Prescribing practices, patterns, and potential harms in patients receiving palliative care: A systematic scoping review

Background Patients receiving palliative care often have existing comorbidities necessitating the prescribing of multiple medications. To maximize quality of life in this patient cohort, it is important to tailor prescribing of medication for preventing and treating existing illnesses and those for controlling symptoms, such as pain, according to individual specific needs. Objective(s) To provide an overview of peer-reviewed observational research on prescribing practices, patterns, and potential harms in patients receiving palliative care. Methods A systematic scoping review was conducted using four electronic databases (PubMed, EMBASE, CINAHL, Web of Science). Each database was searched from inception to May 2020. Search terms included ‘palliative care,’ ‘end of life,’ and ‘prescribing.’ Eligible studies had to examine prescribing for adults (≥18 years) receiving palliative care in any setting as a study aim or outcome. Studies focusing on single medication types (e.g., opioids), medication classes (e.g., chemotherapy), or clinical indications (e.g., pain) were excluded. The review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines for scoping reviews, and the findings were described using narrative synthesis. Results Following deduplication, 16,565 unique citations were reviewed, and 56 studies met inclusion criteria. The average number of prescribed medications per patient ranged from 3 to 23. Typically, prescribing changes involved decreases in preventative medications and increases in symptom-specific medications closer to the time of death. Twenty-one studies assessed the appropriateness of prescribing using various tools. The prevalence of patients with ≥1 potentially inappropriate prescription ranged from 15 to 92%. Three studies reported on adverse drug events. Conclusions This scoping review provides a broad overview of existing research and shows that many patients receiving palliative care receive multiple medications closer to the time of death. Future research should focus in greater detail on prescribing appropriateness using tools specifically developed to guide prescribing in palliative care and the potential for harm.


Introduction
As a consequence of population aging, there is an ever-increasing demand for palliative care for individuals with limited life expectancy. 1 Palliative care is defined by the World Health Organization (WHO) as "an approach that improves the quality of life of patients and their families facing the problems associated with life-threatening illness, through the prevention and relief of suffering by means of early identification and impeccable assessment and treatment of pain and other problems, physical, psychosocial and spiritual." 2 Historically, palliative care was synonymous with end-of-life care provided through hospices. 3 Increasingly it is recognized that palliative care is applicable in the early stages of a life-limiting illness, in conjunction with other treatments intended to prolong life, and is not limited to hospice settings. 4 Therefore, the scope of palliative care encompasses the care provided to individuals from the point of diagnosis of any life-limiting illness through to end of life. 5 Adequate provision of palliative care is recognized as a major public health issue, and dedicated strategies are required to ensure effective integration of palliative care into healthcare systems. 6 A key component of any such strategy involves ensuring the availability of necessary medications intended to treat existing conditions and relieve symptoms experienced by patients at the end of life, such as pain, and ensure that they are prescribed appropriately.
Ensuring appropriate prescribing for patients receiving palliative care is a major challenge to improving quality of life and is an under-researched area. [7][8][9][10] Patients with limited life expectancy often have existing comorbidities necessitating the use of polypharmacy which is commonly defined as the prescribing of five or more medications. 8,9 Optimising medication regimens requires clinicians to consider whether each medication is appropriate in relation to patients' context, treatment goals, and life expectancy. 9,11,12 Under these circumstances, the goal of prescribing moves from preventing and treating existing illnesses to controlling symptoms, such as pain, and improving patients' quality of life.
In recent years, discussion regarding opportunities for deprescribing across healthcare settings has been presented, primarily in the context of older adults (≥65 years). 13 Deprescribing is defined as a systematic process involving identifying and discontinuing medicines in cases where potential or existing harms outweigh benefits. 14 This process is conducted within the context of the individual patient's care goals, values, preferences, and current level of functioning, and life expectancy. 14 Previous reviews have examined prescribing for patients with life-limiting illnesses and focused on preventative medications (i.e., chronic medication used to treat or prevent further worsening of a disease state). 10,15 This has helped to characterize the commonly prescribed types of preventative medicines, as well as the methods used to identify potentially inappropriate medications (PIMs) and opportunities for deprescribing. However, it remains unclear whether patients were receiving key palliative care medications required for optimal symptom control towards end of life, such as appropriate analgesia. [16][17][18] This scoping review aimed to provide an overview of observational research on prescribing practices, patterns, and potential harms in patients receiving palliative care. The objectives were: 1. To examine the number and types of medications prescribed (i.e., preventative and symptom-specific medications) for patients receiving palliative care; 2. To investigate the methods used to assess the appropriateness of medication prescribing for patients receiving palliative care; 3. To examine the risk factors/determinants of potentially inappropriate prescribing for patients receiving palliative care; 4. To establish the types of potential harms (i.e., adverse drug events, drug interactions) associated with prescribing for patients receiving palliative care; 5. To examine changes in medication prescribing for patients receiving palliative care over time.

