An Examination of Adherence to Pain Medication Plans in Older Cancer Patients in Hospice Care

Abstract

Context

Timely and appropriate management of pain is essential to promote comfort at the end of life.

Objectives

To determine if pain-related factors and non-pharmacological interventions affect medication adherence in older cancer patients in community-based hospices.

Methods

The study involved cancer patients 55 years and older, newly admitted to one of 13 community-based hospices in the Midwest U.S. A descriptive design with patients or their proxies providing information during two telephone interviews and review of their hospice medical records was used.

Results

A total sample of 65 patients was obtained, with data directly from 32 patients during interview one (T¹); 25 at interview two (T²), and proxy reports for 33 (T¹) and 30 (T²) patients. The overall mean pain medication adherence scores (maximum 9) for all patients were 8.43 (T¹) and 8.38 (T²). For component analysis (three components; maximum of three points each), patients were the least adherent with opioid orders at both time points (2.65). Patients were the most adherent to nonsteroidal anti-inflammatory/acetaminophen orders at T¹ (2.91) and to medications for neuropathic pain at T² (2.89). Data provided statistical evidence that patients with more hours of controlled pain in the past 24 hours were more likely to have had better adherence, whereas patients with higher levels of comfort over the last few days were more likely to have had worse adherence.

Conclusion

This study identified that pain medication adherence among older adults with cancer receiving hospice care is high. However, hospices must be alert to the fact that even as patients become more comfortable, adherence must continue to be emphasized to ensure pain does not re-develop or exacerbate, if pain relief is a patient priority.

Introduction

Pain is a common occurrence in many advanced illnesses and adherence to an assessment-based pain management plan, consistent with a patient’s goals of care, is essential to ensure comfort at all stages of the disease. Adherence is particularly important in the final or end-stages of disease, a period in which pain frequently increases as death approaches and at a time when caregiver support or control of the pain management plan may be required.

The timely and appropriate management of acute and chronic pain is a critical component to promoting patients’ health, quality of life (QOL), and economic interest. During the past decades, there have been great advances in the management of pain, resulting in clinical guidelines that encapsulate evidence-based treatment approaches and consensus-based best practices in order to provide practical recommendations to guide care, for example, the American Pain Society/American Academy of Pain Medicine’s Clinical Guidelines for the Use of Chronic Opioid Therapy in Chronic Noncancer Pain¹ and the American Geriatrics Society’s Clinical Practice Guideline, Pharmacological Management of Persistent Pain in Older Persons.² These have been widely disseminated to assist in both the assessment and treatment of chronic and acute pain, as well as more educational initiatives, such as Principles of Effective Pain Management at the End of Life, available through Medscape. These guidelines inform clinical practices and also recommend the involvement of the patient and/or family in the planning and implementing of the treatment plan.

Educating the patient and family has been shown to be an important factor in adherence.³ Despite the availability of such resources, adherence to the prescribed medication regimen remains a significant problem in the management of pain.

Adherence can be defined as the extent to which the recommendations and prescriptions of a medical professional are followed by the patient or, for some, the person serving in the caregiving role.⁴ In looking at a general population of patients, only 50% adhere to the recommendations and prescriptions of their providers,⁵ suggesting that adherence is a critical variable in the successful management of pain, and it needs to be a focus of ongoing counseling.

The literature on adherence in older adults shows great variance in adherence rates. Some research has identified that older patients, when compared with their younger counterparts, are more motivated to be adherent to both pharmacological and non-pharmacological interventions^6–8 because of greater trust in the health care system and clinicians, a greater focus on becoming and remaining healthy, and greater acceptance of the recommendations of their health care professionals.⁷ Other research, however, has demonstrated the opposite, with older patients – and particularly those with chronic health conditions – having non-adherence rates between 26%–75%, with the reasons for non-adherence being associated with admission to residential care or a hospital, the number of medicines prescribed, the costs of the medicines, and functional limitations of older adults.^9–10 Lewis and colleagues found that patients reported additional reasons for under-utilizing medications such as opioids for pain, including financial concerns and the belief that medicines should only be taken if severe symptoms were present.¹¹ Pound et al. noted the main reason people were not adherent to their medications was the result of concerns about the medications themselves and a preference to take as little medication as possible.¹² Non-adherence creates many challenges for medical staff as it makes treatment decisions, such as efficacy of a given drug and respective dosing adjustments, more challenging when it may not be discernible exactly what the patient is taking and when.

