Priorities of patients with multimorbidity and of clinicians regarding treatment and health outcomes: a systematic mixed studies review

Objectives To identify studies that have investigated the health outcome and treatment priorities of patients with multimorbidity, clinicians or both, in order to assess whether the priorities of the two groups are in alignment, or whether a disparity exists between the priorities of patients with multimorbidity and clinicians. Design Systematic review. Data sources MEDLINE, EMBASE, CINHAL and Cochrane databases from inception to May 2019 using a predefined search strategy, as well as reference lists containing any relevant articles, as per Preferred Reporting Items for Systematic Reviews and Meta-Analyses and Cochrane guidelines. Eligibility criteria We included studies reporting health outcome and treatment priorities of adult patients with multimorbidity, defined as suffering from two or more chronic conditions, or of clinicians in the context of multimorbidity or both. There was no restriction by study design, and studies using quantitative and/or qualitative methodologies were included. Data synthesis We used a narrative synthesis approach to synthesise the quantitative findings, and a meta-ethnography approach to synthesise the qualitative findings. Results Our search identified 24 studies for inclusion, which comprised 12 quantitative studies, 10 qualitative studies and 2 mixed-methods studies. Twelve studies reported the priorities of both patients and clinicians, 10 studies reported the priorities of patients and 2 studies reported the priorities of clinicians alone. Our findings have shown a mostly low level of agreement between the priorities of patients with multimorbidity and clinicians. We found that prioritisation by patients was mainly driven by their illness experiences, while clinicians focused on longer-term risks. Preserving functional ability emerged as a key priority for patients from across our quantitative and qualitative analyses. Conclusion Recognising that there may be a disparity in prioritisation and understanding the reasons for why this might occur, can facilitate clinicians in accurately eliciting the priorities that are most important to their patients and delivering patient-centred care. PROSPERO registration number CRD42018076076.

I, the Submitting Author has the right to grant and does grant on behalf of all authors of the Work (as defined in the below author licence), an exclusive licence and/or a non-exclusive licence for contributions from authors who are: i) UK Crown employees; ii) where BMJ has agreed a CC-BY licence shall apply, and/or iii) in accordance with the terms applicable for US Federal Government officers or employees acting as part of their official duties; on a worldwide, perpetual, irrevocable, royalty-free basis to BMJ Publishing Group Ltd ("BMJ") its licensees and where the relevant Journal is co-owned by BMJ to the co-owners of the Journal, to publish the Work in this journal and any other BMJ products and to exploit all rights, as set out in our licence.
The Submitting Author accepts and understands that any supply made under these terms is made by BMJ to the Submitting Author unless you are acting as an employee on behalf of your employer or a postgraduate student of an affiliated institution which is paying any applicable article publishing charge ("APC") for Open Access articles. Where the Submitting Author wishes to make the Work available on an Open Access basis (and intends to pay the relevant APC), the terms of reuse of such Open Access shall be governed by a Creative Commons licence -details of these licences and which Creative Commons licence will apply to this Work are set out in our licence referred to above.
Other than as permitted in any relevant BMJ Author's Self Archiving Policies, I confirm this Work has not been accepted for publication elsewhere, is not being considered for publication elsewhere and does not duplicate material already published. I confirm all authors consent to publication of this Work and authorise the granting of this licence. Multi-morbidity, defined as the co-existence of two or more long-term conditions [1] is a global problem [2], which has become the norm across high-income countries [2,3] [4,5] and becoming increasingly prevalent in middle and low-income countries [6][7] [2]. Guidelines for the management of chronic diseases are often single disease-orientated, and can lead to confusion and complications when applied to patients with multi-morbidity [8]. Multi-morbid patients have an increased risk of adverse drug-related events as a result of high levels of polypharmacy and receiving un-coordinated care from multiple healthcare providers [9]. These patients have a poorer health-related quality of life [10], poorer functional status [11] and greater psychological distress [12]. As a result, understanding and finding better strategies to facilitate the management of multi-morbid patients has been identified as a priority for health research [13] .
Key to the effective management of multi-morbidity is using patient-centred care and shared decision-making to set management goals that are acceptable to both the patient and the clinician [14]. Incorporating the priorities of patients in relation to treatments and health outcomes is integral to this process [15][16][17]. However, previous research has shown that whilst doctors recognise the importance of eliciting and incorporating the priorities of their multi-morbid patients, they do not always engage with this process in real world settings, and find eliciting patients' priorities to be difficult [18] [19]. Previous research, completed in a single disease context, has shown that the treatment and health outcome priorities of patients and clinicians can differ [20][21][22], and some studies have highlighted a gap between what doctors' perceive to be the priorities of their patients, and the actual priorities of their patients [23][24][25].
This systematic review aims to identify studies that have investigated the health outcome and treatment priorities of multi-morbid patients, clinicians, or both, in order to assess whether the priorities of the two groups are in alignment, or whether there is a disparity between the priorities of multi-morbid patients and clinicians.

Search strategy
This systematic review has been registered on PROSPERO (ID: CRD42018076076). A comprehensive search strategy (Appendix 1), was developed using guidance for best practice [26] and input from academic librarians at the University of Leicester. The search strategy was used to search MEDLINE, EMBASE, CINHAL and COCHRANE databases from inception to May 2019, as well as searching reference lists for any relevant articles based on PRISMA and Cochrane guidelines [26][27][28]. Citations were stored using Refworks. We have presented our process of article selection in Figure 1.
We included studies reporting the health outcome and treatment priorities of adult patients with multimorbidity [1] and/or clinicians, in relation to patients with multi-morbidity. Studies which did not specify the definition of multi-morbidity as "two or more chronic conditions" [1] in their inclusion criteria, but had a sample patients representative of being diagnosed with multi-morbidity (i.e. with a minimum of two chronic conditions), were also included. There was no restriction by study design, and we included studies using quantitative and/or qualitative methodologies. We excluded studies not published in English language, studies with participants aged under 18 years, and studies focusing on a single disease area.

Study selection
The titles and abstracts of all articles identified by the literature search were assessed independently and in duplicate by two reviewers (HS and RF). Studies that did not meet inclusion criteria were discarded. Full text of selected articles were retrieved and assessed to determine if they met the inclusion criteria, and those studies which met the inclusion criteria were included in the review. Any discrepancies regarding eligibility of an article were discussed, and consensus reached with MS and SS.

Methodological quality assessment and data extraction
Data was extracted using standardised data extraction forms by a single reviewer (HS), and these were checked independently for accuracy by a second reviewer (SS). The reported health outcome and treatment priorities of study participants were the key outcomes that were extracted.
Quality assessment was carried out in parallel with the data extraction process. For the quantitative studies, due to the heterogeneity of study design, we used the AXIS tool for assessment for the cross-sectional studies [29], the Newcastle-Ottawa scale for assessment of the longitudinal observational and cohort studies [30], and the Cochrane collaboration's risk of bias tool for assessment of randomised controlled trials [31]. For the qualitative studies, we used the CASP checklist for appraisal of qualitative research [32]. For the two mixed-methods studies, we used the AXIS tool [29] to assess the quantitative aspects of the study (both cross-sectional in study design), and the CASP checklist for qualitative research [32], to assess the qualitative aspects of these studies.

Data synthesis
We decided a priori not to carry out a meta-analysis due to the heterogeneity of the quantitative studies. Therefore, we have taken a narrative synthesis approach, described by Popay et al [33] to synthesise our quantitative findings. Our approach consists of three key steps: 1) Development of a preliminary synthesis in which study characteristics and descriptions are collated and findings presented in a summary table 2) Exploring relationships in the data between study characteristics and their findings, as well as between the reported findings across different studies with explanations considered where relationships were identified. c) Assessing the robustness of the synthesis using quality assessment tools to guide conclusions and identify directions for clinical practice.
Qualitative studies were synthesised using a meta-ethnography approach [34,35], which consisted of careful reading of the papers, extracting information regarding the context of the study and findings. Key concepts arising from each paper were also identified, with preservation of the terminology used by the authors where possible to ensure accurate representation of the findings of the original studies. The key concepts across the papers were then translated using a table summarising the studies, their findings in relation to the key concepts and the second order interpretations of the authors, which enabled the exploration of any relationships and differences. The translations were then synthesised using a table containing the first order and second order interpretations for the key concepts across the studies, which then led to the development of further, third order interpretations by reviewers [34,35].

Cross-sectional survey
To determine the feasibility of using a simple tool to elicit the preferences of older persons based on their prioritization of universal outcomes Patients aged 65 and over with a known diagnosis of hypertension or use of anti-hypertensive medications, and having a known risk of falls (n=81) > Rankings given by participants to 4 universal health outcomes in the outcome prioritisation tool: --keeping alive -maintaining independence -reducing or eliminating pain -reducing or eliminating other symptoms > Feasibility of the use of outcome prioritisation tool Fried et al, 2006 [42] USA-subspeciality outpatient practices, a community hospital, a university teaching hospital and a veterans administration hospital. Patients aged 69 or over with two or more chronic diseases (one of which had to be cardiovascular disease) and daily use of five or more medications. (n=59) General practitioners (n=17) >Patients' priority rankings of the four health outcomes in the outcome prioritisation tool: -Maintaining independence -Remaining alive -Reducing other symptoms -Reducing pain >Medication changes proposed by the GP, and observed in the patient records following incorporation of the priority rankings given by patients, into a medication review consultation.
Junius-Walker et al 2012 [53] Germany-General practice centres Quantitative: Randomised controlled trial To investigate whether a structured priority-setting consultation reconciles the often-differing doctor-patient views on the importance of problems.
Patients aged 70 or over (n=317) General practitioners (n=40) -Baseline importance rankings given by patients and clinicians to a list of problems generated from a geriatric assessment for each patient.
-Importance rankings given again after a structured consultation incorporating the baseline problem list and importance rankings and degree of reconciliation in doctorpatient agreement after the structured consultation Junius-Walker et al, 2011 [54] Germany-General practice centres General practitioners (n=11) Importance rankings given by patients and clinicians to a list of problems generated from a geriatric assessment for each patient.
Voigt et al, 2010 [55] Germany-General practice centres -Importance rankings given to problems generated from a geriatric assessment by patients and clinicians -Degree of agreement between patients and clinicians on the above
Patients enrolled by General practitioners during scheduled consultations.

