Use of Palliative Care and Hospice Among Surgical and Medical Specialties in the Veterans Health Administration

Importance  Many hospitals have undertaken initiatives to improve care during the end of life, recognizing that some individuals have unique needs that are often not met in acute inpatient care settings. Studies of surgical patients have shown this population to receive palliative care at reduced rates in comparison with medical patients.

Objective  To determine differences in the use of palliative care and hospice between surgical and medical patients in an integrated health care system.

Design, Setting, and Participants  Veterans Health Administration (VHA) enrollment data and administrative data sets were used to identify 191 280 VHA patients who died between October 1, 2008, and September 30, 2012, and who had an acute inpatient episode in the VHA system in the last year of life. Patients were categorized as surgical if at any time during the year preceding death they underwent a surgical procedure (n = 42 143) or medical (n = 149 137) if the patient did not receive surgical treatment in the last year of life.

Main Outcomes and Measures  Receipt of palliative or hospice care and the number of days from palliative or hospice initiation to death were determined using VHA administrative inpatient, outpatient, and fee-based encounter-level data files.

Results  Surgical patients were significantly less likely than medical patients to receive either hospice or palliative care (odds ratio = 0.91; 95% CI, 0.89-0.94; P < .001). When adjusting for demographics and medical comorbidities, this difference was even more pronounced (odds ratio = 0.84; 95% CI, 0.81-0.86). Yet, among patients who received hospice or palliative care, surgical patients lived significantly longer than their medical counterparts (a median of 26 vs 23 days, respectively; P < .001) yet had similar relative use of these services after risk adjustment.

Conclusions and Relevance  In the VHA population, surgical patients are less likely to receive either hospice or palliative care in the year prior to death compared with medical patients, yet surgical patients have a longer length of time in these services. Determining criteria for higher-risk medical and surgical patients may help with increasing the relative use of these services. Potential barriers and differences may exist among surgical and medical services that could impact the use of palliative care or hospice in the last year of life.

Palliative care incorporates many clinical areas other than end-of-life care, including cultural, spiritual, social, psychological, and physical aspects of care for patients with chronic life-limiting illness.¹ Several studies and multiple clinical trials demonstrate that hospice and palliative care greatly reduce health care costs while concurrently providing better quality and improved access to care.^2,3 Earlier palliative care consultation decreases the use of costly intensive care unit (ICU) care without changing mortality.⁴ One quantitative study has shown differences in end-of-life care among medical and surgical specialties, focusing only on care received in the ICU.⁵ The authors found that patients in the ICU with surgical attending physicians had lower ratings of quality of dying than those cared for by medical attending physicians in the ICU. They also found surgical patients had less documented palliative care than medical patients.⁵

With advancements in cancer treatment, palliative care has shown reduction in morbidity and mortality.⁶ As a result, many health care systems are focusing on clinical services offered by hospice and palliative medicine. New education curricula now incorporate palliative medicine into family medicine, internal medicine, and surgery to suit the needs of the changing older patient population.^7,8

Early communication and structured discussions about end-of-life care may be provided in inpatient settings to reduce suffering in the last days of life. Such communication and goal setting are fundamental to patient-centered care. This issue may be especially salient for the Veterans Health Administration (VHA), given the unique needs of veterans at the end of life. For example, veterans are more likely than nonveterans to experience anxiety and sadness at the end of life.⁹ Nevertheless, inpatient health care professionals who traditionally emphasize curative care may fail to recognize patients who would benefit from end-of-life discussions.

In 2002, the VHA established initiatives to improve clinical programs for end-of-life care. The Best Practices for End-of-Life Care for Our Nation's Veterans Trial intervention was developed to improve health care professional skills for identifying patients in acute-term or long-term inpatient settings who were entering the dying process. That trial produced modest yet statistically significant increases in the use of advanced directives and pain management.¹⁰ Since this trial, the VHA has developed further interventions to improve end-of-life care further, reflected in the 2008 VHA directive.