In looking specifically at what is known about adherence in cancer patients, concerns about medical adherence are well-founded, given the high rates of pain among this population and the risks of pain crises. Although as many as 70%–90% of cancer patients experience pain by the final stages of the disease,^13–14 cancer patients in general do not always adhere to the pain management plan, with non-adherence rates ranging from 20%–33%.^15–16 Non-adherence in cancer patients has been linked to many factors including fear of addiction,^17–20 concerns about side effects from the pain medication,^8,17 poor communication between patient and physician or the health care team,¹⁹ lack of knowledge about pain from the health care team,¹¹ and views that pain is “just part of the disease.”^17,20 Musi indicated that non-adherence in cancer patients also is related to how patients view pain.²¹ For some, the presence of pain starts to represent a state of normalcy. Thus, patients start to have ambivalent feelings about not having pain, as they recognize that even without pain, they are unable to achieve pre-disease QOL and level of functioning. Although there is literature on adherence in cancer pain in general, there is a lack of research on adherence among older adults with cancer. A major focus of hospice care is on ensuring a pain-free death, to the extent determined by each patient’s unique circumstances and goals of care. However, this focus, even when strongly endorsed by patient and family, can easily become compromised when patients and caregivers agree on the treatment plan and then do not adhere to the pain management plan that has been developed by the hospice interdisciplinary team. Regardless of the education and support provided by the hospice nurse and other members of the hospice team, deeply rooted fears associated with the “what ifs” of pain management remain, with many patients and caregivers concerned about how opioids in particular might impact cognition and ability for continued communication as death approaches.

This paper reports on a study evaluating adherence to a pain management plan in older adults with cancer who are under hospice care. We particularly examined how pain-related factors (average pain, worst pain, hours of controlled pain [none/mild] in the past 24 hours, level of comfort, QOL, and symptom control) and non-pharmacological interventions for pain impact adherence, as well as changes in adherence over two points in time (48–72 hours after admission and 7–10 days following the initial time point). Recommendations for improving patient adherence also are discussed.

We sought to answer the following research questions:

1. Do pain-related factors and non-pharmacological interventions impact pain-treatment plan adherence in older cancer patients in a community-based hospice setting?

2. Is there a difference in pain-treatment plan adherence of older cancer patients between the time of admission to community-based hospice care and after care has begun?

Methods

Study Design

The study population involved patients from 13 community-based home hospice programs in the Midwest U.S. A descriptive correlational design was used to answer the questions regarding the degree of adherence to the pain management plan for older adults with cancer. According to Burns and Grove, a descriptive correlational design is most appropriate when examining relationships between variables, particularly when there is no attempt to control or manipulate the variables.²² Descriptive correlational designs help to facilitate the identification of interrelationships that exist in a situation over the course of a short period of time.

The hospices were all participating in a larger randomized, controlled, experimental study testing the effect of a multifaceted Translating Research into Practice (TRIP) intervention to promote the adoption of evidence based practice (EBP) for pain assessment and management in older adults with cancer at the end of life; they represent small, medium, and large hospices, for profit and non-profit, and both rural and urban settings. Details on the larger study are reported elsewhere.^23–24 Human subjects approval was obtained from the Institutional Human Subjects Review Board (IRB) at the University of Iowa, which served as the IRB of record for hospices without an internal IRB. Approval was obtained from the corresponding human subjects review boards at participating hospices with an internal IRB. All older adults or their legal guardians provided informed consent to participate. In the consent process, capacity for decision making was determined through a series of structured questions that documented understanding of the study risks and benefits. If decision-making capacity was acceptable, the patient was judged able to reliably complete the study instruments. If not, permission for the primary caregiver to complete the instruments as a proxy for the patient was obtained from the older adult’s legal guardian. All participants or their legal guardians provided verbal informed consent and completion of the study interviews were verification of their implied consent. They also provided written informed consent to obtain copies of their hospice medical records for the time frame corresponding with the telephone interviews.

