Abstract
Background:
Poor patient pain management in home hospice is associated with low family caregiver adherence to analgesic regimens. Health care technology can improve caregiver access to education and communication to hospice nurses.
Objective:
The study purpose was to (1) compare the effects of the e-PainSupport intervention for family caregivers on change in patient pain intensity from baseline to 14 days to the effects of a usual care control condition and (2) examine mediating effects of pain management knowledge, self-efficacy, and adherence to analgesic regimens on change in pain intensity, controlling for study condition and patient gender.
Methods:
Utilizing a 2-group, 2-week randomized controlled trial with dyads (N = 44) of patients (52% female, mean age 74.1 years) and their caregivers (75% female, mean age 55.2 years), dyads were randomly assigned to either the e-PainSupport intervention or usual care control condition. The e-PainSupport intervention included caregiver pain education, a pain assessment and management tracker, and communication to nurses. Participants were recruited from 4 hospice agencies in a large metropolitan area. Outcome measures included caregiver knowledge, self-efficacy, medication adherence, and patient-reported pain intensity.
Results:
The e-PainSupport intervention produced a small positive effect on reducing pain intensity (d = 0.27) and statistically significant increase in adherence (P = .003), compared with usual care. Hierarchical regression models showed a significant mediating effect of increased caregiver knowledge on reduced pain intensity (P < .01), regardless of condition.
Conclusions:
Caregiver use of the e-PainSupport app is feasible and may contribute to decreasing hospice patient pain.
Clinical trial registration:
The study was registered at ClinicalTrials.gov on May 3, 2021, NCT04869085. The first participant was enrolled on April 21, 2021.
Keywords: hospice, family caregivers, pain, palliative care
Dying free from pain is recognized as a primary quality indicator for a good death by patients, family caregivers (FCGs), and health care providers.1 Although professional hospice providers are specialists in effective pain management at the end of life, most direct care for hospice patients is provided at home2 by untrained, unpaid caregivers who are the patient’s family or friends, resulting in suboptimal pain management.3 One of the factors strongly associated with poor hospice pain management is lack of caregiver adherence to analgesic regimens,4 which often occurs due to a lack of knowledge regarding pain reporting and the use of pain medications.5 Hospice nurse home visits are typically intermittent, occurring just once or twice a week, which can limit early detection of FCGs’ nonadherence to prescribed analgesic regimens.6
Interventions for Hospice FCGs
Many FCGs lack confidence regarding pain management and are unsure of when and how much analgesic medication they should administer.7 In a large hospice quality assessment survey, FCGs of patients in pain strongly agreed that they needed help with pain management.8 Additionally, many factors complicate pain management and contribute to under-treatment, including patient gender. For example, women may be viewed as having greater sensitivity to painful stimuli, and health care providers may underestimate their analgesic needs.9
An integrative review of behavioral interventions designed to improve patient pain control in hospice and palliative care settings identified 7 studies that addressed pain knowledge.10 Of these, 5 improved patient and caregiver knowledge and reporting, and the 2 that measured adherence to patients’ medication regimens reported improvements in adherence. Although findings suggest that these behavioral interventions were successful in improving both patient and caregiver outcomes, all of the interventions required nurses’ in-person monitoring of adherence and education delivery. This reduces their feasibility when hospice resources are limited and nurses are unable to devote adequate time to teaching FCGs about pain management.11 Further, following the COVID-19 pandemic, US hospice agencies have faced unprecedented nursing workforce shortages.12,13
There is a need for the development of practical and user-friendly technology-based interventions that can be integrated into clinical practice. A recent integrative review reported the feasibility of implementing smartphone and tablet interventions to improve patient-clinician communication regarding symptom/pain management in the home setting.14 Overall, these interventions decreased patient distress, improved pain management, and decreased caregiver pain misconceptions.14 Only one of the studies, however, assessed real-time medication administration by FCGs, and it targeted patients only.15
We created e-PainSupport, a digital pain application (app) for FCGs, to address this issue by promoting better pain management in home hospice. The development of the app was guided by our Pain Management in Home Hospice Care Framework. This conceptual framework was based on the Health Belief Model, a cognitive learning model widely used to explain or predict health behavior, including adherence to analgesic regimens.16 Our framework suggests that the impact of patient and caregiver factors on improvement in patient pain intensity is mediated by caregiver knowledge, self-efficacy, and adherence to the analgesic regimens—the mechanisms of action (Figure 1).17 In addition, the improvement of patient pain is moderated by patient gender.
Figure 1.
e-PainSupport framework.
Initially, using advisory focus group methods, 6 hospice nurses and 3 informal caregivers helped our team convert a paper-based pain and analgesic diary into a digital app called e-PainSupport.18 The app included a tracking function, in which caregivers reported their patients’ pain and how it was managed. The app relayed pain assessment and management information to offsite hospice nurses in real time by text or e-mail. Pilot testing of the app with 12 home hospice patients and their caregivers was conducted for 9 days. Completion time of each pain report averaged about 5 minutes, and 84.6% of the participants reported being satisfied or very satisfied with the app.17 However, there was a lack of improvement in caregiver knowledge about pain reporting and adherence to analgesic regimens.17
Therefore, we sought to enhance e-PainSupport by integrating an education element that included aspects of the Tailored Barriers Intervention,19 an evidence-based telephone intervention to educate patients and caregivers on pain reporting and management. An advisory board of 2 experts in hospice care and 2 hospice nurses assisted in adapting the Tailored Barriers Intervention for digital use.20 The education element was intended to increase knowledge by reducing barriers to analgesic management and improve self-efficacy.
Purpose
The purpose of this study was to (1) compare the effect of the enhanced e-PainSupport intervention to the effects of a usual care control condition on change in patient pain intensity from baseline to 14 days and (2) examine mediating effects of FCG pain management knowledge, self-efficacy, and adherence to analgesic regimens on change in patient pain intensity, controlling for study condition and patient gender. We hypothesized that: (1) compared to the usual care control condition, patients in the e-PainSupport intervention condition would report significant improvement in pain intensity scores (primary outcome); and (2) FCG knowledge, self-efficacy, and adherence to analgesic regimens would mediate change in pain intensity, controlling for study condition and patient gender (secondary outcome).