Material and methods
This scoping review was conducted and reported in accordance with relevant methodological guidance and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines for scoping reviews (PRISMA-ScR) [Appendix A]. 19,20 The review protocol is available from the authors on request.
For the purpose of this review, palliative care was defined using the WHO's definition, as outlined in the introduction above. 2 In order to meet inclusion criteria, studies must have examined medication prescribing for adult patients (≥18 years) receiving palliative care for any life-limiting illness in any setting as a study aim or outcome. This could encompass one or more of the following: (1) assessments of prescribed medications; (2) assessments of the appropriateness of medication prescribing and/or associated risk factors for potential harms; (3) assessments of changes in medication prescribing over time. At a minimum, studies must have provided a summary statistic regarding the number of medications that patients were receiving and information on the types of medication prescribed. Studies that also examined prescribing in patient groups that were not specifically receiving palliative care were eligible for inclusion, provided that data for the palliative care group were reported separately. Eligible study designs consisted of cross-sectional, case-control, and cohort studies. Any assessment of the appropriateness of prescribing was acceptable, including clinicians' professional judgment and validated assessment criteria (e.g., Beers criteria for older adults). 21 Only full-text manuscripts published in English were eligible for inclusion. Studies were excluded in each of the following instances: medications that patients were receiving; • Studies focusing on single medication types, medication classes, or clinical indications as they did not provide a holistic overview of prescribing practices for sample populations; • Studies of patients with life-limiting illnesses that were not explicitly receiving palliative care; • Non-English language publications; • Published conference abstracts due to a lack of sufficient information.

Search strategy and data extraction
Electronic searches were conducted using PubMed, EMBASE, CINAHL, and Web of Science from the date of inception to May 2020 using established search methods for scoping reviews (Appendix B). 19 Briefly, preliminary searches of each electronic database were undertaken to identify keywords and index terms for articles relating to the review topic. This informed the development of a comprehensive search strategy developed with the assistance of a research librarian using all identified keywords and index terms for each electronic database. Key search terms included: palliative care, end-of-life care, life-limiting illness, and prescribing. Following completion of the electronic database searches, reference lists of all studies meeting inclusion criteria were screened for additional studies.
All abstracts were screened for inclusion by one author (CC). A 20% sample of abstracts was double screened by a second author (MM). If a study appeared to meet inclusion criteria, full-text articles were retrieved and assessed for inclusion by two authors working independently (CC, MB). Any disagreements were resolved by discussion with other members of the research team.
One author (CC) performed data extraction using a data extraction form that was developed in accordance with relevant methodological guidance. 19 The data extraction form was piloted on a sample of three included studies and refined accordingly. Data were extracted relating to each of the following key headings: The data extraction process was intended to enable a logical and descriptive summary of the review findings to be presented that aligned with the review objectives.

