Vol. 50 No. 2 August 2015
Journal of Pain and Symptom Management 147
Original Article
Cost Savings Associated With an Inpatient Palliative Care Unit: Results From the First Two Years Jeremy D. Nathaniel, BA, Melissa M. Garrido, PhD, Emily J. Chai, MD, Gabrielle Goldberg, MD, and Nathan E. Goldstein, MD Icahn School of Medicine at Mount Sinai (J.D.N.), New York, New York; Geriatrics Research, Education and Clinical Center (M.M.G, N.E.G.), James J. Peters Veterans Affairs Medical Center, Bronx, New York; Lilian and Benjamin Hertzberg Palliative Care Institute of the Brookdale Department of Geriatrics and Palliative Medicine (M.M.G., E.J.C., N.E.G.), Icahn School of Medicine at Mount Sinai, New York, New York; and Hofstra North Shore-LIJ School of Medicine (G.G.), Hempstead, New York, USA
Abstract Context. Palliative care consultation services (PCCS) decrease costs for patients by matching treatments received to patients’ and families’ goals of care. However, few studies have examined the costs of a specialized palliative care unit (PCU). Objectives. To quantitatively describe Mount Sinai Hospital’s PCU’s first two years of operation; to examine how patientrelated costs changed in the days before and after transfer to PCU; and to compare cost savings of PCU to those of PCCS. Methods. Cost and administrative data from PCU patients from the first 24.5 months of our PCU’s operation were analyzed. To compare costs between PCU and PCCS patients, we matched PCU patients to similar PCCS patients and used propensity scores to adjust for differences across groups. Results. The PCU admitted 1107 patients in its first 24.5 months. Over this time frame, there was a statistically significant (P < 0.001) decrease in average daily direct costs per patient. The mean of patients’ average cost per day was $687 less while on the PCU than before transfer to PCU. Among patients who died in the hospital, average daily direct cost per patient in the days after transfer to PCU was $240 lower as compared with patients being followed by PCCS on the general hospital wards (SE ¼ $45, P < 0.001). Conclusion. Among patients who died in the hospital, transfer to a PCU is associated with significant cost savings as compared with patients on hospital wards who are seen by a PCCS. J Pain Symptom Manage 2015;50:147e154. Published by Elsevier Inc. on behalf of American Academy of Hospice and Palliative Medicine. Key Words Palliative medicine, palliative care unit, cost savings, cost, longitudinal study
Introduction Palliative care consultation service (PCCS) teams have been shown to improve symptom control, satisfaction, and quality of life for patients with advanced illness and their families.1e4 In recent years, the benefits of PCCS also have extended to reducing hospitalization costs for patients with advanced illness.5 Owing to its proven benefits to both hospital and patient, availability of palliative care in the U.S. hospitals
Address correspondence to: Nathan E. Goldstein, MD, Department of Geriatrics and Palliative Medicine, Mount Sinai Medical Center, One Gustave L. Levy Place, Box 1070, New York, NY 10029, USA. E-mail: [email protected] Published by Elsevier Inc. on behalf of American Academy of Hospice and Palliative Medicine.
more than doubled from 2001 to 2011.6 As of 2011, 63% of hospitals with 50 or more beds had palliative care teams and that number is growing.6 Along with the increase in palliative care consult services, there has been an increasing number of hospitals investing in palliative care units (PCUs), which are specialized hospital floors devoted specifically to providing palliative care.7,8 Numerous studies of PCU patient outcomes have affirmed the importance of PCU as a palliative care modality.9e14 These units
Accepted for publication: February 19, 2015.
0885-3924/$ - see front matter http://dx.doi.org/10.1016/j.jpainsymman.2015.02.023
148
Nathaniel et al.
are founded on the belief that a specialty-level trained palliative medicine team working together in the same location can best meet the needs of palliative care patients with advanced illness. These units require considerable start-up costs,15 including specialty training for nurses and other staff who work on these units, and there is often a large upfront capital expense to renovate a traditional hospital floor to become a more home-like environment for patients and their families. Although many studies have demonstrated the cost-savings of PCCS teams,16e20 few have examined whether PCUs lead to the same lowered health care costs as PCCS teams, and to our knowledge, none have explicitly examined the cost savings of a PCU compared with a PCCS team. PCUs have been shown to provide higher quality of care than PCCS teams, as measured by patients’ family members’ satisfaction and PC team satisfaction.13,15,21 However, further information is needed about the relationship of these units to hospital costs. In this article, we describe the demographics and patterns of care for the first two years of our unit’s existence, analyze how patient-related costs changed over time, and determine the cost differences between patients cared for on our PCU and patients cared for by our PCCS team.
Methods The Palliative Care Unit The Mount Sinai PCU is a physically distinct unit, comprising part of one floor of the hospital. Mount Sinai’s PCCS has been in existence since July 1997, but our 13-bed PCU did not open until June 2011. The majority of patients are referred to PCU from other units throughout the hospital, although some admissions come directly from the Emergency Department or home. Criteria for admission to PCU include patients with: difficult to manage symptoms who need rapid titration of medications; progressive illness who need special assistance in establishing goals of care; a solely comfort-oriented plan of care and who are awaiting discharge to an appropriate care setting; and advanced illness who are in an intensive care unit but who will not be receiving further escalation of medical treatments.
Data Collection The Palliative Care Department employs a data manager who records and tracks data on all patients who are seen by PCCS or who are admitted to PCU. To ensure accuracy, the data are checked daily with the census on the PCU and any discrepancies between the two are resolved within one day. These
Vol. 50 No. 2 August 2015
characteristics include demographics, diagnosis, dates of admission and discharge to PCU, and other relevant clinical data. We merged our internal database tracking PCCS and PCU patients with cost data captured in the Transitions Systems Inc. system at Mount Sinai. Using the Transitions Systems Inc. system we obtained a report of daily direct costs incurred by every PCU and selected PCCS patients. This work was approved by the Program for the Protection of Human Subjects at Mount Sinai.
Data Analysis We first compiled demographic and clinical characteristics for the 1107 patients admitted to PCU from June 2011 to June 2013. To examine changes in demographic characteristics over this time period, we graphed several of the PCU’s daily patient demographics and clinical characteristics over time. There were 36 patients for whom cost data were unavailable, so they were excluded from all cost analyses. All cost data were adjusted by the medical Consumer Price Index to 2013 dollars.22 To examine changes in average patient costs over PCU’s first 24.5 months of operation, we calculated the mean cost per PCU patient for each day and then grouped these daily means into half-month increments. This created 49 data points (two points per month x 12 months per year x 2 years and half of a month) that could be used to examine changes in costs over time. To examine the relationship between a given period of operation and PCU’s mean daily costs, we ran a generalized linear model (GLM) with gamma distribution and log link. We ran both an unadjusted model, as well as a model that adjusted for other covariates that might influence mean costs (percent male, mean age, percent with Do Not Resuscitate orders, percent with cancer diagnoses, percent white patients, and average daily census). Marginal effects (change in mean daily cost for each additional month PCU had been in operation, holding all other covariates at original values) are presented. To examine the change in daily direct costs in the days before and after patient transfer to PCU, we mapped the aggregated mean daily costs for all patients who were transferred to PCU over the two-year course, with Day 0 set as day of transfer. To determine cost savings of PCU vs. PCCS, we compared the patients who died on PCU in its first two years of operation (n ¼ 712) with patients who died while under PCCS care in the 18 months before PCU opened (n ¼ 456). We chose this methodology because we could not match patients transferred to the unit (after it opened) to those who were not transferred in the same time period, as the opening of the
Vol. 50 No. 2 August 2015
Cost Savings of a Palliative Care Unit
unit eliminates an appropriate comparative group during the concurrent time period. Patients on PCCS who are expected to die are almost exclusively transferred to PCU, thus the PCCS patient population became fundamentally different from the PCU patient population after PCU opened. Our method of matching PCU patients to historical PCCS controls creates the most appropriate reference group for our cost comparison. We included only patients who died in the hospital and excluded patients who were discharged to best control for severity of illness between the two groups.5 Palliative care has been previously proven to save costs for patients discharged alive;5 nevertheless, we felt we could not perform an appropriate cost analysis between PCU and PCCS survivor groups. We know from our experiences that at Mount Sinai Hospital, PCCS patients who do not die have a wider range of severity of illness than do PCU patient who do not die. Thus, PCU and PCCS survivors are much more dissimilar than PCU and PCCS decedents in terms of both observed and unobserved factors. For these reasons, we felt that comparison of survivors would not be appropriate regardless of how we matched or tried to control with propensity scores. During the 18 months before PCU opened, 456 PCCS patients died in the hospital. To adjust for confounding because of observable factors associated both with the use of PCU care (vs. PCCS care) and hospital costs, we created a propensity score to reflect probability of PCU use. The propensity score was calculated via logistic regression of PCU use on age at hospitalization, gender, race/ethnicity (white, black, Hispanic, or other), primary diagnosis, and interaction terms for age and diagnoses (age x hepatic diagnosis and age x cardiovascular diagnosis). Balance of the covariates within blocks of the propensity score and across treatment (PCU) and comparison (PCCS) groups was assessed with standardized differences. We then created a kernel weight based on the propensity score to match PCCS and PCU patients. With kernel weights, each PCU patient received a weight of one. For each PCU patient, a match was created by taking a weighted average of information from PCCS patients whose propensity score was within 0.06 of the PCU patient’s propensity score.23 PCCS patients with propensity scores closer to that of PCU patient are given higher weights. This type of weighting strategy enabled us to retain all observations in our dataset while reducing bias resulting from poor matches between the PCU and PCCS groups. After weighting our sample by the propensity score, there was less than 10% standardized bias between PCU and PCCS patients on all covariates included in the propensity score.