Sample

Patients meeting the following inclusion criteria were invited to participate in the telephone interviews: 1) 55 years of age or older; 2) diagnosis of cancer; 3) newly admitted to a participating hospice; 4) receiving community-based hospice services from one of thirteen hospices in the U.S. states of Iowa, Nebraska or Missouri. “Community-based hospice” was defined as a setting where patients received hospice care in an environment that allowed the patient or their family caregiver to oversee the implementation of the pain management plan (e.g., personal home or assisted living facility). Four hundred thirty-five patients were identified by the participating hospices as meeting the established inclusion criteria. Sixty-five patients (15% of eligible patients) admitted during the study period participated or had their caregivers serve as proxy. A total of 341 patients refused participation in the study. An additional 29 patients gave verbal consent and participated in the telephone interviews but did not return a signed consent document allowing access to their medical record data necessary for determining adherence; therefore, they were not included in this analysis. Reasons for refusing to participate include: lack of interest in participation (n=76); health condition too severe (n=70); caregiver unwilling to assist patient with reporting (n=45); patient was actively dying or died before recruitment call (n=42); patient too fatigued to participate (n=23); patient in the midst of a crisis situation (n=16); patient confused (n=12); patient lacked adequate time for participation (n=11); patient unable to speak English (n=2); hearing difficulties made a telephone interview impossible (n=1); admitted to long-term care facility (n=1); or, patient did not answer repeated attempts to contact via telephone (n=42).

Study Instruments

Brief Pain Inventory (BPI)

The BPI is a valid and reliable multidimensional pain instrument assessing pain history, intensity, location, and quality with excellent reliability across a large number of different cancer pain samples and is relatively free of cultural and linguistic bias. It takes less than five minutes to complete. It also elicits information regarding pain treatment effectiveness and pain-related interference with daily activities, such as sleep, mood, and mobility, among others. Although developed for use with cancer patients, the BPI has been validated and is recommended for palliative care and geriatric patients.^25–26 Because the BPI depends on patient report, it was used in patients whose cognitive abilities enabled them to provide reliable responses. The BPI uses a numeric rating scale (NRS) approach that has been demonstrated reliable and valid for use with older adults, including those with mild to moderate cognitive impairment.^27–28

Cognitive impairment, often present in advanced illness with patients receiving opioid treatment, may interfere with use of the BPI. The family caregiver was asked to serve as the proxy for patients unable to provide a rating of pain intensity. Thus, either the patient provided these data or the data were obtained from the proxy. Although some research suggests that family caregivers can provide valid reports of pain intensity at the end of life, other reports of overestimation of pain intensity by family caregivers suggest consideration of potential differences in the interpretation of proxy reports.^29–30 In this study the Cronbach’s alpha for the BPI completed by the patient was 0.89 and by the caregiver, 0.90. Pain intensity was evaluated using the BPI items “average pain” and “’worst pain” during the past 24 hours.

Brief Hospice Inventory (BHI)

The BHI is a valid and reliable multidimensional instrument developed to assess outcomes among hospice patients that is simple and minimizes subject burden.³¹ The BHI includes pain and non-pain symptom assessment (including depression, anxiety, tiredness, loss of appetite, nausea, shortness of breath, and distress as a result of functional changes), QOL, and symptom control, using a 0–10 rating scale demonstrated as effective for obtaining self-report in frail hospice patients and older adults. In a study of 145 hospice patients, 80% of whom were 70 years or older, two subscales were identified in the BHI: a symptom subscale and a QOL subscale. Internal consistency of the BHI patient survey symptom subscale was 0.88 and for the QOL subscale, 0.94. Test-retest reliabilities between week 1 and 2 ranged from 0.58 to 0.63, with lower reliabilities expected because of changing health status of hospice patients. Correlations between patient and caregiver reports were strong, ranging from 0.71–0.83. For this study, the total score of the BHI was not used; instead, only individual items were used in analyses, rather than the two subscales. In this study, the Cronbach’s alpha for the BHI was 0.80 for patients and 0.78 for caregivers. In the event a patient was unable to complete the BHI, similar procedures as discussed above under the BPI were used.

Patient Pain Management Log (PPML)

The PPML, an adaptation of the published Pain Diary adapted by the grant team, provided data from the patient or the caregiver.³² Information collected related to the patient’s pain experience over the past 24 hours, dosages of medication they had taken in the past 24 hours, as well as extra medications taken, medications skipped, and non-pharmacologic therapies used. This provided information on patient goals and choices regarding their pain and pain management.