Qualitative differences in the medical and surgical specialties are known regarding end-of-life and advanced care planning.¹¹ However, the use of palliative care services among surgical and medical specialties in acute inpatient settings has received little attention, and its use among these specialties is unknown.

This study examines the use of end-of-life care in the VHA health system among surgical and medical patients in the last year of life. Specifically, we used VHA national administrative data files to determine the initiation and timing of palliative and hospice services for veterans hospitalized in acute surgery and medical services in VHA hospitals prior to death.

Enrollment data and patient treatment files from the VHA were used to identify VHA patients who died between fiscal years 2009 and 2012 and had an acute inpatient admission (either ICU or acute inpatient) in the VHA system within 365 days prior to death. Patients were excluded if the initial palliative care or hospice point was more than 1 year prior to death. A total of 191 280 patients were included in the analysis. Patients were categorized as surgical if the patient underwent a major surgical procedure in the last year of life (n = 42 143) and medical if the patient did not receive a surgical procedure in the last year of life (n = 149 137). The classifications of major therapeutic or diagnostic surgeries were defined by the Healthcare Cost and Utilization Project (HCUP) Procedure Classes.¹² This study was approved by the institutional review board of the University of Iowa and the Iowa City Veterans Affairs Research and Development Committee. A waiver of participation consent was obtained through the institutional review board.

Receipt of palliative or hospice care and the number of days from palliative or hospice initiation to death were determined using VHA administrative inpatient, outpatient, and fee-based encounter-level data files. Palliative care was designated based on International Classification of Diseases, Ninth Revision diagnostic code V66.7. Hospice care was determined from VHA inpatient, outpatient, and fee-based care files and included home hospice as well as hospice provided in inpatient settings. In general, patients in hospice must have a prognosis of 6 months or less, while patients in palliative care are not restricted by prognostic information. Because hospice is a subset of palliative care, we identified the following 3 service categories: nonhospice palliative care, hospice care, and either nonhospice or hospice palliative care. The number of days in hospice or palliative care was also calculated by subtracting the initial date of hospice or palliative from the date of death.

Additional patient characteristics were identified from VHA claims and included demographics (age, sex, and race/ethnicity), veteran eligibility category (service-connected disability, low income, or other), and comorbid conditions. Specific comorbid conditions were identified by categorizing International Classification of Diseases, Ninth Revision, Clinical Modification codes available on VHA claims during the 12 months prior to death according to the algorithm described by Quan et al.¹³ Using these comorbid conditions and patient age, we calculated a Charlson Comorbidity Index (CCI) for each patient.^14,15

First, differences between medical and surgical patients in sex, race/ethnicity, and comorbid conditions were determined using the χ² test.

Second, differences in the proportion of patients receiving palliative or hospice care during the year prior to death were examined in unadjusted analyses using the χ² test and in risk-adjusted analysis using logistic regression. Logistic regression models controlled for patient demographics (age, sex, and race/ethnicity), veteran eligibility category, and comorbid conditions. The logistic models were estimated as generalized estimating equations with an exchangeable working correlation matrix to account for the clustering of patients in facilities. The regression coefficient associated with surgical patients was exponentiated to provide the relative odds of palliative or hospice care for surgery patients, relative to medical patients.

Third, differences in the number of days in end-of-life care (ie, days between palliative or hospice care initiation and death) were examined. Because the days of care were highly skewed, unadjusted differences in days of care were determined using the Wilcoxon rank test as well as the t test. Risk-adjusted differences were determined using a generalized linear model with a log link and Poisson distribution. In these models, the exponentiated value of the regression coefficient associated with surgical patients provides the relative number of days for surgical patients vs medical patients.

Finally, trends in the receipt of palliative or hospice services by medical and surgical patients during the study period were examined. Statistical analysis was performed using SAS version 9.3 (SAS Institute Inc).

