Abstract
Background:
Palliative care (PC) services expanded rapidly to meet the needs of coronavirus disease 2019 (COVID-19) patients, yet little is known about which patients were referred for PC consultation during the pandemic.
Objective:
Examine factors predictive of PC consultation for COVID-19 patients.
Design:
Retrospective cohort study of COVID-19 patients discharged from four hospitals (March 1–June 30, 2020).
Exposures:
Patient demographic, socioeconomic, and clinical factors and hospital-level characteristics.
Outcome Measurement:
Inpatient PC consultation.
Results:
Of 4319 hospitalized COVID-19 patients, 581 (14%) received PC consultation. Increasing age, serious illness (cancer, chronic obstructive pulmonary disease, and dementia), greater illness severity, and admission to the quaternary hospital were associated with receipt of PC consultation. There was no association between PC consultation and race/ethnicity, household crowding, insurance status, or hospital-factors, including inpatient, emergency department, and intensive care unit census.
Conclusions:
Although site variation existed, the highest acuity patients were most likely to receive PC consultation without racial/ethnic or socioeconomic disparities.
Background
The coronavirus disease 2019
Although hospital-based PC consultation was widely utilized during the pandemic, it is unclear which patients were referred to PC.4–6 Prepandemic studies of patients with cancer suggest clinician variability in the timing of PC referral during an illness (i.e., at the time of serious illness diagnosis, life-prolonging treatment decision, or approaching end of life).7–11 Although studies from the inpatient setting are limited, there is evidence of racial/ethnic disparities in prepandemic access to PC.7–11
In our study, we sought to identify predictors of PC consultation for adults hospitalized with COVID-19 during the initial pandemic surge in New York City. Given the widely recognized prepandemic disparities in access to PC and the disproportionate impact of COVID-19 on Hispanic and Black patients and those living in poverty,12–20 we examined disparities in PC referral during a crisis. Understanding these referral patterns is crucial to ensuring patients have equitable access to PC consultation.
Methods
Analytic file and data sources
We used an analytic file of electronic medical record (EMR) data, administrative billing data, including ICD-10 codes, and skilled nursing facility (SNF) to hospital transfer data. We used patient zip code to link to census tract-level data on average household crowding from the U.S. Census Bureau's 2014–2018 American Community Survey 5-Year Estimates file.
Setting
We included four of eight Mount Sinai Health System hospitals with dedicated inpatient PC consultation teams (i.e., social work, chaplain, physician, and advance practice nurse) during the initial COVID-19 surge, including a 1100-bed quaternary hospital (MSA), two mid-size 330-bed community hospitals (MSB, MSC), and one 100-bed community hospital (MSD). Because MSD faced high consult volumes, an extra physician who served as “inpatient extender” supported the dedicated team.
Sample
We identified all hospitalized patients aged ≥18 years discharged (both alive n = 3251 and deceased n = 1068) between March 1 and June 30, 2020 with (1) an ICD-10 diagnosis of COVID-19 (if they were discharged after April 1, 2020), (2) a polymerase chain reaction (PCR) confirmed severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection, +/− (3) an EMR COVID-19 flag. We excluded admitting services of cardiac catheterization (n = 17), dentistry (n = 1), obstetrics (n = 251), psychiatry (n = 34), rehabilitation medicine (n = 1), addiction services (n = 21), and hospice (n = 6), and elective procedure admissions with a prior PCR-confirmed SARS-CoV-2 infection (n = 45). We also excluded encounters with missing covariates (n = 249): unweighted Elixhauser Comorbidity Index (n = 3), household crowding (n = 226), insurance (n = 17), emergency department (ED) census (n = 6), and missing ICD-10 diagnoses for the encounter (n = 3). Because the patient was our unit of analysis, we analyzed the index COVID-19 hospitalization.