Patient Adherence Tool (PAT)

The PAT was adapted from a cancer pain algorithm study, and contains three subscales to measure adherence to analgesic use including: 1) opioid use, 2) nonsteroidal anti-inflammatory drug (NSAID) and acetaminophen use, and 3) neuropathic pain adjuvant use.³³ The PAT combines provider medication prescription information abstracted from the medical record with the patient interview data. The medical record is used only for the completion of the PAT, not as a separate measure. Research assistants (RAs) received 10 hours of training on the use of the PAT and Medical Record Abstraction tool as it related to completing the PAT. Inter-rater reliability of the PAT was established by two trained reviewers on a randomly selected group of ten patient medical records. The two raters were experienced with the medical record abstraction process and common issues arising in interpretation of the evidence-based practices. Initial inter-rater reliability of the PAT was established at 82%, with intra-rater reliability enhanced by having one trained abstractor re-abstract 10 records after a two-month period, demonstrating 93% concurrence.

All medication doses were recorded as they were prescribed, and only NSAIDs, acetaminophen, opioids, and neuropathic pain medications used specifically for pain control were included. For example, if a subject was taking aspirin on a daily basis for cardiovascular prophylaxis, this would not be included as a medication used for pain. The most recent medication orders were used, which excluded multiple orders, and if a combination drug was prescribed, the total milligrams of opioid only were included. Additionally, fentanyl transdermal patch doses were converted from mcg per hour (mcg/hr) to mg per 24 hours (mg/24hr). Dosages from the PPML and the patient medical record were documented on a medication sheet included in the PAT, and then totaled for use on the subscales.

Each of the three subscales included in the PAT contains a scoring system based on the ratio of medications prescribed to medications taken, and a scale representing the patient’s pain score from the BPI. The subscales were divided between medications prescribed “around the clock” (ATC) and “as needed” (PRN). Scores for the ATC medications can range from 0–2 points based on the ratio of total medications taken in 24 hours and total medications prescribed for 24 hours. If there was no order for an ATC medication and the subject had not taken any, the score would be a 2. If the subject had taken 100% of the ordered ATC medication dose, the score also would be a 2. A score of 1 would be counted if the subject had taken ≥50%–≤99% of the prescribed dose, and a score of 0 would be applied if the subject had taken ≤50% of the prescribed dose. There are a possible three points to obtain on each subscale. Step 1 is added to step 2 and then the scores from all three of the subscales are summed and then averaged for the final adherence score. The PAT, with detailed instructions on its use, is available upon request from the author.

Data Collection Procedures

Data were collected during telephone interviews with patients 65 and older or their proxy (caregiver) at two time points after admission to a community-based hospice setting. The first interviews were conducted within 72 hours of admission (T¹) and the follow-up interviews were completed 7–10 days following the initial interview (T²). Whenever possible, patients completed the interviews independently. If unable, their primary caregiver served as a proxy reporter for all sources of data. Sixty-five patients from 13 unrelated Midwestern hospices or their proxies completed the initial interview, with 55 patients or their proxies completing the follow-up interviews. Additional information for each subject was collected from the medical records received from the participating hospices. As outlined above, the instruments used to gather data included the BPI, the BHI, the PPML and the PAT. Interviewers received over five hours of training on the scripts for the patient/caregiver interviews. Prior to interviewing real patients, role plays with volunteer older adult patients were conducted. Patients and/or their proxies received copies of all study instruments prior to the interview so as to be able to follow along as a trained RA asked each question on the instruments and documented responses on a duplicate copy in the project office.

Medical Records Abstraction

Medication data were obtained from patients’ hospice medical records to determine the pharmacologic pain treatment plan. Medical records for the first two weeks following hospice admission were reviewed for this study. Data on the pain medication orders corresponding to the dates of the patient interviews were collected. Medical record data were abstracted and entered directly into an ACCESS database developed specifically for grant use by trained RAs, who were nurses with experience in acute care, home care, hospice care and geriatrics.