Characteristics of patients who had surgery at the end of life vs medical (those who did not have surgery) are shown in Table 1. For our VHA population, 22% (n = 42 143) of patients had at least 1 major surgical procedure in the last year of life. When compared with medical patients, surgical patients were younger and slightly more likely to be black individuals. They had more comorbid conditions, including a higher prevalence of anemia, diabetes mellitus, acute myocardial infarction, malignant or metastatic cancer, arrhythmias, cerebral vascular disease, pulmonary circulatory disorder, peripheral vascular disease, and renal disease. Furthermore, surgical patients tended to have higher CCI than their medical counterparts (median, 7 vs 9; P < .001).

Overall, surgical patients were significantly less likely than medical patients to receive palliative or hospice services in the last year of life (38.3% vs 41%; P < .001). This difference was also present in a separate analysis of palliative care (36.5%, surgical vs 38.9%, medical) and hospice (21.2%, surgical vs 23.8%, medical) as shown in Table 2. Moreover, differences in the use of palliative or hospice care were intensified after adjusting for patient characteristics. For example, the odds of receiving any palliative or hospice services in the year prior to death for surgical relative to medical patients was 0.91 (odds ratio = 0.91; 95% CI, 0.89-0.94; P < .001) without risk adjustment and 0.84 (odds ratio = 0.84; 95% CI, 0.81-0.86; P < .001) after risk adjustment. This relationship between medical and surgical patients was also seen individually in both palliative care and hospice groups.

In addition to the lower percentage of surgical patients receiving palliative or hospice care, modest but statistically significant differences were noted in the number of days between palliative or hospice care initiation and date of death for medical and surgical patients. (Table 3). For example, the median time between the initiation of either palliative or hospice care and death was slightly higher for surgical patients, with 23 days for medical patients and 26 days for surgical patients (P < .001). Yet when controlling for patient demographics and comorbidities, surgical patients had decreased relative use of either service, with the odds of use for surgical patients 0.95 (95% CI, 0.93-0.98; P < .001) times less than the odds of use for medical patients.

Finally, the use of palliative services increased over the study period, from 29% to 47% for medical patients and from 27% to 45% for surgical patients (Figure). Hospice use increased from 21% to 27% and from 19% to 24% for medical and surgical patients, respectively. Similarly, the median time between palliative or hospice initiation increased over time, from 22 to 25 days for medical patients and from 22 to 30 days for surgical patients.

Our study examined differences in the receipt of end-of-life services among VHA patients. Similar to studies with Medicare data,¹⁶ we found that slightly fewer than 1 in 3 individuals in the last year of life will undergo a major surgical procedure. We found surgical patients significantly less likely than medical patients to have palliative care or hospice before death. These differences were accentuated further when controlling for other factors, including demographics and comorbidities. Both surgical and medical subspecialties have continued to increase use of palliative care and hospice at similar rates during the past several years. In comparison with the number of days spent in palliative care or hospice, surgical patients receive these services for significantly longer periods of time than medical patients.

A previous study described differences in end-of-life care for patients in the ICU who were cared for by surgical and medical subspecialties. Those authors found that patients who had attending physicians from surgery service care for them in the ICU at the end of life had less documentation of palliative care when compared with patients cared for by medical attending physicians.⁵ This is similar to the results of our study, which showed decreased likelihood of receiving either palliative care or hospice for surgical patients. Perhaps surgical patients receive fewer consultations based on the qualitative differences between medicine and surgery. Schwarze et al¹¹ discuss this complicated relationship, which may play a large role in surgical buy-in.