Methods
Design
This was a 2-group, 2-week randomized controlled trial with 44 dyads of patients and their FCGs randomly assigned to either the e-PainSupport intervention condition (e-PainSupport; n = 23 dyads) or the usual care control condition (usual care; n = 21 dyads). An earlier hospice study found a 25% mortality rate 2 weeks after enrollment.21 Thus, we considered a 2-week follow-up period to be feasible, giving patients and caregivers needed time to complete the education element.
The Consolidated Standards of Reporting Trials (CONSORT) guidelines directed the reporting of the intervention.22 The study was approved by the authors’ Institutional Review Board. Consent was obtained during the period from March 9, 2021, to July 6, 2023, from patient-caregiver dyads and their direct care nurses. The first patient-caregiver dyad was screened on August 11, 2021, and the last participant completed follow-up on July 28, 2023. The detailed study protocol was previously published.23
Power Analysis
The initial sample size was calculated using data from 4 RCTs with behavioral interventions, identified in our previous literature review, that were designed to improve pain intensity in hospice/palliative care patients.10 A mean effect size (d) of 0.61 was calculated. This implied that a final sample of 92 (46 per group), assuming a 2-tailed α of .05 and an effect size of 0.61, would produce a power estimate of 0.72.23 Due to recruitment complications related to the COVID-19 pandemic, however, we obtained a sample of only 44 dyads.
Setting, Sample, and Recruitment
Participants were recruited from 4 metropolitan area hospice agencies with which the study team had established relationships.20 Direct-care hospice nurses were recruited by study staff during regular scheduled team meetings. Nurses interested in having their patients participate in the study signed a consent form via REDCap.24 Inclusion criteria were that they were registered nurses with responsibility for direct care of a panel of hospice patients. Nurses could have more than 1 patient in the study.
Inclusion criteria for patients were that they (a) were enrolled in home hospice care, (b) received analgesics for pain, (c) spoke and read English, (d) were aged 18 years or older, (e) had an FCG available for the 2 weeks of the study, (f) were expected to survive at least 2 weeks, (g) were cognitively intact, (h) were able to verbalize pain, and (i) were assigned to a direct-care hospice nurse who agreed to participate in the study. Expected survival of 2 weeks and cognition were assessed with the Palliative Performance Scale, which assesses functional status, ambulation, activity level, self-care, oral intake, and consciousness level.25 A total score of ≥30% was used for expected survival and ≥70% on the consciousness item was used to assess cognition.
Inclusion criteria for FCGs were that they (a) spoke and read English, (b) were aged 18 years or older, (c) cared for an enrolled patient at home, and (d) were available for the two-week study period. FCGs without an Internet-enabled device were provided with a tablet and data plan. Hospice admission nurses, who see patients only at the intake visit and routinely administer the Palliative Performance Scale (PPS)25 as a part of their initial assessment, were oriented to the study. Initially, they distributed a study flyer to patients who met the PPS and medication inclusion criteria and to their caregivers. Direct-care nurses who provided routine care were oriented to the study and were encouraged to discuss the study with their patients. Patients and FCGs could give verbal consent to the admission nurse to share their names with a study staff member. Alternatively, they could contact the study staff directly by phone or e-mail.
Later, several changes were made to boost recruitment. First, an introductory letter signed by the PI (a certified hospice and palliative care nurse) explaining the study along with the study flyer were given to all newly admitted patients, informing them that they could contact a study staff member if they met study criteria and they could opt out by phone or e-mail if they did not want to receive a call. Second, hospice agency staff members were hired as research assistants to identify eligible patients from the electronic health record (EHR) and send names to study staff via encrypted e-mail. A study staff member contacted all potentially eligible patients by phone to assess interest and schedule a follow-up home visit.
e-PainSupport Intervention Condition
The e-PainSupport digital app is self-administered by the caregiver on a cellphone or tablet that tracks all user activity.18,20 The patient’s medication list is available to the caregiver through the app. In hospice agencies able to link the app to their EHR, the patient’s medication list was updated directly to the patient dashboard. For hospice agencies without EHR integration, study staff entered the medications manually on the e-PainSupport patient portal. FCGs had no access to the patient EHR. e-PainSupport has 3 elements: (1) caregiver pain education, (2) a pain assessment and management tracker, and (3) communication (alerts) to nurses. Caregivers (and optionally, patients) have access to the education and tracker elements, while direct-care nurses have access to all 3 elements.
Caregiver pain education.
The caregiver education element was described in detail earlier.20 When a caregiver opens e-PainSupport on their device, they touch an icon titled “Pain Education.” The introduction to the app appears and includes instructions, encouragement to cover all topics, and an explanation of common pain terms. A voiceover option is available to accommodate users with low literacy. Next, caregivers are prompted to click on a list of 5 statements that address pain management barriers relevant to hospice care: a sense of fatalism (pain is beyond ones’ control), fear of causing patient addiction, fear of being seen as a complainer, fear of drug tolerance, and fear of side effects.20 For each barrier, the caregiver selects “yes” if they believe the statement is correct or “no” if they believe the statement is incorrect. For misconceptions, the app provides accurate information with the best strategies for overcoming the associated barrier. Each statement takes about 3 to 4 minutes to complete, for a total of 15 minutes. FCGs are encouraged to finish all 5 statements by the end of the intervention. Additional components include general pain management tips and a link to selected Internet pain management resources from the National Cancer Institute (eg, pain and quality of life, pain assessment, and treatments).26 Users can review the content at any time during the study.
Pain assessment and management tracker.
Every time the patient experiences moderate to severe intensity pain (greater than 3 on 0–10 scale) the caregiver opens e-PainSupport and touches the “I am in pain now” icon to begin reporting an assessment of patient pain (pain intensity; pain quality from 21 descriptors). Pain reports are obtained from the patient; if the patient is unable to report pain, the caregiver is instructed not to estimate pain.
e-PainSupport also contains a “What was done for pain?” section, where actions used to manage pain, including analgesics and dosage and non-pharmacological interventions, are recorded. At the end of each day, prompted by a reminder alarm at a time chosen by the caregiver, caregivers complete an “End of Day Report” describing the patient’s worst pain intensity in the past 24 hours on a 0 to 10 scale. They also record the daily frequency and average duration (in hours) of moderate to severe pain (>3 on a 0–10 scale) in the last 24 hours.