Outcome and covariates
Our primary outcome was receipt of PC consultation. We included the following covariates groups: demographic data (age, gender, and race/ethnicity); socioeconomic data (insurance status, household crowding defined as the percent of households with >1 occupant per room in a census tract given the association of crowding and household spread of COVID-19); clinical characteristics (unweighted Elixhauser comorbidity index,21,22 serious illnesses, including cancer, chronic obstructive pulmonary disease [COPD], dementia, and coronary artery disease [CAD]23,24 using ICD-10 codes, SNF transfer data to identify which patients transferred to the hospital from an SNF); illness severity (worst Sequential Organ Failure Assessment [SOFA] score in the first week of admission,25,26 maximal oxygen support received with the following categories: nasal cannula, high-flow nasal cannula [HFNC], Bilevel Positive Airway Pressure [BiPAP] or ventilator); and hospital-level factors (daily site-specific ED, intensive care unit [ICU], and inpatient census data in quartiles, and hospital discharge month to account for clinical learning that may have occurred over time).
Statistical analysis
We conducted descriptive analyses using t tests and χ 2 tests to examine bivariate associations between receipt of PC consultation and demographic, socioeconomic, clinical, illness severity, and hospital-level characteristics. Subsequently, we conducted multivariable logistic regression to estimate odds ratios (ORs) and 95% confidence intervals (CIs) to compare those who did and did not receive PC consultation. All analyses were conducted using STATA 16.0 (Stata Corp.). Two-sided p values <0.05 were considered statistically significant. The study was approved by the Icahn School of Medicine at Mount Sinai Institutional Review Board.
Results
Sample characteristics
We identified 4319 hospitalized patients with COVID-19 from March 1 to June 30, 2020 of whom 581 (14%) received PC consultation. Patients who received PC consultation had a median age of 75 years (IQR 66–85) versus 65 years (IQR 54–76) for those who did not. Furthermore, the prevalence of serious illness was higher in the group who received PC consultation (cancer, 2% vs. 8%; COPD, 66% vs. 84%; dementia, 9% vs. 25%; CAD 30% vs. 49%). Of those who received PC consultation, 24% were Black and 27% Hispanic, as compared with those who did not, 23% were Black and 34% Hispanic. The majority (73%) who received PC consultation had Medicare, and 14% were hospitalized from a SNF (Table 1).
Characteristics of Hospitalized Coronavirus Disease 2019 Patients by Receipt of Inpatient Palliative Care Consultation, and Patient, Clinical, and Hospital Predictors of Palliative Care Consultation
p value <0.05.
Crowding is measured by the percent of households with >1 occupants per room in a census tract.
BiPAP, Bilevel Positive Airway Pressure; CAD, coronary artery disease; CI, confidence interval; COPD, chronic obstructive pulmonary disease; COVID-19, coronavirus disease 2019; ED, emergency department; HFNC, high-flow nasal cannula; ICU, intensive care unit; OR, odds ratio; PC, palliative care; PPO, preferred provider organization; SNF, skilled nursing facility; SOFA, sequential organ failure assessment.
Predictors of palliative care consultation
Patient demographic, socioeconomic, and clinical characteristics
Increasing age (65–74 years, OR: 2.10, 95% CI: 1.36–3.28; 75–84 years, OR: 3.77, 95% CI: 2.36–6.03; 85+ years, OR: 7.29, 95% CI: 4.40–12.10) and number of comorbidities (OR: 1.14, 95% CI: 1.08–1.19) were associated with PC consultation. Serious illness diagnosis was associated with PC consultation (cancer, OR: 5.25, 95% CI: 3.31–8.33; COPD, OR: 1.39, 95% CI: 1.04–1.84; dementia, OR: 2.39, 95% CI: 1.78–3.21). Patients admitted from a SNF were more likely to receive PC consultation (OR: 2.10, 95% CI: 1.47–3.01). There was no association between race/ethnicity, household crowding, or insurance status and PC consultation (Table 1).
COVID-19 illness severity
Increasing illness severity predicted PC consultation (maximal oxygen support: HFNC, OR: 2.00, 95% CI: 1.26–3.16; BiPAP, OR: 3.28, 95% CI: 2.26–4.75; ventilator, OR: 2.41, 95% CI: 1.63–3.58; SOFA score: 15–50%, OR: 2.64, 95% CI: 1.95–3.57; >50%, OR: 6.07, 95% CI: 4.16–8.86). Longer hospital length of stay was associated with receipt of PC consultation (OR: 1.83, 95% CI: 1.55–2.14) (Table 1).
Hospital-level factors
Patients hospitalized at the mid-size hospitals as compared with the quaternary hospital were less likely to receive PC consultation (MSB, OR: 0.44, 95% CI: 0.33–0.60; MSC, OR: 0.67, 95% CI: 0.47–0.95). ED, ICU, and inpatient census at the time of admission and hospital discharge month were not associated with PC consultation (Table 1).
