Abstract
Abstract
Background:
The quality of end-of-life (EOL) care at Veterans Affairs Medical Centers (VAMC) has improved. To date, however, the quality and outcomes of end-of-life care delivered to women veterans have not been examined.
Objective:
The goal of this study was to evaluate gender differences in the quality of EOL care received by patients in VAMCs nationwide.
Design:
The study was conducted via retrospective medical chart review and telephone survey with next of kin of recently deceased inpatients.
Setting/subjects:
The chart review included records for all patients who died in acute and long-term care units in 145 VAMCs nationwide (n=36,618). For the survey, the documented next of kin were invited to respond on behalf of the deceased veteran; a total of 25,638 next of kin completed the survey.
Measurements:
Chart review measures included five indicators of optimal end-of-life care. Bereaved family survey items included one global and nine specific items (e.g., bereavement care, pain management) describing care in the last month of life.
Results:
Receipt of optimal end-of-life care did not differ significantly between women and men with respect to frequency of discussion of treatment goals with a family member, receipt of palliative consult, bereavement contact, and chaplain contact with a family member. Family members of women were more likely than those of men to report that the overall care provided to the veteran had been “excellent” (adjusted proportions: 63% versus 56%; odds ratio (OR)=1.33; 95% confidence interval (CI) 1.10–1.61; p=0.003).
Conclusions:
In this nationwide study of all inpatient deaths in VAMCs, women received comparable and on some metrics better quality EOL care than that received by male patients.
Introduction
Gender disparities have become increasingly important for the Veterans Health Administration, the nation's largest health system. Although most VA patients are men, the number of women who receive health care from the VA nearly doubled from 2000 to 2009. 10 This increase reflects higher numbers of women serving in the military coupled with a higher proportion of women veterans seeking care at VA facilities. 11 In 2008 there were approximately 1.8 million women veterans, 12 and the population of women served by the VA is expected to increase substantially in the future.
Almost three decades ago the VA created the Women Veterans Health Program to streamline services and provide high quality care for women veterans. 13 Several initiatives to address women's specific health care needs have been implemented, including comprehensive primary care at separate women's health clinics, reproductive health services, and onsite child care. Despite these efforts, studies have documented gender disparities in the receipt, quality, and perceptions of care among veterans. 14 Washington and colleagues 12 reported that almost one in five women veterans either did not obtain the health care that they needed or delayed seeking health care during the 12 months prior to the survey. In a random, national sample of VA users, Wright and colleagues 15 found that women reported less satisfaction with inpatient VA care than did men, a finding that remained unchanged after controlling for covariates.
In addition to its efforts to improve women's health services, the VA launched a comprehensive program to enhance both access to and quality of hospice and palliative care delivered to veterans. Since the program's inception in 2008, the quality of end-of-life (EOL) care significantly improved.16,17 To date, gender differences in EOL care quality and outcomes for veterans receiving care at Veterans Affairs Medical Centers (VAMCs) have not been examined.
This paper examines gender differences in the quality of EOL care delivered to VA inpatients nationwide. This is the first study to examine the following questions: (1) How does EOL care delivered to women veterans compare to EOL care delivered to men? (2) How do next of kin evaluate the EOL care for men and women veterans? This study fills an important gap in the literature on gender disparities in the delivery of health care, and in particular, EOL care for women veterans.
Methods
Data sources and procedures
The data used to address the study aims were collected as part of a national VA quality improvement program called PROMISE (Performance Reporting and Outcomes Measurement to Improve the Standard of Care at EOL). VAMCs joined the quality improvement project on a rolling basis beginning in July 2008. By October 2009, 145 VAMCs had joined the project. This study included patients who died in VA inpatient facilities between July 2008 and June 2011. Inpatient deaths were retrieved from national VA databases derived from the electronic medical record; this method identifies over 95% of decedents.18,19 Human subjects approval for this secondary analysis of PROMISE data was obtained from the Philadelphia VAMC institutional review board.
Data collection involved both an extensive medical record review and a phone survey of bereaved family members (Bereaved Family Survey, described below) that focuses on their perceptions of the EOL care received by the deceased veteran in the 31 days prior to death. The chart review involved extraction of demographic and health-related variables as well as data related to several performance measures indicative of high-quality EOL care. These data were collected for nearly every veteran who died in a VAMC (including acute care units and long-term care settings called community living centers) during the study period. We excluded only patients who died as a result of suicide or accident and those who died within 24 hours of admission and/or in the emergency department. The latter, however, were included if they had also been admitted to a VA facility in the preceding month.