Communication to nurse.
A pain alert is sent by the text message to the nurse if an “End of Day Pain Report” indicates severe pain (≥7 on a 0–10 scale). In addition, the “Pain Summary for Nurses” (seen only by the nurse) provides an ongoing graph of the patient’s “End of Day Reports” for worst pain intensity.
Usual Care Control Condition
Usual care was standard across all 4 hospice agencies. Direct-care home hospice nurses visit their patients once or twice per week when patients are stable. Nurses assess the patient’s self-reported pain intensity, select appropriate analgesics from the hospice-prescribed analgesic regimens, and enter the analgesic regimens into the EHR on a laptop computer while in the patient’s home. When pain is not controlled, FCGs can contact an on-call hospice nurse at any time. On-call nurses manage patient and caregiver needs by phone or send a team member (eg, registered nurse, social worker, grief counselor) to the patient’s home. Additionally, all patients enrolled in hospice receive a comfort medication kit (eg, pain medication, antinausea medication, antiseizure medication, constipation medication) and a hospice booklet. Booklets across the 4 agencies included hospice contact information, guidance on how to use medications in the comfort medication kit, and general information about pain management.
In addition to the comfort kit, one agency provides a pain assessment scale, but without instructions. Another agency includes a medication tracking sheet. None of the agencies provide FCG-specific information about pain management barriers. In this study, usual care control condition participants received a link to the same Internet resources from the National Cancer Institute26 regarding pain management on paper included in the e-PainSupport app.
Fidelity
Monitoring of treatment fidelity using design, training, delivery, and receipt components was previously reported.23 Briefly, to minimize contamination, the project director, a master’s-prepared nurse specializing in critical care, held separate staff training for baseline data collectors and post-intervention data collectors. Baseline data collectors were experienced registered nurses with hospice and critical care experience. They completed a 3-day training (18 hours total) that covered re-administering the PPS25 to reconfirm eligibility, obtaining consent from patients and FCGs, collecting baseline data, informing participants of their condition assignment, and orienting intervention condition participants on the use of e-PainSupport.
The post-intervention data collectors were all graduate nursing students who had completed course work on end-of-life care. They were blinded to the participant’s condition. They received a 1-day training (6-hours total) on the data collection procedure, which included administering the post-intervention questionnaire. Before and after each data collection, all data collectors met with the project director and were debriefed using the research assistant checklist of activities.
The receipt component of fidelity, reported here, is represented by usage data for the 3 e-PainSupport elements downloaded by staff directly from the app. Data were obtained for (1) opening the educational element one or more times (yes/no); (2) mean number of days the educational element was opened; (3) number of times caregivers recorded pain assessments; (4) number of days caregivers recorded pain management; and (5) number of days caregivers recorded an “End of Day Report.”
Measures
Considering the rapid decline in the functional status of hospice patients, all measures used in this study were kept brief (under 5 minutes to complete), with the exception of the Barrier Questionnaire (BQII) which is composed of 27 items. Although the mean time to complete the BQII is 12 minutes,27 it has been tested and has been shown to be acceptable to hospice caregivers.17,28
Demographics.
The BRICS NINR Demographics Form29 was used to collect participants’ age, gender, education, race, ethnicity, marital status, employment status, and caregivers’ relationship to patient.
Pain.
The PROMIS Pain Intensity-Short Form 3a v2.0 was used to measure self-reported patient pain intensity at baseline and 14 weeks.30 It consists of 3 items: worst and average pain in the past 7 days, and current pain. Pain intensity is measured using a numeric rating scale from 1 (no pain) to 5 (very severe pain). Items are summed for a range of 3 to 15 and raw scores are translated into an item response theory-based T-score, with higher T-scores representing greater pain intensity. A T-score of 50 represents the US general population mean. Using Spearman rho, convergent validity was 0.68 with the PROMIS global pain scale and 0.61 with the PROMIS global physical health scale.30
A minimally important clinical improvement was assessed by examining the score for each of the 3 items. According to the 2008 Initiative on Methods, Measurement, and Pain Assessment in Clinical Trials Committee, a 10% change in an 11-point pain intensity scale indicates minimally important clinical improvement.31 Since we used a 5-point scale, we set our criterion for clinical improvement at 20% change for each of the 3 items.
Pain management knowledge.
The 27-item Barriers Questionnaire II was used to measure caregiver knowledge barriers about reporting pain and using analgesics at baseline and 14 days.27 The 4 knowledge barrier subscales include physiological effects (12 items), fatalism (3 items), communication (6 items), and harmful effects (6 items). Each item is rated on a 6-point scale from 0 (do not agree) to 5 (agree very much). Items are summed and averaged for a range of 0 to 5 with higher scores indicating more barriers to pain management or lower pain management knowledge. Prior Cronbach’s alpha was 0.89, and patients with adequate analgesics had lower scores than patients with inadequate analgesics.27 Cronbach’s alpha for this study was 0.88.
Self-efficacy.
The 5-item Chronic Pain Self-Efficacy Scale was used to measure caregiver self-efficacy for pain management at baseline and 14 days.32 Each item is rated on a 10-point scale from 10 (very uncertain) to 100 (very certain). Items are summed and averaged. Prior Cronbach’s alpha was 0.80, and concurrent validity was demonstrated by a high association with lower strain, decreased negative mood, and increased positive mood.32 Cronbach’s alpha for this study was 0.75.
Adherence.
The four-item Morisky-Green-Levine Medication Adherence Scale33 was used to measure caregiver adherence to analgesic regimens at baseline and 14 days. Two items were slightly modified to be answered by caregivers rather than patients. Each answer was scored as 1 (yes) or 0 (no). Items were summed with a range of 0 to 4. A score of 0 to 1 indicated poor adherence; 2 to 3 indicated moderate adherence; and 4 indicated good adherence. Cronbach’s alpha has been reported34 as 0.86, and Spearman rho convergent validity has been reported35 as 0.79. Cronbach’s alpha for this study was 0.39.
