INTRODUCTION
In the United States, approximately 45 million people live with serious illness, and nearly half of those individuals are over 65 years of age.1 Serious illness is defined as a life-limiting condition that negatively impacts daily functioning or quality of life and/or involves burdensome symptoms, treatments, or caregiving.2 Individuals with serious illness constitute high healthcare utilizers and, as such, incur significant healthcare costs.3 Older adults with serious illness, including those with frailty or cognitive impairment, who undergo major surgery have high rates of mortality, complications, and functional debility post-operatively.4,5 Palliative care is interdisciplinary care which aims to improve quality of life, and as such, should be delivered alongside curative-intent treatments, including surgery, for all patients with serious illness.6 Previously-endorsed quality indicators of palliative care delivery include processes such as goals of care conversations, code status determination, palliative care consultation, assessment for hospice, and surrogate decision maker designation.7,8 Palliative care processes, including those provided by non-specialists, delivered alongside surgical care are associated with improved communication, reduced symptom burden, and lower healthcare utilization.9 However, it remains unknown how often patients with serious illness undergoing major scheduled surgery (i.e., planned or elective, rather than urgent or emergent surgery) receive palliative care processes, or if there are downstream associations between palliative care processes and postoperative healthcare utilization.
Both the National Consensus Guidelines for Palliative Care and the American College of Surgeons Geriatric Surgery Verification Program recommend the integration of palliative care processes as a standard of high-quality surgical care for older seriously ill patients undergoing surgery, irrespective of the type of procedure.6,10 Despite these recommendations, prior studies demonstrate deficits in palliative care delivery to surgical patients, but these are often single center studies focused on isolated processes, patients at the end-of-life, patients with specific diseases, or only high-risk procedures.11–14 Little work has been done to measure outcomes associated with palliative care processes documented during a surgical episode among patients with preexisting serious illness undergoing major scheduled surgery, which represent a uniquely vulnerable patient group with high healthcare utilization. To address this gap, we sought to determine associations between the delivery of palliative care processes and healthcare utilization in the year after discharge. We hypothesized that older adults with serious illness who received any palliative care processes during their surgical admission would have had reduced post-discharge healthcare utilization compared to those who did not receive any palliative care.
METHODS
Data Source & Study Population
The primary data source was the Mass General Brigham Research (MGB) Patient Data Registry (RPDR). RPDR is a repository of clinical and administrative data from all hospitals in the Mass General Brigham Healthcare System, and we selected five hospitals (two academic, three community) for study, all of which have specialty palliative care services. RPDR captures all inpatient encounters and contains mortality data, billing and procedure codes, and electronic health record (EHR) data including notes, procedure reports, and discharge summaries. We specifically collected discharge summaries, ambulatory visit notes, history and physical notes, progress reports, and longitudinal medical record ambulatory visit notes. These data were linked to Medicare claims.
The study cohort included seriously ill older adults 66 years of age and older with fee-for-service (FFS) Medicare coverage who had undergone a qualifying major elective operation during an in-hospital admission at MGB between January 1, 2016 and December 31, 2018. Serious illness, a life-limiting condition that negatively impacts daily functioning or quality of life and/or involves burdensome symptoms, treatments, or caregiving,2 was defined by consensus-based definitions of nine qualifying conditions or health states15–17 (Appendix A). Major scheduled surgeries included five procedures: colectomy, coronary artery bypass graft (CABG), lower extremity peripheral arterial revascularization, hip replacement, or knee replacement. These procedures were chosen because they are frequently performed, specifically in older adults – many with serious illness. For example, abdominal surgery such as colectomy and joint replacements are among the most common scheduled operations in older adults.18 Additionally, these procedures are performed by surgeons in various subspecialties, typically required a hospital stay during the years examined, and are deployed for a variety of indications including symptom management, cure, and extending life. Individuals undergoing procedures which were urgent, emergent, or due to trauma, were excluded.
Use of Natural Language Processing (NLP) to Capture Palliative Care Processes
NLP refers to methods that enable computers to process and analyze written text. Using NLP, we screened the EHR of our study cohort for specific terms pertaining to palliative care processes. We employed a previously validated NLP method and codebook of terms that best capture these processes (Appendix B).7,19 Processes were pursued directly from the medical record, instead of administrative codes, as prior studies have demonstrated administrative codes have poor sensitivity for detecting palliative care delivery.20,21
We identified the documentation of five palliative care processes endorsed by the National Consensus Project and the American College of Surgeons, including goals of care conversations, code status limitation, palliative care consultation, assessment for hospice, and surrogate decision maker designation.6,7 Our definition of each process is included in Appendix C. We defined surrogate decision maker present when a named individual was documented in the medical record. We abstracted documented code status limitations (e.g., do not intubate/resuscitate), rather than any documented code status because documented limitations implied that code status was discussed, whereas documented full code did not imply a discussion because a full code order could be a default setting in the electronic health record.