Our study is unique from previous studies reviewing use of palliative care because those studies focused primarily on ICU patients.^2,4,5,9,17 Our study includes data not only on patients admitted to the ICU but also on patients admitted to general hospital floors, thus expanding the generalizability of the results. By using national VHA administrative data, our study includes a much larger number of patients compared with other studies that have been based on medical record reviews of single institutions or small multi-institutions.⁵

Half of medical and surgical patients who receive either palliative care or hospice in the VHA spend no more than 3 to 4 weeks in those respective services. Surgical patients may receive fewer referrals to hospice or palliative care yet tend to live longer in these services than medical patients. This phenomenon may be a result of several interactions. Surgical patients may be more likely to undergo aggressive treatment, either curative or symptom-focused, or some procedures may be performed with palliative intent; thus, there is better symptom management for these patients. Perhaps surgeons are also able to identify appropriate patients earlier. Involving surgeons more could help increase access and earlier referrals to quality palliative care for surgical patients. Future research needs to address this further.

Earlier palliative care intervention improves quality of life and survival,¹⁸ yet one study found that patients who were in hospice for 1 week or less had the same quality of life as those who did not receive hospice.¹⁹ Thus, if the patients in our study had been identified earlier and palliative care or hospice initiated earlier in their illness, they may have benefited from improved quality of life and survival. An indicator variable that is relatively easy to calculate, such as the CCI, may be of a benefit in this population. Surgical patients tended to have slightly higher CCI than their medical counterparts; selecting a cutoff point could be a good trigger to initiate palliative care and hospice. The CCI has its limitations and does not include functional status, an important predictor of mortality.²⁰

Studies show palliative care specialists provide patients with more accurate prognostic information in comparison with nonspecialists; thus, it is important to involve them earlier with chronic or life-limiting illnesses.²¹ A prognosis of a 6-month life expectancy is necessary to be enrolled in hospice, but the same is not true for palliative care; patients can be considered palliative years prior to anticipated death. This is important because despite the prognostic distinction between the groups, the difference in magnitude of the median or mean days spent in palliative care or hospice is no more than 2 to 3 weeks.

Our results show differences among surgical and medical patients that may indicate the need for use of different triggers for each group. Many hospitals and health care systems have initiated triggers for palliative care consultation. One study reviewed trigger criteria for ICU patients (ICU admission with hospitalization of >10 days, multisystem organ failure of >3 systems, stage IV malignancy, admission following cardiac arrest, or intracerebral hemorrhage requiring ventilator support) and found that 1 in 7 patients met at least 1 of the criteria for consultation.²² Yet these criteria are based on an ICU population and not applicable to a general acute inpatient admission or an outpatient setting. Using triggers associated with mortality and hospital readmission may be a possible way to identify earlier patients who may benefit from hospice or palliative care,²³ but they need to be used in the appropriate patient population.

Finally, we explored whether the increase in the use of palliative and hospice services during our study period is similar to trends in the use of these services outside the VHA. During our study period from 2009 through 2012, the use of hospice by VHA patients increased roughly 27%, while the number of hospitals with 1 or more patients referred to hospice increased by 18%. Similarly, the use of palliative care increased by 52% overall, and the number of hospitals providing palliative care increased by 21%. These increases appear to be larger than those experienced in US hospitals outside the VHA. For example, the number of US hospitals with palliative care was estimated to increase by 14%,²⁴ and the number of individuals using hospice prior to death was estimated to increase by 14%²⁵ during the same period. It is possible that the larger increases in the VHA may be attributed to the VHA’s ongoing initiative to improve access to end-of-life services in VHA hospitals.

There are a few limitations of our study. First, although we had data on all veteran deaths from fiscal years 2009 to 2012, it is possible patients received end-of-life services through other sources that would not be included in VHA administrative data. Therefore, our findings likely underestimate the rates of receipt of services. Additionally, using administrative data may not provide the most clinically relevant information or provide background circumstances for a surgical procedure or referral to palliative care or hospice. For example, many surgical patients are seen by the surgeon in an outpatient clinic and scheduled for elective surgery. From our administrative data, we were unable to determine circumstances (elective vs emergent) leading to surgical intervention. Also, we were unable to distinguish surgery as a palliative procedure. In our study, all surgical patients had at least 1 major surgery prior to either hospice or palliative consultation, with only a small group undergoing a major surgical procedure following initiation of palliative care and hospice. Some studies found that procedures performed with palliative intent can improve quality of life and increase longevity,²⁶ but our study cannot speak to this.