Quality assessment of included studies
As the aim of a scoping review is to provide a broad overview of the existing literature relating to the research question, formal assessments of the methodological quality of included studies are not routinely undertaken. 19 However, in summarizing, synthesizing, and interpreting the body of literature identified in this review, critical appraisal was conducted focusing specifically on the generalisability of study findings.

Data analysis and synthesis
Palliative care populations can differ extensively with respect to age, diagnoses, functional status, symptom burden, and survival. 22 In light of this and observed heterogeneity in previous related reviews, 10,15 the findings of this review were described using narrative synthesis, which involved the following key steps. 22 (1) A preliminary synthesis of the findings of included studies was developed in which study characteristics and findings were tabulated to summarise key information. (2) Extracted study data were reviewed to explore any relationships in the data. (3) The review team critically reviewed the findings of the synthesis process in terms of the available evidence and potential limitations of the evidence sources, and any discrepancies and uncertainties identified relating to the review questions.
NVivo QSR 12 was used to manage the extracted data. This involved coding the extracted data under key headings from the data extraction form (outlined above) and performing a content analysis of this data to identify key similarities and differences across included studies.

Search results
Following deduplication, the electronic searches identified 16,565 unique citations. Following title and abstract screening, 754 full-text articles were reviewed for eligibility. In total, 56 studies met inclusion criteria ( Fig. 1).  Three studies had more than one reference. 63,64,66,67,79,80 All other articles did not meet the inclusion criteria. Table 1 provides an overview of the characteristics of included studies. Study designs consisted primarily of observational cohort studies (52 studies), 14 of which were conducted prospectively. Four studies were based on data collected as part of cross-sectional surveys. 24,30,68,74 The studies were conducted across 25 countries. Two studies were multinational, involving three and 12 countries, respectively. 56,58 Studies were primarily conducted across hospice settings (n = 16) and dedicated palliative care centers, units, and/or services (n = 22). Other settings included general practice (n = 1), hospitals (n = 12), nursing homes (n = 3) and long-term care facilities (n = 1). One study was conducted across academic and communitybased clinical sites that formed part of a clinical trial led by a palliative care research group. The number of study sites varied (range 1 to 1174), and 27 studies were conducted within a single site.

Study design and participants
Sample sizes ranged from 25 to 88,957 patients (Table 1). Four studies involved nationally representative samples of palliative care patients using surveys 24,30 or population datasets. 26,79 Across included studies, patients' gender profiles varied (25 studies had a majority of female patients, 27 studies had a majority of male patients) with an average age ranging from 39 to 86.7 years. Cancer was the most commonly reported life-limiting illness across studies, with 19 studies focusing specifically on patients with cancer. The time points over which assessments occurred varied and included referral/admission to palliative care and over the last one to two weeks of life (Appendix C). Eleven studies involved cross-sectional assessments of patients receiving palliative care without any clearly identifiable time point.

Prescribing appropriateness
Twenty-one studies assessed the appropriateness of prescribing. Summary details of each assessment tool (n = 14) are provided in Table 2, which included established tools for assessing appropriate prescribing in the general older population (i.e., Beers criteria, Medication Appropriateness Index), as well as study-specific tools for defined patient populations (e.g., cancer, dementia). The prevalence of patients with ≥1 PIM ranged from 15 to 92% (Table 3). Commonly identified PIMs across studies included lipid-modifying agents, antihypertensives, anti-thrombotic agents, and drugs for peptic ulcer and gastro-oesophageal reflux disease. Four studies examined patient factors associated with PIMs. 34,37,46,80 One study found that PIMs more commonly occurred in patients who were bedbound, had the shortest life expectancy, or were discharged from the hospital and admitted to the hospice. 34 Another study found a significant association between polypharmacy (≥5 medications) and PIM use. 37 The third study reported various demographics (e.g., increased age, residing in nursing or assisted living facilities) that increased the likelihood of continuing medication with limited benefit after hospice admission. 80 The remaining study found no patient-specific factors associated with the incidence of PIMs. 42 Only one study reported on under-prescribing. 34 This study reported concomitant drug deficiency (e.g., absence of laxatives in the cases of regular administration of strong opioids) in 31.5% of patients and an absence of drugs for specific symptoms (i.e. pain, seizures, depression, delirium, thrombosis) in 2.1% of patients.