Procedures
Baseline data collectors made home visits to interested dyads to screen patients with the PPS to ensure they still met physical and cognitive criteria and to obtain signed consent. They immediately contacted the PI by phone to review the patient’s condition and PPS score if the score had changed. Once eligibility was confirmed by the PI, they conducted baseline data collection via REDCap.24 Next, they opened a sealed envelope with the dyad’s random assignment to either usual care or intervention conditions. Random assignments were made by the data manager prior to visits using the randomization module in REDCap.24 Randomization was conducted in blocks of 10 pairs to minimize possible historical effects.
Dyads assigned to usual care were informed that they would continue usual care provided by the hospice agency. The dyads assigned to the e-PainSupport intervention condition received an orientation on using e-PainSupport and performed return demonstrations. Also, they received a manual, which included screenshots of each component of the e-PainSupport app, a troubleshooting guide, and contact information for technology support.
The project director informed the patient’s direct-care hospice nurse after dyad assignment. Direct-care nurses caring for dyads in the e-PainSupport intervention condition received a 30-minute orientation by a staff member and were given a secure Internet connection to allow them to access e-PainSupport on their phone, tablet, or laptop and given access to e-mail/telephone technical support.
Post-intervention data collectors were independent from those who collected baseline data and were also blinded to condition assignment. Patients and caregivers each received a $10 gift card after the baseline data collection and a $30 gift card after completing the study. Nurses whose patients were enrolled in the intervention condition received a $20 gift card after they completed the study. Data regarding nurses’ experiences in the study were collected and will be reported elsewhere.
Statistical Analyses
SPSS for Windows (v. 24; IBM Corp, Armonk, NY, USA) was used for data management and statistical analysis. A threshold of .05 significance level was used for all statistical tests. All analyses were performed on an intent-to-treat basis.36 Missing data were imputed using Hot Decking methods,37 using the LAG function in SPSS, which assigns values based on the nearest neighbor principle. These methods are most suited for small samples of non-normally distributed data.
We compared descriptive statistics for demographics to identify significant differences between the 2 conditions and between complete case dyads (neither caregiver nor patient missing at 14 weeks) and partial complete/lost dyad cases at 14 weeks. Descriptive statistics were also calculated for fidelity measures. Using imputed data, we conducted a series of repeated measures ANOVA to assess changes from baseline to 14 days in pain intensity, pain management knowledge, self-efficacy, and adherence, to identify change from baseline to 14 days. In addition, due to the small sample size, we changed our analytic strategy to emphasize effect sizes (Cohen’s d),38 which are independent of sample size, rather than using probability statements, which are dependent on sample size.39 Because of missing data, we cross-validated our results by examining the data using the complete case dyads (unimputed).
A theoretical hierarchical regression analysis model40 was used to estimate the mediation effects of pain management knowledge, self-efficacy, and adherence on change in patient pain intensity, controlling for study condition (intervention and control) and patient gender by examining the amount of variance added to the overall model by each potential mediator. If the added variance was significant, it indicated that the mediator may have had a significant effect. In our regression model, we analyzed each variable in the order in which we expected it would affect pain. Model 1 included only patient gender and study condition, while Models 2 to 4 added pain management knowledge, self-efficacy, and adherence, respectively. The addition of each variable in the model resulted in a change in model mean square; thus, we examined the change in mean square variance associated with the addition of each variable. This represented a relatively unbiased estimate of the variables’ impact on pain. Because of missing data, we also cross-validated these results by examining the data using the complete case dyads (unimputed).
We conducted a complete case (no missing data for patients) analysis of response to e-PainSupport based on minimally significant clinical improvement in pain (20%).31 We estimated effect sizes in terms of the risk ratios between the improvement rates of the 2 conditions.
Results
Of the 618 dyads referred to the study, 61% (377/618) declined to participate, 16% (101/618) were ineligible, and 16% (96/618) died prior to being contacted by a study staff member (Figure 2). A total of 44 of 618 (7%) dyads completed baseline assessments and were randomized to one of the 2 study conditions. Of the 44 dyads, 36 complete case dyads (patient and caregiver both retained) and 4 partial-case dyads (patient or caregiver) finished follow-up at 14 days, while 4 dyads were lost to follow-up at 14 days. The primary reason for partial-case dyads or lost-to-follow-up dyads was patient death (n = 6). Of the 50 hospice nurses who consented and agreed to have their patients participate if eligible and interested, 28 had participating dyads (number of patients per nurse ranged from 1 [20 nurses] to 5 [1 nurse]).
Figure 2.
CONSORT flow diagram of participant enrollment.
The mean age for patients was 74.1 (SD = 13.4) years (Table 1). Approximately half (52%) were female. Patients were predominately white (71%), married (61%), and completed some college or less (59%). The 2 most common patient diagnoses were cancer (n = 15) and cardiovascular disease (n = 11). Thirteen patients had 7 other diagnoses, and 7 patients were missing diagnoses. The mean age for caregivers was 55.2 (SD = 12.5) years. Caregivers were predominately female (75%), white (73%), married (62%), and had less than a college degree (52%). There were no significant demographic differences by condition. Caregivers in the complete cases were significantly older than those in the partial complete/lost cases (57.2 years vs 45.2 years respectively).
Table 1.
Demographic Characteristics of Family Caregivers and Patients.