Another limitation is that there is no gold standard to measure receipt of end-of-life services; thus, precise estimates of the desired rates or number of days of care are not available for comparison. Although the VHA is an integrated health care system, regional and individual hospital variation exists. Some hospitals use an integrative model, focusing on incorporating principles and interventions of palliative care into daily practice, while others use a consultative model, focusing on involvement and effectiveness of palliative care practitioners with high-risk patients, or some combination of the 2 to identify patients with potential unmet palliative care needs.²⁷ These methods can be very effective; however, there may be concerns by surgeons of losing individual control of patient care, especially in the integrative model. Yet, identification of the appropriate patient for palliative care or hospice remains one of the largest challenges because an optimal trigger or screening criteria has not been identified.

Regardless, our objective was to describe differences between medical and surgical specialties in patients in the last year of life. There may be other differences among patients (cultural and religious barriers) and physicians (moral and ethical) not reviewed in this study, which may influence the use of services for end-of-life care.^28,29 In particular, patients who prefer more aggressive, life-saving interventions may be more likely to be admitted to the surgery service compared with other patients, which may explain the lower use of palliative care for surgical patients found in this study.

In 2008, the VHA health system mandated the creation of palliative care consult teams in individual facilities to improve the quality of end-of-life planning for patients with advanced disease that is life-limiting. The increase in use of palliative care and hospice services may be related to this initiative given the increasing use we found in our analysis.

The VHA has made substantial improvements in the provision of end-of-life care to hospitalized veterans, although the use of palliative services by surgical patients lags somewhat compared with medical patients. Further studies need to be conducted to see if these differences are clinically relevant and affect patient outcome.

The hospice and palliative care movement has focused on improving quality of life, which pairs well with the rising importance and growth of quality improvement in health care. This study highlights differences and disparities in end-of-life care between surgical and medical patients, which may be a potential area of improvement. Further studies need to be conducted to determine the clinical significance of the differences in these groups and to establish if earlier intervention in surgical patients is possible using triggers.

Corresponding Author: Courtney L. Olmsted, BSE, MD, Iowa City Veterans Affairs Healthcare System, 601 Highway 6 West, Iowa City, IA 52246 (courtney.olmsted@va.gov).

Accepted for Publication: June 19, 2014.

Published Online: September 24, 2014. doi:10.1001/jamasurg.2014.2101.

Author Contributions: Dr Olmsted had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Olmsted, Kaboli, Cullen.

Acquisition, analysis, or interpretation of data: Olmsted, Johnson, Kaboli, Vaughan-Sarrazin.

Drafting of the manuscript: Olmsted, Cullen.

Critical revision of the manuscript for important intellectual content: Olmsted, Johnson, Kaboli, Vaughan-Sarrazin.

Statistical analysis: Olmsted, Johnson, Vaughan-Sarrazin.

Obtained funding: Kaboli.

Administrative, technical, or material support: Kaboli, Vaughan-Sarrazin.

Study supervision: Kaboli, Cullen, Vaughan-Sarrazin.

Conflict of Interest Disclosures: None reported.

Funding/Support: This work was supported by the Department of Veterans Affairs, VHA, and the Health Services Research and Development Service through grant CIN 13-412 from the Comprehensive Access and Delivery Research and Evaluation Center. Dr Olmsted is supported by the Veterans Affairs Quality Scholars Fellowship Program and Veterans Affairs Office of Academic Affiliation.

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Previous Presentation: This paper was presented at the 38th Annual Surgical Symposium of the Association of VA Surgeons; April 7, 2014; New Haven, Connecticut.