Quantitative: Cross-sectional survey
To describe FPs' medical priority ranking of conditions relative to their prevalence in patients with multimorbidity Patients suffering from at least 3 of 75 chronic conditions on a predefined list (based on the International classification of primary care 2 (n=888) General Practitioners (n=100) Importance rankings given by family practitioners to the list of chronic conditions that each patient had on the day of their inclusion in the study Mantelli et al, 2018[58] Switzerland-General practitioners working in Switzerland who had previously taken part in case-vignette studies Quantitative: cross-sectional survey To determine whether, how and why GPs deprescribe in frail oldest-old patients with multimorbidity and polypharmacy, and to identify factors that influenced their decision to deprescribe General Practitioners (n=157) -Percentage of GPs willing to de-prescribe at least one medication in the case of frail older patients with CVD and compared to frail older patients without CVD -Reasons for de-prescribing -Importance ratings given to factors influencing decision to de-prescribe Déruaz-Luyet et al, 2018 [59] Switzerland-Primary data was from "Multimorbidity in Family medicine" study [61].
Patients enrolled by General practitioners during scheduled consultations.

Quantitative: Cross-sectional survey
To evaluate whether GPs could identify the condition that their patients with multimorbidity considered most important.
Patients suffering from at least 3 of 75 chronic conditions on a predefined list (based on the International classification of primary care 2, and receiving follow-up from their GP for at least the preceding 6 months (n= 572 for main analysis, 585 for sensitivity analysis)

General Practitioners (n=100)
Whether there is agreement between what patients considered to be their most important health condition and what GPs thought patients considered to be their most important health condition Van Summeren et al, 2016 [46] Netherlands-General practice centres Mixed-methods: Cross-sectional survey pilot and qualitative interviews to assess acceptability (semi-structured and indepth)  Naik et al, 2016 [43] USA-Qualitative data from the VETCARES study [62]

Summary of quality assessment
The outcome of quality assessment based on each of the afore-mentioned tools is summarised in Appendix 2. The majority of the quantitative studies were cross-sectional in design [36,39,40,46,47,54,55,[57][58][59] [48], including the quantitative elements of the two mixed-methods studies. The other studies included one longitudinal observational study [42], one cohort study [45] and one randomised controlled trial [53]. The cross-sectional studies were of moderate quality, with a number of studies having small sample sizes [40,46,47,55]. The sample sizes of clinicians in most of the cross-sectional studies were particularly small, ranging from of 9 to 157 clinicians [46,47,55,58], which impacts upon the generalisability and application of their findings. We noted that a number of the studies did not use pre-validated questions and tools to ascertain priorities [36,55,[57][58][59], leading to a degree of subjectivity in the way in which priorities were ascertained, and the risk of measurement bias which again impacts on the generalisability of their findings.
The majority of the qualitative studies, including the qualitative aspects of the two mixed-methods studies, used interviews for data collection (n=8). Two studies used focus groups [41,56], one study used a combination of focus groups and interviews [50] and one study used the nominal group technique [49]. The qualitative studies were of good quality, with appropriate use of qualitative methodology and transparent descriptions of the data analysis processes. Three studies only gave a limited description of their analytic process [48,49], with two of these studies not presenting any quotes [48,49].

QUANTITATIVE SYNTHESIS
Within our quantitative synthesis, we found that the studies focused either on the overall state of the patients' health, the problems posed by different chronic disease groups, or the patients' treatment regimens. Some of the quantitative studies elicited patient and/or clinician priorities as part of an intervention [53] [47]. Therefore, in order to reduce the risk of bias from the interventions, we included only the pre-intervention results from these studies.

Health outcome priorities
Four studies reported patient priorities of overall health outcomes using a "health outcome prioritisation tool" [39,40,46], which is a visual analogue scale requiring the following health outcomes to be given a score out of 100: "Maintaining independence"; "Staying alive"; "Pain relief"; "Symptom relief". Maintaining independence was the outcome that had the highest importance after a pooling of the most important rankings from the four studies, followed by "Staying alive" ( Table 2). For clinicians' priorities, one study reported that 98% of a sample of 157 general practitioners identified the "quality of life for the patient", and 96% identified the "life expectancy of the patient", as the most important factors in influencing their clinical decision-making to deprescribe for elderly, multi-morbid patients [58].

Priorities based on health problems
Three studies reported patient and general practitioners' priorities based on various health problems, following a geriatric assessment [53][54][55]. These problems were then categorised into domains, and the importance rankings for each of the domains were presented. Problems in the domains of "Social" "Mood" and "Function" recurrently featured in the top four of the most highly ranked priorities by patients across all three studies. In terms of the importance rankings by clinicians, problems in the domains of "Mood" and "Function" also featured in the top four importance rankings across all three studies, whilst "Social" problems were rated highly in one study [54] and problems in the domain of "Medication" were ranked highly in the other two studies [53,55]. Interestingly, the authors in one study [54] found that patients feeling "Emotionally affected" was the strongest predictor for a problem being rated as important (OR 11.1 CI 6.73 to 18.33), whereas "Poor prognosis" was the strongest predictor for clinicians (OR 6.39 CI 4.61 TO 8.87)

Disease-specific priorities
Two studies reported patient priorities in relation to specific diseases or disease groups [45,59]. Zulman et al. reported that "Diabetes/glycaemic control" was most frequently ranked as "most important", with "Hypertension" coming second [45]. However, the sample of patients included in this study were all diabetic, hypertensive patients. Deruaz-luyet et al. found that musculoskeletal conditions including back pain, were most frequently reported to be the most important conditions for their patients, however endocrine/metabolic conditions (including obesity) were second and cardiovascular conditions were third [59].   [57], and found that "multiple sclerosis", "mental retardation", and "bronchus lung neoplasm" were all highly prioritised by their participants. Zulman et al. reported the priorities of "primary care providers" who consisted of physicians, physician assistants or nurse practitioners [45], and found that diabetes was the top priority for primary care providers, with hypertension coming second, in alignment with their previously described patient priorities [45]. Moore et al. examined the priorities of different types of clinicians, including family physicians, geriatricians and nurse practitioners [36], and as with Zulman et al., found that diabetes was the top priority for family physicians and also nurse practitioners, whereas dementia was the top priority for geriatricians [36]. In addition, heart failure, atrial fibrillation and hypertension formed three of the top five conditions considered to be most important by the family practitioners in the study [36].

Treatment priorities
Fried et al. examined patient-ratings as acceptable vs unacceptable for four adverse outcomes from treatment, namely 'unable to leave house'; 'only able to get from bed to chair'; 'severe memory problems'; and 'daily pain', at baseline and over the course of a two year follow-up period [42]. They found that participants were more likely to find the health states relating to functional ability (i.e. 'unable to leave house' and 'only able to get from bed to chair') as acceptable at baseline and throughout the study, whereas 'severe memory problems' and 'daily pain' were more likely to be rated as unacceptable, at baseline and throughout the study.
As part of a study to examine the influence of the risks and benefits of medications on treatment preferences of patients, Caughey et al. also examined the priorities of patients in the face of "competing outcomes" [48]. They found that 80% of participants would not be willing to take medication to reduce "joint pain", if the medication increased their risk of a myocardial infarction by 10%. However, this was deduced from a sample of only 15 patients [48].

Agreement between patients and clinicians
Five of the included studies investigated the level of agreement in priority rankings between patients and their clinicians [45,[53][54][55]59]. Three studies reported a low level of agreement between patient and clinicians' priority rankings [53][54][55]. Two of these studies used a Cohen's Kappa calculation to estimate the degree of agreement between the importance ratings of patients and clinicians, and the values of which were 0.18 and 0.11 respectively, indicating "slight agreement" after allowing for chance [54] [55]. One study used a weighted kappa calculation to measure the degree of agreement, which, at a pre-intervention point in this study, was low at 6% [53].
Two studies reported that there was a "high" level of agreement [45,59]. Deruaz-Luyet et al. found that in the case of 54.9% (n=314) of their patients, the condition that their GP had considered to be either the first or second most important, was in the same disease-group as the condition that the patient considered to be most important [59].
Zulman et al. reported that 60% of "patient-provider pairs" had a "high concordance", meaning that the same three conditions had been rated as top three priorities by both parties, or that two of the same conditions had been rated in the top three priorities by both parties [45]. In this case, given that the sample of patients were all diabetic and hypertensive could have led to a narrowing of the range of chronic diseases across the sample, which in turn could have led to an increased likelihood of agreement. However, the participant characteristics reported by the authors state that the patients had a mean of eight health conditions (SD 3.00), suggesting that the patients did not have a narrow range of chronic diseases.

QUALITATIVE SYNTHESIS
Whilst our quantitative synthesis allowed us to investigate which health outcomes, diseases or treatments were important to multi-morbid patients and their clinicians, our qualitative analysis enabled us to explore how prioritisation occurs. Below, we describe the key findings from our qualitative analysis.

Mechanisms of prioritisation
In the qualitative studies that approached prioritisation from a disease-specific perspective, patients were able to identify an illness as their main priority [50,51]. For many patients, prioritisation appeared to be driven by their experience of the illness, which formed part of its "meaning as consequence" [51] as phrased by Cheraghi-Sohi et al. The 'consequences' of an illness consisted of the impact that the illness was having on the patients' everyday lives, which included functional limitation and the symptomatic burden of the illness, including its "unpredictability" (Table 3) [50]. For others, prioritisation appeared to be driven by their perception of the risk now and in the future with respect to functional deterioration and mortality.
In other studies, patients framed their priorities between quality of life vs length of life (Table 3) [43]. Patients in the study by Naik et al. who were multi-morbid adults with cancer, prioritised "quality of life" more highly than "length of life" [43]. This was also reflected in the findings of Fried et al., who found that when considering medication with competing outcomes in terms of extending life compared to quality of life, participants appeared to prioritise preserving quality of life [41].
Van Summeren et al. found that prioritisation was "difficult" when there was no "specific need" for a treatment decision to be made [46]. This concept of a difference in prioritisation based on hypothetical, or experiential levels, was also shared in the findings of Elliott et al [44] and Fried et al [41].
Where clinicians' perspectives were explored alongside patients, clinicians reported that exploring patients' priorities was "extremely important" when managing "competing interests" [48] and beneficial in providing patient-centred care [46]. Some clinicians in the mixed-methods study carried out by Van Summeren et al. reported that exploring their patients' priorities allowed them to have a "deeper understanding" of the patient, helped with making patient-centred treatment decisions and advance care planning (Table 3) [46]. However, other clinicians in the same study found exploring patient priorities to be difficult due its "novelty" and the fact that it represented a change to their usual consultations [46].