Four human coders (H.D., O.F., C.C., L.S.) used the text annotation software Clinical Regex developed by the Lindvall Lab at Dana-Farber Cancer Institute to facilitate review of the NLP output, meaning the EHR documentation that potentially captured a palliative care process.22 These coders received annotation training and subsequently reviewed the clinical notes that were flagged by NLP to determine if the text met criteria for documented delivery of one of the five palliative care processes we examined. An initial sample of 100 patients (9.2% of the total cohort) was reviewed by all four coders independently to assess for agreement, with a calculated kappa of 0.815, indicating high inter-coder agreement. Remaining patient data were divided and reviewed by individual coders.
One Year Post-Operative Healthcare Utilization
The primary outcome was healthcare utilization within the year following index surgical admission. Healthcare utilization was determined through Medicare claims and included ED visits, hospital days, intensive care unit (ICU) days, days at home, and enrollment in hospice. Days at home were determined by days alive in the one year after hospital discharge from index operation excluding inpatient admissions, ED visits, and days at skilled nursing facilities or other non-acute institutional stays.
Statistical Analysis and Covariates
Patients were divided into two exposure groups: those who had any (at least one) documented palliative care process during their index surgical admission and those who had none. We performed descriptive statistics using median and interquartile range (IQR) for continuous variables and counts and percentage for categorical variables. We measured differences between groups using Mann-Whitney for continuous variables, Pearson chi-square tests for categorical variables, and Fisher exact tests for categorical measures when cell counts were less than five. We used a Poisson regression to account for overdispersion to examine the association between palliative care processes and one-year post-operative healthcare utilization. The regression model was not conducted for some palliative care processes (goals of care conversation, palliative consultation, hospice assessment) and one-year post-operation hospice enrollment due to insufficient sample size. Model was adjusted for age, race (White vs. non-White), marital status (married and partnered vs. other), surgery type (colectomy, hip replacement, knee replacement, lower extremity peripheral artery bypass), Charlson Comorbidity Index (<2 vs. ≥2), ICU stay during index admission, length of stay during index admission, and serious illness qualifiers (1 vs. ≥2). All analyses were conducted using SAS version 9.4 (SAS Institute) and R version 4.2.1.23 A p-value less than 0.05 was considered statistically significant.
Sensitivity Analysis for Pre-Operative Goals of Care Conversations
To account for the possibility that patients in the cohort had documented palliative care processes in their pre-operative visits rather than at the time of their index admission, we performed a sensitivity analysis using NLP to scan the electronic health record specifically for goals of care conversations in the 30 days prior to surgery. We did not perform the sensitivity analysis using code status limitations or surrogate decision maker designation as these processes should accurately appear in the electronic health record at the time of hospitalization. Furthermore, specialist palliative care consultation and hospice assessment were infrequently performed in patients at the time of index admission and, therefore, unlikely to be present in the pre-operative period, without subsequent mention during the index admission.
RESULTS
Patient Characteristics
In total, 1,082 older adults with serious illness who had undergone major scheduled surgery were identified. Among these, the median age was 78 years; 57.0% were female and 93.3% were White. The most frequent qualifying serious illness was vulnerable elders (adults over 84 years of age or older than 64 years with any functional or cognitive disability) (62.4%), and 35.7% of the total cohort had more than one serious illness. The most common operation was colectomy (27.4%). Over half of patients (56.2%) had at least one documented palliative care process during their index admission (Table 1).
Table 1. Characteristics from Index Hospitalization.
Comparative characteristics for older adults with serious illness in MGB RPDR linked to Medicare Claims after major elective surgery from 2016 to 2018 (N=1082), stratified by those who received versus did not receive any palliative care processes.