149
Within the propensity score-weighted sample, we compared the difference in costs between these two groups for the hospital stay before and after initiation of palliative care. We compared mean daily costs before and after transfer to PCU/initiation of care by PCCS with adjusted Wald tests. To evaluate whether PCU is associated with lower costs than PCCS in the days following palliative care initiation, we used a GLM model with gamma family and log link function with and without additional adjustment for covariates included in the propensity score. Incremental effects (change in mean daily cost when patient changes from PCCS to PCU, holding all other covariates at original values) for PCU patients are presented.
Results Description of PCU Patients The demographic characteristics of the patients who were admitted to PCU are shown in Table 1. From June 2011 until June 2013, the Mount Sinai PCU admitted 1107 patients. Of those, 57% were female, and the average age was 70.7 years, with a range of 22e107 years. The median hospital length of stay was 12 days, with a median of four of those days spent on PCU (range 0e39 days). Sixty-six percent of PCU patients died there, and 31% were discharged. The primary diagnosis for which 46% of PCU patients were admitted to PCU was cancer, although the percentage of PCU patients whose original hospital admission diagnosis was cancer was much higher. We examined temporal changes in a number of demographic and clinical characteristics of patients admitted to PCU in the first two years of its existence including percent of patients with a Do Not Resuscitate order, percent of patients with primary admitting diagnosis of cancer, average length of stay on PCU, average length of stay in the hospital before transfer to PCU, percent of patients who were male, percent of patients who died while on PCU, average age of patient, and average cost per patient per day. With the exception of cost, discussed in the following sections, none of these patient-level characteristics changed over the PCU’s first 24.5 months of operation in a clinically or statistically significant manner.
Costs for All PCU Patients Mean direct cost per patient-day for all PCU patients (regardless of discharge status) over time is shown in Figure 1. From the beginning of operations through the end of month 6.5 (mid-June 2011, through December 31, 2011), mean direct cost per patient-day was $1397. For the next 12 months (January 1, 2012, to December 31, 2012), it was
Nathaniel et al.
Table 1 Characteristics of Patients Admitted to the Palliative Care Unit PCU Patient Characteristics Patient admissions Race White Black Hispanic Other/unknown Age (yrs) Gender Male Female Average hospital LOSa (d) Average LOS pre-PC (d) Average LOS on PC (d) PCU disposition Died on PCU Discharged from PCU Transferred to other floor Location in hospital before transfer to PCU Medical/surgical floor ICU Emergency Department Home Other Had Do Not Resuscitate Order while on PCU Diagnosisb Cancer Hepatobiliary Lung Gastrointestinal Hematological Gynecological Breast Genitourinary Other/unknown Non-cancer Neurological Cardiovascular Infectious disease Hepatic Pulmonary Renal Gastrointestinal Other
Mean (SD) or N (%) 1107 400 228 242 237 70.7
(36) (21) (22) (21) (16.0)
481 626 12.0 7.0 4.0
(43) (57) (7e21) (2e14) (2e7)
735 (66) 348 (31) 24 (2) 652 272 116 64 3 1027
(59) (25) (10) (6) (0) (93)
506 92 80 73 67 51 35 35 73 601 140 99 94 83 76 36 34 39
(46) (8) (7) (7) (6) (5) (3) (3) (7) (54) (13) (9) (8) (7) (7) (3) (3) (4)
LOS ¼ length of stay; PC ¼ palliative care; PCU ¼ palliative care unit; ICU ¼ intensive care unit. This table reports the demographic characteristics of the entire population of patients who were admitted to the PCU in its first 24.5 months of existence. a LOS reported as median and interquartile range. b Diagnosis refers to the diagnosis for which palliative care was consulted, not necessarily the primary admitting diagnosis.
$901. Then, for the next six months, mean direct cost per patient-day was $1038. In our unadjusted GLM model, mean daily costs decreased significantly over the entire time period (mean incremental effect/ month ¼ $16, SE ¼ $3, P < 0.001). This relationship persisted after adjusting for patient characteristics (mean incremental effect/month ¼ $18, SE ¼ $4, P < 0.001). We next calculated average direct cost for PCU patients in the days before, the day of, and the days after their transfer to PCU (Fig. 2). Figure 2 includes data
Mean Direct Cost per PaƟent per Day
150
Vol. 50 No. 2 August 2015
$1,600 $1,400 $1,200 $1,000 $800 $600 $400 $200 $0.0
5.0
10.0 15.0 Months Since PCU Opened
20.0
25.0
Fig. 1. Average daily cost per patient over the PCU’s first 24.5 months in operation. Mean direct costs per patient were calculated for each discrete day of the PCU’s first 24.5 months of operation. These values were averaged for half-month increments, yielding 49 data points, which are mapped over time (June 20, 2011 to June 30, 2013). All costs have been adjusted to 2013 dollars. PCU ¼ palliative care unit.
for all PCU patients, both those who died on PCU and those who were discharged alive from the hospital. Average direct cost per day began decreasing three days before transfer to PCU and continued decreasing until the patient’s first full day on PCU. Costs then leveled at a steady daily rate. The mean of patients’ average cost per day was $1522 in days before transfer to PCU and $835 in the days after transfer to PCU. Without accounting for confounding factors, such as the frontloading of hospital costs, this cost difference represents a savings of $687 in daily costs per patient.
Comparison of Costs of PCU to PCCS We also compared the cost savings of PCU with those of the PCCS by propensity score matching the 712 patients who died on PCU with 456 PCCS patients who died on PCCS in the 18 months before PCU opened. PCU deaths represented 32% of hospital deaths during its first two years of operations. PCCS deaths over the 18 months we analyzed (before the PCU opening) represented about 32% of inpatient hospital deaths. The demographic characteristics of the patients who died on PCU, and those who died on PCCS services, are shown in Table 2, along with standardized differences before and after weighting. Using the propensity score weighted samples, we then compared average patient costs for each group in the days before and after transfer to PCU or PCCS (Fig. 3). In the propensity score weighted sample, PCU patients had significantly lower mean daily costs than PCCS patients in the days following initiation of palliative care ($753, SE ¼ $20; $1033, SE ¼ $43, respectively, P < 0.001). The average incremental effect of PCU in the unadjusted GLM was e$279 (bootstrapped SE ¼ $48, P < 0.001). This relationship remained statistically significant after including covariates from the propensity score in the regression model (average incremental effect ¼ e$240, bootstrapped SE ¼ $45,
Vol. 50 No. 2 August 2015
Cost Savings of a Palliative Care Unit
151
Mean Direct Cost per Day
$2,000 $1,500 $1,000 $500
$0 -20-19-18-17-16-15-14-13-12-11-10-9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 1011121314151617181920 Day, in RelaƟon to Transfer to PCU (Day 0)
Fig. 2. Patients’ average direct cost per day in relation to PCU transfer. Direct costs for each PCU patient from June 2011 through June 2013 for whom cost data were available (n ¼ 1071) were calculated for each discrete day of his or her hospitalization. Average direct cost for patients on the day of transfer to PCU is designated as Day 0. Average direct costs for patients on each discrete day before and after transfer are also graphed, with Day 1 as the day before transfer, Day 1 the first full day on PCU, etc. Number of cases per day is 690 on Day 5, 798 on Day 3, 1071 on Day 0, 622 on Day 3, and 421 on Day 5. All costs have been adjusted to 2013 dollars. PCU ¼ palliative care unit.