Concept Examples from included studies
Unpredictability of symptoms "My final issue is diverticulitis. In many ways that is the thing that makes the most impact on my life because of the unreliability of it. You make plans to do something to go somewhere and at the last minute you don't dare leave the house because you don't leave the loo. In itself it's not an important medical issue. It's the social problem more than anything else." -Lindsay et al [50]

Factors influencing prioritisation:
Our analysis revealed that there were a number of factors that appeared to influence how both patients and clinicians arrived at their priorities, and which priorities they chose.

i. Functional ability
Preserving functional ability as a priority for patients was a dominant concept across the majority of the qualitative studies [37,38,50,52] [43][48] [41]. Preserving independence emerged as the most significant reason for prioritising functional ability for patients, and maintaining the ability to engage in activities of daily living, mobility, maintaining cognitive ability and wanting to avoid being a "burden" or lacking social support to help them cope with functional deterioration (Table 4) [38,50,51].
Conditions which caused limitation to patients' ability to self-manage their health conditions, led to a "tension" between the patients' expectations of themselves and what they were physically able to do [52]. Lifestyle management, particularly reduced ability to exercise and the adverse impact of this on weight, was cited as part of patients' ability to self-manage [50].
Maintaining patients' functional ability was reported as a priority by some clinicians [37] [48]. Clinicians considered the wider implications of the patients' functional deterioration, particularly cognitive deterioration, and spoke of wanting to reduce the risk of "burnout" for the patients' family members/caregivers [37].

ii. Mortality
Reducing the risk of mortality emerged as a recurrent priority for clinicians [48,56]. Caughey et al found that clinicians prioritised mortality in younger (less than 65 years) multi-morbid patients rather than older multi-morbid patients, as they felt they could be more "aggressive" in their treatment [48]. Reducing the risk of mortality also emerged as a priority for patients across a number of studies [37,38,44,51,52] [43]. Some patients found the asymptomatic nature of hypertension to be concerning; hence, the consequences of hypertension could be unpredictable, compared to some other chronic illnesses where symptoms can give warning of onset and severity (Table 4) [38,44]. iii.

Symptom control
The symptomatic burden of a condition contributed to its "meaning as consequence" for patients [51]. Symptoms were cited as being a cause of functional limitation [38,50], and in some cases their "unpredictability" could cause significant disruption to patients' daily lives [50]. Symptom control was reported to be a priority by some clinicians [37] [48]. However, clinicians in one study considered symptom control to be less important, particularly when there was no risk of mortality [56]. In these cases, clinicians seemed to be aware that patients may still be prioritising symptom control highly, even if the clinicians did not (Table 4). F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   15 Factors related to the treatment burden of an illness appeared to adversely impact prioritisation for patients, leading to de-prioritisation of certain medications and treatments [38,41,44,49]. Elliot et al. reported that cost and distressing side effects, were factors which led patients to stop taking a medication [44]. Similarly, Fried et al. found that patients reported unpleasant side effects to be a "competing outcome", which negatively influenced their decision regarding continuing a medication [41]. However, difficulty with achieving control over the management of an illness, as well as requirement for high levels of engagement with self-management, emerged as factors that contributed to the prioritisation of an illness by some patients (Table 4) [50].

Health outcome and treatment priorities
From our findings, patients' prioritisation appeared to be driven by weighing up the empirical compared to the hypothetical impact of a disease, whereby the empirical impact of a disease, which included its impact on function, symptomatic and treatment burden, was the most dominant driver of prioritisation. This is consistent with the findings of previous literature showing patients with rheumatoid arthritis who had reported experiencing higher levels of pain, were more likely to report pain as a priority [69].
Amongst empirical factors, preserving functionality emerged as most highly prioritised by patients amongst the quantitative studies that took a health outcome approach [39,40,47], whilst "function" was a domain that was prioritised highly by both patients and clinicians in the studies where prioritisation of various health problems were investigated [53][54][55]. From our qualitative findings, functional ability formed a key part of the preservation of various aspects of the patients' ".. I talk [to her] for a quarter of an hour about this and that every time after which she replies, "but my vertigo," and I answer every time, well, unfortunately there is nothing I can do about it, we have already tried and done everything. But it is probably the first diagnosis she will mention: "What are you suffering from?". "Vertigo". For me, this would be somewhere all the way at the bottom." -Hansen et al [56]

Factors influencing prioritisation
Treatment burden "It's the knee that's the most concerning because everything else is controlled by tablets. The knee is a problem because if I have one little slip I'm in plaster again for 6 weeks." Lindsay et al [50]  independence and their quality of life, as well as their ability to self-manage. Existing evidence shows that the prevalence of multi-morbidity is highest in those aged over 65 years [70], and the population for the majority of the included studies were older multi-morbid adults. This could provide an explanation for why preserving functionality was highly prioritised.
Prioritisation was not a static process and was subject to change, based on factors such as illness exacerbations, life events, whether there was a need for a treatment decision to be made, and whether the priority related to retrospective or prospective healthcare [50,52]. When considering the hypothetical impact of an illness, perceptions of future risk came into play, and in particular, the risk of mortality [44]. This was particularly evident in relation to cardiovascular disease, where patients appeared to perceive the risk of mortality to be high [38].
Risk of mortality was a dominant driver for prioritisation amongst clinicians. This was shown in our quantitative synthesis, where amongst studies assessing disease-specific priorities, conditions with a higher risk of mortality, such as cardiovascular disease and diabetes, recurrently emerged as being highly prioritised by clinicians [36,45,57] and differentiated by age [48]. This age-based consideration could explain why clinicians prioritised "quality of life for the patient" as higher, albeit marginally, than "life expectancy of the patient" in their clinical decision-making for de-prescribing for elderly, multi-morbid patients [58].
Our findings show a varying degree of agreement between the priorities of multi-morbid patients and clinicians. Previous studies carried out in the context of diabetes [71], and psoriasis [72] have found a low level of agreement on health outcome and treatment priorities between patients and clinicians, which correlates with the findings of some studies included in this review [53][54][55], but not others [45]. The nature of the patients' illnesses emerged as a factor for concordance or discordance of priorities with their clinicians [37]. Patients and clinicians were in agreement in situations where patients were currently experiencing an exacerbation of a particular condition, or had a "stable" state of health. However, in patients who suffered from illnesses with more complex courses, discordance of priorities tended to occur between patients and clinicians [37].

Strengths and limitations
To our knowledge, this is the first systematic review to assimilate and compare the findings of existing literature on the health outcome and treatment priorities of both patients and clinicians for patients living with multi-morbidities. In this review, we have been able to add a novel line of argument to the ongoing discussion on this subject. By incorporating papers using both qualitative and quantitative methodologies, we have been able to explore patterns and relationships in the findings of a wide range of studies, thus creating a comprehensive and well-rounded systematic review.
There are noteworthy limitations. We did not include the term "comorbidity", in our search terms, and whilst "comorbidity" is distinctive from multi-morbidity, there is also some conceptual overlap between the two terms. We felt that including "comorbidity" in our search strategy would identify studies focusing on a specific condition rather than multi-morbidity.
A number of the quantitative studies did not use pre-validated tools to ascertain priorities [36,55,[57][58][59], leading to a risk of measurement bias, which could limit the generalisability of findings in this review. We also detected a large variation in the tools used to ascertain priorities, which meant that carrying out a meta-analysis to synthesise the findings of the quantitative studies was not possible. Yet, we have tried to mitigate the lack of meta-analysis by using a well-described and wellestablished method of narrative synthesis [33], in order to maintain rigour and transparency.

Recommendations for the future
We recommend that future guidelines developed for clinicians in the management of multimorbidity highlight the need to elicit and consider both short term and long term priorities for their patients', and review these priorities continually, and particularly when exacerbations, changes to illness course or treatment regimens, or other wider socially-contextualised changes occur in their patients' lives.
There was a large variation in how priorities were ascertained, and in the tools used to ascertain priorities. The relative lack of standardised and validated tools for use to ascertain patient priorities in everyday clinical practice has also been described in previous literature [73]. We highlight a need for the development of a standardised and validated tool that is acceptable to both patients and clinicians, and can be used to ascertain patient-priorities in the multiple dimensions described in this review. Such a tool would a valuable aid to treatment decision-making, advance care planning and achieving patient-centeredness for patients living with multi-morbidity.

Conclusion
The findings from this review show the priorities of patients and clinicians can have varying degrees of concordance, being mostly low [53,55], in alignment with previous findings in single disease contexts [71,72]. We have found that the mechanisms of prioritisation can also differ between our two groups, in that patients are driven by illness experiences, whereas clinicians may be focused on managing longer term risks. Understanding these differences can help clinicians to better recognise situations where the patients' priorities may be different to theirs and elicit the most important priorities for their patients.   Schoenber g et al [2] Cheraghi -Sohi et al [3] Naik et al [4] Lindsay et al [5] Hansen et al [6] Morris et al [7] Elliott et al [8] Fried et al [9] Turner et al [10] Van Summeren et al [11] Caughey et al [12] Was Fried et al [16] Fried et al [17] Moore et al [18] Van Summeren et al [19] Voigt et al [20] Van Summeren et al [11] Caughey et al [21] Mantelli et al [22] Deruaz-Luyet et al [23] Herzig et al [24]  Yes-in the parent study [25] Yes-in the parent study [25] Study Junius-Walker et al [27] Bias Authors' judgement Support for judgement Random sequence generation (selection bias) Low risk "Participating doctors were allocated 1:1 into the intervention and control group using random block sizes of four" Allocation concealment (selection bias)

The Cochrane Collaboration's tool for assessing risk of bias in randomised controlled trials summary table
Unclear risk No information given regarding any efforts to conceal the allocation sequence Blinding of participants and researchers (performance bias)

Low risk
Participants were only informed of the procedures of their own trial arm. Blinding of outcome assessment (detection bias) Low risk Participants were blinded to the pre-intervention importance ratings, when completing the final importance ratings. Incomplete outcome data (attrition bias) High risk 25 patients dropped out prior to baseline ratings and 5 further patients dropped out prior to final ratings, these patients were excluded from analysis, however intention to treat analysis cannot be carried out in this context due to the nature of the intervention Selective reporting (reporting bias) Low risk Adequate reporting on all of the specified outcomes Other bias None detected   Multi-morbidity, defined as the co-existence of two or more long-term conditions [1] is a global problem [2], which has become the norm across high-income countries [2,3] [4,5] and becoming increasingly prevalent in middle and low-income countries [6][7] [2]. Guidelines for the management of chronic diseases are often single disease-orientated, and can lead to confusion and complications when applied to patients with multi-morbidity [8]. Multi-morbid patients have an increased risk of adverse drug-related events as a result of high levels of polypharmacy and receiving un-coordinated care from multiple healthcare providers [9]. These patients have a poorer health-related quality of life [10], poorer functional status [11] and greater psychological distress [12]. As a result, understanding and finding better strategies to facilitate the management of multi-morbid patients has been identified as a priority for health research [13] .
Key to the effective management of multi-morbidity is using patient-centred care and shared decision-making to set management goals that are acceptable to both the patient and the clinician [14]. Incorporating the priorities of patients in relation to treatments and health outcomes is integral to this process [15][16][17]. However, previous research has shown that whilst doctors recognise the importance of eliciting and incorporating the priorities of their multi-morbid patients, they do not always engage with this process in real world settings, and find eliciting patients' priorities to be difficult [18] [19]. Previous research, completed in a single disease context, has shown that the treatment and health outcome priorities of patients and clinicians can differ [20][21][22], and some studies have highlighted a gap between what doctors' perceive to be the priorities of their patients, and the actual priorities of their patients [23][24][25].
This systematic review aims to identify studies that have investigated the health outcome and treatment priorities of multi-morbid patients, clinicians, or both, in order to assess whether the priorities of the two groups are in alignment, or whether there is a disparity between the priorities of multi-morbid patients and clinicians.