Discussion
In this retrospective cohort study of hospitalized COVID-19 patients, we found that those who received PC consultation were older, had more comorbidities, required more oxygen support, and had higher predicted mortality. There was no association between race/ethnicity, household crowding, or insurance status and PC consultation.
Despite robust PC services, there were disparities in access to PC consultation across hospital sites, suggesting that the reallocation of resources could not meet rapidly changing clinical demand. Of note, no differences were found in receipt of PC consultation between the quaternary hospital and the small community hospital that received an inpatient physician extender. However, COVID-19 patients hospitalized at the mid-size hospitals, where additional staffing was not provided, were less likely to receive PC consultation even after adjusting for patient volume as measured by daily census. Given that resource allocation is a crucial part of pandemic responsiveness,6,27,28 our experience with COVID-19 highlights ongoing learning required to ensure equitable access to PC services among hospitalized patients.
Our negative findings regarding racial/ethnic and socioeconomic disparities in access to PC consultation should be interpreted with caution. Although we adjusted for markers of socioeconomic status (insurance and household crowding), we could not comprehensively measure social determinants of health (e.g., education, primary language, and household income) nor assess consult quality. Prior studies suggest that Black and Hispanic patients experience inferior symptom management and serious illness communication.12,16,20 Implicit bias, or the unconscious reliance on negative cultural stereotypes, may account for these disparities in PC referral.29–32 Implicit bias is known to be exacerbated by time pressure and clinical uncertainty, which were ubiquitous during the pandemic. 31 Understanding disparities in the quality of PC consultation (e.g., time spent on goals of care discussion, consultation timing, symptom management, and patient and caregiver satisfaction) would provide a more complete picture of the care received by hospitalized COVID-19 patients.
Limitations
Our study has several limitations. First, generalizability may be limited as we assessed predictors of PC consultation across a single urban health system with a mature well-resourced PC program. In addition, our results reflect the peak of the pandemic in New York City, which experienced extremely high COVID-19 patient volume at a time when the evidence base for the treatment of COVID-19 did not yet exist. Furthermore, we were not able to describe whether there were disparities in the PC provided by frontline providers without the involvement of the specialized PC consultation team. In light of the rapidly changing clinical status of patients, our data (worst SOFA score during the first week of admission and highest oxygen support required during the hospitalization) may not adequately capture illness severity on the day of PC referral. Finally, the race/ethnicity data may not be self-reported in the EMR.
Conclusions and Future Directions
Our study provides new insights about the predictors of PC consultation for adults hospitalized with COVID-19. Differences in PC consultation access across hospital sites point to ongoing work required to optimally allocate these resources. Encouragingly, at a given site, the highest acuity patients were most likely to be seen by PC consultants without racial/ethnic or socioeconomic disparities. Access, however, is only one component of assessing disparities for COVID-19 patients. Future studies should examine disparities in the quality of PC consultation to ensure that all hospitalized patients with serious illness received high-quality symptom management and serious illness communication during the crisis.
Footnotes
Disclaimer
The content is solely the responsibility of the authors and does not necessarily reflect the official view of the NIH or the Veterans Administration. There are no relationships with industry.
Authors' Contributions
Concept, design, statistical analysis, and drafting of the article by J.L.F. and L.P.G. Design and editing of the article by E.C. and J.S. Methodology and statistical analysis by J.M. Design, statistical analysis, and editing of the article by M.D.A. and R.S.M.
Acknowledgments
We would like to acknowledge the extraordinary work of PC clinicians in the Brookdale Department of Geriatrics and Palliative Medicine who provided crucial support to patients with COVID-19 and their families during the peak of the pandemic in New York City.
Funding Information
L.P.G. received support from the National Institute on Aging (K23AG049930) and Cambia Health Foundation as Cambia Sojourns Leadership Scholar. R.S.M. received support from National Institute on Aging (P30AG027841 and R33AG065726) and the National Palliative Care Research Center. J.L.F. received support from the Mount Sinai Claude D. Pepper Older Americans Independence Center (P30AG027841) and the National Institute on Aging of the National Institutes of Health (T32AG066598).
Author Disclosure Statement
No competing financial interests exist.