Potential harms
Three studies reported on ADEs (i.e., harms caused by medication use). 37,54,57 One study examined prescribing for end-of-life care patients     within an acute care unit for older adults in a secondary care hospital. 54 On admission, ADEs were identified in 21% of patients. The most commonly identified ADEs were symptomatic hypotension, blood disorders, falls, and hypoglycemia. The study reported a significant positive correlation between the number of prescribed medications and the incidence of ADEs and a significantly higher prevalence of ADEs in patients with inappropriate prescriptions compared to patients with appropriate drug therapy (37.7% vs. 5.35%, p < 0.001). However, the severity, causality, and preventability of identified ADEs were not assessed. Another study examined adverse drug reactions (ADRs) in patients receiving palliative care during an inpatient hospital admission over one year. 57 The study reported that 57.4% of patients experienced at least one ADR. The most commonly affected organ systems were gastrointestinal, neurological, and dermatological. The medications most commonly associated with ADRs were antimicrobials, opioids, and anticoagulants. The remaining study reported on chemotherapyrelated toxicity, which was observed in 53.8% of older patients with cancer receiving first-line palliative chemotherapy. 37 Forty-one percent of patients visited an emergency room or were hospitalized due to chemotherapyrelated toxicity. A significant association was identified between polypharmacy (≥5 medications) and hospitalization or emergency room visits in these patients. Eight studies reported on drug-drug interactions. 32,34,35,37,59,60,63,73 Interaction detection relied primarily on different computer software systems which classified drug interactions according to potential severity. In three studies, these software systems were supplemented by reviews and classification by healthcare professionals. The proportion of patients with at least one potential drug interaction ranged from 12% to 64%, with further subcategories according to severity level. Identified risk factors for drug interactions included advanced age, presence of comorbid illness, and an increasing number of medications. One study examined the prognostic impact of potential drug interaction on overall survival in patients with advanced cancer receiving palliative care. 35 This study found that major-risk drug interactions were not significantly associated with overall survival in the study population. Another study involving older patients with cancer receiving first-line palliative chemotherapy reported no significant association between drug interactions and chemotherapy-related toxicity. 37

Prescribing changes over time
Thirty-two studies reported on changes in prescribing over time (Appendix C). Reported details of the prescribing changes varied, with some studies reporting on specific types of medications and others focusing more broadly on changes in the numbers of medications prescribed. The time points over which changes were assessed and reported also varied, which precluded a detailed synthesis. Commonly reported assessment time points included: during transition to palliative care, from admission/ referral to palliative care to death, and over the last one to two weeks of life. Typically, prescribing changes involved decreases in preventative medications and increases in medications for symptom control as the time of death approached.
Two studies examined prescribing trends longitudinally using repeated cross-sectional analyses. 75,79 Twycross et al. reported changes in the most commonly prescribed medications within a single palliative care unit between 1988 and 1992. 75 Morphine and co-danthrusate were consistently identified as the most commonly prescribed medications across all study years. Zueger et al. used a nationally representative population database to examine the most commonly dispensed medications to patients as part of a health insurance program (Medicare Part D) after hospice admission between 2008 and 2013. 79 The study reported little observed variation in the prevalence of the preventative drug classes (e.g., lipid-modifying agents, antihypertensive agents) examined.
Across the eleven studies that examined changes in the prevalence of PIMs over time, decreases in the prevalence of PIMs were typically reported as death approached (Table 3). One notable exception to this was the study by Currow et al., 27 which used Beers criteria to assess both symptomspecific medications and medications for comorbid conditions. The study found that over the assessment period (from patient referral to the palliative care service until death) the proportion of high-risk, symptomspecific PIMs increased (29% to 48%) whereas the proportion of high-risk PIMs for comorbid conditions remained stable (13% to 15%).