Search strategy
This systematic review has been registered on PROSPERO (ID: CRD42018076076). A comprehensive search strategy (Appendix 1), was developed using guidance for best practice [26] and input from academic librarians at the University of Leicester. The search strategy was used to search MEDLINE, EMBASE, CINHAL and COCHRANE databases from inception to May 2019, as well as searching reference lists for any relevant articles based on PRISMA and Cochrane guidelines [26][27][28]. We undertook a scoping search using google scholar using our key terms (Patient*; Priorit*; Clinician, Physician, Doctor, General-practitioner, Family-practitioner; Multi-morbid*) to identify relevant grey literature. Citations were stored using Refworks. We have presented our process of article selection in Figure 1.
We included studies reporting the health outcome and treatment priorities of adult patients with multimorbidity [1] and/or clinicians, in relation to patients with multi-morbidity. Studies which did not specify the definition of multi-morbidity as "two or more chronic conditions" [1] in their inclusion criteria, but had a sample patients representative of being diagnosed with multi-morbidity (i.e. with a minimum of two chronic conditions), were also included. There was no restriction by

Patient and Public Involvement
Patient and public involvement was not applicable in the design, conduct or reporting of this review.

Study selection
The titles and abstracts of all articles identified by the literature search were assessed independently and in duplicate by two reviewers (HS and RF). Studies that did not meet inclusion criteria were discarded. Full text of selected articles were retrieved and assessed to determine if they met the inclusion criteria, and those studies which met the inclusion criteria were included in the review. Any discrepancies regarding eligibility of an article were discussed, and consensus reached with MS and SS.

Methodological quality assessment and data extraction
Data was extracted using standardised data extraction forms by a single reviewer (HS), and these were checked independently for accuracy by a second reviewer (SS). The reported health outcome and treatment priorities of study participants were the key outcomes that were extracted.
Quality assessment was carried out in parallel with the data extraction process. For the quantitative studies, due to the heterogeneity of study design, we used the AXIS tool for assessment for the cross-sectional studies [29], the Newcastle-Ottawa scale for assessment of the longitudinal observational and cohort studies [30], and the Cochrane collaboration's risk of bias tool for assessment of randomised controlled trials [31]. For the qualitative studies, we used the CASP checklist for appraisal of qualitative research [32]. For the two mixed-methods studies, we used the AXIS tool [29] to assess the quantitative aspects of the study (both cross-sectional in study design), and the CASP checklist for qualitative research [32], to assess the qualitative aspects of these studies.

Data synthesis
We decided a priori not to carry out a meta-analysis due to the heterogeneity of the quantitative studies. Therefore, we have taken a narrative synthesis approach, described by Popay et al [33] to synthesise our quantitative findings. Our approach consists of three key steps: 1) Development of a preliminary synthesis in which study characteristics and descriptions are collated and findings presented in a summary table 2) Exploring relationships in the data between study characteristics and their findings, as well as between the reported findings across different studies with explanations considered where relationships were identified. c) Assessing the robustness of the synthesis using quality assessment tools to guide conclusions and identify directions for clinical practice.
Qualitative studies were synthesised using a meta-ethnography approach [34,35], which consisted of careful reading of the papers, extracting information regarding the context of the study and findings. Key concepts arising from each paper were also identified, with preservation of the terminology used by the authors where possible to ensure accurate representation of the findings of The key concepts across the papers were then translated using a table summarising the studies, their findings in relation to the key concepts and the second order interpretations of the authors, which enabled the exploration of any relationships and differences. The translations were then synthesised using a table containing the first order and second order interpretations for the key concepts across the studies, which then led to the development of further, third order interpretations by reviewers [34,35].
From the results of our narrative synthesis of the quantitative studies and meta-ethnography of the qualitative studies, we considered how the findings of the two syntheses complement one another, particularly where our qualitative findings may provide possible explanations for our quantitative findings. The outcome of this process is described in the discussion section.

Overall study characteristics
Our search resulted in the identification of 24 studies for inclusion, which comprised of 12 quantitative studies, 10 qualitative studies and two mixed-methods studies. The characteristics of all of the included studies are described in -Baseline importance rankings given by patients and clinicians to a list of problems generated from a geriatric assessment for each patient.
-Importance rankings given again after a structured consultation incorporating the baseline problem list and importance rankings and degree of reconciliation in doctorpatient agreement after the structured consultation -Patient rankings given in terms of their most important health concerns and providers rankings in terms of conditions most likely to affect each patient's outcomes -Concordance between the importance ratings of patientprovider "pairs" Herzig et al, 2019 [56] Switzerland-Primary data was from "Multimorbidity in Family medicine" study [60].

Summary of quality assessment
The outcome of quality assessment based on each of the afore-mentioned tools is summarised in Appendix 2. The majority of the quantitative studies were cross-sectional in design [36,39,40,45,46,53,54,[56][57][58] [47], including the quantitative elements of the two mixed-methods studies. The other studies included one cohort study [44] and one randomised controlled trial [52]. The crosssectional studies were of moderate quality, with a number of studies having small sample sizes [40,45,46,54]. The sample sizes of clinicians in most of the cross-sectional studies were particularly small, ranging from of 9 to 157 clinicians [45,46,54,57], which impacts upon the generalisability and application of their findings. We noted that a number of the studies did not use pre-validated questions and tools to ascertain priorities [36,54,[56][57][58], leading to a degree of subjectivity in the way in which priorities were ascertained, and the risk of measurement bias which again impacts on the generalisability of their findings.
The majority of the qualitative studies, including the qualitative aspects of the two mixed-methods studies, used interviews for data collection (n=8). Two studies used focus groups [41,55], one study used a combination of focus groups and interviews [49] and one study used the nominal group technique [48]. The qualitative studies were of good quality, with appropriate use of qualitative methodology and transparent descriptions of the data analysis processes. Three studies only gave a limited description of their analytic process [47][48][49], with two of these studies [47,48] and one mixed-methods study [45], not presenting any quotes.

QUANTITATIVE SYNTHESIS
Within our quantitative synthesis, we found that the studies focused either on the overall state of the patients' health, the problems posed by different chronic disease groups, or the patients' treatment regimens. Some of the quantitative studies elicited patient and/or clinician priorities as part of an intervention [52] [46]. Therefore, in order to reduce the risk of bias from the interventions, we included only the pre-intervention results from these studies.

Health outcome priorities
Four studies reported patient priorities of overall health outcomes using a "health outcome prioritisation tool" [39,40,45], which is a visual analogue scale requiring the following health outcomes to be given a score out of 100: "Maintaining independence"; "Staying alive"; "Pain relief"; "Symptom relief". Maintaining independence was the outcome that had the highest importance after a pooling of the most important rankings from the four studies, followed by "Staying alive" ( Table 2). For clinicians' priorities, one study reported that 98% of a sample of 157 general practitioners identified the "quality of life for the patient", and 96% identified the "life expectancy of the patient", as the most important factors in influencing their clinical decision-making to deprescribe for elderly, multi-morbid patients [57].

Priorities based on health problems
Three studies reported patient and general practitioners' priorities based on various health problems, following a geriatric assessment [52][53][54]. These problems were then categorised into domains, and the importance rankings for each of the domains were presented. Problems in the domains of "Social" "Mood" and "Function" recurrently featured in the top four of the most highly ranked priorities by patients across all three studies. In terms of the importance rankings by clinicians, problems in the domains of "Mood" and "Function" also featured in the top four importance rankings across all three studies, whilst "Social" problems were rated highly in one study [53] and problems in the domain of "Medication" were ranked highly in the other two studies [52,54]. Interestingly, the authors in one study [53] found that patients feeling "Emotionally affected" was the strongest predictor for a problem being rated as important (OR 11.1 CI 6.73 to 18.33), whereas "Poor prognosis" was the strongest predictor for clinicians (

Condition-focused priorities
Two studies reported patient priorities in relation to specific conditions or groups of conditions [44,58], in the context of multi-morbidity. Zulman et al. reported that "Diabetes/glycaemic control" was most frequently ranked as "most important", with "Hypertension" coming second [44]. However, the sample of patients included in this study were all diabetic, hypertensive patients. Deruaz-luyet et al. found that musculoskeletal conditions including back pain, were most frequently reported to be the most important conditions for their patients, however endocrine/metabolic conditions (including obesity) were second and cardiovascular conditions were third [58].
Three studies reported condition-focused priorities of clinicians in the context of multi-morbidity. Herzig et al. reported the priorities of general practitioners alone [56], and found that "multiple sclerosis", "mental retardation", and "bronchus lung neoplasm" were all highly prioritised by their participants. Zulman et al reported the priorities of "primary care providers" who consisted of physicians, physician assistants or nurse practitioners [44], and found that diabetes was the top priority for primary care providers, with hypertension coming second, in alignment with their previously described patient priorities [44]. Moore et al. examined the priorities of different types of clinicians, including family physicians, geriatricians and nurse practitioners [36], and as with Zulman et al., found that diabetes was the top priority for family physicians and also nurse practitioners, whereas dementia was the top priority for geriatricians [36]. In addition, heart failure, atrial fibrillation and hypertension formed three of the top five conditions considered to be most important by the family practitioners in the study [36].

Treatment priorities
As part of a study to examine the influence of the risks and benefits of medications on treatment preferences of patients, Caughey et al. also examined the priorities of patients in the face of "competing outcomes" [47]. They found that 80% of participants would not be willing to take medication to reduce "joint pain", if the medication increased their risk of a myocardial infarction by 10%. However, this was deduced from a sample of only 15 patients [47].

