: control groups, e.g., standard care, no treatment, waitlist control, active attention control. (4) Outcomes: anxiety, depression, and distress. (5) Study designs: randomised controlled trial (RCT), clinical controlled trial, or waitlistcontrolled trial.
This systematic review and meta-analysis focused on psychological interventions that are feasible for PIP. We defined psychological short-term interventions as interventions that include from one to twelve sessions delivered over a time interval of a maximum of eight weeks. We also included studies that reported outcomes at a first assessment within the eight-week time interval.
The exclusion criteria were as follows: (1) Studies of complementary therapies such as physical (e.g., yoga, physiotherapy), music, art, or aromatherapy as well as group, partner, or caregiver-delivered therapies. (2) Interventions delivered online and solely by telephone.
(3) Studies written in a language other than English or German. (4) Grey literature, conference proceedings, abstracts, posters, editorials, protocols, unavailable full texts, or studies that report only qualitative results. (5) Studies with insufficient information on outcome data or inclusion criteria (studies comparing only two equal interventions with no additional control condition were excluded).

Search Strategy
A systematic literature search was conducted in the following five databases: Embase, PubMed, PsycINFO, Web of Science, and CINAHL (see Appendix A for search terms used). Records were considered from the inception of the databases to 10th of September 2021. In addition, reference lists of previously published systematic reviews were manually screened.

Study Selection
All records obtained from the databases and hand search were exported to Excel and screened by title by two reviewers (AS and RS). At the initial screening stage, the authors read the titles and abstracts of the candidate studies and decided to include or exclude each study from the review. Any duplicates were removed. Relevant full texts were examined according to inclusion and exclusion criteria. In cases where it was unclear from the title or abstract whether a paper was relevant, a copy of the full text of the study was sought for consideration. Any uncertainty was discussed between the two reviewers (AS and RS) and, if consensus could not be reached, another author assisted with final decision-making (SE or SW). Corresponding authors of eligible studies were contacted if information on inclusion and exclusion criteria could not be obtained from the full text.

Analysis Method
The network meta-analysis (NMA) was performed using the BUGSnet package on R [20]. All NMA models were based on a Bayesian approach through the Markov Chain Monte Carlo (MCMC) simulation [21]. The parameters assessed in the NMA models were treatment effect compared to other treatment arms and the likelihood function was depen-dent on the outcome. Continuous data were pooled as mean difference (MD) with posterior median and 95% credible intervals displayed for the outcomes HADS total, depression, anxiety, and distress. Surface Under the Cumulative Ranking Curve (SUCRA) analysis was performed to rank treatment arms according to their efficacy. SUCRA plot and score are presented in the results. Model selection and goodness-of-fit were evaluated through deviance information criteria (DIC). Adequacy of the model fit was assessed through a comparison of the residual deviance of the models, where a close match between both models was considered an adequate fit. The Brooks-Gelman-Rubin diagnostic was employed to assess the convergence through visual inspection, with a potential scale reduction factor (PSRF) value of <1.05 considered as an indication that the simulation is valid.

Result of the Literature Search
The literature search and study selection are shown in Figure 1. A total of 12,339 records were identified in five databases. In addition, two records were added that were found by hand searching, yielding a total of 12,341 records. Of these, 12,120 were excluded after title screening, leaving 221 records. After removing 96 duplicates, 125 full texts were screened. As Figure 1 shows, 116 full texts were excluded, as they did not fulfil the eligibility criteria, and nine studies were excluded because the authors did not respond to our request for more information about the eligibility criteria or outcomes. Nine studies were included in the final analysis.

Data Collection Process
The detailed information of the selected studies is summarised in Table 2. The extracted sample sizes, means, and standard deviations of the outcomes were assessed by two reviewers (AS and RS) under the Appendix A.

Data Collection Process
The detailed information of the selected studies is summarised in Table 2. The extracted sample sizes, means, and standard deviations of the outcomes were assessed by two reviewers (AS and RS) under the Appendix A.

Anxiety
All included studies that assessed anxiety symptoms used the anxiety subscale of the HADS [29] except for Steinhauser et al. [16], who used the Profile of Mood States (POMS, Ref. [34]).

Distress
Three of the included studies [22,27,28] assessed distress using a single-item rating scale.