| Total (N=1082) | No Palliative Care (N=474) | Received any Palliative Care (N=608) | p-value | |
|---|---|---|---|---|
| Age (years), Median (IQR) | 78 (72-85) | 76 (72-84) | 78 (73-85) | 0.006 |
| Sex (female), N (%) | 617 (57.0%) | 279 (58.9%) | 338 (55.6%) | 0.28 |
| Race/Ethnicity, N (%) | 0.22 | |||
| White | 1009 (93.3%) | 437 (92.2%) | 572 (94.1%) | |
| Non-White1 | 73 (6.75%) | 37 (7.81%) | 36 (5.92%) | |
| Serious Illness, N (%) | ||||
| Vulnerable Elder | 675 (62.4%) | 277 (58.4%) | 398 (65.5%) | 0.02 |
| Advanced Cancer | 168 (15.5%) | 80 (16.9%) | 88 (14.5%) | 0.28 |
| Pulmonary Disease | 44 (4.1%) | 19 (4.0%) | 25 (4.1%) | 0.93 |
| Heart Failure | 284 (26.2%) | 108 (22.8%) | 176 (28.9%) | 0.02 |
| Cirrhosis | 81 (7.5%) | 36 (7.6%) | 45 (7.40%) | 0.90 |
| ESRD | 88 (8.1%) | 27 (5.7%) | 61 (10.0%) | 0.010 |
| Dementia | 33 (3.0%) | <11 (<2.3%) | >22 (>3.6%) | 0.003 |
| Frailty | 300 (27.7%) | 123 (25.9%) | 177 (29.1%) | 0.25 |
| Nursing Home Resident | 20 (1.8%) | <11 (<2.3%) | >11 (>1.8%) | 0.42 |
| >1 Serious Illness Qualifier | 386 (35.7%) | 143 (30.2%) | 243 (40.0%) | <0.001 |
| >2 Serious Illness Qualifier | 157 (14.5%) | 46 (9.7%) | 111 (18.3%) | <0.001 |
| Major Surgery Type, N (%) | ||||
| Colectomy | 297 (27.4%) | 122 (25.7%) | 175 (28.8%) | 0.27 |
| Hip Replacement | 254 (23.5%) | 127 (26.8%) | 127 (20.9%) | 0.02 |
| Knee Replacement | 294 (27.2%) | 133 (28.1%) | 161 (26.5%) | 0.56 |
| Coronary Artery Bypass Graft | 141 (13.0%) | 56 (11.8%) | 85 (14.0%) | 0.29 |
| Lower Extremity Revascularization | 99 (9.1%) | 36 (7.6%) | 63 (10.4%) | 0.12 |
| Duration of Index Admission (Days), Median (IQR) | 4 (3-6) | 3 (3-6) | 4 (3-7) | <0.001 |
| Any ICU Stay During Index Admission, N (%) | 211 (19.5%) | 82 (17.3%) | 129 (21.2%) | 0.11 |
| Days in ICU During Index Admission, Median (IQR) | 3 (1-7) | 2 (1-4.75) | 4 (2-9) | <0.001 |
| Palliative Care (PC) Process2, N (%) | ||||
| Goals of Care | 28 (2.6%) | -- | 28 (4.6%) | -- |
| Palliative Care Consultation | <11 (<1.0%) | -- | <11 (<1.8%) | -- |
| Hospice Assessment | 14 (1.3%) | -- | 14 (2.3%) | -- |
| Code Status Limitation | 47 (4.3%) | -- | 47 (7.7%) | -- |
| Surrogate Decision Maker Designation | 585 (54.1%) | -- | 585 (96.2%) | -- |
| Number of Palliative Care Processes, N (%) | ||||
| 0 | 474 (43.8%) | -- | 474 (43.8%) | -- |
| 1 | 559 (51.7%) | -- | 559 (51.7%) | -- |
| 2 | 31 (2.87%) | -- | 31 (2.87%) | -- |
| >2 | 18 (1.66%) | -- | 18 (2.96%) | -- |
| Discharge Location after Surgery, N (%) | <0.001 | |||
| Home | 519 (48.0%) | 243 (51.3%) | 276 (45.4%) | |
| Hospital | <11 (<1.0%) | <11 (<2.3%) | 0 (0%) | |
| Acute Inpatient | 126 (11.6%) | 46 (9.7%) | 80 (13.2%) | |
| Long Term Care | 16 (1.5%) | <11 (<2.3%) | >11 (>1.8%) | |
| Hospice | <11 (<1.0%) | 0 (0%) | <11 (<1.8%) | |
| Other3 | <11 (<1.0%) | <11 (<2.3%) | 0 (0%) | |
| Deceased | 20 (1.8%) | <11 (<2.3%) | >11 (>1.8%) | |
| 30-Day Readmission4, N (%) | 241 (22.7%) | 86 (18.2%) | 155 (26.3%) | 0.002 |
| In-Hospital Mortality, N (%) | 20 (1.8%) | <11 (<2.3%) | >11 (>1.8%) | 0.002 |
| 30-Day Mortality4, N (%) | <11 (<1.0%) | 0 (0%) | <11 (<1.8%) | 0.07 |
| 90-Day Mortality4, N (%) | 18 (1.7%) | <11 (<2.3%) | >11 (>1.8%) | 0.06 |
| 1-Year Mortality4, N (%) | 77 (7.3%) | 27 (5.7%) | 50 (8.5%) | 0.09 |
Includes Black, Asian, North American Native, Hispanic, unknown, other
See Appendix B for codebook defining each palliative care processes
Left against medical advice or discontinued care or discharged/transferred to a critical access hospital
Missing data due to patient death within the relevant timeframe of given outcome
Patients who received palliative care processes were older (78 vs. 76 years, p = 0.006) and more likely to have greater than one qualifying serious illness (40.0% vs. 30.2%, p = <0.001). More patients who received palliative care processes qualified as seriously ill as vulnerable elders (65.5% vs. 58.4%, p = 0.02) or with diagnoses of heart failure (28.9% vs. 22.8%, p = 0.02), end-stage renal disease (10.0% vs. 5.7%, p = 0.010), and dementia (>3.6% vs. <2.3%, p = 0.003). As compared to those without documented processes, those with documented processes had longer hospitalizations during their index admission (median 4 vs. 3 days, p <0.001), spent more days in the ICU (median 4 vs. 2 days, p <0.001), were less likely to be discharged to home (45.4% vs 51.3%, p <0.001), and had higher 30-day readmission rates (26.3% vs. 18.2%, p = 0.002). There were no significant differences in one-year mortality rates. In-hospital, 30-day, and 90-day mortality rates were too low to detect significant differences between groups (Table 1, Supplementary Table 1 for additional characteristics).