Data Analysis

The data management and statistical analyses were performed using SAS 9.2 (SAS Institute Inc., Cary, NC). A 0.05 level of significance was used for all tests. Demographic characteristics of patients were analyzed using descriptive statistics. Statistical differences among the three PAT subscales, their means and totals between the admission interview (T¹) and second interview (T²), were determined using paired t-tests. Seven explanatory variables were collected and considered for inclusion in the logistic regression model to examine factors associated with pain medicine adherence. The variables include: average pain in the past 24 hours, worst pain in the past 24 hours, hours of controlled pain (none/mild) in the past 24 hours, level of comfort in the past 24 hours, QOL in the past 24 hours, degree of symptoms in the past 24 hours, and non-pharmacologic therapies used in the past 24 hours. The seven variables were analyzed by comparing the difference between the two interviews for the patient self-report group and caregiver proxy report group, respectively. Eight patients were excluded from the paired t-test analysis because of their change of report type (patient vs. proxy report) from the first interview to the second interview.

Logistic regression was applied to answer our first research question, whether pain-related factors and non-pharmacological interventions impact pain medicine adherence, where the dependent variable is the dichotomized patients’ adherence total score: whether or not the sum of the three subscales is equal to nine, the perfect score. As mentioned earlier, data were collected for two interviews at T¹ and T². Therefore, correlations existed among patients within the same hospice, as well as between the two interviews of each patient. We used generalized estimating equations (GEE) for two-level clusters and assumed an exchangeable working correlation structure (e.g., the correlation between any two interviews of each patient is the same). For a better interpretation, numeric explanatory variables with uneven distributions of their possible outcomes were categorized before they were included in the model: hours of controlled pain was dichotomized based on whether the proportion of hours of controlled pain was greater than 0.75; the category of non-pharmacologic therapies was dichotomized based on whether any type of non-pharmacologic therapies were used. To determine the final model, backward selection was performed on the initial model with the seven explanatory variables plus the type of report (i.e., self-report or proxy report). A final regression model was then fit based on the significant variables identified by the backward selection procedure with a significance level of P=0.01.

Results

Sample Description

Thirty-two patients were able to provide self-reports during the first interview (T¹) and 25 patients provided self-reports for the second interview (T²). Proxy (caregiver) reports were collected for 33 patients at T¹ and for 30 patients at T². The overall sample gender distribution was fairly equal, with slightly more males than females participating. The age criteria for inclusion was reduced from 65 years to 55 years in order to improve recruitment and increase the sample, resulting in 10.8% of participants between the ages of 55–64, 29.2% ages 65–74, 36.9% ages 75–84, and 23.1% greater than 85 years of age. The sample was predominantly Caucasian (89.2%). Table 1 shows the specific demographic data for the patient sample (n=65). Because 10 patients were not able to complete both interviews, data from both interview periods is provided for comparison in Table 1.

Table 1.

Demographic Characteristics of Patients Who Completed Interviews 1 and 2 as Reported by Either Patients or Caregivers

	1st Interview	2nd Interview
Number of Patients	65	55
Gender
Total Males	36 (55.4%)	32 (58.2%)
Total Females	29 (44.6%)	23 (41.8%)
Age
55–64	7 (10.8%)	6 (10.9%)
65–74	19 (29.2%)	17 (30.9%)
75–84	24 (36.9%)	19 (34.5%)
>85	15 (23.1%)	13 (23.7%)
Race
Black	3 (4.6%)	2 (3.6%)
White	58 (89.2%)	51 (92.7%)
Other	4 (6.2%)	2 (3.6%)

Note: There were no statistically significant differences between patients at both time points

Medical Record Data

Data from patient medical records, used to complete the PAT, indicated that the majority of patients had opioids ordered to manage their pain, with 78.1% having an opioid order at the time of their first interview (T¹) and 75.9% at the second interview (T²). Additionally, a much lower percentage of patients had orders for NSAIDs/acetaminophen, (T¹=34.4%; T²=35.2%) and neuropathic pain medications (T¹=10.9%; T²=9.3%) for the same time periods.

Brief Pain Inventory (BPI)

Data from the BPI provided by either the patient or their proxy (caregiver) indicated that the “average pain” in the past 24 hours reported for all patients was in the mild range (3.34 or less) for both T¹ and T². “Average pain” reports both for patients who were able to self-report and those who had proxy reports decreased from T¹ to T², with patient self-reports decreasing from 2.34 to 2.04 and proxy reports decreasing from 3.34 to 2.63. Neither decrease was significant. Patient self-reports of “worst pain” in the past 24 hours were higher than their reports of “average pain,” as expected, and decreased from T¹ to T², diminishing from 3.94 to 3.28. For patients with proxy reports, “worst pain” in the last 24 hours was reported at a moderate level, 5.94 at T¹ and decreased to 4.0 at T². This change was significant (P-value 0.05). In addition, the difference between the “worst pain” in the past 24 hours across reporting groups (patient self-report vs. proxy report) at T¹ was significant (P = 0.01) (Table 2).