| Characteristic | Patients (N = 44) | Family caregivers (N = 44) | ||||
|---|---|---|---|---|---|---|
| Control (n = 21) |
Intervention (n = 23) |
Total (N = 44) |
Control (n = 21) |
Intervention (n = 23) |
Total (N = 44) |
|
| Age (years) (M, SD) | 75.7 (13.0) | 72.7 (13.8) | 74.1 (13.4) | 57.2 (11.7) | 53.3 (13.2) | 55.2 (12.5) |
| Gender (n, %) | ||||||
| Female | 10 (48) | 13 (57) | 23 (52) | 15 (71) | 18 (78) | 33 (75) |
| Male | 11 (52) | 10 (43) | 21 (48) | 6 (29) | 5 (22) | 11 (25) |
| Race (n, %) | ||||||
| White | 17 (81) | 15 (65) | 32 (73) | 17 (81) | 14 (61) | 31 (70) |
| Black | 3 (14) | 3 (13) | 6 (14) | 2 (10) | 5 (22) | 7 (16) |
| Other | 1 (5) | 5 (22) | 6 (14) | 1 (5) | 4 (17) | 5 (11) |
| Missing | 0 (0) | 0 (0) | 0 (0) | 1 (5) | 0 (0) | 1 (2) |
| Ethnicity, Hispanic or Latino (n, % yes) | 0 (0) | 1 (4) | 1 (2) | 1 (5) | 3 (13) | 4 (9) |
| Married/Domestic partner (n, % yes) | 15 (11) | 13 (57) | 27 (61) | 14 (67) | 13 (57) | 27 (62) |
| Education (n, %) | ||||||
| Some college or less | 11 (52) | 15 (65) | 26 (59) | 12 (57) | 11 (48) | 23 (52) |
| College degree or more | 9 (43) | 8 (35) | 17 (39) | 9 (43) | 11 (48) | 20 (46) |
| Missing | 1 (5) | 0 (0) | 1 (2) | 0 (0) | 1 (4) | 1 (2) |
| Relationship to the patient (n, %) | ||||||
| Child | 9 (42) | 13 (58) | 21 (49) | |||
| Spouse or partner | 10 (48) | 6 (26) | 16 (36) | |||
| Other family/friend | 2 (10) | 4 (16) | 7 (14) | |||
| Employed (n, % yes) | 9 (42) | 13 (58) | 21 (49) | |||
Examination of fidelity for caregiver receipt (usage) of the e-PainSupport app revealed that 60.87% (14/23) used the education element at least once during the study period. Caregivers who used the educational element used it a mean of 5.43 times (1–13 times). Pain assessment was recorded in the tracker by 91.30% (21/23) of caregivers. On average, caregivers reported assessing pain intensity a mean of 3.74 times (1–6.42 times). Pain management was recorded by 87.0% (20/23) of caregivers. On average, caregivers reported pain management a mean of 46.05 times (0–173). “End of Day” summary was recorded in the tracker by 87.0% (20/23) of caregivers on at least one of the 14 days. On average, of the caregivers who completed the end of day summary, it was reported a mean of 4.9 days out of a possible 14 days.
Condition Differences in Patient Pain, Caregiver Knowledge, Self-efficacy, and Adherence
ANOVA results for pain intensity, knowledge barriers, self-efficacy, and adherence to analgesic regimens are presented in Table 2. There was a nonsignificant decrease in pain from baseline to 14 days for both conditions. There was a small effect size associated with the e-PainSupport intervention condition (d = 0.27). For both conditions, there was a nonsignificant decrease in knowledge barriers (increase in pain management knowledge) from baseline to 14 days. However, the intervention condition had a slightly larger improvement in pain management knowledge than usual care, and the effect size was in the expected direction (d = −0.07). For self-efficacy, both conditions demonstrated a nonsignificant decrease from baseline to 14 days. Intervention participants had a slightly larger but nonsignificant decrease in self-efficacy than the control condition.
Table 2.
Change in Patient Pain Intensity and Family Caregiver Knowledge (Barriers to Pain Management), Self-Efficacy, and Adherence to Analgesic Regimen for Control and Intervention from Baseline to 14 days and Difference in Change Scores (N = 44).
| Variable | Control (n = 21) | Intervention (n = 23) | Interaction term | Effect size | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Baseline | 14 days | Difference | Baseline | 14 days | Difference | ||||||
| Mean (SD) | Mean (SD) | Mean (SD) | Mean (SD) | Mean (SD) | Mean (SD) | Difference in change | 95% Cl | P | P CC | Cohen’s d | |
| Pain intensity (T-score) | 58.31 (13.32) | 56.73 (11.11) | 1.59 (14.88) | 63.25 (13.59) | 57.37 (10.53) | 5.89 (17.14) | 4.30 | (0.43, 9.03] | .38 | .50 | 0.27 |
| Knowledge barrier | 2.23 (0.80) | 2.02 (0.83) | −0.21 (0.67) | 1.83 (0.58) | 1.56 (0.71) | −0.27 (1.01) | −0.48 | (−0.74, −0.22) | .81 | .88 | −0.07 |
| Self-efficacy | 65.71 (19.21) | 65.33 (24.28) | −0.38 (18.75) | 68.26 (22.04) | 65.39 (23.49) | −2.87 (24.77) | −3.25 | (−9.70, 3.18) | .71 | .80 | −0.11 |
| Adherence | 1.38 (0.50) | 1.14 (0.35) | −0.24 (0.43) | 1.13 (0.34) | 1.35 (0.49) | 0.22 (0.52) | 0.46 | (0.32, 0.60)* | <.01 | .04 | 0.97 |
Abbreviation: CC: complete cases.
This interaction effect is significant using confidence intervals but is nonsignificant in the ANOVA model. These sorts of differences are anomalies of using an ANOVA model.
For adherence to analgesic regimens, there was a nonsignificant drop for the control condition, and a nonsignificant increase for the intervention condition from baseline to 14 days. However, there was a significant interaction effect for adherence in which the intervention condition (M = 0.22, SD = 0.52) had a larger increase than the control condition (M = −0.24, SD = 0.43) (P = .003) (Table 2). This interaction was associated with a large effect size (d = 0.97); the intervention condition grew more adherent, and the control condition became less adherent. Complete case analyses for pain intensity, knowledge, self-efficacy, and adherence revealed no difference in results.
Mediators of Pain Intensity Change
Using hierarchical regression, Model 1 showed no significant effect of caregiver gender or study condition on decreasing patient pain from baseline to 14 days (Table 3). Model 2 showed a significant independent positive effect of increased pain management knowledge on decreased pain (MS = 648.04, df = 3, F = 4.07, P < .01). Models 3 and 4 introduced self-efficacy and adherence, and neither showed a significant independent effect on patient pain. Complete case analyses revealed no difference in results.
Table 3.