Palliative Care Processes
Surrogate decision maker designation was documented in 54.1% of the total cohort and in 96.2% of patients who had any palliative care processes; assessment for hospice and palliative care consults were documented for 1.3% and 1.0% of patients respectively (Table 1, Figure 1). Patients who underwent lower extremity arterial revascularizations and CABG had significantly higher numbers of documented palliative care processes than other types of surgery (Supplemental Tables 2-7). For example, for those who underwent arterial revascularization, 11.1% had more than two documented palliative care processes as compared to only 3.7% of patients who underwent knee replacement surgery.
Figure 1. Number of Palliative Care Processes.
For 1,082 patients with serious illness who underwent major surgery between 2016 and 2018, 474 (44%) had zero palliative care processes documented during this index admission, while 559 (52%) had one documented palliative care process.
One Year Post-Operative Healthcare Utilization
In univariate analysis (Table 2), patients with any documented palliative care process spent fewer days at home in the year after their index admission compared to those who did not have documented palliative care processes (351 vs. 353 days, p = 0.008). There were no significant differences in ICU utilization, total hospital days, hospital readmissions, emergency department visits, or hospice enrollments in the year after surgery.
Table 2. Unadjusted Healthcare Utilization within One Year of Index Surgical Admission.
Healthcare utilization data from Medicare claims for outcomes of interest, stratified by presence of any documented palliative care process vs. none. Twenty patients excluded from one-year outcomes due to death during index hospitalization.
| Total (N=1062) | No Palliative Care (N=472) | Received any Palliative Care (N=590) | p-value | |
|---|---|---|---|---|
| Any ICU Utilization, N (%) | 121 (11.4%) | 46 (9.7%) | 75 (12.7%) | 0.13 |
| ICU days, Median (IQR) | 4 (2-6) | 3.50 (2-7) | 4 (2-5) | 0.39 |
| Total Hospital Days, Median (IQR) | 3 (0-13) | 3 (0-10) | 4 (0-14) | 0.07 |
| Total Days at Home, Median (IQR) | 352 (334-362) | 353 (341-362) | 351 (326-362) | 0.008 |
| Any Hospital Readmissions, N (%) | 534 (50.3%) | 228 (48.3%) | 306 (51.9%) | 0.25 |
| Number of Hospital Readmissions, Median (IQR) | 2 (1-3) | 1 (1-3) | 2 (1-3) | 0.12 |
| Any ED Visits, N (%) | 648 (61.0%) | 290 (61.4%) | 358 (60.7%) | 0.80 |
| Number of ED visits, median (IQR) | 2 (1-3) | 2 (1-3) | 2 (1-4) | 0.45 |
| Hospice Enrolment, N (%) | 51 (4.8%) | 18 (3.8%) | 33 (5.6%) | 0.18 |
After adjusting for covariates, there were no significant differences in one-year healthcare utilization for patients with and without documented palliative care processes. However, patients with code status limitations as their only documented palliative care process spent significantly fewer days at home in the year after surgery (314.9 vs. 339.6, p = 0.004) (Table 3).
Table 3. Adjusted One-Year Healthcare Utilization for Seriously Ill Older Adults Undergoing Major Elective Surgery by Receipt of Palliative Care.