Table 2.

Reports of “Average Pain” and “Worst Pain” on the Brief Pain Inventory: Patient Self-Report or Proxy Report at Interview 1 and Interview 2

	Patient Report						Proxy Report						Patient Self-report vs. Proxy Report
	1st Interview		2nd Interview		Paired t-test		1st Interview		2nd Interview		Paired t-test		1st Interview (T1)	2nd Interview (T2)
	n	Mean (SD)	n	Mean (SD)	n	P-value	n	Mean (SD)	n	Mean (SD)	n	P-value	Two Sample t-test P-value
Average Pain	32	2.34 (2.25)	25	2.04 (1.97)	23	0.4873	33	3.34 (2.07)	30	2.63 (2.57)	24	0.2052	0.0694	0.3367
Worst Pain	32	3.94 (3.0)	25	3.28 (2.78)	23	0.2082	33	5.94 (3.2)	30	4.0 (3.41)	24	0.0509	0.0116	0.3923
Hours of none/mild pain in the past 24 hrs	32	20.28 (6.89)	25	20.96 (5.34)	22 a	0.45	33	19.81 (5.73)	30	21.04 (5.47)	20 a	0.207	0.7859	0.9589

^aPatients were excluded in the paired t-tests because reports at T¹ and T² were provided by different reporters (patient vs. proxy).

Brief Hospice Inventory

Data on the BHI indicated that the level of comfort, QOL, and degree of symptom control all decreased slightly for patients able to self-report from T¹ to T². None of the changes noted from T¹ to T² were significant, reinforcing previous determinations of connectivity among these clinical variables and pain.³¹ For those patients with proxy reports, level of comfort and degree of symptom control also decreased, whereas QOL increased slightly from T¹ to T². Again, none of the changes were significant. However, there were significant differences noted when comparing the level of comfort and QOL items between the two reporting groups for each time period. In both instances, proxy reports were significantly higher (i.e., perception that patients’ symptoms were worse) (Table 3).

Table 3.

Reports of Level of Comfort, Quality of Life, and Symptom Control on the Brief Hospice Inventory: Patient Self-Report or Proxy Report at Interview 1 and Interview 2

	Patient Report						Proxy Report						Patient Self-report vs. Proxy Report
	1st Interview		2nd Interview		Paired t-test		1st Interview		2nd Interview		Paired t-test		1st Interview (T1)	2nd Interview (T2)
	n	Mean (SD)	n	Mean (SD)	n	P- value	n	Mean (SD)	n	Mean (SD)	n	P-value	Two Sample t-test P-value
Level of comfort	32	2.94 (2.65)	25	2.76 (2.01)	23	0.7918	33	4.58 (2.09)	30	4.39 (2.7)	21	0.286	0.0082	0.0152
Quality of life	32	4.41 (3.26)	25	4.2 (2.96)	23	0.5831	33	6.57 (2.69)	30	6.78 (2.86)	21	0.5356	0.0059	0.0025
Symptom Control	32	3.72 (2.96)	25	3.32 (2.67)	23 a	0.3726	33	4.39 (3.02)	30	4.15 (3.02)	20 a	0.5045	0.3789	0.2995

^aPatients were excluded in the paired t-tests because reports at T¹ and T² were provided by different reporters (patient vs. proxy).

Patient Pain Management Log

PPML data provided by either patient self-report or proxy (caregiver) report indicated that 28.1% of patients took PRN medications in the 24 hours prior to T¹ and 26.4% took extra pain medications in the 24 hours prior to T². Conversely, 4.7% of patients skipped “ordered” pain medications in the 24 hours prior to T¹ and 1.9% skipped “ordered” pain medications in the 24 hours prior to T². Non-pharmacologic therapies were used very sparingly during the 24 hours prior to each telephone interview. Patients able to self-report at T¹ indicated use of any non-pharmacologic therapies at less than one time per patient (0.75) during the 24 hours prior to their telephone interview. This increased just slightly by T² to 1.33. Similarly, use of non-pharmacologic therapies as reported by proxies also was low, 1.21 at T¹ and down slightly to 0.78 by T². None of these changes were noted as significant (Table 4). A variety of non-pharmacologic therapies were used, including heat, ice, massage, rest, change position, physical therapy, music, relaxation, prayer/meditation, and other (distraction, healing touch, oxygen, vaporizer, and whirlpool bath). Among these therapies, changing position, resting, and prayer/meditation were used the most often (23%, 22% and 22%, respectively) in the 24 hours prior to the telephone interviews.