P < 0.001). Meanwhile, the mean daily costs for PCU and PCCS patients from hospital admission to day before palliative care initiation did not differ significantly ($1667, SE ¼ 34, $1778, SE ¼ 98, respectively,
P ¼ 0.29). These PCU cost savings, however, did not manifest themselves in the first six months of PCU operations but only became apparent after this start-up period.
Table 2 Characteristics of Patients Who Died on PCU and Under PCCS Care Standardized
Patient Characteristics
a
Difference Before
Difference After
Died on PCU
Died on PCCS
Propensity Score
Propensity Score
Mean (SD) or N (%)
Mean (SD) or N (%)
Weighting
Weighting
1.3 0.7 15.4 12.9
0.2 1.7 2.1 0.4
8.1
0.6
19.8 3.3 23.6 6.9 5.3 17.5 8.5 35.3
5.1 1 0.3 2.9 1.5 11.5 2.5 32.3
9.2 0.4 6.2 24 5.3
13.9 6.7 2.8 24.2 5.3
Patient admissions 712 Race White 264 (37) Black 140 (20) Hispanic 151 (21) Other/unknown 157 (22) Age (yrs) 70.2 (16.0) Gender Male 332 (47) Female 380 (53) Diagnosisb Cancer 342 (48.0) Neurological 97 (14) Cardiovascular 69 (10) Infectious disease 65 (9) Pulmonary 55 (8) c 12.0 (7e22) Median hospital LOS (d) Median LOS pre-PC (d) 8.0 (3e16) Median LOS on PC (d) 3.0 (1e6) Location in hospital before transfer to PCU/induction of PCCS Medical/surgical floor 402 (56) ICU 234 (33) Emergency Department 55 (8) Home 20 (3) Other 1 (0)
456 193 92 98 73 68.0
(42) (20) (22) (16) (15.8)
231 (51) 225 (49) 173 57 81 33 29 13.0 6.0 4.0
(38) (13) (18) (7) (6) (6e26) (2e17) (1e8)
278 149 28 0 0
(61) (33) (6) (0) (0)
PCU ¼ palliative care unit; PCCS ¼ palliative care consultation service; LOS ¼ length of stay; PC ¼ palliative care; ICU ¼ intensive care unit. The ‘‘Died on PCU’’ and ‘‘Died on PCCS’’ columns report characteristics for patients who died on the PCU during its first 24.5 months of existence and patients who died under PCCS care in the 18 months before the PCU opened, respectively. The first ‘‘Standardized Difference’’ column reports the standardized differences for covariates of the two groups before propensity score matching and weighting. The right-most ‘‘Standardized Difference’’ column reports standardized differences for covariates after propensity score matching and weighting. ‘‘Length of Stay’’ and ‘‘Location before Transfer’’ data were not included in the propensity score. a Excludes 23 patients for whom cost data were not available. b Diagnosis refers to the diagnosis for which palliative care was consulted, not necessarily the primary hospital admitting diagnosis. c Length of stay raw figures reported as median and interquartile range; standardized differences shown are for mean.
152
Nathaniel et al.
Vol. 50 No. 2 August 2015
Mean Direct Cost per Day
$3,000 $2,500 $2,000 $1,500
PCU paƟents
$1,000
PCCS paƟents
$500 $0
-15
-10
-5
0
5
10
15
Day, in RelaƟon to Transfer to PCU/ PCCS
Fig. 3. Comparison of average direct daily costs of patients on the PCU as compared with propensity score matched PCCS patients. The two lines represent the patients who died on the PCU during its first 24.5 months of operation and propensity score weighted historical control PCCS patients who died in the main hospital while under the care of PCCS care in the 18 months before the PCU opened. Direct costs for each patient for whom cost data were available were calculated for each discrete day of hospitalization. Day 0 is designated as the day of transfer to PCU for PCU patient group and day of induction of PCCS for PCCS patient group. Average direct costs for patients on each discrete day before and after transfer are also graphed, with Day 1 as the day before transfer, Day 1 the first full day on PCU or PCCS, etc. All costs have been adjusted to 2013 dollars. PCU ¼ palliative care unit; PCCS ¼ palliative care consultation service.
Discussion In this study, we characterized direct costs of a PCU in its first two years of operation and sought to determine whether PC modality (unit or consult service) was associated with differences in mean daily direct hospitalization costs. We found that mean cost per patient-day held steady at $1397 per patient for the first seven months, then dropped to $901, where they remained for 12 months, then rose to $1038 for the duration of the period under study. Daily cost savings began before the patients were transferred to PCU and then continued after transfer. Among patients who died in the hospital, average daily direct costs per patient in the days after transfer to PCU were $240 lower as compared with patients being followed by PCCS on the general hospital wards. To our knowledge, this is the first report to definitively demonstrate that the cost-savings of patients on a PCU is even greater than that of patients on a PCCS.
PCU Costs Over Time There were three distinct ranges of cost that emerged when we analyzed data from these two years. The two cost inflection points coincide with the ends of the 2011 and 2012 fiscal years. These cost changes can be explained in part by changes in hospital accounting. Most significantly, daily room costs for PCU were different in 2011, 2012, and the first six months of 2013, $1305, $793, and $990, respectively. The major factors determining room cost for a given floor are staff costs and occupancy rates. These have direct and indirect causative relationships, respectively, to room cost. These costs fluctuated widely for the first years of our unit as the hospital adjusted to the patient population, their needs related
to staffing, and occupancy. Room costs for the more well-established surgical and medical floors fluctuated minimally over these same time periods. While these exact costs are unique to our system, it may be helpful for administrators considering opening new PCUs to be aware of these variations in daily costs in the first years when a unit opens. Our analysis of changes in average daily cost per patient before and after transfer to PCU shows that PCU is associated with cost savings that begin before the patients are transferred and continue once they are on the unit. These findings corroborate and extend those by Smith et al., Albanese et al., and Eti et al., who had similar findings,24e26 although we incorporate data from a far larger sample of patients than have previous authors. These findings continue to confirm the overall cost savings related to palliative care.27
PCU vs. PCCS Costs Along with receiving higher quality of care,13,21 PCU patients who die in the hospital also incur lower endof-life costs than similar patients on other units who are cared for by a PCCS. These results may be related to the fact that standardizing the delivery of palliative care has been shown to lower costs.28 At Mount Sinai Hospital, patients in the PCU are more likely than patients under PCCS care to be seen exclusively by physicians and nurses who specialize in palliative care. This results in continued clarification of goals and allows treatments to continue to be refined so any therapies not concordant with the patient’s and family’s goals can be eliminated. Because on the PCU the palliative care team is most often the team of record, the team has the ability to make decisions that assure patient symptom
Vol. 50 No. 2 August 2015
Cost Savings of a Palliative Care Unit
management while also minimizing costs. Specific initiatives that help to encourage cost-saving include avoiding use of more expensive synthetic opioids like hydromorphone in patients with no contraindications to morphine and avoiding use of intravenous acetaminophen when other routes of delivery are possible. These cost savings may have large implications for a hospital system. PCU patients who died in the hospital cost an average of at least $240 per day less than similar PCCS patients in the days after PC initiation and average daily census for PCU in its first two years was 9.1 patients. This may represent a savings of over $1.6 million in direct patient costs in the first two years of our PCU. By demonstrating that PCU decedent daily costs are lower than costs for similar patients receiving PCCS care, our study adds significantly to the understanding of the cost-savings associated with PCU units. In addition, our results align with and build on results published by Albanese et al., which show that PCU patient costs are lower than costs for similar patients on other floors.25 They also support results published by Eti et al., which demonstrate that hospital charges were lower for patients managed directly by a palliative medicine attending, compared with patients who had a palliative medicine physician as a consultant.26
153
symptom severity and end-of-life care preferences. Still, it is possible that there remain unmeasured confounders between the two groups. Also, limiting our sample to only decedents is itself a limitation of this study, as palliative care has been shown to save costs in patients discharged alive as well.5 Our overall hospital length-of-stay means are significantly higher for palliative care patients than those published in many previous articles.25,26 We did not include length of stay in our propensity score calculations, as inclusion of a variable affected by treatment can obscure the treatment effect of interest.30
Conclusion Previous studies have demonstrated that both PCCS teams and PCUs are associated with cost savings for hospitals.5,24 We confirm that transfer to PCU is associated with decreasing direct costs. This is also the first study to demonstrate via a controlled comparison that PCUs provide more cost savings than PCCS for patient who die in the hospital. This has implications for hospital administrators who are weighing the potential benefits of a PCU against its capital and operational costs. Hospitals can use the data presented in this study when analyzing the cost/benefit ratios of opening an inpatient PCU.