Poisson model adjusted for age, sex (female vs. male), race (White vs. non-White), marital status (married and partnered vs others), surgery type (colectomy, hip replacement, knee replacement, lower extremity peripheral artery revascularization), Charlson Comorbidity Index (<2 vs. ≥2), ICU stay during index admission (yes vs. no), length of stay for index admission, serious illness qualifiers (1 vs. >1). There was insufficient power to run regression for remaining palliative care processes (goals of care conversation, palliative care consultation, assessment for hospice) and insufficient power to run regression for outcome of hospice enrollment.
| Hospital Days | Mean (Days) | Ratio of Means (IRR) | p-value |
|---|---|---|---|
| Any Palliative Care vs. None | 10.57 vs. 8.99 | 1.18 (0.91 - 1.52) | 0.21 |
| Code Status Limitation vs. None | 15.08 vs. 9.70 | 1.55 (0.90 – 2.69) | 0.12 |
| Surrogate Decision Maker vs. None | 10.66 vs. 8.96 | 1.19 (0.93 - 1.52) | 0.17 |
| Days at Home | Mean (Days) | Ratio of Means (IRR) | p-value |
| Any Palliative Care vs. None | 336.69 vs. 339.55 | 0.99 (0.98 - 1.01) | 0.27 |
| Code Status Limitation vs. None | 314.88 vs. 338.63 | 0.93 (0.89 - 0.98) | 0.004 |
| Surrogate Decision Maker vs. None | 336.65 vs. 339.51 | 0.99 (0.98 - 1.01) | 0.27 |
| ED Visits | Mean (N) | Ratio of Means (IRR) | p-value |
| Any Palliative Care vs. None | 1.51 vs. 1.57 | 0.96 (0.82 - 1.13) | 0.64 |
| Code Status Limitation vs. None | 2.14 vs. 1.52 | 1.41 (0.94 – 2.11) | 0.10 |
| Surrogate Decision Maker vs. None | 1.49 vs. 1.58 | 0.95 (0.80 - 1.11) | 0.51 |
Sensitivity Analysis
A sensitivity analysis to assess for goals of care conversations documented within the 30 days prior to the index admission showed that 3.6% of patients had goals of care conversations in the 30 days prior to surgery (Supplemental Table 8). Patients who had goals of care conversations in a pre-operative visit had greater odds of having goals of care conversation during their surgical admission (OR 10.6 [95% CI 4.2-27.0], p <0.001).
DISCUSSION
Our findings demonstrate low rates of documentation of palliative care processes in this cohort of older seriously ill patients undergoing major scheduled surgery. Surrogate decision maker was the most frequently documented process, yet only slightly more than half of patients in the cohort had evidence of a designated surrogate decision maker during their index admission for a major operation where most were expected to undergo general anesthesia. We also found that goals of care conversations were infrequently documented in the 30 days prior to index admission, demonstrating significant needs for improvement in practice, documentation, or both, to meet national standards like the American College of Surgeons Geriatric Verification Program.10 Contrary to our hypothesis, there were no differences in one year healthcare utilization between seriously ill older patients who did, and did not, receive palliative care processes proximate to major scheduled surgery.
This study corroborates that rates of documented palliative care among surgical patients are generally low.11,24–27 For example, Kim et al. found that, in a cohort of over 190,000 veterans at one VA hospital, only 3.8% of patients who underwent procedures had documented pre-operative goals of care conversations, and that high-risk surgery was not associated with a greater likelihood of goals of care documentation.24 While stratifying by surgical risk is important, even low-risk operations can be burdensome to patients who are themselves high-risk due to underlying serious illness. As such, we examined a diverse set of processes and started from the premise that all patients in our cohort were eligible for goals of care conversations because they met accepted criteria for serious illness.25 Our findings are consistent with prior efforts to quantify rates of palliative care processes. However, our multicenter study goes further by utilizing NLP methods to identify palliative care documentation, aggregating multiple processes at the index surgical admission, focusing on older patients with preexisting serious illness, and assessing for the relationship between the documentation of palliative care processes and one-year post-operative outcomes. Furthermore, unlike prior efforts, our study uniquely examines not only specialty palliative care delivery but also palliative care processes that can be provided by non-specialists. In addition, linking EHR with Medicare data offers new insights into inpatient palliative care processes and one-year outcomes among seriously ill older surgical patients.