Regression of Change in Pain Intensity on Change in Caregiver Knowledge, Self-efficacy, and Adherence controlling for Patient Gender and Treatment Condition.
| Step and predictor variables | Sums of squares | df | Mean squares | F | p | Delta sums of squares | Delta df | Delta mean squares | Mean squares error | Delta F | Delta p |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Step 1 | |||||||||||
| Patient gender | 13.02 | 2.00 | 6.51 | 0.03 | .97 | ||||||
| Treatment condition | |||||||||||
| Step 2 | |||||||||||
| Patient gender | 1944.12 | 3.00 | 648.04 | 4.07 | .01* | 1931.10 | 1.00 | 1931.10 | 159.17 | 12.13 | <.01** |
| Treatment condition | |||||||||||
| Knowledge change | |||||||||||
| Step 3 | |||||||||||
| Patient gender | 2143.43 | 4.00 | 535.86 | 3.39 | .02* | 199.30 | 1.00 | 199.30 | 158.14 | 1.26 | .27 |
| Treatment condition | |||||||||||
| Knowledge change | |||||||||||
| Self-efficacy change | |||||||||||
| Step 4 | |||||||||||
| Patient gender | 2164.73 | 5.00 | 432.95 | 2.68 | .04* | 21.30 | 1.00 | 21.30 | 161.74 | 0.13 | .72 |
| Treatment condition | |||||||||||
| Knowledge change | |||||||||||
| Self-efficacy change | |||||||||||
| Adherence change |
For gender, 0 = Male and 1 = Female. For treatment condition, 0 = Control, 1 = Intervention.
p < .05.
p < .01.
Response to Treatment
Examination of complete case data for patients revealed substantial differences between the 2 conditions in response to treatment rates, defined as a 20% drop in worst, average, and current pain from baseline to 14 days (Table 4). Patients in the intervention condition were 2.50 times more likely to have a decrease in pain for worst pain, 3.38 times more likely to have a decrease in average pain, and 2.50 times more likely to have a decrease in current pain than patients in the control condition over the 2-week study period.
Table 4.
Rates of Minimally Important Clinical Improvement in Patient Pain Intensity (20% Reduction in Pain Intensity) From Baseline to 14 days.
| Pain intensity | Control improved | Intervention improved | Risk ratio |
|---|---|---|---|
| n (%) | n (%) | ||
| Worst pain | 2/20 (10.0) | 4/16 (25.0) | 2.50 |
| Average pain | 1/18 (5.6) | 3/16 (18.8) | 3.38 |
| Pain now | 2/20 (10.0) | 3/16 (25.0) | 2.50 |
Discussion
This is one of the first studies to examine the effects of a digital app intervention (e-PainSupport) enhanced with an educational component for FCGs. It uniquely addresses the mediating effects of intervention targets (mechanisms of action; pain management knowledge barriers, self-efficacy, adherence to analgesic regimens) on change in patient pain intensity. Using a small sample, collected at the height of the COVID-19 pandemic, we found that e-PainSupport was feasible for use by FCGs and had a small but positive effect on patient pain. Notably, patients whose caregivers were in the e-PainSupport intervention condition were 2 to 4 times more likely to have a decrease in worst, average, and current pain than those in the usual care control condition.
We found a small improvement in knowledge related to pain management barriers for both conditions with the change being somewhat better in the e-PainSupport intervention condition. This finding provides some support for adding the Tailored Barrier Intervention, which earlier reported improvements in knowledge of pain barriers in patients with cancer and their caregivers.41 FCG knowledge in the intervention condition, however, did not improve as much as we anticipated, possibly because the educational element was intentionally made brief to minimize caregiver burden. Unlike the Tailored Barrier Intervention, there was no interventionist available to address concerns. Enhancing the educational content by converting written materials into video format with patient, caregiver, and provider actors could allow the addition of more content without increasing FCG burden.
An unexpected finding in this study was a decrease in self-efficacy from baseline to 14 days for both conditions. Although nonsignificant, self-efficacy decreased more in the e-PainSupport condition. When beginning hospice care, FCGs often have limited experience with end-of-life patient pain, and as a result, they may overestimate their ability to manage pain. This is consistent with our earlier work showing that FCGs had high self-efficacy in managing patient pain despite having many misconceptions about pain management.17 Therefore, for caregivers in the e-PainSupport condition, having access to the educational element of the app might have heightened their awareness of a deficit in their ability to manage patient pain.
Improvement in FCG adherence to analgesic regimens was higher for the e-PainSupport intervention condition compared to the control condition. This is consistent with an earlier study finding that patient adherence to the analgesic regimen improved using a digital device.15 Our findings suggest that the educational element of the e-PainSupport intervention may have been instrumental in driving better adherence. However, this needs to be further evaluated with a comparison group that receives only the pain assessment and management tracker element.
Overall, examination of all 3 mediators (mechanisms of action) and pain outcomes revealed that caregiver knowledge may be instrumental in reducing patient pain intensity. The improvement in caregiver knowledge found in the e-PainSupport intervention condition suggests that caregiver education may be instrumental to the improvement in patient pain. However, due to the small sample size these results should be interpreted with caution.
A limitation of the study was that, consistent with other studies, recruitment was largely affected by the COVID-19 pandemic. Patients and caregivers were hesitant to participate in research studies because they were afraid of contracting COVID-19.42,43 In addition, hospice agencies placed restrictions on visitation from outside organizations,12 which limited our access. The pandemic resulted in an unprecedented nursing shortage, which placed significant economic pressure on hospice agencies as they hired contract nurses to fill vacant spots at a higher cost.12,13 Of our 4 participating agencies, 3 underwent mergers with other organizations, which resulted in disruptions to study recruitment. These factors all contributed to our not reaching the proposed sample size and the need for modifications to the original data analysis plan as noted in the Statistical Analyses section. Due to the small sample, to avoid overfitting the model we only examined the potential moderating effects of patient gender. Future studies with larger samples should explore other moderating factors, such as patient and caregiver demographics and relationship dynamics to provide deeper insights into the effects of the intervention on patient pain.
In conclusion, FCGs assume major responsibility during hospice due to patients’ increased symptom burden and decreased functional status prior to death.44 An important characteristic of this study was the focus on providing FCGs, rather than on patients, with educational information to overcome beliefs that can act as barriers to optimal pain management. The study demonstrated that e-PainSupport is feasible for use by FCGs in home hospice, and outcomes suggest that its use may contribute to pain reduction in hospice patients, but broader scale research with a larger sample is needed before final conclusions can be drawn.