Risk of Bias Assesment
The authors RS and AS assessed and independently rated the quality of the studies using the Cochrane Risk of Bias Assessment Tool [35]. The tool comprises seven criteria (see Table 2). It is generally known that the quality criteria "blinding of participants and personnel" is not conferrable in studies that deliver psychological interventions to patients [36]. However, for the sake of completeness we did not exclude this criterion and rated studies that did not fulfil this criterion as "high" (see Table 2). If information was missing or incomplete, the field was judged as "unclear". Table 1 presents the results of the risk of bias assessment. Seven studies [22][23][24][25][26]28,32] fulfilled five out of seven criteria with "low bias". One study [37] had to be excluded, as all criteria were rated as "high risk" or "unclear".

Study Characteristics
The study characteristics of the final nine studies are listed in Table 2. Most were conducted in North America (n = 3; Refs. [16,26,32]), Asia (n = 3; Refs. [22,25,27]), Europe (n = 2; Refs. [24,28]), and South America (n = 1; Ref. [23]). All studies were published in peerreviewed journals between 2006 and 2020. All were randomised-controlled trials, apart from one randomised crossover trial [28]. Two studies applied a three-arm RCT design [16,23]; only two of these three arms (targeted intervention and control condition) were included in the analysis. The total sample size of all studies used for the meta-analysis consisted of N = 543 at baseline. The patients' age ranged between 19 to 96 years and the proportion of female patients varied from 38.10% to 100%. The patients' remaining life expectancy varied between 1 to 24 months. Inclusion and accrual rates across the included studies were described as generally low (with approximately 5-20%) and most studies reported high to very high attrition rates of up to 55% due to worsened medical conditions throughout the study course. Patients in six studies [22,24,[26][27][28]32] were recruited exclusively in inpatient PC settings, while three studies [16,23,25] recruited in both inpatient and outpatient settings. However, these three studies fulfilled our eligibility criteria of at least 50% of inpatients. All patients suffered from life-limiting diseases. Five studies focused on patients with a terminal cancer diagnosis (n = 3 of all cancer types [22,23,27] and n = 2 [26,32] of homogenous cancer diagnoses) and four studies included a minority of patients with other life-limiting diseases [16,24,25,28]. The patients' life expectancy varied between 1 to 24 months [22,27,28,32]. Studies investigated CBT-based interventions (n = 2), short-term LRIs (n =3), mindfulness-based interventions (n = 3), outlook intervention (n = 1), and emotion-and symptom-focused engagement (n = 1; an intervention based on traumafocused therapy and supportive psychotherapy). Five studies reported outcomes on the Hospital Anxiety and Depression Scale (HADS) total, which reports aggregated data on anxiety and depression in a hospital setting. Further outcomes that were assessed by the studies were anxiety and depression, as well as distress (assessed by distress scales and thermometers). For an overview of the different assessment tools that were used to measure the outcomes, see Table 2. To allow for comparison between the different outcome measures, we applied a Bayesian meta-analysis approach.

Anxiety and Depression (as Measured by the HADS Total)
Five studies [22][23][24][25][26] provide data on the HADS total score and include five groups for comparison (control groups included) (Figure 2, Network plot). LRI was the best ranked treatment for improving HADS total score (SUCRA = 81.6), followed by CBT-based psychosocial intervention (SUCRA = 59.3) and waitlist control group (SUCRA = 30.7). Standard PC was the lowest ranked treatment for improving HADS total (SUCRA = 28.3); (Figure 3, SUCRA plot). In pairwise direct comparison, there was no significant difference between the network treatment arms. Hospital Anxiety and Depression Scale (HADS) total, which reports aggregated data on anxiety and depression in a hospital setting. Further outcomes that were assessed by the studies were anxiety and depression, as well as distress (assessed by distress scales and thermometers). For an overview of the different assessment tools that were used to measure the outcomes, see Table 2. To allow for comparison between the different outcome measures, we applied a Bayesian meta-analysis approach.

Anxiety and Depression (as Measured by the HADS Total)
Five studies [22][23][24][25][26] provide data on the HADS total score and include five groups for comparison (control groups included) (Figure 2, Network plot). LRI was the best ranked treatment for improving HADS total score (SUCRA = 81.6), followed by CBTbased psychosocial intervention (SUCRA = 59.3) and waitlist control group (SUCRA = 30.7). Standard PC was the lowest ranked treatment for improving HADS total (SUCRA = 28.3); (Figure 3, SUCRA plot). In pairwise direct comparison, there was no significant difference between the network treatment arms.
To assess the convergence, a Gelman--Rubin--Brooks plot was constructed. It showed that the Potential Scale Reduction Factor (PSRF) was below 1.05, indicating that the simulations performed are valid (Supplementary file S1. Gelman convergence plot). The deviance report showed the contribution of each study arm to the residual deviance, where all points were around 1, close to the equality line, and there were no extreme values, indicating that no study had a high residual deviance (Supplementary file S2. Deviance plot). As for the contour plot, no leverage values were outside the contour of 3 in the REM, indicating that no study is considered poorly fitting (Supplementary file S3. Leverage plot).