Prior work has shown that, in patients with advanced cancer, specialist palliative care delivery is associated with decreased healthcare utilization including evidence of lower hospital charges,28,29 shorter hospital length of stay,29 and fewer emergency department visits.30 In other, non-surgical patient groups – largely with advanced cancer, policymakers have advocated for palliative care to reduce non-beneficial care and therefore, to reduce costs.31 These findings shaped our hypothesis that seriously ill older surgical patients with documented palliative care processes would have lower healthcare utilization compared to those without palliative care process documentation. However, our findings underscore differences in outcomes associated with palliative care delivery to patients with serious illness undergoing scheduled surgery as compared to palliative care provision to patients with advanced cancer specifically. Given high rates of post-operative survival, healthcare utilization may well be within goals of care for this cohort of patients with serious illness eligible for major surgery, unlike patients who are nearing the end of life who may desire less healthcare utilization. Further research is needed to identify the outcomes that specifically matter to seriously ill surgical patients undergoing scheduled operations. Patient-reported outcome measures of quality of life, pain, and depression may offer more appropriate outcomes for future studies, although there remain challenges with integrating these into standard practice.32,33
Similar questions about the effects of palliative care on surgical patients arose in two recent randomized control trials. A study by Shinall et al. randomized patients undergoing curative-intent abdominal surgery for cancer to palliative care consultation and follow-up.34 Shinall and colleagues found that the intervention was not associated with any difference in the number of days alive at home in the three-months after surgery, suggesting that post-operative hospital use was unassociated with a palliative care consultation for patients undergoing scheduled cancer operations. Another randomized control trial by Aslakson and colleagues did not find improvements in patient-reported outcomes such as quality of life or mental health for surgical oncology patients receiving palliative care team comanagement prior to pursuing curative intent operations.35 These studies differ from ours in that they focus on a specialty palliative care intervention, while we focus on broad process measures of palliative care as they are already practiced as part of routine surgical care.. Taken together, these studies and ours suggest that seriously patients undergoing scheduled surgery have illness trajectories that are distinct from other seriously ill hospitalized patients, and that perioperative palliative care does not reduce downstream healthcare utilization. This raises the critical question of how and when to better address the palliative care needs among elective surgical patients, including identifying which patients may meaningfully benefit from specialist palliative care and which may be supported through non-specialist palliative care processes. These studies suggest that outcomes associated with palliative care utilization for seriously ill surgical patients must be evaluated in the context of surgical patients’ unique life expectancy and goals.
While overall we did not observe significant differences in healthcare utilization between patients with documented palliative care processes – specifically with the most frequently documented process, surrogate decision maker designation – and those without, it is notable that patients for whom code status limitations were the only documented palliative care process, spent on average 25 fewer days at home in the year after surgery. This finding suggests that clinicians likely intuited or estimated that this subgroup of patients had a higher risk of postoperative mortality, prompting code status discussions and limitations. However, a retrospective approach hinders our ability to understand exactly what prompted code status limitations for these patients. Nonetheless, it is evident that clinicians are appropriately discerning which patients are at highest risk for mortality. Additionally, our data support that patients who are generally sicker are, in fact, more likely to receive palliative care processes: patients who had documented palliative care processes had longer index admissions, spent more days in the ICU, were less likely to discharge home, and had higher readmission rates – all of which suggests that clinicians are targeting the most seriously ill patients for palliative care processes. Nonetheless, low rates of documented palliative care processes among a cohort of patients with life-limiting and burdensome illness suggest that surgical clinicians are not including palliative care as a routine part of care for patients with serious illness as national guidelines suggest they should.
Importantly, our study cohort had high rates of post-operative survival, yet substantial post-operative burden, with more than half not discharged home after surgery and nearly one in four readmitted within 30 days. Our findings suggest that surgical clinicians directed palliative care processes toward the most seriously ill patients, a pattern that also reflects the very low rate of documented pre-operative goals-of-care conversations. The combination of high survival and high post-operative burden in this cohort highlights the value of pre-operative conversations that extend beyond survival, code status and discussions about life sustaining treatment to address what matters most to patients and families (e.g., functional recovery, symptom management, and quality of life) and to support shared and informed surgical decision-making. The findings also underscore that, while national consensus guidelines may be broadly applicable to promote high quality care for patients with serious illness, patients undergoing surgery have unique health trajectories and may have fewer palliative care needs than other groups of patients with serious illness.
Several limitations exist in this study. First, the documentation of palliative care processes is not equivalent to the delivery of palliative care. Inversely, the absence of documentation does not indicate that palliative care processes were absent from the clinical encounter. However, at present, documentation is the only way to capture evidence of palliative care process delivery for quality measurement. Our findings highlight opportunities for further improvement in how palliative care processes are documented. Second, this study was conducted in one, resource-rich regional health system and therefore the findings are not generalizable. However, this study did include five distinct hospitals with different patients, surgeons, and palliative care capacity. Furthermore, as this health system is resource-rich, we imagine that palliative care delivery is likely lower at other institutions. The patients in this study are largely White and enrolled in fee-for-service Medicare further limiting the generalizability of these findings. A third limitation is that the study period (2016-2018) predates national quality improvement efforts such as the American College of Surgeons Geriatric Surgery Verification Program, which launched in 2019.10 We selected this study period because it was the most recent data available at the time of analysis that linked electronic health records with Medicare claims, which is a novel method, but recognize the ways in which this specific timeframe again limits the findings’ generalizability. A fourth limitation is that we examined select processes and did not capture other important domains of palliative care such as symptom management. However, we included palliative care processes that are relevant to hospital admission irrespective of specialist palliative care consultation.