Acknowledgments
The authors thank Dr. Sandra Ward, PhD, FAAN, for her advice.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Research reported in this publication was supported by the National Institute of Nursing Research of the National Institutes of Health under Award Number R21NR018952.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Ethical Approval
This project was approved by Rush University Medical Center Institutional Review Board (ORA#19111205). Written informed consent forms were obtained from all study participants.
References
- 1.Bhadelia A, Oldfield LE, Cruz JL, Singh R, Finkelstein EA. Identifying core domains to assess the “quality of death”: a scoping review. J Pain Symptom Manage. 2022;63(4):e365–e386. doi: 10.1016/j.jpainsymman.2021.11.015 [DOI] [PubMed] [Google Scholar]
- 2.National Hospice and Palliative Care Organization. 2024 Edition: Hospice Facts and Figures. National Hospice and Palliative Care Organization; 2024. Accessed October 2, 2024. https://www.nhpco.org/wp-content/uploads/NHPCO-Facts-Figures-2024.pdf [Google Scholar]
- 3.Coyne P, Lowry S, Mulvenon C, Paice JA. American Society for Pain Management Nursing and Hospice and Palliative Nurses Association position statement: pain management at the end of life. Pain Manag Nurs. 2024;25(4):327–329. doi: 10.1016/j.pmn.2024.03.020 [DOI] [PubMed] [Google Scholar]
- 4.Mayahara M, Paice J, Wilbur J, Fogg L, Foreman M. Common errors in using analgesics by home-based nonprofessional hospice caregivers. J Hosp Palliat Nurs. 2014;16(3):134–140. doi: 10.1097/NJH.0000000000000040 [DOI] [Google Scholar]
- 5.Makhlouf SM, Pini S, Ahmed S, Bennett MI. Managing pain in people with cancer—a systematic review of the attitudes and knowledge of professionals, patients, caregivers and public. J Cancer Educ. 2020;35(2):214–240. doi: 10.1007/s13187-019-01548-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.U.S. Department of Health and Human Services; Office of Inspector General (OIG). Memorandum Report: “Medicare Hospice Care: Services Provided to Beneficiaries Residing in Nursing Facilities,” OEI-02-06-00223. Accessed April 2, 2025. https://oig.hhs.gov/documents/evaluation/2502/OEI-02-06-00223-Complete%20Report.pdf
- 7.Chi NC, Fu YK, Nakad L, et al. Family caregiver challenges in pain management for patients with advanced illnesses: a systematic review. J Palliat Med. 2022;25(12):1865–1876. doi: 10.1089/jpm.2020.0806 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Parast L, Tolpadi AA, Teno JM, Elliott MN, Price RA. Hospice care experiences among cancer patients and their caregivers. J Gen Intern Med. 2021;36(4):961–969. doi: 10.1007/s11606-020-06490-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Zhang L, Losin EAR, Ashar YK, Koban L, Wager TD. Gender biases in estimation of others’ pain. J Pain. 2021;22(9):1048–1059. doi: 10.1016/j.jpain.2021.03.001 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Mayahara M, Wilbur J, Fogg L, Breitenstein SM. Behavioral pain intervention for hospice and palliative care patients: an integrative review. Am J Hosp Palliat Care. 2018;35(9):1245–1255. doi: 10.1177/1049909118775421 [DOI] [PubMed] [Google Scholar]
- 11.Ferrell BR. Family caregiving and cancer pain management. Anesth Analg. 2019;129(5):1408–1413. doi: 10.1213/ane.0000000000003937 [DOI] [PubMed] [Google Scholar]
- 12.Kates J, Gerolamo A, Pogorzelska-Maziarz M. The impact of COVID-19 on the hospice and palliative care workforce. Public Health Nurs. 2021;38(3):459–463. doi: 10.1111/phn.12827 [DOI] [PubMed] [Google Scholar]
- 13.Rogers JEB, Constantine LA, Thompson JM, Mupamombe CT, Vanin JM, Navia RO. COVID-19 pandemic impacts on U.S. hospice agencies: a national survey of hospice nurses and physicians. Am J Hosp Palliat Care. 2021;38(5):521–527. doi: 10.1177/1049909121989987 [DOI] [PubMed] [Google Scholar]
- 14.Ansari N, Wilson CM, Heneghan MB, Supiano K, Mooney K. How technology can improve communication and health outcomes in patients with advanced cancer: an integrative review. Support Care Cancer. 2022;30(8):6525–6543. doi: 10.1007/s00520-022-07037-y [DOI] [PubMed] [Google Scholar]
- 15.Wilkie DJ, Yao Y, Ezenwa MO, et al. A stepped-wedge randomized controlled trial: effects of eHealth interventions for pain control among adults with cancer in hospice. J Pain Symptom Manage. 2020;59(3):626–636. doi: 10.1016/j.jpainsymman.2019.10.028 [DOI] [PubMed] [Google Scholar]
- 16.Rosenstock IM, Strecher VJ, Becker MH. Social learning theory and the health belief model. Health Educ Q. 1988;15(2):175–183. doi: 10.1177/109019818801500203 [DOI] [PubMed] [Google Scholar]
- 17.Mayahara M, Wilbur J, Fogg L, Breitenstein SM, Miller AM. Feasibility of e-Pain Reporter: a digital pain management tool for informal caregivers in home hospice. J Hosp Palliat Nurs. 2019;21(3):193–199. doi: 10.1097/njh.0000000000000548 [DOI] [PubMed] [Google Scholar]
- 18.Mayahara M, Wilbur J, O’Mahony S, Breitenstein S. E-Pain Reporter: a digital pain and analgesic diary for home hospice care. J Palliat Care. 2017;32(2):77–84. doi: 10.1177/0825859717722466 [DOI] [PubMed] [Google Scholar]
- 19.Ward SE, Wang KK, Serlin RC, Peterson SL, Murray ME. A randomized trial of a tailored barriers intervention for Cancer Information Service (CIS) callers in pain. Pain. 2009;144(1–2):49–56. doi: 10.1016/j.pain.2009.02.021 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Mayahara M, Wilbur J, Fogg L, Paice JA, Miller AM. e-PainSupport: a digital pain management application for home hospice care. Am J Hosp Palliat Care. 2024;41(10):1120–1126. doi: 10.1177/10499091231211493 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Cagle JG, Zimmerman S, Cohen LW, Porter LS, Hanson LC, Reed D. EMPOWER: an intervention to address barriers to pain management in hospice. J Pain Symptom Manage. 2015;49(1):1–12. doi: 10.1016/j.jpainsymman.2014.05.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Schulz KF, Altman DG, Moher D. CONSORT 2010 statement: updated guidelines for reporting parallel group randomized trials. Ann Intern Med. 2010;152(11):726–732. doi: 10.7326/0003-4819-152-11-201006010-00232 [DOI] [PubMed] [Google Scholar]