Limitations There are a few limitations that must be considered. Mount Sinai has had a Palliative Care program for 15 years, so the Mount Sinai PCU was well-positioned to be an effective unit from the start. Also, end-of-life hospital costs in New York City are historically higher than national averages.29 Therefore, the magnitude of these cost savings may not be as easily achieved at other institutions. Additionally, our cost analyses include data from the very first day that PCU became operational. The higher PCU costs seen in early months of PCU operations are likely not representative of actual costs of a fully hospital-integrated PCU. Therefore, we considered excluding data from those first six months. However, we chose to include early data so as to best paint the full picture of fluctuations in cost from the opening of a unit so as to be instructive to other groups who may be planning to open their own unit. Although we adjusted for sociodemographic and clinical characteristics that might be associated both with PCU use and hospital costs, we were unable to observe symptom severity and end-of-life care preferences. Unobserved variables are beyond the capabilities of propensity score analyses. We attempted to control for these differences in our analysis of PCU vs. PCCS costs by limiting our sample to only deceased patients, who were more likely to have similar
Disclosures and Acknowledgments Mr. Nathaniel is supported by the Medical Student Training in Aging Research (MSTAR) Program through the American Geriatrics Society. Dr. Garrido is supported by CDA 11-201/CDP 12-255 from the Department of Veterans Affairs, Veterans Health Administration, Office of Research and Development, Health Services Research and Development Service. The authors have no disclosures. 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 States government. The funding sources had no role in the in study design; in the collection, analysis, and interpretation of data; in the writing of the report; and in the decision to submit the article for publication. The authors acknowledge PCU and PCCS staff, as well as our patients and their families.
References 1. Christakis NA, Iwashyna TJ. The health impact of health care on families: a matched cohort study of hospice use by decedents and mortality outcomes in surviving, widowed spouses. Soc Sci Med 2003;57:465e475.
154
Nathaniel et al.
2. Teno JM, Clarridge BR, Casey V, et al. Family perspectives on end-of-life care at the last place of care. JAMA 2004;291:88e93. 3. Jordhoy MS, Fayers P, Loge JH, Ahlner-Elmqvist M, Kaasa S. Quality of life in palliative cancer care: results from a cluster randomized trial. J Clin Oncol 2001;19: 3884e3894. 4. Ringdal GI, Jordhoy MS, Kaasa S. Family satisfaction with end-of-life care for cancer patients in a cluster randomized trial. J Pain Symptom Manage 2002;24:53e63. 5. Morrison RS, Penrod JD, Cassel JB, et al. Cost savings associated with US hospital palliative care consultation programs. Arch Intern Med 2008;168:1783e1790. 6. Morrison RS, Augustin R, Souvanna P, Meier DE. America’s care of serious illness: a state-by-state report card on access to palliative care in our nation’s hospitals. J Palliat Med 2011;14:1094e1096. 7. Rigby A, Krzyzanowska M, Le LW, et al. Impact of opening an acute palliative care unit on administrative outcomes for a general oncology ward. Cancer 2008;113:3267e3274. 8. Kellar N, Martinez J, Finis N, Bolger A, von Gunten CF. Characterization of an acute inpatient hospice palliative care unit in a U.S. teaching hospital. J Nurs Adm 1996;26:16e20. 9. Hui D, Elsayem A, Palla S, et al. Discharge outcomes and survival of patients with advanced cancer admitted to an acute palliative care unit at a comprehensive cancer center. J Pain Symptom Manage 2010;13:49e57. 10. Shin SH, Hui D, Chisholm GB, et al. Characteristics and outcomes of patients admitted to the acute palliative care unit from the emergency center. J Pain Symptom Manage 2014;47:1028e1034. 11. Digwood G, Lustbader D, Pekmezaris R, et al. The impact of a palliative care unit on mortality rate and length of stay for medical intensive care unit patients. Palliat Support Care 2011;9:387e392. 12. Lagman R, Rivera N, Walsh D, LeGrand S, Davis MP. Acute inpatient palliative medicine in a cancer center: clinical problems and medical interventionsda prospective study. Am J Hosp Palliat Care 2007;24:20e28. 13. Gaertner J, Frechen S, Sladek M, Ostgathe C, Voltz R. Palliative care consultation service and palliative care unit: why do we need both? Oncologist 2012;17:428e435. 14. Gaertner J, Drabik A, Marschall U, Schlesiger G, Voltz R, Stock S. Inpatient palliative care: a nationwide analysis. Health Policy 2013;109:311e318. 15. Casarett D, Johnson M, Smith D, Richardson D. The optimal delivery of palliative care: a national comparison of the outcomes of consultation teams vs inpatient units. Arch Intern Med 2011;171:649e655. 16. Back AL, Li Y-F, Sales AE. Impact of palliative care case management on resource use by patients dying of cancer
Vol. 50 No. 2 August 2015
at a veterans affairs medical center. J Palliat Med 2005;8: 26e35. 17. Cowan JD. Hospital charges for a community inpatient palliative care program. Am J Hosp Palliat Care 2004;21: 177e190. 18. Campbell M, Frank R. Experience with an end-of-life practice at a university hospital. Crit Care Med 1997;25: 197e202. 19. Ciemins EL, Blum L, Nunley M, Lasher A, Newman JM. The economic and clinical impact of an inpatient palliative care consultation service: a multifaceted approach. J Palliat Med 2007;10:1347e1355. 20. Higginson IJ, Finlay IG, Goodwin DM, et al. Is there evidence that palliative care teams alter end-of-life experiences of patients and their caregivers? J Pain Symptom Manage 2003;25:150e168. 21. Roza K. Comparison of quality of care of palliative care unit, palliative care consult team and usual care patients. New York: Icahn School of Medicine at Mount Sinai, 2014. 22. Bureau of Labor Statistics. 2013 Consumer Price Index detailed report tables. Washington, DC: Bureau of Labor Statistics, 2013. 23. Garrido MM, Kelly AS, Paris J, et al. Methods for constructing and assessing propensity scores. Health Serv Res 2014;49:1701e1720. 24. Smith TJ, Coyne P, Cassel B, Penberthy L, Hopson A, Hager MA. A high-volume specialist palliative care unit and team may reduce in-hospital end-of-life care costs. J Palliat Med 2003;5:699e705. 25. Albanese TH, Radwany SM, Mason H, Gayomali C, Dieter K. Assessing the financial impact of an inpatient acute palliative care unit in a tertiary care teaching hospital. J Palliat Med 2013;16:289e294. 26. Eti S, O’Mahony S, McHugh M, Guilbe R, Blank A, Selwyn P. Outcomes of the acute palliative care unit in an academic medical center. Am J Hosp Palliat Med 2014;31: 380e384. 27. Elsayem A, Swint K, Fisch MJ, et al. Palliative care inpatient service in a comprehensive cancer center: clinical and financial outcomes. J Clin Oncol 2004;22:2008e2014. 28. Smith TJ, Hillner BE. Ensuring quality cancer care by the use of clinical practice guidelines and critical pathways. J Clin Oncol 2001;19:2886e2897. 29. The Dartmouth Atlas of Healthcare. End-of-life care: data by region.2012; Hanover, NH: The Dartmouth Institute for Health Policy & Clinical Practice. Available from: http:// www.dartmouthatlas.org/data/topic/topic.aspx?cat=18. Last accessed April 25, 2015. 30. Garrido MM. Propensity scores: a practical method for assessing treatment effects in pain and symptom management research. J Pain Symptom Manage 2014;48:711e718.