Conclusion
Contrary to our hypothesis, palliative care process documentation was not generally associated with decreased healthcare utilization in the year after surgery, although patients with documented goals of care conversations were found to have fewer days at home post-discharge. As compared to other seriously ill hospitalized patients, reduced healthcare utilization should not be an anticipated outcome of palliative care delivery for seriously ill patients undergoing scheduled surgery.
Supplementary Material
Funding Sources:
Orly N. Farber received funding from the NIDDK T32 Grant #T32DK007754. Zara Cooper receives funding from NIA Grant #R01AG070252A. Hiba Dhanani received funding from the NIH Grant #3R01AG070252-02S1. Tamryn F. Gray receives funding from the RWJF Harold Amos Medical Faculty Development Program and the Betty Irene Moore Fellows for Nurse Leaders and Innovators.
Footnotes
Conflicts of Interest Statement: The authors report no potential conflicts of interest to disclose.
References
- 1.Bell JF, Whitney RL, Young HM. Family Caregiving in Serious Illness in the United States: Recommendations to Support an Invisible Workforce. J Am Geriatr Soc. 2019;67(S2):S451–s456. [DOI] [PubMed] [Google Scholar]
- 2.Kelley AS. Defining “Serious Illness”. Journal of Palliative Medicine. 2014;17(9):985–985. [DOI] [PubMed] [Google Scholar]
- 3.Aldridge MD, Kelley AS. The Myth Regarding the High Cost of End-of-Life Care. Am J Public Health. 2015;105(12):2411–2415. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Makary MA, Segev DL, Pronovost PJ, et al. Frailty as a Predictor of Surgical Outcomes in Older Patients. Journal of the American College of Surgeons. 2010;210(6):901–908. [DOI] [PubMed] [Google Scholar]
- 5.Chen L, Au E, Saripella A, et al. Postoperative outcomes in older surgical patients with preoperative cognitive impairment: A systematic review and meta-analysis. Journal of Clinical Anesthesia. 2022;80:110883. [DOI] [PubMed] [Google Scholar]
- 6.Clinical Practice Guidelines for Quality Palliative Care. 4th ed. Richmond, VA: National Coalition for Hospice and Palliative Care; 2018. [Google Scholar]
- 7.Lilley EJ, Lindvall C, Lillemoe KD, Tulsky JA, Wiener DC, Cooper Z. Measuring Processes of Care in Palliative Surgery: A Novel Approach Using Natural Language Processing. Annals of surgery. 2018;267(5):823–825. [DOI] [PubMed] [Google Scholar]
- 8.ACS TQIP: PALLIATIVE CARE BEST PRACTICES GUIDELINES.; 2017. American College of Surgeons. https://www.facs.org/-/media/files/quality-programs/trauma/tqip/palliative_guidelines.ashx. Accessed February 3, 2025.
- 9.Lilley EJ, Khan KT, Johnston FM, et al. Palliative Care Interventions for Surgical Patients: A Systematic Review. JAMA Surgery. 2016;151(2):172–183. [DOI] [PubMed] [Google Scholar]
- 10.Optimal Resources for Geriatric Surgery. American College of Surgeons. Geriatric Surgery Verification Quality Improvement Program Web site. https://www.facs.org/media/yldfbgwz/19_re_manual_gsv-standards_digital-linked-pdf-1.pdf. Published 2019. Accessed December 12, 2024. [Google Scholar]
- 11.Kalbfell E, Kata A, Buffington AS, et al. Frequency of Preoperative Advance Care Planning for Older Adults Undergoing High-risk Surgery: A Secondary Analysis of a Randomized Clinical Trial. JAMA Surg. 2021;156(7):e211521. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Evans BA, Turner MC, Gloria JN, Pickett LC, Galanos AN. Palliative Care Consultation Is Underutilized in Critically Ill General Surgery Patients. American Journal of Hospice and Palliative Medicine®. 2020;37(2):149–153. [DOI] [PubMed] [Google Scholar]
- 13.Olmsted CL, Johnson AM, Kaboli P, Cullen J, Vaughan-Sarrazin MS. Use of Palliative Care and Hospice Among Surgical and Medical Specialties in the Veterans Health Administration. JAMA Surgery. 2014;149(11):1169–1175. [DOI] [PubMed] [Google Scholar]
- 14.Udelsman BV, Lilley EJ, Qadan M, et al. Deficits in the Palliative Care Process Measures in Patients with Advanced Pancreatic Cancer Undergoing Operative and Invasive Nonoperative Palliative Procedures. Ann Surg Oncol. 2019;26(13):4204–4212. [DOI] [PubMed] [Google Scholar]
- 15.Kelly MT, Sturgeon D, Harlow AF, Jarman M, Weissman JS, Cooper Z. Using Medicare Data to Identify Serious Illness in Older Surgical Patients. Journal of Pain and Symptom Management. 2020;60(2):e101–e103. [DOI] [PubMed] [Google Scholar]