Table 4.

Number of Hours Pain Was Reported as None/Mild in the Past 24 Hours and Number of Non-Pharmacologic Therapies Used in the Past 24 hours as Reported on the Patient Pain Management Log by Patient Self-Report or Proxy Report at Interview 1 and Interview 2

	Patient Report						Proxy Report						Patient Self-Report vs. Proxy Report
	1st Interview		2nd Interview		Paired t-test		1st Interview		2nd Interview		Paired t-test		1st Interview (T1)	2nd Interview (T2)
	n	Mean (SD)	n	Mean (SD)	n a	P-value	n	Mean (SD)	n	Mean (SD)	n a	P-value	Two Sample t-test P -value
Hours of none/mild pain in the past 24 hours	32	20.28 (6.89)	25	20.96 (5.34)	22	0.45	33	19.81 (5.73	30	21.04 (5.47)	20	0.2607	0.7859	0.9589
Number of Non-pharmacologic therapies used in the past 24 hours	32	0.75 (1.08)	25	1.33 (2.37)	19	0.3755	33	1.21 (1.35)	30	0.78 (1.22)	21	0.6289	0.1621	0.3369

^aPatients were excluded in the paired t-tests because reports at T¹ and T² were provided by different reporters (patient vs. proxy).

Patient Adherence Tool

Data from the PAT indicated high levels of adherence to pharmacologic pain management plans. There were no significant differences noted between patient self-report and proxy reports on the PAT at T¹ and T²; therefore, for reporting purposes, all have been combined. The PAT allows for a possible total score of nine points based upon a maximum of three points on each of three subscales: 1) opioids; 2) NASID/acetaminophen; and 3) neuropathic pain medications. The overall mean adherence for all patients at T¹ was calculated at 8.43 and for T² at 8.38. Patients were the least adherent with their opioid orders at both T¹ and T², as indicated by a mean adherence for both time periods of 2.65 of the maximum three points on the opioid subscale. Patients were the most adherent to NSAID/acetaminophen orders at T¹ (2.91) and to orders for neuropathic medications at T² (2.89). No significance was noted between overall reports at T¹ and T² (Table 5).

Table 5.

Patient Adherence Tool Data for Interview 1 (T¹) and Interview 2 (T2): Patient Self-Report and Proxy (Caregiver) Report Combined

Patient Adherence Tool Data	Interview 1 (T1)				Interview 2 (T2)
	n	Mean	Range	SD	n	Mean	Range	SD
Subscale 1 (opioids) max=3	65	2.65	(1,3)	0.67	55	2.65	(1,3)	0.62
Subscale 2 (NSAIDs/Tylenol)	65	2.91	(1,3)	0.34	55	2.84	(0,3)	0.57
Subscale 3 (neuropathic)	65	2.88	(1,3)	0.48	55	2.89	(1,3)	0.46
Subscale Total max=9	65	8.43	(7,9)	0.83	55	8.38	(6,9)	0.95

No significance noted.

Final Model (Logistic Regression Model With GEE)

Two variables, hours of controlled pain and level of comfort, stayed in the final model after the backward selection procedure, with a significance level 0.1. Results show that, on average, when the remaining variable in the model is fixed, 1) the patients with controlled pain (defined as none or mild pain) more than 75% of the time during the last 24 hours were 30.2% more likely to have perfect adherence scores when all the other six explanatory variables were controlled (P-value: 0.0278); 2) patients were 68.4% less likely to have a perfect adherence score when level of comfort increased by 1 (P-value: 0.0034). In conclusion, of all the variables evaluated, the one that stands out the most, and for which there is statistical significance, is that patients with more hours of controlled pain in the last 24 hours were more likely to have better adherence and that patients with a higher level of comfort over the past few days were more likely to have worse adherence. This finding suggests that as patients become more comfortable through good pain management, adherence to pain relief regimens must continue to be emphasized to ensure pain does not re-emerge.