Agreement between patients and clinicians
Five of the included studies investigated the level of agreement in priority rankings between patients and their clinicians [44,[52][53][54]58]. Three studies reported a low level of agreement between patient and clinicians' priority rankings [52][53][54]. Two of these studies used a Cohen's Kappa calculation to estimate the degree of agreement between the importance ratings of patients and clinicians, and the values of which were 0.18 and 0.11 respectively, indicating "slight agreement" after allowing for chance [53][54]. One study used a weighted kappa calculation to measure the degree of agreement, which, at a pre-intervention point in this study, was low at 6% [52].
Two studies reported that there was a "high" level of agreement [44,58]. Deruaz-Luyet et al. found that in the case of 54.9% (n=314) of their patients, the condition that their GP had considered to be either the first or second most important, was in the same disease-group as the condition that the patient considered to be most important [58].
Zulman et al. reported that 60% of "patient-provider pairs" had a "high concordance", meaning that the same three conditions had been rated as top three priorities by both parties, or that two of the same conditions had been rated in the top three priorities by both parties [44]. In this case, given that the sample of patients were all diabetic and hypertensive could have led to a narrowing of the range of chronic diseases across the sample, which in turn could have led to an increased likelihood of agreement. However, the participant characteristics reported by the authors state that the patients had a mean of eight health conditions (SD 3.00), suggesting that the patients did not have a narrow range of chronic diseases. Furthermore, the questions posed to patients and providers were phrased differently, in that providers were asked to choose the top three most important medical concerns "that are likely to affect health outcomes for this patient", whereas patients were asked to choose their top three most important health concerns. The authors acknowledge this in their paper, and justify this difference as being due to their aim of exploring the concordance in priorities about the "most important problems facing the patient", rather which problems "providers thought the patient would have prioritised", which, they argue, is a different concept to their aim [44].

QUALITATIVE SYNTHESIS
Whilst our quantitative synthesis allowed us to investigate which health outcomes, diseases or treatments were important to multi-morbid patients and their clinicians, our qualitative analysis enabled us to explore how prioritisation occurs. Below, we describe the key findings from our qualitative analysis.

Mechanisms of prioritisation
In the qualitative studies that approached prioritisation from a disease-specific perspective, patients were able to identify an illness as their main priority [49,50]. For many patients, prioritisation appeared to be driven by their experience of the illness, which formed part of its "meaning as consequence" [50] as phrased by Cheraghi-Sohi et al. The 'consequences' of an illness consisted of the impact that the illness was having on the patients' everyday lives, which included functional limitation and the symptomatic burden of the illness, including its "unpredictability" (Table 3) [49]. For others, prioritisation appeared to be driven by their perception of the risk now and in the future with respect to functional deterioration and mortality.
In other studies, patients framed their priorities between quality of life vs length of life (Table 3) [42]. Patients in the study by Naik et al. who were multi-morbid adults with cancer, prioritised "quality of life" more highly than "length of life" [42]. This was also reflected in the findings of Fried et al., who found that when considering medication with competing outcomes in terms of extending life compared to quality of life, participants appeared to prioritise preserving quality of life [41].
Van Summeren et al. found that prioritisation was "difficult" when there was no "specific need" for a treatment decision to be made [45]. This concept of a difference in prioritisation based on hypothetical, or experiential levels, was also shared in the findings of Elliott et al [43] and Fried et al [41].
Where clinicians' perspectives were explored alongside patients, clinicians reported that exploring patients' priorities was "extremely important" when managing "competing interests" [47] and beneficial in providing patient-centred care [45]. Some clinicians in the mixed-methods study carried out by Van Summeren et al. reported that exploring their patients' priorities allowed them to have a "deeper understanding" of the patient, helped with making patient-centred treatment decisions and advance care planning (Table 3) [45]. However, other clinicians in the same study found exploring patient priorities to be difficult due its "novelty" and the fact that it represented a change to their usual consultations [45].

Mechanisms of prioritisation
Facilitating clinicians' decision making "In future, I'll be happier to be more decisive in keeping an eye on what we do and do not do as regards this patient." Van Summeren et al [45]

Factors influencing prioritisation:
Our analysis revealed that there were a number of factors that appeared to influence how both patients and clinicians arrived at their priorities, and which priorities they chose.

i. Functional ability
Preserving functional ability as a priority for patients was a dominant concept across the majority of the qualitative studies [37,38,49,51] [42][47] [41]. Preserving independence emerged as the most significant reason for prioritising functional ability for patients, and maintaining the ability to engage in activities of daily living, mobility, maintaining cognitive ability and wanting to avoid being a "burden" or lacking social support to help them cope with functional deterioration (Table 4) [38,49,50].
Conditions which caused limitation to patients' ability to self-manage their health conditions, led to a "tension" between the patients' expectations of themselves and what they were physically able to do [51]. Lifestyle management, particularly reduced ability to exercise and the adverse impact of this on weight, was cited as part of patients' ability to self-manage [49].
Maintaining patients' functional ability was reported as a priority by some clinicians [37] [47]. Clinicians considered the wider implications of the patients' functional deterioration, particularly cognitive deterioration, and spoke of wanting to reduce the risk of "burnout" for the patients' family members/caregivers [37].

ii. Mortality
Reducing the risk of mortality emerged as a recurrent priority for clinicians [47,55]. Caughey et al found that clinicians prioritised mortality in younger (less than 65 years) multi-morbid patients rather than older multi-morbid patients, as they felt they could be more "aggressive" in their treatment [47]. Reducing the risk of mortality also emerged as a priority for patients across a number of studies [37,38,43,50,51] [42]. Some patients found the asymptomatic nature of hypertension to be concerning; hence, the consequences of hypertension could be unpredictable, compared to some other chronic illnesses where symptoms can give warning of onset and severity (Table 4) [38,43]. iii.

Symptom control
The symptomatic burden of a condition contributed to its "meaning as consequence" for patients [50]. Symptoms were cited as being a cause of functional limitation [38,49], and in some cases their "unpredictability" could cause significant disruption to patients' daily lives [49]. Symptom control was reported to be a priority by some clinicians [37][47]. However, clinicians in one study considered symptom control to be less important, particularly when there was no risk of mortality [55]. In these cases, clinicians seemed to be aware that patients may still be prioritising symptom control highly, even if the clinicians did not (Table 4).

iv. Treatment burden
Factors related to the treatment burden of an illness appeared to adversely impact prioritisation for patients, leading to de-prioritisation of certain medications and treatments [38,41,43,48]. Elliot et al. reported that cost and distressing side effects, were factors which led patients to stop taking a medication [43]. Similarly, Fried et al. found that patients reported unpleasant side effects to be a "competing outcome", which negatively influenced their decision regarding continuing a medication [41]. However, difficulty with achieving control over the management of an illness, as well as requirement for high levels of engagement with self-management, emerged as factors that contributed to the prioritisation of an illness by some patients (Table 4) [49].

DISCUSSION
Prioritisation as a concept is broad, context-dependent and difficult to confine into a single definitive definition. As a result, determining what can be interpreted as a health outcome or treatment priority as part of our study selection in this review, was inherently difficult. We excluded some studies that investigated the preferences of multi-morbid patients or clinicians, in contexts that we judged to be different to the aim of this review. These included patient preferences for healthcare

Concept Examples from included studies
Functional ability "I mean, because I have to be mobile, I am living on my own, no one is going to take care of me, I have got to look after myself.." Cheraghi-Sohi et al [50] Mortality "Well I really do worry the most about the high blood pressure. 'Cause see you know you got arthritis and you can tell when it's coming on. But you can't hardly tell about high blood pressure. It can just hit you like that [snaps fingers] …."" Lindsay et al [49] Symptom control "I would not want to live with pain. I won't allow that to happen"-Naik et al [42] Disparity in prioritisation of symptom control ".. I talk [to her] for a quarter of an hour about this and that every time after which she replies, "but my vertigo," and I answer every time, well, unfortunately there is nothing I can do about it, we have already tried and done everything. But it is probably the first diagnosis she will mention: "What are you suffering from?". "Vertigo". For me, this would be somewhere all the way at the bottom." -Hansen et al [55]

Factors influencing prioritisation
Treatment burden "It's the knee that's the most concerning because everything else is controlled by tablets. The knee is a problem because if I have one little slip I'm in plaster again for 6 weeks." Lindsay et al [49] [73]. Whilst these studies represent very important areas of research, they were not within the scope of our aim in this review i.e. identifying studies that report the health outcome and treatment priorities of multi-morbid patients or those of clinicians in relation to multi-morbid patients. A discussion from our synthesis of findings of the included studies in this review is presented below.

Health outcome and treatment priorities
From our findings, patients' prioritisation appeared to be driven by weighing up the empirical compared to the hypothetical impact of a disease, whereby the empirical impact of a disease, which included its impact on function, symptomatic and treatment burden, was the most dominant driver of prioritisation. This is consistent with the findings of previous literature showing patients with rheumatoid arthritis who had reported experiencing higher levels of pain, were more likely to report pain as a priority [74].
Amongst empirical factors, preserving functionality emerged as most highly prioritised by patients amongst the quantitative studies that took a health outcome approach [39,40,46], whilst "function" was a domain that was prioritised highly by both patients and clinicians in the studies where prioritisation of various health problems were investigated [52][53][54]. From our qualitative findings, functional ability formed a key part of the preservation of various aspects of the patients' independence and their quality of life, as well as their ability to self-manage. Existing evidence shows that the prevalence of multi-morbidity is highest in those aged over 65 years [75], and the population for the majority of the included studies were older multi-morbid adults. This could provide an explanation for why preserving functionality was highly prioritised.
Prioritisation was not a static process and was subject to change, based on factors such as illness exacerbations, life events, whether there was a need for a treatment decision to be made, and whether the priority related to retrospective or prospective healthcare [49,51]. When considering the hypothetical impact of an illness, perceptions of future risk came into play, and in particular, the risk of mortality [43]. This was particularly evident in relation to cardiovascular disease, where patients appeared to perceive the risk of mortality to be high [38].
Risk of mortality was a dominant driver for prioritisation amongst clinicians. This was shown in our quantitative synthesis, where amongst studies assessing disease-specific priorities, conditions with a higher risk of mortality, such as cardiovascular disease and diabetes, recurrently emerged as being highly prioritised by clinicians [36,44,56] and differentiated by age [47]. This age-based consideration could explain why clinicians prioritised "quality of life for the patient" as higher, albeit marginally, than "life expectancy of the patient" in their clinical decision-making for de-prescribing for elderly, multi-morbid patients [57].
Smith et al previously developed a "Core Outcome Set" [76] in which a Delphi consensus panel formed of 26 international health experts, identified and prioritised a set of outcomes tailored for application to research studies targeting multi-morbid patients. Mortality, mental health outcomes and quality of life featured most highly in their list of prioritised outcomes, which also emerged in this review. However, we found that relatively few studies reported the prioritisation of mental health outcomes, with the exception of the studies that took a problem-based approach to prioritisation, where problems with regard to "Mood" were prioritised highly by both patients and clinicians [52][53][54]. Our findings show a varying degree of agreement between the priorities of multi-morbid patients and clinicians. Previous studies carried out in the context of diabetes [77], and psoriasis [78] have found a low level of agreement on health outcome and treatment priorities between patients and clinicians, which correlates with the findings of some studies included in this review [52][53][54], but not others [44]. The nature of the patients' illnesses emerged as a factor for concordance or discordance of priorities with their clinicians [37]. Patients and clinicians were in agreement in situations where patients were currently experiencing an exacerbation of a particular condition, or had a "stable" state of health. However, in patients who suffered from illnesses with more complex courses, discordance of priorities tended to occur between patients and clinicians [37].
In recent times, the traditional paternalistic model for the doctor-patient relationship has given way to an egalitarian model [79], where doctors and patients each play an equitable role in a shareddecision making process, which places the patient at its core and thus achieving greater patientcentred care [80] [79]. A shared agreement between patients and doctors on treatment priorities have been highlighted to play an important part in achieving patient-centred care and creating a therapeutic alliance, the benefits of which can include improved treatment adherence [79,80]. Indeed, Jowsey et al found that agreement between patients and clinicians in the formulation of care plans promoted adherence to these plans, whereas a lack of agreement led to disengagement with care plans by patients [81].