To collect data on family members' perceptions of EOL care, we identified the person listed as next of kin from the medical record according to the VA policy of death notification. In most cases, next of kin were spouses, children, or siblings, although approximately 14% of decedents named friends, live-in partners, and others. Four weeks after the patient's death, an introductory letter that described the survey was sent to the designated next of kin. Interviewers then made up to three call attempts. We excluded potential respondents who did not have a working telephone number or did not speak English or Spanish. We also excluded family members who stated that they could not evaluate the care that the patient received in the last month of life; we asked them to identify a knowledgeable informant, who was then contacted. If after three attempts interviewers were still unable to reach the family member, we mailed them the survey. These methods yielded a 70% response rate for the Bereaved Family Survey. Per protocol, interviews occurred between 6–10 weeks post-veterans' death.
Trained staff conducted interviews and chart reviews, both of which were protocol-driven. All staff were required to meet a minimal level of agreement and accuracy prior to collecting data independently. Supervisors conducted regular data quality audits.
Study variables
Demographic and patient clinical data
Patient age, gender, and race were collected through medical chart review. Diagnoses were collected from discharge summaries.
Chart review indicators of high-quality EOL care
The five indicators of optimal palliative care were positive answers to these questions: (1) Did the patient receive a palliative care consult within the last month of life? (2) Did the patient die in a designated hospice/palliative care unit? (3) Was there at least one contact between a chaplain and a family member within the last month of life? (4) Was there a discussion about treatment options and/or goals of care between a provider and a family member within the last month of life? (5) Within two weeks after the patient's death, was there bereavement contact with a family member? These factors have been shown in earlier studies to be associated with better quality care and higher bereaved family satisfaction.18,20,21 We also ascertained whether there was a do-not-resuscitate (DNR) order, because earlier analyses showed significant, positive associations between DNR orders and family members' perceptions of care quality. 21 A DNR may be a proxy measure for a “goals of care” conversation having taken place between the care team and patients or families.
Family perceptions of care at the EOL
The Bereaved Family Survey, 22 administered in either English or Spanish, was used to evaluate family members' perceptions of EOL care. It was derived from the Family Assessment of Treatment at EOL – Short Form, whose development and psychometric properties have been described.18,21,23,24 Ten core items that appear in both surveys were used to evaluate EOL care. Nine items measured specific aspects of care; for each item, scores were dichotomized as “1” (“always”) or as “0” for all other responses (“usually,” “sometimes,” “never”). The tenth, global item, which asked families to evaluate the overall care that the patient received in the last month of life (“poor,” “fair,” “good,” “very good,” or “excellent”), was dichotomized as “excellent” versus all other responses. 22
Statistical analysis
Veteran characteristics were compared by gender for the chart review and the survey samples. We used the χ2 test for categorical variables and the Wilcoxon rank sum test for age, which was coded as a continuous variable.
To compare quality of EOL indicators between female and male decedents, we fit unadjusted logistic regression models for each indicator, and then accounted for clustering within facilities with robust variance estimates. We report odds ratios for receipt of each indicator, comparing female and male veterans, along with corresponding 95% confidence intervals.
To reduce the potential impact of nonresponse bias on our estimates of gender differences in family satisfaction, we adjusted for nonresponse in our survey analyses, using response propensity weights. We first fit a logistic regression model in which the outcome variable was Bereaved Family Survey completion, and included as explanatory variables patient characteristics associated with likelihood of response, including patient age, race, and relationship to survey respondent. Using this model, we calculated a weight for each patient that was equal to the reciprocal of that patient's probability of having a completed Bereaved Family Survey. We applied these calculated weights to the sample of patients who completed the survey, and checked that the characteristics of the weighted sample were close to those of the full sample of eligible patients, using the standardized difference for each characteristic. 25 We then used this weighted sample of Bereaved Family Survey completers in all subsequent analyses.
To compare Bereaved Family Survey scores for men and women veterans, we fit a series of response propensity weighted logistic regression models for the overall satisfaction score and for each individual item score. We accounted for clustering of observations within facilities by using the Generalized Estimating Equations (GEE) method with robust Huber-White standard error estimates. We then conducted multivariable adjusted analyses controlling for patient age, ethnicity, and respondents' relationship to the decedent. We compared the global item scores and individual item scores between female and male patients. Individual survey items were analyzed using all available data, and we report the proportion missing for each item in completed surveys. We report odds ratios for the likelihood of reporting excellent care for each survey item, comparing females to males.
Stata statistical software version 11.2 (StataCorp., College Station, TX) was used for all analyses. 26
Results
Sample characteristics
We identified 42,951 patients who died in one of 145 VA facilities between July 2008 and June 2011. Of these, 6333 were excluded randomly from further data collection to manage staff workload, leaving 36,618 (85.3% of eligible veterans) in the chart review sample (see Table 1). Of these 36,618 decedents, 1921 family members could not be contacted, either because they did not have a working phone number or valid address (n=1820) or there was no next of kin listed (n=101). Of the remaining sample, Bereaved Family Surveys could not be completed for 4323 veterans, because the family member did not answer the phone or return a mail survey, and 3388 family members refused to participate. In addition, 638 potential respondents did not have adequate knowledge about care received in the veteran's last month of life, 190 were reluctant to talk about the veteran's death, and 121 were ill and unable to complete a survey. Thirty-five respondents were excluded because they spoke neither English nor Spanish. Of the remaining 26,002 respondents, 364 family members did not complete the entire survey, leaving 25,638 in the final Bereaved Family Survey.
differences between groups evaluated with Wilcoxon Rank-Sum Test for continuous variables, χ2 for categorical variables.