Journal of Pain and Symptom Management 147
Original Article
Cost Savings Associated With an Inpatient Palliative Care Unit: Results From the First Two Years Jeremy D. Nathaniel, BA, Melissa M. Garrido, PhD, Emily J. Chai, MD, Gabrielle Goldberg, MD, and Nathan E. Goldstein, MD Icahn School of Medicine at Mount Sinai (J.D.N.), New York, New York; Geriatrics Research, Education and Clinical Center (M.M.G, N.E.G.), James J. Peters Veterans Affairs Medical Center, Bronx, New York; Lilian and Benjamin Hertzberg Palliative Care Institute of the Brookdale Department of Geriatrics and Palliative Medicine (M.M.G., E.J.C., N.E.G.), Icahn School of Medicine at Mount Sinai, New York, New York; and Hofstra North Shore-LIJ School of Medicine (G.G.), Hempstead, New York, USA
Abstract Context. Palliative care consultation services (PCCS) decrease costs for patients by matching treatments received to patients’ and families’ goals of care. However, few studies have examined the costs of a specialized palliative care unit (PCU). Objectives. To quantitatively describe Mount Sinai Hospital’s PCU’s first two years of operation; to examine how patientrelated costs changed in the days before and after transfer to PCU; and to compare cost savings of PCU to those of PCCS. Methods. Cost and administrative data from PCU patients from the first 24.5 months of our PCU’s operation were analyzed. To compare costs between PCU and PCCS patients, we matched PCU patients to similar PCCS patients and used propensity scores to adjust for differences across groups. Results. The PCU admitted 1107 patients in its first 24.5 months. Over this time frame, there was a statistically significant (P < 0.001) decrease in average daily direct costs per patient. The mean of patients’ average cost per day was $687 less while on the PCU than before transfer to PCU. Among patients who died in the hospital, average daily direct cost per patient in the days after transfer to PCU was $240 lower as compared with patients being followed by PCCS on the general hospital wards (SE ¼ $45, P < 0.001). Conclusion. Among patients who died in the hospital, transfer to a PCU is associated with significant cost savings as compared with patients on hospital wards who are seen by a PCCS. J Pain Symptom Manage 2015;50:147e154. Published by Elsevier Inc. on behalf of American Academy of Hospice and Palliative Medicine. Key Words Palliative medicine, palliative care unit, cost savings, cost, longitudinal study
Introduction Palliative care consultation service (PCCS) teams have been shown to improve symptom control, satisfaction, and quality of life for patients with advanced illness and their families.1e4 In recent years, the benefits of PCCS also have extended to reducing hospitalization costs for patients with advanced illness.5 Owing to its proven benefits to both hospital and patient, availability of palliative care in the U.S. hospitals
Address correspondence to: Nathan E. Goldstein, MD, Department of Geriatrics and Palliative Medicine, Mount Sinai Medical Center, One Gustave L. Levy Place, Box 1070, New York, NY 10029, USA. E-mail: [email protected] Published by Elsevier Inc. on behalf of American Academy of Hospice and Palliative Medicine.
more than doubled from 2001 to 2011.6 As of 2011, 63% of hospitals with 50 or more beds had palliative care teams and that number is growing.6 Along with the increase in palliative care consult services, there has been an increasing number of hospitals investing in palliative care units (PCUs), which are specialized hospital floors devoted specifically to providing palliative care.7,8 Numerous studies of PCU patient outcomes have affirmed the importance of PCU as a palliative care modality.9e14 These units
Accepted for publication: February 19, 2015.
0885-3924/$ - see front matter http://dx.doi.org/10.1016/j.jpainsymman.2015.02.023
148
Nathaniel et al.
are founded on the belief that a specialty-level trained palliative medicine team working together in the same location can best meet the needs of palliative care patients with advanced illness. These units require considerable start-up costs,15 including specialty training for nurses and other staff who work on these units, and there is often a large upfront capital expense to renovate a traditional hospital floor to become a more home-like environment for patients and their families. Although many studies have demonstrated the cost-savings of PCCS teams,16e20 few have examined whether PCUs lead to the same lowered health care costs as PCCS teams, and to our knowledge, none have explicitly examined the cost savings of a PCU compared with a PCCS team. PCUs have been shown to provide higher quality of care than PCCS teams, as measured by patients’ family members’ satisfaction and PC team satisfaction.13,15,21 However, further information is needed about the relationship of these units to hospital costs. In this article, we describe the demographics and patterns of care for the first two years of our unit’s existence, analyze how patient-related costs changed over time, and determine the cost differences between patients cared for on our PCU and patients cared for by our PCCS team.
Methods The Palliative Care Unit The Mount Sinai PCU is a physically distinct unit, comprising part of one floor of the hospital. Mount Sinai’s PCCS has been in existence since July 1997, but our 13-bed PCU did not open until June 2011. The majority of patients are referred to PCU from other units throughout the hospital, although some admissions come directly from the Emergency Department or home. Criteria for admission to PCU include patients with: difficult to manage symptoms who need rapid titration of medications; progressive illness who need special assistance in establishing goals of care; a solely comfort-oriented plan of care and who are awaiting discharge to an appropriate care setting; and advanced illness who are in an intensive care unit but who will not be receiving further escalation of medical treatments.
Data Collection The Palliative Care Department employs a data manager who records and tracks data on all patients who are seen by PCCS or who are admitted to PCU. To ensure accuracy, the data are checked daily with the census on the PCU and any discrepancies between the two are resolved within one day. These
Vol. 50 No. 2 August 2015
characteristics include demographics, diagnosis, dates of admission and discharge to PCU, and other relevant clinical data. We merged our internal database tracking PCCS and PCU patients with cost data captured in the Transitions Systems Inc. system at Mount Sinai. Using the Transitions Systems Inc. system we obtained a report of daily direct costs incurred by every PCU and selected PCCS patients. This work was approved by the Program for the Protection of Human Subjects at Mount Sinai.
Data Analysis We first compiled demographic and clinical characteristics for the 1107 patients admitted to PCU from June 2011 to June 2013. To examine changes in demographic characteristics over this time period, we graphed several of the PCU’s daily patient demographics and clinical characteristics over time. There were 36 patients for whom cost data were unavailable, so they were excluded from all cost analyses. All cost data were adjusted by the medical Consumer Price Index to 2013 dollars.22 To examine changes in average patient costs over PCU’s first 24.5 months of operation, we calculated the mean cost per PCU patient for each day and then grouped these daily means into half-month increments. This created 49 data points (two points per month x 12 months per year x 2 years and half of a month) that could be used to examine changes in costs over time. To examine the relationship between a given period of operation and PCU’s mean daily costs, we ran a generalized linear model (GLM) with gamma distribution and log link. We ran both an unadjusted model, as well as a model that adjusted for other covariates that might influence mean costs (percent male, mean age, percent with Do Not Resuscitate orders, percent with cancer diagnoses, percent white patients, and average daily census). Marginal effects (change in mean daily cost for each additional month PCU had been in operation, holding all other covariates at original values) are presented. To examine the change in daily direct costs in the days before and after patient transfer to PCU, we mapped the aggregated mean daily costs for all patients who were transferred to PCU over the two-year course, with Day 0 set as day of transfer. To determine cost savings of PCU vs. PCCS, we compared the patients who died on PCU in its first two years of operation (n ¼ 712) with patients who died while under PCCS care in the 18 months before PCU opened (n ¼ 456). We chose this methodology because we could not match patients transferred to the unit (after it opened) to those who were not transferred in the same time period, as the opening of the
Vol. 50 No. 2 August 2015
Cost Savings of a Palliative Care Unit
unit eliminates an appropriate comparative group during the concurrent time period. Patients on PCCS who are expected to die are almost exclusively transferred to PCU, thus the PCCS patient population became fundamentally different from the PCU patient population after PCU opened. Our method of matching PCU patients to historical PCCS controls creates the most appropriate reference group for our cost comparison. We included only patients who died in the hospital and excluded patients who were discharged to best control for severity of illness between the two groups.5 Palliative care has been previously proven to save costs for patients discharged alive;5 nevertheless, we felt we could not perform an appropriate cost analysis between PCU and PCCS survivor groups. We know from our experiences that at Mount Sinai Hospital, PCCS patients who do not die have a wider range of severity of illness than do PCU patient who do not die. Thus, PCU and PCCS survivors are much more dissimilar than PCU and PCCS decedents in terms of both observed and unobserved factors. For these reasons, we felt that comparison of survivors would not be appropriate regardless of how we matched or tried to control with propensity scores. During the 18 months before PCU opened, 456 PCCS patients died in the hospital. To adjust for confounding because of observable factors associated both with the use of PCU care (vs. PCCS care) and hospital costs, we created a propensity score to reflect probability of PCU use. The propensity score was calculated via logistic regression of PCU use on age at hospitalization, gender, race/ethnicity (white, black, Hispanic, or other), primary diagnosis, and interaction terms for age and diagnoses (age x hepatic diagnosis and age x cardiovascular diagnosis). Balance of the covariates within blocks of the propensity score and across treatment (PCU) and comparison (PCCS) groups was assessed with standardized differences. We then created a kernel weight based on the propensity score to match PCCS and PCU patients. With kernel weights, each PCU patient received a weight of one. For each PCU patient, a match was created by taking a weighted average of information from PCCS patients whose propensity score was within 0.06 of the PCU patient’s propensity score.23 PCCS patients with propensity scores closer to that of PCU patient are given higher weights. This type of weighting strategy enabled us to retain all observations in our dataset while reducing bias resulting from poor matches between the PCU and PCCS groups. After weighting our sample by the propensity score, there was less than 10% standardized bias between PCU and PCCS patients on all covariates included in the propensity score.