Discussion
To provide an overview of the impact of short-term psychotherapies performed by trained healthcare personnel for PIP, we conducted a systematic Bayesian network metaanalysis (N = 543). We defined a "short-term intervention" as one that would rely on a maximum of twelve sessions and a treatment duration of less than eight weeks. The included studies investigated CBT-based interventions (n = 2), short-term LRI (n =3), mindfulness-based interventions (n = 3), outlook intervention (n = 1), and emotion-and symptom-focused engagement (n = 1). Overall, we found a positive effect of short-term psychotherapies on patient outcomes of anxiety, depression, and distress. Furthermore, we demonstrated that they are feasible in inpatient settings.
Considering the results of our meta-analysis, we found that LRI was the best ranked treatment for improving overall HADS scores and for improving anxiety and distress in PIP, while the top ranked treatment for reducing depression was outlook intervention that contains components of LRI. That LRI /dignity therapy was the highest rated treatment overall is consistent with previous review findings that imply that LRI interventions are effective against anxiety and depression in PC patients [38] and improve overall quality of life [39,40]. CBT-based therapies appeared to be the second most helpful intervention for anxiety and depression, while mindfulness-based interventions turned out to be the second-best treatment option for distress. This finding is in accordance with a previous meta-analysis [41] that demonstrated that CBT-based interventions, among others, were effective in reducing symptoms of depression and anxiety in adults suffering from serious illness. In our meta-analysis, the "standard PC" control group achieved good results for anxiety and depression as compared to no care. This was expected according to recent literature, as standard PC usually entails some form of psychosocial support for patients and their relatives [9]. To assess the convergence, a Gelman-Rubin-Brooks plot was constructed. It showed that the Potential Scale Reduction Factor (PSRF) was below 1.05, indicating that the simulations performed are valid (Supplementary file S1. Gelman convergence plot). The deviance report showed the contribution of each study arm to the residual deviance, where all points were around 1, close to the equality line, and there were no extreme values, indicating that no study had a high residual deviance (Supplementary file S2. Deviance plot). As for the contour plot, no leverage values were outside the contour of 3 in the REM, indicating that no study is considered poorly fitting (Supplementary file S3. Leverage plot).