Discussion
This review provides an overview of existing observational research on prescribing practices, patterns, and potential harms in patients receiving palliative care. The broad, scoping nature of the review was intended to overcome limitations of previous reviews, which focused solely on preventative medications among patients with any life-limiting illness irrespective of the type of care received. 10,15 Despite the inclusion of 56 studies, the review highlights a limited assessment of prescribing appropriateness, potential harms, and prescribing trends across included studies.

Prescribing in palliative care
The review shows that many patients with palliative care needs received a considerable number of medications at various time points towards the end of life. A number of studies referred to the term 'polypharmacy,' which has been widely discussed in the context of prescribing for the general older population. 82 The studies mirrored previously used definitions for the older population in terms of the numerical thresholds and overall variation. 83 However, a key challenge in critically reviewing the numbers of medications prescribed was that studies often did not clearly differentiate according to treatment intention (i.e., preventative versus symptomatic relief) or examine prescribing changes over time. One study did find that the total number of medications increased closer to death due to the continuation of medications for comorbid conditions and the addition of symptom-specific medications. 51 This highlights the importance of classifying medications according to treatment intention in order to review the medications prescribed critically. It is also important to recognize that the number of medications is only one factor contributing to overall treatment burden (i.e., the work that patients must do to take care of their health). 84 Other medication-related factors that may exacerbate treatment burden include challenges with taking the medication due to the complexity of treatment regimens and any medication-related side effects.

Prescribing appropriateness
The importance of ensuring appropriate prescribing in patients with limited life expectancy is increasingly recognized. 85,86 Various frameworks, tools, and classification systems have been developed to assist in identifying   An unnecessary medication was defined as any medication that did not result in significant patient benefit in terms of survival or symptom control; lacked evidence to support its use (e.g., unproven efficacy); or where treatment goals were only expected with long-term chronic use (e.g., statins for hypercholesterolemia).

Riechelmann 2009 63
Study-specific assessment criteria List of unnecessary medications identified based on a previous systematic review and list of essential medications identified based on recommendations of three different healthcare organizations. Both lists were reviewed by three clinicians.  Progress has been made in developing tools that focus on prescribing in palliative care. For example, the OncPal deprescribing guideline 46 and PEACE Programme criteria 36 have been developed to guide prescribing in palliative care for patients with cancer and dementia, respectively. However, existing tools are primarily focused on the deprescribing of unnecessary medications. This is evidenced by the review findings whereby the reported cases of potentially inappropriate prescribing primarily involved medications that were deemed inappropriate or futile. There is growing evidence to support the discontinuation of preventative medications, such as statins, towards end of life. 92 However, it is important to recognize that the concept of potentially inappropriate prescribing is broader than overprescribing (i.e., prescribing where no clinical indication exists) and misprescribing (i.e., prescribing incorrect doses, frequencies, or durations of treatment that significantly increase the risk of adverse events). 93 It also includes underprescribing (i.e., the omission of medications for specific clinical indications aimed at prevention or treatment). This is an important issue as patients with palliative care needs experience variable levels of symptoms, and underprescribing of analgesics and other symptomspecific medications has been documented in palliative care populations. [94][95][96][97][98][99] This was evident in the only included study that reported on underprescribing which encompassed concomitant drug deficiency (e.g. absence of laxatives in the cases of regular administration of strong opioids) and an absence of drugs for specific symptoms (e.g. pain). 34 However, exact details of how underprescribing was assessed were not reported and no formal assessment tool was cited. The International Association of Hospice and Palliative Care previously developed a list of essential medicines for treating commonly encountered symptoms in palliative care. 17 However, this list is intended to guide decisions regarding medication availability for palliative care within healthcare systems in satisfying the healthcare needs of the population as opposed to the appropriateness of the individual medications for use in specific populations (e.g. older patients with advanced cancer).