Strengths and limitations
To our knowledge, this is the first systematic review to assimilate and compare the findings of existing literature on the health outcome and treatment priorities of both patients and clinicians for patients living with multi-morbidities. In this review, we have been able to add a novel line of argument to the ongoing discussion on this subject. By incorporating papers using both qualitative and quantitative methodologies, we have been able to explore patterns and relationships in the findings of a wide range of studies, thus creating a comprehensive and well-rounded systematic review.
There are noteworthy limitations. We did not include the term "comorbidity", in our search terms, and whilst "comorbidity" is distinctive from multi-morbidity, there is also some conceptual overlap between the two terms. We felt that including "comorbidity" in our search strategy would identify studies focusing on a specific condition rather than multi-morbidity.
A number of the quantitative studies did not use pre-validated tools to ascertain priorities [36,54,[56][57][58], leading to a risk of measurement bias, which could limit the generalisability of findings in this review. We also detected a large variation in the tools used to ascertain priorities, which meant that carrying out a meta-analysis to synthesise the findings of the quantitative studies was not possible. Yet, we have tried to mitigate the lack of meta-analysis by using a well-described and wellestablished method of narrative synthesis [33], in order to maintain rigour and transparency.
Another limitation is that in our inclusion criteria we chose to also include studies which did not explicitly specify a definition of multi-morbidity as "two or more chronic conditions" in their inclusion criteria but had a sample of participants that were reflective of multi-morbidity (i.e. with a minimum of two chronic conditions which could be identified from participant demographic data). We chose to do this as in the absence of a universally accepted and uniform definition of multimorbidity, we sought to base our judgement on the inclusivity of each paper on its value in answering our review question. This, along with the previously discussed difficulty in defining prioritisation, may have introduced a degree of subjective interpretation in the process of study

Recommendations for the future
We recommend that future guidelines developed for clinicians in the management of multimorbidity highlight the need to elicit and consider both short term and long term priorities for their patients', as our review has shown that patients' priorities for their current illness experiences and future risks posed by illnesses, may differ. In accordance with current NICE guidance, we also reiterate the need to review these priorities continually, and particularly when exacerbations, changes to illness course or treatment regimens, or other wider socially-contextualised changes occur in their patients' lives.
There was a large variation in how priorities were ascertained, and in the tools used to ascertain priorities. The relative lack of standardised and validated tools for use to ascertain patient priorities in everyday clinical practice has also been described in previous literature [82]. We highlight a need for the development of a standardised and validated tool that is acceptable to both patients and clinicians, and can be used to ascertain patient-priorities in the multiple dimensions described in this review. Such a tool would a valuable aid to treatment decision-making, advance care planning and achieving patient-centeredness for patients living with multi-morbidity.

INTRODUCTION
Multimorbidity, defined as the co-existence of two or more long-term conditions [1] is a global problem [2], which has become the norm across high-income countries [2,3] [4,5] and becoming increasingly prevalent in middle and low-income countries [6][7] [2]. Guidelines for the management of chronic diseases are often single disease-orientated, and can lead to confusion and complications when applied to patients with multimorbidity [8]. Patients with multimorbidity have an increased risk of adverse drug-related events as a result of high levels of polypharmacy and receiving uncoordinated care from multiple healthcare providers [9]. These patients have a poorer healthrelated quality of life [10], poorer functional status [11] and greater psychological distress [12]. As a result, understanding and finding better strategies to facilitate the management of patients with multimorbidity has been identified as a priority for health research [13] .
Key to the effective management of multimorbidity is using patient-centred care and shared decision-making to set management goals that are acceptable to both the patient and the clinician [14]. Incorporating the priorities of patients in relation to treatments and health outcomes is integral to this process [15][16][17]. However, previous research has shown that whilst doctors recognise the importance of eliciting and incorporating the priorities of their patients with multimorbidity, they do not always engage with this process in real world settings, and find eliciting patients' priorities to be difficult [18] [19]. Previous research, completed in a single disease context, has shown that the treatment and health outcome priorities of patients and clinicians can differ [20][21][22], and some studies have highlighted a gap between what doctors' perceive to be the priorities of their patients, and the actual priorities of their patients [23][24][25].
This systematic review aims to identify studies that have investigated the health outcome and treatment priorities of patients with multimorbidity, clinicians, or both, in order to assess whether the priorities of the two groups are in alignment, or whether there is a disparity between the priorities of patients with multimorbidity and clinicians.

Search strategy
This systematic review has been registered on PROSPERO (ID: CRD42018076076). A comprehensive search strategy (Appendix 1), was developed using guidance for best practice [26] and input from academic librarians at the University of Leicester. The search strategy was used to search MEDLINE, EMBASE, CINHAL and COCHRANE databases from inception to May 2019, as well as searching reference lists for any relevant articles based on PRISMA and Cochrane guidelines [26][27][28]. We undertook a scoping search using google scholar using our key terms (Patient*; Priorit*; Clinician, Physician, Doctor, General-practitioner, Family-practitioner; Multimorbidit*; Multi morbid*) to identify relevant grey literature. Citations were stored using Refworks. We have presented our process of article selection in Figure 1.
We included studies reporting the health outcome and treatment priorities of adult patients with multimorbidity [1] and/or clinicians, in relation to patients with multimorbidity. Studies which did not specify the definition of multimorbidity as "two or more chronic conditions" [1] in their inclusion criteria, but had a sample patients representative of being diagnosed with multimorbidity (i.e. with a minimum of two chronic conditions), were also included. There was no restriction by study design, and we included studies using quantitative and/or qualitative methodologies. We excluded studies not published in English language, studies with participants aged under 18 years, and studies focusing on a single disease area.

Patient and Public Involvement
Patient and public involvement was not applicable in the design, conduct or reporting of this review.

Study selection
The titles and abstracts of all articles identified by the literature search were assessed independently and in duplicate by two reviewers (HS and RF). Studies that did not meet inclusion criteria were discarded. Full text of selected articles were retrieved and assessed to determine if they met the inclusion criteria, and those studies which met the inclusion criteria were included in the review. Any discrepancies regarding eligibility of an article were discussed, and consensus reached with MS and SS.

Methodological quality assessment and data extraction
Data was extracted using standardised data extraction forms by a single reviewer (HS), and these were checked independently for accuracy by a second reviewer (SS). The reported health outcome and treatment priorities of study participants were the key outcomes that were extracted.
Quality assessment was carried out in parallel with the data extraction process. For the quantitative studies, due to the heterogeneity of study design, we used the AXIS tool for assessment for the cross-sectional studies [29], the Newcastle-Ottawa scale for assessment of the longitudinal observational and cohort studies [30], and the Cochrane collaboration's risk of bias tool for assessment of randomised controlled trials [31]. For the qualitative studies, we used the CASP checklist for appraisal of qualitative research [32]. For the two mixed-methods studies, we used the AXIS tool [29] to assess the quantitative aspects of the study (both cross-sectional in study design), and the CASP checklist for qualitative research [32], to assess the qualitative aspects of these studies.

Data synthesis
We decided a priori not to carry out a meta-analysis due to the heterogeneity of the quantitative studies. Therefore, we have taken a narrative synthesis approach, described by Popay et al [33] to synthesise our quantitative findings. Our approach consists of three key steps: 1) Development of a preliminary synthesis in which study characteristics and descriptions are collated and findings presented in a summary table 2) Exploring relationships in the data between study characteristics and their findings, as well as between the reported findings across different studies with explanations considered where relationships were identified. c) Assessing the robustness of the synthesis using quality assessment tools to guide conclusions and identify directions for clinical practice.
Qualitative studies were synthesised using a meta-ethnography approach [34,35], which consisted of careful reading of the papers, extracting information regarding the context of the study and findings. Key concepts arising from each paper were also identified, with preservation of the terminology used by the authors where possible to ensure accurate representation of the findings of The key concepts across the papers were then translated using a table summarising the studies, their findings in relation to the key concepts and the second order interpretations of the authors, which enabled the exploration of any relationships and differences. The translations were then synthesised using a table containing the first order and second order interpretations for the key concepts across the studies, which then led to the development of further, third order interpretations by reviewers [34,35].
From the results of our narrative synthesis of the quantitative studies and meta-ethnography of the qualitative studies, we considered how the findings of the two syntheses complement one another, particularly where our qualitative findings may provide possible explanations for our quantitative findings. The outcome of this process is described in the discussion section.

Overall study characteristics
Our search resulted in the identification of 24 studies for inclusion, which comprised of 12 quantitative studies, 10 qualitative studies and two mixed-methods studies. The characteristics of all of the included studies are described in Table 1. The included studies had all been conducted in high income developed countries, including Canada [36,37], USA [38][39][40][41][42][43][44], Netherlands [45,46], Australia [47,48], UK [49][50][51], Germany [52][53][54][55]and Switzerland [56][57][58]. Sample sizes ranged from 15 to 1169 patients and 5 to 92 clinicians in the quantitative studies, and 15 to 146 patients and 4 to 19 clinicians in the qualitative studies. Patients aged 69 or over with two or more chronic diseases (one of which had to be cardiovascular disease) and daily use of five or more medications. (n=59) General practitioners (n=17) >Patients' priority rankings of the four health outcomes in the outcome prioritisation tool: -Maintaining independence -Remaining alive -Reducing other symptoms -Reducing pain >Medication changes proposed by the GP, and observed in the patient records following incorporation of the priority rankings given by patients, into a medication review consultation.