149
Within the propensity score-weighted sample, we compared the difference in costs between these two groups for the hospital stay before and after initiation of palliative care. We compared mean daily costs before and after transfer to PCU/initiation of care by PCCS with adjusted Wald tests. To evaluate whether PCU is associated with lower costs than PCCS in the days following palliative care initiation, we used a GLM model with gamma family and log link function with and without additional adjustment for covariates included in the propensity score. Incremental effects (change in mean daily cost when patient changes from PCCS to PCU, holding all other covariates at original values) for PCU patients are presented.
Results Description of PCU Patients The demographic characteristics of the patients who were admitted to PCU are shown in Table 1. From June 2011 until June 2013, the Mount Sinai PCU admitted 1107 patients. Of those, 57% were female, and the average age was 70.7 years, with a range of 22e107 years. The median hospital length of stay was 12 days, with a median of four of those days spent on PCU (range 0e39 days). Sixty-six percent of PCU patients died there, and 31% were discharged. The primary diagnosis for which 46% of PCU patients were admitted to PCU was cancer, although the percentage of PCU patients whose original hospital admission diagnosis was cancer was much higher. We examined temporal changes in a number of demographic and clinical characteristics of patients admitted to PCU in the first two years of its existence including percent of patients with a Do Not Resuscitate order, percent of patients with primary admitting diagnosis of cancer, average length of stay on PCU, average length of stay in the hospital before transfer to PCU, percent of patients who were male, percent of patients who died while on PCU, average age of patient, and average cost per patient per day. With the exception of cost, discussed in the following sections, none of these patient-level characteristics changed over the PCU’s first 24.5 months of operation in a clinically or statistically significant manner.
Costs for All PCU Patients Mean direct cost per patient-day for all PCU patients (regardless of discharge status) over time is shown in Figure 1. From the beginning of operations through the end of month 6.5 (mid-June 2011, through December 31, 2011), mean direct cost per patient-day was $1397. For the next 12 months (January 1, 2012, to December 31, 2012), it was
Nathaniel et al.
Table 1 Characteristics of Patients Admitted to the Palliative Care Unit PCU Patient Characteristics Patient admissions Race White Black Hispanic Other/unknown Age (yrs) Gender Male Female Average hospital LOSa (d) Average LOS pre-PC (d) Average LOS on PC (d) PCU disposition Died on PCU Discharged from PCU Transferred to other floor Location in hospital before transfer to PCU Medical/surgical floor ICU Emergency Department Home Other Had Do Not Resuscitate Order while on PCU Diagnosisb Cancer Hepatobiliary Lung Gastrointestinal Hematological Gynecological Breast Genitourinary Other/unknown Non-cancer Neurological Cardiovascular Infectious disease Hepatic Pulmonary Renal Gastrointestinal Other
Mean (SD) or N (%) 1107 400 228 242 237 70.7
(36) (21) (22) (21) (16.0)
481 626 12.0 7.0 4.0
(43) (57) (7e21) (2e14) (2e7)
735 (66) 348 (31) 24 (2) 652 272 116 64 3 1027
(59) (25) (10) (6) (0) (93)
506 92 80 73 67 51 35 35 73 601 140 99 94 83 76 36 34 39
(46) (8) (7) (7) (6) (5) (3) (3) (7) (54) (13) (9) (8) (7) (7) (3) (3) (4)
LOS ¼ length of stay; PC ¼ palliative care; PCU ¼ palliative care unit; ICU ¼ intensive care unit. This table reports the demographic characteristics of the entire population of patients who were admitted to the PCU in its first 24.5 months of existence. a LOS reported as median and interquartile range. b Diagnosis refers to the diagnosis for which palliative care was consulted, not necessarily the primary admitting diagnosis.
$901. Then, for the next six months, mean direct cost per patient-day was $1038. In our unadjusted GLM model, mean daily costs decreased significantly over the entire time period (mean incremental effect/ month ¼ $16, SE ¼ $3, P < 0.001). This relationship persisted after adjusting for patient characteristics (mean incremental effect/month ¼ $18, SE ¼ $4, P < 0.001). We next calculated average direct cost for PCU patients in the days before, the day of, and the days after their transfer to PCU (Fig. 2). Figure 2 includes data
Mean Direct Cost per PaƟent per Day
150
Vol. 50 No. 2 August 2015
$1,600 $1,400 $1,200 $1,000 $800 $600 $400 $200 $0.0
5.0
10.0 15.0 Months Since PCU Opened
20.0
25.0
Fig. 1. Average daily cost per patient over the PCU’s first 24.5 months in operation. Mean direct costs per patient were calculated for each discrete day of the PCU’s first 24.5 months of operation. These values were averaged for half-month increments, yielding 49 data points, which are mapped over time (June 20, 2011 to June 30, 2013). All costs have been adjusted to 2013 dollars. PCU ¼ palliative care unit.
for all PCU patients, both those who died on PCU and those who were discharged alive from the hospital. Average direct cost per day began decreasing three days before transfer to PCU and continued decreasing until the patient’s first full day on PCU. Costs then leveled at a steady daily rate. The mean of patients’ average cost per day was $1522 in days before transfer to PCU and $835 in the days after transfer to PCU. Without accounting for confounding factors, such as the frontloading of hospital costs, this cost difference represents a savings of $687 in daily costs per patient.
Comparison of Costs of PCU to PCCS We also compared the cost savings of PCU with those of the PCCS by propensity score matching the 712 patients who died on PCU with 456 PCCS patients who died on PCCS in the 18 months before PCU opened. PCU deaths represented 32% of hospital deaths during its first two years of operations. PCCS deaths over the 18 months we analyzed (before the PCU opening) represented about 32% of inpatient hospital deaths. The demographic characteristics of the patients who died on PCU, and those who died on PCCS services, are shown in Table 2, along with standardized differences before and after weighting. Using the propensity score weighted samples, we then compared average patient costs for each group in the days before and after transfer to PCU or PCCS (Fig. 3). In the propensity score weighted sample, PCU patients had significantly lower mean daily costs than PCCS patients in the days following initiation of palliative care ($753, SE ¼ $20; $1033, SE ¼ $43, respectively, P < 0.001). The average incremental effect of PCU in the unadjusted GLM was e$279 (bootstrapped SE ¼ $48, P < 0.001). This relationship remained statistically significant after including covariates from the propensity score in the regression model (average incremental effect ¼ e$240, bootstrapped SE ¼ $45,
Vol. 50 No. 2 August 2015
Cost Savings of a Palliative Care Unit
151
Mean Direct Cost per Day
$2,000 $1,500 $1,000 $500
$0 -20-19-18-17-16-15-14-13-12-11-10-9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 1011121314151617181920 Day, in RelaƟon to Transfer to PCU (Day 0)
Fig. 2. Patients’ average direct cost per day in relation to PCU transfer. Direct costs for each PCU patient from June 2011 through June 2013 for whom cost data were available (n ¼ 1071) were calculated for each discrete day of his or her hospitalization. Average direct cost for patients on the day of transfer to PCU is designated as Day 0. Average direct costs for patients on each discrete day before and after transfer are also graphed, with Day 1 as the day before transfer, Day 1 the first full day on PCU, etc. Number of cases per day is 690 on Day 5, 798 on Day 3, 1071 on Day 0, 622 on Day 3, and 421 on Day 5. All costs have been adjusted to 2013 dollars. PCU ¼ palliative care unit.