Multiple diagnoses were coded for each patient; proportions sum to >100%.
BFS, Bereaved Family Survey; PTSD, posttraumatic stress syndrome; SD, standard deviation.
Table 1 describes the characteristics of men and women decedents from the chart review and Bereaved Family Survey samples. Women veterans were younger than male veterans, in both samples. Men were significantly more likely than women to have kidney disease, pneumonia, coronary artery disease, and heart failure. For male veterans, the most likely Bereaved Family Survey respondents were spouses; for women veterans, the most likely respondents were their sons or daughters.
Chart review indicators of high-quality EOL care
Receipt of optimal EOL care did not differ significantly between women and men veterans with respect to frequency of treatment goals discussion with a family member, palliative consult, bereavement contact, and chaplain contact with a family member. Among patients who died in facilities with inpatient hospice units, women were more likely than men to die in the hospice unit. Women were also more likely than men to have completed a DNR order. Results are presented in Table 2.
These results are adjusted for clustering within facilities.
Less than 5% missing data. Incomplete cases were dropped.
Only includes deaths in facilities with an inpatient hospice unit N=26,382; men=25,840; women=542.
CI, confidence interval; DNR, do not resuscitate; EOL, end of life; VA, Veterans Administration.
Family perceptions of care at the EOL
The following patient characteristics were associated positively with survey completion and were included in the response propensity model: white race, older age of veteran, and having a spouse identified as the next of kin. 25 After response weights were applied, we compared the weighted sample of veterans with completed surveys with the full study sample. The standardized differences between the weighted sample of survey responders and the full sample were all less than 10%, which suggests that our weighted sample was adequately representative of the population of interest with respect to measured characteristics.
Of the 25,638 completed surveys, 559 were from family members of women veterans and 25,079 were from family members of male veterans (see Table 1). After adjusting for nonresponse, patient age, family relationship, and ethnicity, and clustering by facility, family members of women veterans were more likely than those of male veterans to report that overall care in the last month of life had been “excellent” (adjusted proportions: 63% versus 56%; OR=1.33; 95% CI 1.1–1.61; p=0.003). A significant gender difference was found for three of the nine specific items. First, more family members of female veterans than male veterans agreed that providers always gave enough spiritual support to the veteran in the last month of life. Second, more family members of female veterans than male veterans reported that providers always gave enough emotional support in the month prior to the veteran's death. And third, family members of women veterans were more likely to agree that the VA always provided enough emotional support after the veteran's death. No significant gender differences were observed for the remaining six items (see Table 3).
Gender comparisons adjusted for age, race, family relationship, and clustered by facility.
Sample size varies to reflect missing data for individual items.
Includes only those patients whose family responded that the veteran experienced pain, N=20,886.
CI, confidence interval.
Of note, a small minority of family members reported that the veteran's pain never made him or her uncomfortable in the last month of life (adjusted proportions: women, 10%; men, 11%). We did not observe a significant difference between genders on this metric.
Discussion
This study documented relatively high levels of bereaved family satisfaction with EOL care for veterans, with care in the last month of life rated as “excellent” by the family members of 63% of female and 56% of male veterans. These are higher levels of overall satisfaction with inpatient EOL care than were reported in an earlier, nationally representative survey of family members. 20 Nonetheless, there is room for improvement; in a previous national study, Teno and colleagues 20 reported bereaved family member overall satisfaction score for patients receiving hospice care—the gold standard for end-of-life care—as 70.7%.
Our study findings also indicate that, on most metrics, female and male veterans received similar care at EOL. The few significant differences between the two groups suggested that women veterans received slightly higher quality of EOL care. For example, significantly more women veterans died in designated inpatient hospice units. Based on other nationwide evidence, family satisfaction with EOL care is typically higher when the death occurs in an inpatient hospice rather than in an acute care hospital or a nursing home. 20 In this study, family members of women veterans also reported significantly higher satisfaction with spiritual and emotional support provided to the veteran in the last month of life, and to the family after the death.