Discussion
To provide an overview of the impact of short-term psychotherapies performed by trained healthcare personnel for PIP, we conducted a systematic Bayesian network metaanalysis (N = 543). We defined a "short-term intervention" as one that would rely on a maximum of twelve sessions and a treatment duration of less than eight weeks. The included studies investigated CBT-based interventions (n = 2), short-term LRI (n =3), mindfulness-based interventions (n = 3), outlook intervention (n = 1), and emotion-and symptom-focused engagement (n = 1). Overall, we found a positive effect of short-term psychotherapies on patient outcomes of anxiety, depression, and distress. Furthermore, we demonstrated that they are feasible in inpatient settings.
Considering the results of our meta-analysis, we found that LRI was the best ranked treatment for improving overall HADS scores and for improving anxiety and distress in PIP, while the top ranked treatment for reducing depression was outlook intervention that contains components of LRI. That LRI /dignity therapy was the highest rated treatment overall is consistent with previous review findings that imply that LRI interventions are effective against anxiety and depression in PC patients [38] and improve overall quality of life [39,40]. CBT-based therapies appeared to be the second most helpful intervention for anxiety and depression, while mindfulness-based interventions turned out to be the second-best treatment option for distress. This finding is in accordance with a previous meta-analysis [41] that demonstrated that CBT-based interventions, among others, were effective in reducing symptoms of depression and anxiety in adults suffering from serious illness. In our meta-analysis, the "standard PC" control group achieved good results for anxiety and depression as compared to no care. This was expected according to recent literature, as standard PC usually entails some form of psychosocial support for patients and their relatives [9].
For symptoms of distress, mindfulness-based interventions proved to be a good treatment option. This finding is in line with a recent review and meta-analysis which found that mindfulness-based interventions appeared efficacious in reducing psychological distress and other symptoms in cancer patients and survivors [42].
Our results provide compelling evidence that life-review intervention is particularly helpful in reducing anxiety and depression in PC inpatients. That LRI can be delivered in as few as two sessions is crucial for inpatient settings, where the condition of palliative patients can deteriorate rapidly, and hospitalisation is of short duration [38]. Furthermore, LRI therapy is particularly cost-and resource-efficient, with relatives also benefiting from the documents the patients leave to them [39,40,43]. We therefore recommend that specialised inpatient palliative care settings invest in the LRI training of healthcare professionals. Ultimately, however, the choice of therapy must always consider the individual preferences of the patient and the patient's needs.
There are some limitations that should be considered for the interpretation of our results. First, two studies that otherwise fulfilled all inclusion criteria could not be included in our meta-analyses due to insufficient data that would have limited the generalisability of our results. Second, risk of bias was high in all studies. This may be due to the use of the Cochrane tool for assessment, which was originally developed for general randomised controlled trials. Even though we used a version that was adjusted for psychotherapy research [36], all included studies performed rather poorly in the assessment. This phenomenon seems to be the norm and was shown in other meta-analyses with similar patient populations [4], as patient populations nearing their end-of-life are particularly vulnerable and clinical trials are usually not feasible under the highest quality standards.
Another critical issue refers to the small number of studies included, especially in the meta-analysis on distress (n = 3). Some of the included studies had very small sample sizes. To anticipate an underpowered analysis, we specifically applied the method of network meta-analysis. However, future studies in this field with higher sample sizes are warranted. According to Do Carmo et al. [23], future studies to be conducted with this population group need to revise the eligibility criteria and make them less restrictive, so that higher participation rates can be achieved. However, as previously discussed, this will further increase the study's risk of bias.
Future reviews and meta-analyses are encouraged to investigate the categorisation of effects in different treatment phases of palliative care, which would enable a more precise overview of the patients' needs in each phase and benefit therapeutic intervention design. This was conducted by Warth et al. [4], who investigated the effect of music therapy in different stages of cancer patients.
Regarding moderating factors, interesting factors such as pre-existing psychological conditions or experiences with/preferences for psychotherapy could not be considered, as they were rarely assessed in the included studies. Moreover, this meta-analysis did not consider which profession delivered the intervention or the training of the therapists. Future research should consider examining these aspects as potential modulating variables. Another possible influence which studies often do not report is PIPs' intake of psychopharmaceuticals and pain medication, especially opioids. PIPs often receive high doses of pain medication and a wide range of psychopharmaceuticals to ease their physical and emotional discomfort. These are control variables of high interest, since their impact is possibly confounded with the effect of psychotherapeutic interventions on anxiety, distress, and depression. With respect to future research, studies on short-term psychotherapies in inpatient PC settings are encouraged to reduce the risk of bias, for example through assessing confounding variables, publishing primary and secondary outcomes in a respective study protocol, and documenting longer lasting effects of the applied interventions. As this paper was focused on mostly inpatient samples (patients of papers [22,24,[26][27][28]32] were recruited exclusively in inpatient PC settings), it is unclear to what extent the results are generalisable to an ambulatory or home care setting. Three of the included papers included a smaller proportion of ambulant patients, or patients who were referred back and forth between outpatient and inpatient settings. Swift referrals from one setting to the other are very common, as often inpatient facilities can keep patients only for a limited time due to cost and capacity constraints. However, with the inclusion criterion of at least 50% inpatients per paper, we prove that the investigated short-term psychotherapies are indeed feasible in the inpatient setting. Future research on the effect of different psychotherapy treatments exclusively on PIP is highly warranted.
With this analysis we open a new discourse which is yet missing in terms of the effectiveness of short-term psychological therapies specifically offered to PIP. To date, shortterm PC interventions applied in a clinical setting are heterogeneous in content and delivery. Furthermore, there is no best practice for treatment for anxiety, depression, and distress in PIP. However, this meta-analysis demonstrates that different short-term psychological interventions for PIP are feasible, are beneficial, and that there is a need to make short-term psychotherapies in hospitals and hospices available to palliative patients.

Conclusions
The findings of this meta-analysis show that it is feasible and beneficial to grant PIP access to psychological short-term interventions. LRIs consisting of only two to four sessions seem to have great potential to help PC patients cope with anxiety and depression. However, results should be interpreted with caution due to the limited number of randomised controlled trials and associated methodological weaknesses. Further rigorously designed randomised controlled trials with psychological short-term interventions are warranted in inpatient palliative care settings.