P < 0.001). Meanwhile, the mean daily costs for PCU and PCCS patients from hospital admission to day before palliative care initiation did not differ significantly ($1667, SE ¼ 34, $1778, SE ¼ 98, respectively,
P ¼ 0.29). These PCU cost savings, however, did not manifest themselves in the first six months of PCU operations but only became apparent after this start-up period.
Table 2 Characteristics of Patients Who Died on PCU and Under PCCS Care Standardized
Patient Characteristics
a
Difference Before
Difference After
Died on PCU
Died on PCCS
Propensity Score
Propensity Score
Mean (SD) or N (%)
Mean (SD) or N (%)
Weighting
Weighting
1.3 0.7 15.4 12.9
0.2 1.7 2.1 0.4
8.1
0.6
19.8 3.3 23.6 6.9 5.3 17.5 8.5 35.3
5.1 1 0.3 2.9 1.5 11.5 2.5 32.3
9.2 0.4 6.2 24 5.3
13.9 6.7 2.8 24.2 5.3
Patient admissions 712 Race White 264 (37) Black 140 (20) Hispanic 151 (21) Other/unknown 157 (22) Age (yrs) 70.2 (16.0) Gender Male 332 (47) Female 380 (53) Diagnosisb Cancer 342 (48.0) Neurological 97 (14) Cardiovascular 69 (10) Infectious disease 65 (9) Pulmonary 55 (8) c 12.0 (7e22) Median hospital LOS (d) Median LOS pre-PC (d) 8.0 (3e16) Median LOS on PC (d) 3.0 (1e6) Location in hospital before transfer to PCU/induction of PCCS Medical/surgical floor 402 (56) ICU 234 (33) Emergency Department 55 (8) Home 20 (3) Other 1 (0)
456 193 92 98 73 68.0
(42) (20) (22) (16) (15.8)
231 (51) 225 (49) 173 57 81 33 29 13.0 6.0 4.0
(38) (13) (18) (7) (6) (6e26) (2e17) (1e8)
278 149 28 0 0
(61) (33) (6) (0) (0)
PCU ¼ palliative care unit; PCCS ¼ palliative care consultation service; LOS ¼ length of stay; PC ¼ palliative care; ICU ¼ intensive care unit. The ‘‘Died on PCU’’ and ‘‘Died on PCCS’’ columns report characteristics for patients who died on the PCU during its first 24.5 months of existence and patients who died under PCCS care in the 18 months before the PCU opened, respectively. The first ‘‘Standardized Difference’’ column reports the standardized differences for covariates of the two groups before propensity score matching and weighting. The right-most ‘‘Standardized Difference’’ column reports standardized differences for covariates after propensity score matching and weighting. ‘‘Length of Stay’’ and ‘‘Location before Transfer’’ data were not included in the propensity score. a Excludes 23 patients for whom cost data were not available. b Diagnosis refers to the diagnosis for which palliative care was consulted, not necessarily the primary hospital admitting diagnosis. c Length of stay raw figures reported as median and interquartile range; standardized differences shown are for mean.
152
Nathaniel et al.
Vol. 50 No. 2 August 2015
Mean Direct Cost per Day
$3,000 $2,500 $2,000 $1,500
PCU paƟents
$1,000
PCCS paƟents
$500 $0
-15
-10
-5
0
5
10
15
Day, in RelaƟon to Transfer to PCU/ PCCS
Fig. 3. Comparison of average direct daily costs of patients on the PCU as compared with propensity score matched PCCS patients. The two lines represent the patients who died on the PCU during its first 24.5 months of operation and propensity score weighted historical control PCCS patients who died in the main hospital while under the care of PCCS care in the 18 months before the PCU opened. Direct costs for each patient for whom cost data were available were calculated for each discrete day of hospitalization. Day 0 is designated as the day of transfer to PCU for PCU patient group and day of induction of PCCS for PCCS patient group. Average direct costs for patients on each discrete day before and after transfer are also graphed, with Day 1 as the day before transfer, Day 1 the first full day on PCU or PCCS, etc. All costs have been adjusted to 2013 dollars. PCU ¼ palliative care unit; PCCS ¼ palliative care consultation service.
Discussion In this study, we characterized direct costs of a PCU in its first two years of operation and sought to determine whether PC modality (unit or consult service) was associated with differences in mean daily direct hospitalization costs. We found that mean cost per patient-day held steady at $1397 per patient for the first seven months, then dropped to $901, where they remained for 12 months, then rose to $1038 for the duration of the period under study. Daily cost savings began before the patients were transferred to PCU and then continued after transfer. Among patients who died in the hospital, average daily direct costs per patient in the days after transfer to PCU were $240 lower as compared with patients being followed by PCCS on the general hospital wards. To our knowledge, this is the first report to definitively demonstrate that the cost-savings of patients on a PCU is even greater than that of patients on a PCCS.
PCU Costs Over Time There were three distinct ranges of cost that emerged when we analyzed data from these two years. The two cost inflection points coincide with the ends of the 2011 and 2012 fiscal years. These cost changes can be explained in part by changes in hospital accounting. Most significantly, daily room costs for PCU were different in 2011, 2012, and the first six months of 2013, $1305, $793, and $990, respectively. The major factors determining room cost for a given floor are staff costs and occupancy rates. These have direct and indirect causative relationships, respectively, to room cost. These costs fluctuated widely for the first years of our unit as the hospital adjusted to the patient population, their needs related
to staffing, and occupancy. Room costs for the more well-established surgical and medical floors fluctuated minimally over these same time periods. While these exact costs are unique to our system, it may be helpful for administrators considering opening new PCUs to be aware of these variations in daily costs in the first years when a unit opens. Our analysis of changes in average daily cost per patient before and after transfer to PCU shows that PCU is associated with cost savings that begin before the patients are transferred and continue once they are on the unit. These findings corroborate and extend those by Smith et al., Albanese et al., and Eti et al., who had similar findings,24e26 although we incorporate data from a far larger sample of patients than have previous authors. These findings continue to confirm the overall cost savings related to palliative care.27
PCU vs. PCCS Costs Along with receiving higher quality of care,13,21 PCU patients who die in the hospital also incur lower endof-life costs than similar patients on other units who are cared for by a PCCS. These results may be related to the fact that standardizing the delivery of palliative care has been shown to lower costs.28 At Mount Sinai Hospital, patients in the PCU are more likely than patients under PCCS care to be seen exclusively by physicians and nurses who specialize in palliative care. This results in continued clarification of goals and allows treatments to continue to be refined so any therapies not concordant with the patient’s and family’s goals can be eliminated. Because on the PCU the palliative care team is most often the team of record, the team has the ability to make decisions that assure patient symptom
Vol. 50 No. 2 August 2015
Cost Savings of a Palliative Care Unit
management while also minimizing costs. Specific initiatives that help to encourage cost-saving include avoiding use of more expensive synthetic opioids like hydromorphone in patients with no contraindications to morphine and avoiding use of intravenous acetaminophen when other routes of delivery are possible. These cost savings may have large implications for a hospital system. PCU patients who died in the hospital cost an average of at least $240 per day less than similar PCCS patients in the days after PC initiation and average daily census for PCU in its first two years was 9.1 patients. This may represent a savings of over $1.6 million in direct patient costs in the first two years of our PCU. By demonstrating that PCU decedent daily costs are lower than costs for similar patients receiving PCCS care, our study adds significantly to the understanding of the cost-savings associated with PCU units. In addition, our results align with and build on results published by Albanese et al., which show that PCU patient costs are lower than costs for similar patients on other floors.25 They also support results published by Eti et al., which demonstrate that hospital charges were lower for patients managed directly by a palliative medicine attending, compared with patients who had a palliative medicine physician as a consultant.26
153
symptom severity and end-of-life care preferences. Still, it is possible that there remain unmeasured confounders between the two groups. Also, limiting our sample to only decedents is itself a limitation of this study, as palliative care has been shown to save costs in patients discharged alive as well.5 Our overall hospital length-of-stay means are significantly higher for palliative care patients than those published in many previous articles.25,26 We did not include length of stay in our propensity score calculations, as inclusion of a variable affected by treatment can obscure the treatment effect of interest.30
Conclusion Previous studies have demonstrated that both PCCS teams and PCUs are associated with cost savings for hospitals.5,24 We confirm that transfer to PCU is associated with decreasing direct costs. This is also the first study to demonstrate via a controlled comparison that PCUs provide more cost savings than PCCS for patient who die in the hospital. This has implications for hospital administrators who are weighing the potential benefits of a PCU against its capital and operational costs. Hospitals can use the data presented in this study when analyzing the cost/benefit ratios of opening an inpatient PCU.