For more than 25 years, federal agencies, including the VA, have stipulated that women must be included in research studies. However, since women represent a minority of VA health care users, obtaining adequate sample sizes among women veterans is a persistent methodological challenge. Findings from this study are particularly compelling, given that we found limited evidence of gender disparities in EOL care in this large, nationally representative sample—one that (for the chart review component) approximates the population of all veterans who died as VA inpatients during the study period. For the survey component of the study, the response rate for eligible decedents was 70%; attempts were made to elicit data from next of kin of nearly all inpatient decedents.
Veterans, and particularly women veterans, may be reluctant to seek VA care, because they perceive that VA care is of poor quality. 27 However, this study and others27,28 document relatively high quality of care at the VA. A study by Lu and colleagues 17 reported that family members of VA decedents rated inpatient EOL care significantly higher than did families of persons dying in non-VA facilities.
The overwhelming majority of veterans had DNR orders in place at the time of death and had conversations with their health care providers regarding their care needs—essential components of EOL care planning. However, among both men and women, there remain important areas for attention, including room for improvement in the area of family members' overall assessment of the EOL care. A specific area of marked need is in pain control for both women and men. Few family members reported that pain was never an issue during the last month of the veteran's life. Pain control should be a care goal whenever feasible. According to family members' reports, pain was inadequately controlled for the majority of patients in this study, as has also been reported in a range of studies outside the VA.20,29
One key priority for the Women Veterans Health Program is the call to “accurately represent and evaluate women veterans' health and health care needs through data and analysis.” This report responds to that call, offering, for the first time, a comparison of next of kin perceptions of care for both men and women veterans at EOL. Although the PROMISE Center does not collect gender-specific performance measures, and so it may be that we have not measured key areas of concern to female veterans, the PROMISE staff routinely seeks open-ended responses regarding satisfaction with care in all follow-up telephone calls with family members. Family members have occasionally indicated that there were inadequate gender-specific accommodations for women. However, this finding was uncommon in the open-ended responses, and any dissatisfaction with privacy, if present, did not result in lower overall satisfaction scores reported by family members of deceased female veterans.
Family members' perceptions of care are widely used to evaluate EOL care.20,30,31 However, families' perceptions of care may not accurately reflect patients' own experiences and satisfaction with care. This difference may in part explain the contradictory findings of Wright and colleagues, 15 who reported that women inpatients at VA hospitals were less satisfied than men with the care they received. Prior research conducted among almost two million recently hospitalized adults suggests that women are generally less satisfied with their health care than are men 7 —a phenomenon that also may be operative in family members' reports of satisfaction. Since nearly 40% of respondents for male veterans were spouses (and thus likely to be women), the lower ratings for male veterans on the support items may be attributable more to respondents' gender rather than the gender of the veterans themselves. Evaluating the quality of care using chart review data also has limitations. Rarely does documentation in the medical record reflect the depth of conversations or the thoroughness or complexity of interdisciplinary assessment and interventions. Moreover, we did not evaluate the timing of specific interventions, such as discussion of goals of care or palliative care consults, which may be more effective earlier rather than shortly before death. Nonetheless, our aim was to compare the care between male and female veterans. Using two types of outcomes, that is, family perceptions of care and documentation of care processes, we have provided strong evidence that the care for the two groups was similar in this population, on most metrics.
Another limitation is that this study incorporated response propensity weights to account for possible nonresponse bias in the analysis of survey data. This technique is widely accepted in health services research. Nevertheless, its effectiveness depends upon the accurate measurement of variables of interest, and there remains the possibility that other, unmeasured characteristics may have influenced the results reported here. One characteristic that may have influenced our results is the gender of the respondent, which was unmeasured.
Finally, there may be limited generalizability of these findings to women who were not veterans. Nonetheless, there is scant evidence about gender differences in the delivery and outcomes of end-of-life care, and thus our study contributes important information about the care of women in hospice and palliative care settings within the nation's largest health system. Unlike other investigators,6,7 we did not find gender differences in pain assessment or management, although our data suggested inadequate treatment for pain in both men and women. Nor did we find, as reported by Wachterman and Sommers, 8 that women were less likely than men to be cared for in a dedicated hospice/palliative care unit. These findings suggest that, even in a health care system with a predominantly male patient population, high-quality care for women at the end of life as possible. Thus, the VA's care processes, documentation of care, and outcome measurement may serve as an example to other health care systems.
This study provides reassuring evidence that comparable quality of care is provided to men and women veterans who died in VA facilities. In some cases, quality indicators pointed to somewhat higher quality of care received by women, though small absolute differences observed were statistically significant in part because of the large sample size. This study also provides evidence of work that remains to improve EOL care for all veterans, especially in the areas of bereavement counseling, emotional support, and pain management.
Footnotes
Acknowledgements
This material is based upon work supported by the Department of Veterans Affairs, Veterans Health Administration, Office of Research and Development. All authors had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs or the United Stated Government. The authors thank Melissa Dichter, MSW, PhD, for her thoughtful review of the manuscript prior to its submission.