Author and year of publication
Van Summeren et al, 2016 [45] Netherlands-General practice centres Mixed-methods: Cross-sectional survey pilot and qualitative interviews to assess acceptability (semi-structured and in-depth) To explore whether an outcome prioritization tool (OPT) is appropriate in the context of medication review in family practice, focusing on its acceptability and practicality Patients aged 69 or over with two or more chronic diseases (one of which had to be cardiovascular disease) and daily use of five or more medications (n=60) General practitioners (n=13) >Patients' prioritisation of the four domains of the outcome prioritisation tool: -Maintaining independence -Remaining alive -Reducing other symptoms -Reducing pain > Family practitioners views on the acceptability and practicality of using the outcome prioritisation tool for medication review General practitioners (n=11) Importance rankings given by patients and clinicians to a list of problems generated from a geriatric assessment for each patient.

Cross-sectional survey
To ascertain health priorities of older patients and treatment priorities of their general practitioners (GP) on the basis of a geriatric assessment and to determine the agreement between these priorities.
Patients aged 70 or over and at least one contact with the general practitioner in the preceding 3 months (n= 35) General practitioners (n=9) -Importance rankings given to problems generated from a geriatric assessment by patients and clinicians -Degree of agreement between patients and clinicians on the above Primary care providers i.e. physicians, physician assistants or nurse practitioners (n= 92) -Patient rankings given in terms of their most important health concerns and providers rankings in terms of conditions most likely to affect each patient's outcomes -Concordance between the importance ratings of patientprovider "pairs" Herzig et al, 2019 [56] Switzerland-Primary data was from "Multimorbidity in Family medicine" study [60].  Table 1 Characteristics of all of the included studies in order of reference

Summary of quality assessment
The outcome of quality assessment based on each of the afore-mentioned tools is summarised in Appendix 2. The majority of the quantitative studies were cross-sectional in design [36,39,40,45,46,53,54,[56][57][58] [47], including the quantitative elements of the two mixed-methods studies. The other studies included one cohort study [44] and one randomised controlled trial [52]. The crosssectional studies were of moderate quality, with a number of studies having small sample sizes [40,45,46,54]. The sample sizes of clinicians in most of the cross-sectional studies were particularly small, ranging from of 9 to 157 clinicians [45,46,54,57], which impacts upon the generalisability and application of their findings. We noted that a number of the studies did not use pre-validated questions and tools to ascertain priorities [36,54,[56][57][58], leading to a degree of subjectivity in the way in which priorities were ascertained, and the risk of measurement bias which again impacts on the generalisability of their findings.
The majority of the qualitative studies, including the qualitative aspects of the two mixed-methods studies, used interviews for data collection (n=8). Two studies used focus groups [41,55], one study used a combination of focus groups and interviews [49] and one study used the nominal group technique [48]. The qualitative studies were of good quality, with appropriate use of qualitative methodology and transparent descriptions of the data analysis processes. Three studies only gave a limited description of their analytic process [47][48][49], with two of these studies [47,48] and one mixed-methods study [45], not presenting any quotes.

QUANTITATIVE SYNTHESIS
Within our quantitative synthesis, we found that the studies focused either on the overall state of the patients' health, the problems posed by different chronic disease groups, or the patients' treatment regimens. Some of the quantitative studies elicited patient and/or clinician priorities as part of an intervention [52] [46]. Therefore, in order to reduce the risk of bias from the interventions, we included only the pre-intervention results from these studies.

Health outcome priorities
Four studies reported patient priorities of overall health outcomes using a "health outcome prioritisation tool" [39,40,45], which is a visual analogue scale requiring the following health outcomes to be given a score out of 100: "Maintaining independence"; "Staying alive"; "Pain relief"; "Symptom relief". Maintaining independence was the outcome that had the highest importance after a pooling of the most important rankings from the four studies, followed by "Staying alive" ( Table 2). For clinicians' priorities, one study reported that 98% of a sample of 157 general practitioners identified the "quality of life for the patient", and 96% identified the "life expectancy of the patient", as the most important factors in influencing their clinical decision-making to deprescribe for elderly, patients with multimorbidity [57].

Priorities based on health problems
Three studies reported patient and general practitioners' priorities based on various health problems, following a geriatric assessment [52][53][54]. These problems were then categorised into domains, and the importance rankings for each of the domains were presented. Problems in the domains of "Social" "Mood" and "Function" recurrently featured in the top four of the most highly  ranked priorities by patients across all three studies. In terms of the importance rankings by clinicians, problems in the domains of "Mood" and "Function" also featured in the top four importance rankings across all three studies, whilst "Social" problems were rated highly in one study [53] and problems in the domain of "Medication" were ranked highly in the other two studies [52,54]. Interestingly, the authors in one study [53] found that patients feeling "Emotionally affected" was the strongest predictor for a problem being rated as important (OR 11.1 CI 6.73 to 18.33), whereas "Poor prognosis" was the strongest predictor for clinicians (OR 6.39 CI 4.61 TO 8.87)

Condition-focused priorities
Two studies reported patient priorities in relation to specific conditions or groups of conditions [44,58], in the context of multimorbidity. Zulman et al. reported that "Diabetes/glycaemic control" was most frequently ranked as "most important", with "Hypertension" coming second [44]. However, the sample of patients included in this study were all diabetic, hypertensive patients. Deruaz-luyet et al. found that musculoskeletal conditions including back pain, were most frequently reported to be the most important conditions for their patients, however endocrine/metabolic conditions (including obesity) were second and cardiovascular conditions were third [58].
Three studies reported condition-focused priorities of clinicians in the context of multimorbidity. Herzig et al. reported the priorities of general practitioners alone [56], and found that "multiple sclerosis", "mental retardation", and "bronchus lung neoplasm" were all highly prioritised by their participants. Zulman et al reported the priorities of "primary care providers" who consisted of physicians, physician assistants or nurse practitioners [44], and found that diabetes was the top priority for primary care providers, with hypertension coming second, in alignment with their previously described patient priorities [44]. Moore et al. examined the priorities of different types of clinicians, including family physicians, geriatricians and nurse practitioners [36], and as with Zulman et al., found that diabetes was the top priority for family physicians and also nurse practitioners, whereas dementia was the top priority for geriatricians [36]. In addition, heart failure, atrial fibrillation and hypertension formed three of the top five conditions considered to be most important by the family practitioners in the study [36].

Treatment priorities
As part of a study to examine the influence of the risks and benefits of medications on treatment preferences of patients, Caughey et al. also examined the priorities of patients in the face of "competing outcomes" [47]. They found that 80% of participants would not be willing to take medication to reduce "joint pain", if the medication increased their risk of a myocardial infarction by 10%. However, this was deduced from a sample of only 15 patients [47].

Agreement between patients and clinicians
Five of the included studies investigated the level of agreement in priority rankings between patients and their clinicians [44,[52][53][54]58]. Three studies reported a low level of agreement between patient and clinicians' priority rankings [52][53][54]. Two of these studies used a Cohen's Kappa calculation to estimate the degree of agreement between the importance ratings of patients and clinicians, and the values of which were 0.18 and 0.11 respectively, indicating "slight agreement" after allowing for chance [53] [54]. One study used a weighted kappa calculation to measure the degree of agreement, which, at a pre-intervention point in this study, was low at 6% [52]. Two studies reported that there was a "high" level of agreement [44,58]. Deruaz-Luyet et al. found that in the case of 54.9% (n=314) of their patients, the condition that their GP had considered to be either the first or second most important, was in the same disease-group as the condition that the patient considered to be most important [58].
Zulman et al. reported that 60% of "patient-provider pairs" had a "high concordance", meaning that the same three conditions had been rated as top three priorities by both parties, or that two of the same conditions had been rated in the top three priorities by both parties [44]. In this case, given that the sample of patients were all diabetic and hypertensive could have led to a narrowing of the range of chronic diseases across the sample, which in turn could have led to an increased likelihood of agreement. However, the participant characteristics reported by the authors state that the patients had a mean of eight health conditions (SD 3.00), suggesting that the patients did not have a narrow range of chronic diseases. Furthermore, the questions posed to patients and providers were phrased differently, in that providers were asked to choose the top three most important medical concerns "that are likely to affect health outcomes for this patient", whereas patients were asked to choose their top three most important health concerns. The authors acknowledge this in their paper, and justify this difference as being due to their aim of exploring the concordance in priorities about the "most important problems facing the patient", rather which problems "providers thought the patient would have prioritised", which, they argue, is a different concept to their aim [44].

QUALITATIVE SYNTHESIS
Whilst our quantitative synthesis allowed us to investigate which health outcomes, diseases or treatments were important to patients with multimorbidity and their clinicians, our qualitative analysis enabled us to explore how prioritisation occurs. Below, we describe the key findings from our qualitative analysis.

Mechanisms of prioritisation
In the qualitative studies that approached prioritisation from a disease-specific perspective, patients were able to identify an illness as their main priority [49,50]. For many patients, prioritisation appeared to be driven by their experience of the illness, which formed part of its "meaning as consequence" [50] as phrased by Cheraghi-Sohi et al. The 'consequences' of an illness consisted of the impact that the illness was having on the patients' everyday lives, which included functional limitation and the symptomatic burden of the illness, including its "unpredictability" (Table 3) [49]. For others, prioritisation appeared to be driven by their perception of the risk now and in the future with respect to functional deterioration and mortality.
In other studies, patients framed their priorities between quality of life vs length of life (Table 3) [42]. Patients in the study by Naik et al who were adults with multimorbidity and suffering from cancer, prioritised "quality of life" more highly than "length of life" [42]. This was also reflected in the findings of Fried et al., who found that when considering medication with competing outcomes in terms of extending life compared to quality of life, participants appeared to prioritise preserving quality of life [41].
Van Summeren et al. found that prioritisation was "difficult" when there was no "specific need" for a treatment decision to be made [45]. This concept of a difference in prioritisation based on hypothetical, or experiential levels, was also shared in the findings of Elliott et al [43] and Fried et al [41]. Where clinicians' perspectives were explored alongside patients, clinicians reported that exploring patients' priorities was "extremely important" when managing "competing interests" [47] and beneficial in providing patient-centred care [45]. Some clinicians in the mixed-methods study carried out by Van Summeren et al. reported that exploring their patients' priorities allowed them to have a "deeper understanding" of the patient, helped with making patient-centred treatment decisions and advance care planning (Table 3) [45]. However, other clinicians in the same study found exploring patient priorities to be difficult due its "novelty" and the fact that it represented a change to their usual consultations [45].