- 23.Mayahara M, Wilbur J, Miller AM, Fogg L. A study protocol for e-PainSupport: the use of a digital application for reporting pain and pain management in home hospice. Contemp Clin Trials Commun. 2023;36:101071. doi: 10.1016/j.conctc.2023.101071 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Harris PA, Taylor R, Minor BL, et al. The REDCap consortium: building an international community of software platform partners. J Biomed Inform. 2019;95:103208. doi: 10.1016/j.jbi.2019.103208 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Anderson F, Downing GM, Hill J, Casorso L, Lerch N. Palliative performance scale (PPS): a new tool. J Palliat Care. 1996;12(1):5–11. [PubMed] [Google Scholar]
- 26.PDQ Supportive and Palliative Care Editorial Board. Cancer Pain (PDQ®): Patient Version. PDQ Cancer Information Summaries. National Cancer Institute (US); 2002. [Google Scholar]
- 27.Gunnarsdottir S, Serlin RC, Ward S. Patient-related barriers to pain management: the Icelandic Barriers Questionnaire II. J Pain Symptom Manage. 2005;29(3):273–285. doi: 10.1016/j.jpainsymman.2004.06.015 [DOI] [PubMed] [Google Scholar]
- 28.Mayahara M, Foreman MD, Wilbur J, Paice JA, Fogg LF. Effect of hospice nonprofessional caregiver barriers to pain management on adherence to analgesic administration recommendations and patient outcomes. Pain Manag Nurs. 2015;16(3):249–256. doi: 10.1016/j.pmn.2014.07.001 [DOI] [PubMed] [Google Scholar]
- 29.National Library of Medicine; National Institute of Health; U.S. Department of Health and Human Services. Form: BRICS NINR Demographics. Updated December 5, 2016. Accessed April 2, 2025. https://cde.nlm.nih.gov/formView?tinyId=XJo315r5M
- 30.Cella D, Riley W, Stone A, et al. The Patient-Reported Outcomes Measurement Information System (PROMIS) developed and tested its first wave of adult self-reported health outcome item banks: 2005–2008. J Clin Epidemiol. 2010;63(11):1179–1194. doi: 10.1016/j.jclinepi.2010.04.011 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Dworkin RH, Turk DC, Wyrwich KW, et al. Interpreting the clinical importance of treatment outcomes in chronic pain clinical trials: IMMPACT recommendations. J Pain. 2008;9(2):105–121. doi: 10.1016/j.jpain.2007.09.005 [DOI] [PubMed] [Google Scholar]
- 32.Keefe FJ, Ahles TA, Porter LS, et al. The self-efficacy of family caregivers for helping cancer patients manage pain at end-of-life. Pain. 2003;103(1–2):157–162. doi: 10.1016/s0304-3959(02)00448-7 [DOI] [PubMed] [Google Scholar]
- 33.Morisky DE, Green LW, Levine DM. Concurrent and predictive validity of a self-reported measure of medication adherence. Med Care. 1986;24(1):67–74. doi: 10.1097/00005650-198601000-00007 [DOI] [PubMed] [Google Scholar]
- 34.Elhenawy YI, Abdelmageed RI, Zaafar DK, Abdelaziz AW. Adherence to insulin therapy among children with type 1 diabetes: reliability and validity of the Arabic version of the 4-Item Morisky Medication Adherence Scale. Patient Prefer Adherence. 2022;16:1415–1421. doi: 10.2147/ppa.S341061 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Teng CL, Chan CW, Wong PS. Medication adherence of persons with type 2 diabetes in Malaysia: a scoping review and meta-analysis. J ASEAN Fed Endocr Soc. 2022;37(1):75–82. doi: 10.15605/jafes.037.01.14 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.McCoy CE. Understanding the intention-to-treat principle in randomized controlled trials. West J Emerg Med. 2017;18(6):1075–1078. doi: 10.5811/westjem.2017.8.35985 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Little RJA, Rubin DB. Statistical Analysis With Missing Data. 3rd ed. Wiley; 2024. [Google Scholar]
- 38.Cohen MZ, Thompson CB, Yates B, Zimmerman L, Pullen CH. Implementing common data elements across studies to advance research. Nurs Outlook. 2015;63(2):181–188. doi: 10.1016/j.outlook.2014.11.006 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Rosenthal R, Rosnow RL, Rubin DB. Contrasts and Effect Sizes in Behavioral Research: A Correlational Approach. Cambridge University Press; 2000. [Google Scholar]
- 40.Bock RD. Multivariate Statistical Methods in Behavioral Research. McGraw-Hill Series in Psychology. McGraw-Hill; 1975:xiii, 623. [Google Scholar]
- 41.Ward SE, Serlin RC, Donovan HS, et al. A randomized trial of a representational intervention for cancer pain: does targeting the dyad make a difference? Health Psychol. 2009;28(5):588–597. doi: 10.1037/a0015216 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Abdulhussein D, Yap TE, Manzar H, Miodragovic S, Cordeiro F. Factors impacting participation in research during the COVID-19 pandemic: results from a survey of patients in the ophthalmology outpatient department. Trials. 2022;23(1):823. doi: 10.1186/s13063-022-06748-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Cardel MI, Manasse S, Krukowski RA, et al. COVID-19 impacts mental health outcomes and ability/desire to participate in research among current research participants. Obesity. 2020;28(12):2272–2281. doi: 10.1002/oby.23016 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Chow R, Mathews JJ, Cheng EY, et al. Interventions to improve outcomes for caregivers of patients with advanced cancer: a meta-analysis. J Natl Cancer Inst. 2023;115(8):896–908. doi: 10.1093/jnci/djad075 [DOI] [PMC free article] [PubMed] [Google Scholar]