- 16.Lee KC, Walling AM, Senglaub SS, Kelley AS, Cooper Z. Defining Serious Illness Among Adult Surgical Patients. J Pain Symptom Manage. 2019;58(5):844–850.e842. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Kelley AS, Ferreira KB, Bollens-Lund E, Mather H, Hanson LC, Ritchie CS. Identifying Older Adults With Serious Illness: Transitioning From ICD-9 to ICD-10. Journal of pain and symptom management. 2019;57(6):1137–1142. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Berlin NL, Kirch M, Singer DC, Solway E, Malani PN, Kullgren JT. Preoperative Concerns of Older US Adults and Decisions About Elective Surgery. JAMA Network Open. 2024;7(1):e2353857–e2353857. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Lindvall C, Lilley EJ, Zupanc SN, et al. Natural Language Processing to Assess End-of-Life Quality Indicators in Cancer Patients Receiving Palliative Surgery. Journal of Palliative Medicine. 2018;22(2):183–187. [DOI] [PubMed] [Google Scholar]
- 20.Stubbs JM, Assareh H, Achat HM, Greenaway S, Muruganantham P. Verification of administrative data to measure palliative care at terminal hospital stays. Health Information Management Journal. 2023;52(1):28–36. [DOI] [PubMed] [Google Scholar]
- 21.Hua M, Li G, Clancy C, Morrison RS, Wunsch H. Validation of the V66.7 Code for Palliative Care Consultation in a Single Academic Medical Center. J Palliat Med. 2017;20(4):372–377. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.ClinicalRegex. https://lindvalllab.github.io/clinical-regex/. Accessed March 3, 2025.
- 23.R: A language and environment for statistical computing. R Foundation for Statistical Computing Web site. https://www.r-project.org/. Published R Core Team. 2022. Accessed November 14, 2024. [Google Scholar]
- 24.Kim KM, Giannitrapani KF, Garcia A, et al. Patient Characteristics Associated With Occurrence of Preoperative Goals-of-Care Conversations. JAMA Network Open. 2023;6(2):e2255407–e2255407. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Colley A, Lin JA, Pierce L, Finlayson E, Sudore RL, Wick E. Missed Opportunities and Health Disparities for Advance Care Planning Before Elective Surgery in Older Adults. JAMA Surg. 2022;157(10):e223687. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Robbins AJ, Beilman GJ, Ditta T, et al. Mortality After Elective Surgery: The Potential Role for Preoperative Palliative Care. The Journal of surgical research. 2021;266:44–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Marks S, Wanner JP, Cobb AS, Swetz KM, Lange GM. Surgery without a surrogate: the low prevalence of healthcare power of attorney documents among preoperative patients. Hospital Practice. 2019;47(1):28–33. [DOI] [PubMed] [Google Scholar]
- 28.Kim SJ, Patel I, Park C, Shin DY, Chang J. Palliative care and healthcare utilization among metastatic breast cancer patients in US Hospitals. Scientific reports. 2023;13(1):4358–4358. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Chang J, Han K-T, Medina M, Kim SJ. Palliative care and healthcare utilization among deceased metastatic lung cancer patients in US hospitals. BMC palliative care. 2022;21(1):136–138. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Gonzalez R, Srinivas S, Waterman BL, et al. Impact of early vs late palliative care referrals on healthcare utilization in patients with pancreatic cancer. Journal of cancer research and clinical oncology. 2023;149(16):14997–15002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Palliative Care: Key Studies on Cost Savings. Amercain Cancer Society Cancer Action Network. Public Policy Resources Web site. https://www.fightcancer.org/policy-resources/palliative-care-key-studies-cost-savings. Published 2022. Accessed December 13, 2024. [Google Scholar]
- 32.Müller E, Mayer-Steinacker R, Gencer D, et al. Feasibility, use and benefits of patient-reported outcome measures in palliative care units: a multicentre observational study. BMC Palliat Care.2023;22(1):6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Driscoll B, Leonard LD, Kovar A, et al. Surgeon Perceptions of the Integration of Patient-Reported Outcome Measures into Clinical Practice. J Surg Res. 2022;280:486–494. [DOI] [PubMed] [Google Scholar]
- 34.Shinall MC, Martin SF, Karlekar M, et al. Effects of Specialist Palliative Care for Patients Undergoing Major Abdominal Surgery for Cancer: A Randomized Clinical Trial. Archives of surgery (Chicago 1960). 2023;158(7):747–755. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Aslakson RA, Rickerson E, Fahy B, et al. Effect of Perioperative Palliative Care on Health-Related Quality of Life Among Patients Undergoing Surgery for Cancer: A Randomized Clinical Trial. JAMA Network Open. 2023;6(5):e2314660–e2314660. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.