Limitations There are a few limitations that must be considered. Mount Sinai has had a Palliative Care program for 15 years, so the Mount Sinai PCU was well-positioned to be an effective unit from the start. Also, end-of-life hospital costs in New York City are historically higher than national averages.29 Therefore, the magnitude of these cost savings may not be as easily achieved at other institutions. Additionally, our cost analyses include data from the very first day that PCU became operational. The higher PCU costs seen in early months of PCU operations are likely not representative of actual costs of a fully hospital-integrated PCU. Therefore, we considered excluding data from those first six months. However, we chose to include early data so as to best paint the full picture of fluctuations in cost from the opening of a unit so as to be instructive to other groups who may be planning to open their own unit. Although we adjusted for sociodemographic and clinical characteristics that might be associated both with PCU use and hospital costs, we were unable to observe symptom severity and end-of-life care preferences. Unobserved variables are beyond the capabilities of propensity score analyses. We attempted to control for these differences in our analysis of PCU vs. PCCS costs by limiting our sample to only deceased patients, who were more likely to have similar
Disclosures and Acknowledgments Mr. Nathaniel is supported by the Medical Student Training in Aging Research (MSTAR) Program through the American Geriatrics Society. Dr. Garrido is supported by CDA 11-201/CDP 12-255 from the Department of Veterans Affairs, Veterans Health Administration, Office of Research and Development, Health Services Research and Development Service. The authors have no disclosures. 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 States government. The funding sources had no role in the in study design; in the collection, analysis, and interpretation of data; in the writing of the report; and in the decision to submit the article for publication. The authors acknowledge PCU and PCCS staff, as well as our patients and their families.
References 1. Christakis NA, Iwashyna TJ. The health impact of health care on families: a matched cohort study of hospice use by decedents and mortality outcomes in surviving, widowed spouses. Soc Sci Med 2003;57:465e475.
154
Nathaniel et al.
2. Teno JM, Clarridge BR, Casey V, et al. Family perspectives on end-of-life care at the last place of care. JAMA 2004;291:88e93. 3. Jordhoy MS, Fayers P, Loge JH, Ahlner-Elmqvist M, Kaasa S. Quality of life in palliative cancer care: results from a cluster randomized trial. J Clin Oncol 2001;19: 3884e3894. 4. Ringdal GI, Jordhoy MS, Kaasa S. Family satisfaction with end-of-life care for cancer patients in a cluster randomized trial. J Pain Symptom Manage 2002;24:53e63. 5. Morrison RS, Penrod JD, Cassel JB, et al. Cost savings associated with US hospital palliative care consultation programs. Arch Intern Med 2008;168:1783e1790. 6. Morrison RS, Augustin R, Souvanna P, Meier DE. America’s care of serious illness: a state-by-state report card on access to palliative care in our nation’s hospitals. J Palliat Med 2011;14:1094e1096. 7. Rigby A, Krzyzanowska M, Le LW, et al. Impact of opening an acute palliative care unit on administrative outcomes for a general oncology ward. Cancer 2008;113:3267e3274. 8. Kellar N, Martinez J, Finis N, Bolger A, von Gunten CF. Characterization of an acute inpatient hospice palliative care unit in a U.S. teaching hospital. J Nurs Adm 1996;26:16e20. 9. Hui D, Elsayem A, Palla S, et al. Discharge outcomes and survival of patients with advanced cancer admitted to an acute palliative care unit at a comprehensive cancer center. J Pain Symptom Manage 2010;13:49e57. 10. Shin SH, Hui D, Chisholm GB, et al. Characteristics and outcomes of patients admitted to the acute palliative care unit from the emergency center. J Pain Symptom Manage 2014;47:1028e1034. 11. Digwood G, Lustbader D, Pekmezaris R, et al. The impact of a palliative care unit on mortality rate and length of stay for medical intensive care unit patients. Palliat Support Care 2011;9:387e392. 12. Lagman R, Rivera N, Walsh D, LeGrand S, Davis MP. Acute inpatient palliative medicine in a cancer center: clinical problems and medical interventionsda prospective study. Am J Hosp Palliat Care 2007;24:20e28. 13. Gaertner J, Frechen S, Sladek M, Ostgathe C, Voltz R. Palliative care consultation service and palliative care unit: why do we need both? Oncologist 2012;17:428e435. 14. Gaertner J, Drabik A, Marschall U, Schlesiger G, Voltz R, Stock S. Inpatient palliative care: a nationwide analysis. Health Policy 2013;109:311e318. 15. Casarett D, Johnson M, Smith D, Richardson D. The optimal delivery of palliative care: a national comparison of the outcomes of consultation teams vs inpatient units. Arch Intern Med 2011;171:649e655. 16. Back AL, Li Y-F, Sales AE. Impact of palliative care case management on resource use by patients dying of cancer
Vol. 50 No. 2 August 2015
at a veterans affairs medical center. J Palliat Med 2005;8: 26e35. 17. Cowan JD. Hospital charges for a community inpatient palliative care program. Am J Hosp Palliat Care 2004;21: 177e190. 18. Campbell M, Frank R. Experience with an end-of-life practice at a university hospital. Crit Care Med 1997;25: 197e202. 19. Ciemins EL, Blum L, Nunley M, Lasher A, Newman JM. The economic and clinical impact of an inpatient palliative care consultation service: a multifaceted approach. J Palliat Med 2007;10:1347e1355. 20. Higginson IJ, Finlay IG, Goodwin DM, et al. Is there evidence that palliative care teams alter end-of-life experiences of patients and their caregivers? J Pain Symptom Manage 2003;25:150e168. 21. Roza K. Comparison of quality of care of palliative care unit, palliative care consult team and usual care patients. New York: Icahn School of Medicine at Mount Sinai, 2014. 22. Bureau of Labor Statistics. 2013 Consumer Price Index detailed report tables. Washington, DC: Bureau of Labor Statistics, 2013. 23. Garrido MM, Kelly AS, Paris J, et al. Methods for constructing and assessing propensity scores. Health Serv Res 2014;49:1701e1720. 24. Smith TJ, Coyne P, Cassel B, Penberthy L, Hopson A, Hager MA. A high-volume specialist palliative care unit and team may reduce in-hospital end-of-life care costs. J Palliat Med 2003;5:699e705. 25. Albanese TH, Radwany SM, Mason H, Gayomali C, Dieter K. Assessing the financial impact of an inpatient acute palliative care unit in a tertiary care teaching hospital. J Palliat Med 2013;16:289e294. 26. Eti S, O’Mahony S, McHugh M, Guilbe R, Blank A, Selwyn P. Outcomes of the acute palliative care unit in an academic medical center. Am J Hosp Palliat Med 2014;31: 380e384. 27. Elsayem A, Swint K, Fisch MJ, et al. Palliative care inpatient service in a comprehensive cancer center: clinical and financial outcomes. J Clin Oncol 2004;22:2008e2014. 28. Smith TJ, Hillner BE. Ensuring quality cancer care by the use of clinical practice guidelines and critical pathways. J Clin Oncol 2001;19:2886e2897. 29. The Dartmouth Atlas of Healthcare. End-of-life care: data by region.2012; Hanover, NH: The Dartmouth Institute for Health Policy & Clinical Practice. Available from: http:// www.dartmouthatlas.org/data/topic/topic.aspx?cat=18. Last accessed April 25, 2015. 30. Garrido MM. Propensity scores: a practical method for assessing treatment effects in pain and symptom management research. J Pain Symptom Manage 2014;48:711e718.