Mechanisms of prioritisation
Facilitating clinicians' decision making "In future, I'll be happier to be more decisive in keeping an eye on what we do and do not do as regards this patient." Van Summeren et al [45]

Factors influencing prioritisation:
Our analysis revealed that there were a number of factors that appeared to influence how both patients and clinicians arrived at their priorities, and which priorities they chose.

i. Functional ability
Preserving functional ability as a priority for patients was a dominant concept across the majority of the qualitative studies [37,38,49,51] [42][47] [41]. Preserving independence emerged as the most significant reason for prioritising functional ability for patients, and maintaining the ability to engage in activities of daily living, mobility, maintaining cognitive ability and wanting to avoid being a "burden" or lacking social support to help them cope with functional deterioration (Table 4) [38,49,50].
Conditions which caused limitation to patients' ability to self-manage their health conditions, led to a "tension" between the patients' expectations of themselves and what they were physically able to do [51]. Lifestyle management, particularly reduced ability to exercise and the adverse impact of this on weight, was cited as part of patients' ability to self-manage [49].
Maintaining patients' functional ability was reported as a priority by some clinicians [37] [47]. Clinicians considered the wider implications of the patients' functional deterioration, particularly cognitive deterioration, and spoke of wanting to reduce the risk of "burnout" for the patients' family members/caregivers [37].

Symptom control
The symptomatic burden of a condition contributed to its "meaning as consequence" for patients [50]. Symptoms were cited as being a cause of functional limitation [38,49], and in some cases their "unpredictability" could cause significant disruption to patients' daily lives [49]. Symptom control was reported to be a priority by some clinicians [37][47]. However, clinicians in one study considered symptom control to be less important, particularly when there was no risk of mortality [55]. In these cases, clinicians seemed to be aware that patients may still be prioritising symptom control highly, even if the clinicians did not (Table 4).

iv. Treatment burden
Factors related to the treatment burden of an illness appeared to adversely impact prioritisation for patients, leading to de-prioritisation of certain medications and treatments [38,41,43,48]. Elliot et al. reported that cost and distressing side effects, were factors which led patients to stop taking a medication [43]. Similarly, Fried et al. found that patients reported unpleasant side effects to be a "competing outcome", which negatively influenced their decision regarding continuing a medication [41]. However, difficulty with achieving control over the management of an illness, as well as requirement for high levels of engagement with self-management, emerged as factors that contributed to the prioritisation of an illness by some patients (Table 4) [49].

Concept Examples from included studies
Functional ability "I mean, because I have to be mobile, I am living on my own, no one is going to take care of me, I have got to look after myself.." Cheraghi-Sohi et al [50] Mortality "Well I really do worry the most about the high blood pressure. 'Cause see you know you got arthritis and you can tell when it's coming on. But you can't hardly tell about high blood pressure. It can just hit you like that [snaps fingers] …."" Lindsay et al [49] Symptom control "I would not want to live with pain. I won't allow that to happen"-Naik et al [42] Disparity in prioritisation of symptom control ".. I talk [to her] for a quarter of an hour about this and that every time after which she replies, "but my vertigo," and I answer every time, well, unfortunately there is nothing I can do about it, we have already tried and done everything. But it is probably the first diagnosis she will mention: "What are you suffering from?". "Vertigo". For me, this would be somewhere all the way at the bottom." -Hansen et al [55]

Factors influencing prioritisation
Treatment burden "It's the knee that's the most concerning because everything else is controlled by tablets. The knee is a problem because if I have one little slip I'm in plaster again for 6 weeks." Lindsay et al [49] 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 [73]. Whilst these studies represent very important areas of research, they were not within the scope of our aim in this review i.e. identifying studies that report the health outcome and treatment priorities of patients with multimorbidity or those of clinicians in relation to patients with multimorbidity. A discussion from our synthesis of findings of the included studies in this review is presented below.

Health outcome and treatment priorities
From our findings, patients' prioritisation appeared to be driven by weighing up the empirical compared to the hypothetical impact of a disease, whereby the empirical impact of a disease, which included its impact on function, symptomatic and treatment burden, was the most dominant driver of prioritisation. This is consistent with the findings of previous literature showing patients with rheumatoid arthritis who had reported experiencing higher levels of pain, were more likely to report pain as a priority [74].
Amongst empirical factors, preserving functionality emerged as most highly prioritised by patients amongst the quantitative studies that took a health outcome approach [39,40,46], whilst "function" was a domain that was prioritised highly by both patients and clinicians in the studies where prioritisation of various health problems were investigated [52][53][54]. From our qualitative findings, functional ability formed a key part of the preservation of various aspects of the patients' independence and their quality of life, as well as their ability to self-manage. Existing evidence shows that the prevalence of multimorbidity is highest in those aged over 65 years [75], and the population for the majority of the included studies were older adults with multimorbidity. This could provide an explanation for why preserving functionality was highly prioritised.
Prioritisation was not a static process and was subject to change, based on factors such as illness exacerbations, life events, whether there was a need for a treatment decision to be made, and whether the priority related to retrospective or prospective healthcare [49,51]. When considering the hypothetical impact of an illness, perceptions of future risk came into play, and in particular, the risk of mortality [43]. This was particularly evident in relation to cardiovascular disease, where patients appeared to perceive the risk of mortality to be high [38].
Risk of mortality was a dominant driver for prioritisation amongst clinicians. This was shown in our quantitative synthesis, where amongst studies assessing disease-specific priorities, conditions with a higher risk of mortality, such as cardiovascular disease and diabetes, recurrently emerged as being highly prioritised by clinicians [36,44,56] and differentiated by age [47]. This age-based consideration could explain why clinicians prioritised "quality of life for the patient" as higher, albeit marginally, than "life expectancy of the patient" in their clinical decision-making for de-prescribing for elderly, patients with multimorbidity [57].
Smith et al previously developed a "Core Outcome Set" [76] in which a Delphi consensus panel formed of 26 international health experts, identified and prioritised a set of outcomes tailored for  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 application to research studies targeting patients with multimorbidity. Mortality, mental health outcomes and quality of life featured most highly in their list of prioritised outcomes, which also emerged in this review. However, we found that relatively few studies reported the prioritisation of mental health outcomes, with the exception of the studies that took a problem-based approach to prioritisation, where problems with regard to "Mood" were prioritised highly by both patients and clinicians [52][53][54].
Our findings show a varying degree of agreement between the priorities of patients with multimorbidity and clinicians. Previous studies carried out in the context of diabetes [77], and psoriasis [78] have found a low level of agreement on health outcome and treatment priorities between patients and clinicians, which correlates with the findings of some studies included in this review [52][53][54], but not others [44]. The nature of the patients' illnesses emerged as a factor for concordance or discordance of priorities with their clinicians [37]. Patients and clinicians were in agreement in situations where patients were currently experiencing an exacerbation of a particular condition, or had a "stable" state of health. However, in patients who suffered from illnesses with more complex courses, discordance of priorities tended to occur between patients and clinicians [37].
In recent times, the traditional paternalistic model for the doctor-patient relationship has given way to an egalitarian model [79], where doctors and patients each play an equitable role in a shareddecision making process, which places the patient at its core and thus achieving greater patientcentred care [80] [79]. A shared agreement between patients and doctors on treatment priorities have been highlighted to play an important part in achieving patient-centred care and creating a therapeutic alliance, the benefits of which can include improved treatment adherence [79,80]. Indeed, Jowsey et al found that agreement between patients and clinicians in the formulation of care plans promoted adherence to these plans, whereas a lack of agreement led to disengagement with care plans by patients [81].

Strengths and limitations
To our knowledge, this is the first systematic review to assimilate and compare the findings of existing literature on the health outcome and treatment priorities of both patients and clinicians for patients living with multi-morbidities. In this review, we have been able to add a novel line of argument to the ongoing discussion on this subject. By incorporating papers using both qualitative and quantitative methodologies, we have been able to explore patterns and relationships in the findings of a wide range of studies, thus creating a comprehensive and well-rounded systematic review.
There are noteworthy limitations. We did not include the term "comorbidity", in our search terms, and whilst "comorbidity" is distinctive from multimorbidity, there is also some conceptual overlap between the two terms. We felt that including "comorbidity" in our search strategy would identify studies focusing on a specific condition rather than multimorbidity.
A number of the quantitative studies did not use pre-validated tools to ascertain priorities [36,54,[56][57][58], leading to a risk of measurement bias, which could limit the generalisability of findings in this review. All of the included studies were conducted in developed, western countries, which limits the global generalisability of our findings, as the priorities of patients with multimorbidity and of clinicians in developing and/or eastern countries may differ to the findings of this review.
We also detected a large variation in the tools used to ascertain priorities, which meant that carrying out a meta-analysis to synthesise the findings of the quantitative studies was not possible. Yet, we  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   18 have tried to mitigate the lack of meta-analysis by using a well-described and well-established method of narrative synthesis [33], in order to maintain rigour and transparency.
Another limitation is that in our inclusion criteria we chose to also include studies which did not explicitly specify a definition of multimorbidity as "two or more chronic conditions" in their inclusion criteria but had a sample of participants that were reflective of multimorbidity (i.e. with a minimum of two chronic conditions which could be identified from participant demographic data). We chose to do this as in the absence of a universally accepted and uniform definition of multimorbidity, we sought to base our judgement on the inclusivity of each paper on its value in answering our review question. This, along with the previously discussed difficulty in defining prioritisation, may have introduced a degree of subjective interpretation in the process of study selection, despite our attempt to mitigate this by incorporating independent review of the results of our literature searching by two reviewers in duplicate.

Recommendations for the future
We recommend that future guidelines developed for clinicians in the management of multimorbidity highlight the need to elicit and consider both short term and long term priorities for their patients', as our review has shown that patients' priorities for their current illness experiences and future risks posed by illnesses, may differ. In accordance with current NICE guidance, we also reiterate the need to review these priorities continually, and particularly when exacerbations, changes to illness course or treatment regimens, or other wider socially-contextualised changes occur in their patients' lives.
There was a large variation in how priorities were ascertained, and in the tools used to ascertain priorities. The relative lack of standardised and validated tools for use to ascertain patient priorities in everyday clinical practice has also been described in previous literature [82]. We highlight a need for the development of a standardised and validated tool that is acceptable to both patients and clinicians, and can be used to ascertain patient-priorities in the multiple dimensions described in this review. Such a tool would a valuable aid to treatment decision-making, advance care planning and achieving patient-centeredness for patients living with multimorbidity.

Low risk
Participants were only informed of the procedures of their own trial arm. Blinding of outcome assessment (detection bias) Low risk Participants were blinded to the pre-intervention importance ratings, when completing the final importance ratings. Incomplete outcome data (attrition bias) High risk 25 patients dropped out prior to baseline ratings and 5 further patients dropped out prior to final ratings, these patients were excluded from analysis, however intention to treat analysis cannot be carried out in this context due to the nature of the intervention Selective reporting (reporting bias) Low risk Adequate reporting on all of the specified outcomes Other bias None detected      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