Discussion

Poor medication adherence in older adults can lead to adverse consequences for the patients themselves and for the larger health care system, given that increased rates of hospitalization and even death can result.^34–35 The present study was able to identify that pain medication adherence among older adults with cancer receiving hospice care is exceptionally high (96.2%) compared with other adherence data for older adults and those with cancer, particularly when looking at PRN medications.³⁶ This may be a function of their stage of disease, the support and education about pain and pain management provided by hospice, impact of their caregiver, or a characteristic of the patients who are referred and elected to receive hospice care and participate in this study. The PAT assesses adherence, but not effectiveness, which was not the primary focus of this study, but it is clearly important. It can be inferred from low pain scores and other comfort-related variables that the pain control regimens were relatively efficacious.

It is important to note that patient adherence is associated with patient choice, which is consistent with the hospice goal of patient-centered care. Although the present study did not identify if adherence was high because of the patient’s choice to adhere (versus family pressure to adhere) and the sample that elected to participate in the study, the authors recognize through their clinical experience that some patients may choose to not take their pain medicines because of concerns over adverse effects such as sedation, cognitive impairment, or other reasons. A more in depth examination into the reasons for the high adherence rates in this study population would help provide insight into this phenomenon, as well as uncover how patient choice plays a role in determining adherence.

Although we did not collect data on length of hospice stay for the study cohort, the patients who participated were not imminently dying (e.g., not in their last hours to days of life), which may have allowed them to be more active participants in their pain care. Additionally, the effect of caregivers (proxies) and their role in assuring or promoting medication use was not evaluated in this study; thus, it is unknown how this may have impacted the adherence rates noted, as some may have been over- or even underreporting adherence. It is reasonable to conclude that individuals who elect to receive hospice care—a program that encourages involvement and education of both the patients and their families/caregivers—and elect to participate in this type of study, may be highly self-motivated, and have committed caregivers who support and encourage adherence to treatment plans. Regardless, once admitted to hospice care, the older adults with cancer in this study were highly adherent. Future research should compare these outcomes with other health care settings of patients with advanced medical illness and determine more specific reasons for adherence, non-adherence (if applicable) and consequences.

Research on adherence suggests that education alone is not enough to assure high adherence rates; instead, behavioral modification, such as dosing packages or cues to help patients to remember to take their medicines are necessary.³⁷ In examining the findings from the present study, this is particularly important given that patients who reported increased comfort in the past few days showed reduced adherence. When one is comfortable, the physical cue to take the medication is reduced, thus placing the patient at risk for reduced adherence and heightening the risk for a re-emergence of pain. Hospice nurses should consider ways to use memory cues or other strategies besides physical discomfort to help patients remember to take the prescribed pain medicines. Careful monitoring and discussion about the need for continued adherence, especially once the patient is comfortable, is a critical function of the hospice nurse. Additionally, the use of non-pharmacological strategies for pain management was quite low in this study. As attention is given to non-pharmacological treatments for pain in older cancer patients to reduce medication-related adverse effects, it is important to determine how adherent patients will be to these courses of treatment.

In summary, adherence to a pain management plan in older cancer patients in hospice care is important. This study showed high adherence to the pain management plan; however, future research needs to continue to examine this topic with other populations of older adults receiving hospice care to determine if adherence remains this high and to identify additional factors that may impact adherence. Whereas self-determination and patient choice are essentials of hospice care, educating patients and their caregivers about the importance of adherence is equally important, as this is one key step in ensuring a more comfortable death.

Limitations

The results of this study need to be interpreted with caution given the low participation rate from hospice patients and proxies. Although it is well known that hospice patients have lower participation rates in research,³⁸ the lack of participation impacts the ability of these findings to be generalized to other groups of older adults with cancer in hospice care. Response bias is a concern in this study. The patients who did not participate in this study were typically sicker and closer to death than those who did participate. Thus, it is unknown how adherent they were to their pain medicine.

Another limitation of this study is that validity checks were not performed on the proxy reports. Research has determined that family members over-report pain in hospice patients.³⁹ Thus, proxy reports of pain could have been overestimated, impacting the generalizability of the findings; however, this is not known for the present study. Additionally, this study did not analyze other factors that may impact adherence, particularly, what effect the involvement of a caregiver may have on a patient’s adherence to their medication regimen.