Potential harms
In addition to the limited number of assessments of prescribing appropriateness, only three studies examined ADEs, all of which focused on older adults receiving palliative care in inpatient settings. 37,54,57 The findings were consistent with research into medication-related harms in the general older population, whereby a higher risk was associated with an increasing number of medications. 100 It remains unclear how changes in patients' medication profile towards the end of life impact the potential for harm, particularly in terms of the addition of medications for symptomatic relief. The included studies that examined drug interactions highlighted considerable risks of harm. 32,34,35,37,59,60,63,73 However, further research is required to determine the extent to which these risks translate into actual harm as the only two studies that examined the impact of drug interactions on clinical outcomes found that they were not associated with chemotherapy-related toxicity 37 or overall survival in patients with cancer receiving palliative care. 35

Going forward
In advancing research into prescribing in palliative care, it would be important to consider how the synthesis and generalisability of study findings could be enhanced. There were considerable challenges in providing a meaningful synthesis of included studies due to observed heterogeneity. For example, there was variation in terms of the study populations, settings, assessment time points, and information reported for key outcomes of interest to the review, which impacts the applicability and generalisability of the review findings. This is a recognized issue in palliative care research. 22,[101][102][103] For example, an international multicentre study of palliative care centers across Europe identified wide variation in terms of both the services provided and patients receiving care. 103 A basic dataset of patient characteristics and medical variables to describe a palliative care cancer population has been developed to standardize reporting. 22 This tool has undergone pilot testing, and there is scope for adapting it to include details of other medical conditions. 104 Many of the studies were also limited by their single-site design. Population datasets or clinical registries may help to provide more generalizable findings with the potential to examine longitudinal trends over time. 105 It is perhaps unsurprising that cancer was the most common diagnosis across included studies, given the historical focus on cancer in palliative care. However, it is important to look at palliative care requirements in other patient populations to avoid the potential for care inequalities previously reported. 106,107 Further work is needed to examine the appropriateness of prescribing in palliative care. Additional tools may need to be developed or adapted for other populations with life-limiting conditions (e.g., organ failure). Any developed tools should also include assessments of the appropriateness of medications for symptomatic relief and provide a method for systematically and reliably assessing potential underprescribing/omissions of any such medications.

Strengths and limitations
This is the first known scoping review of observational research examining prescribing in palliative care. It provides a broad overview of existing published literature and followed rigorous methods. It identified a sizeable number of studies conducted across 25 countries. However, it must be acknowledged that other studies closely related to the review topic may have been excluded because of the review's exclusion criteria (e.g., population datasets that examined prescribing in the last year of life irrespective of the care received). 108 The inclusion/exclusion criteria were developed to answer the review questions and identify a body of literature that would enable a meaningful synthesis to be provided. The review focused on prescribing of medications for palliative care patients. However, it must be noted that there are other aspects of care towards end of life that may be inappropriate (e.g., diagnostic testing). 109 Additional limitations were that the review only looked at studies published in the English language, and no grey literature searches were conducted, which may have introduced the potential for publication bias.

Conclusions
This scoping review provides a broad overview of observational studies examining prescribing in palliative care. The review shows that many patients with palliative care needs receive considerable numbers of medications, including preventative medications that may provide limited or no therapeutic benefit closer to death. A limited number of studies examined the appropriateness of prescribing or the potential for harm. Future research should look to include assessments of prescribing appropriateness using tools that have been developed specifically to guide prescribing in palliative care. This should also include assessments of the appropriateness of medications to relieve common symptoms experienced by palliative care populations towards the end of life.

Author contributions
CC led on the design of the review protocol and the conduct and writing of the review. MM and MB contributed to study identification. CH, KB, and SMcL each contributed to the development of the review protocol and provided relevant clinical and/or methodological expertise. All authors reviewed drafts of the review and approved the final submission. The above funders had no involvement in this review.

Declaration of Competing Interest
None.