End-of-Life Care in Older Patients After Serious or Severe Traumatic Brain Injury in Low-Mortality Hospitals Compared With All Other Hospitals

Key Points

Question

Does end-of-life care after traumatic brain injury differ at low-mortality hospitals vs all other hospitals?

Findings

In this cohort study that included 34 691 Medicare beneficiaries with serious or severe traumatic brain injury, hospitals with the lowest in-hospital mortality performed fewer high-intensity treatments at the end of life and enrolled more patients in hospice.

Meaning

Low-mortality hospitals provide less aggressive end-of-life care to patients who die shortly after hospital discharge without increasing cumulative 30-day mortality.

Abstract

Importance

More than 80% of older patients die or are seriously impaired within 1 year after severe traumatic brain injury (TBI). Given their poor survival, information about end-of-life care is a relevant marker of high-value trauma care for these patients. In-hospital mortality is commonly used to measure quality of trauma care; however, it is not known what type of end-of-life care hospitals with the best survival outcomes provide to those who die.

Objective

To determine whether end-of-life care for older patients with TBI is correlated with in-hospital mortality.

Design, Setting, and Participants

A retrospective cohort study using 2005-2011 national Medicare claims from acute care hospitals was conducted. Medicare beneficiaries aged 65 years or older who were admitted with serious or severe TBI were included. Transferred patients, those treated at low-volume hospitals, and those who died on the date of admission were excluded. Low-mortality hospitals were those in the lowest quartile for in-hospital mortality using standardized mortality rates adjusting for age, sex, race/ethnicity, comorbidity, and injury severity. Patients at low-mortality hospitals were compared with patients at all other hospitals. The study was conducted from January 2005 to December 2011. Data analysis was conducted between August 2016 and February 2017.

Main Outcomes and Measures

End-of-life care outcomes for patients who died in hospital or 30 days or less after discharge included gastrostomy and tracheostomy placement during the TBI admission and enrollment in hospice.

Results

Of 363 hospitals included in the analysis, 91 (25.1%) were designated as low-mortality. The cohort included 34 691 patients (median age, 79 years; interquartile range, 72-84 years; 40.8% women). Of these patients, 55.8% of those at low-mortality hospitals and 62.5% at all other hospitals died in the hospital or 30 days or less after discharge (P < .01). Among patients who died in the hospital (n = 16 994), end-of-life care was similar at low-mortality hospitals and all other hospitals. For patients who survived the TBI admission and died 30 days or less after discharge (n = 4027), those at low-mortality hospitals underwent fewer gastrostomy (15.9% vs 24.0%; adjusted OR, 0.61; 95% CI, 0.52-0.72) or tracheostomy (18.2% vs 24.9%; adjusted OR, 0.71; 95% CI, 0.60-0.83) procedures and received more hospice care (66.3% vs 52.5%; adjusted OR, 1.72; 95% CI, 1.50-1.96).

Conclusions and Relevance

For older patients with serious or severe TBI, hospitals with the lowest in-hospital mortality perform fewer high-intensity treatments at the end of life and enroll more patients in hospice without increasing cumulative mortality 30 days or less after discharge.

This cohort study compares end-of-life treatment between low-mortality hospitals and all other hospitals in older patients with traumatic brain injury.

Introduction

Patients aged 65 years or older currently represent 23% of trauma admissions, which is projected to reach 39% by 2050. Older patients with trauma have increased mortality and worse recovery outcomes compared with their younger counterparts. This finding is particularly true for those with serious illness and frailty, for whom critical injury is often a terminal diagnosis. Severe traumatic brain injury (TBI) is known to have a poor prognosis for older patients, with major impairment and 1-year mortality rates in excess of 80%. As a result, trauma surgeons are increasingly tasked with delivering end-of-life care to older patients.

Reports from national organizations, including the National Academy of Medicine, have emphasized the need to improve the quality and value (patient outcomes per dollar spent) of end-of-life care; however, little is known about end-of-life care received by patients who die after TBI. This group of patients is unique in that their sudden critical illness and cognitive impairment results in major, immediate medical decisions being made by surrogates with physicians who are typically meeting these patients for the first time. Currently, in-hospital mortality is the standard by which trauma centers are benchmarked; however, use of this single metric does not account for the value some patients place on access to palliative care, avoidance of high-intensity treatments, or appropriate enrollment in hospice.

Measuring the quality of hospital care is of increasing importance for patients and other stakeholders seeking greater value. However, prior research in other areas of hospital care often fail to find agreement between disparate measures. Given the expected increase in TBI among older adults in coming decades and their high mortality, there is a pressing need to evaluate whether hospitals that are skilled at keeping patients alive also provide high-quality end-of-life care for those who do not survive. Therefore, we sought to describe treatment intensity, palliative care consultations, and hospice enrollment among older patients who died after TBI and determine whether end-of-life care in this patient population correlated with hospital rankings based on mortality.

Methods

Data Source and Study Population

This was a retrospective cohort study using a 100% sample of nationwide Medicare claims from January 2005 to December 2011. Billing claims for inpatient care were obtained from the Medicare Provider Analysis and Review file, and hospice claims were identified in the Medicare Hospice file. These claims data were linked to demographic information for beneficiaries in the Master Beneficiary Summary file using a unique beneficiary identification number. This study was approved by the Partners Human Research Committee, with waiver of informed patient consent.

We included Medicare beneficiaries aged 65 years or older who were hospitalized with 1 or more claim code corresponding with TBI (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] diagnosis codes 800-801, 803-804, 850-854, or 959.01) and were continuously enrolled in Medicare Parts A and B without health maintenance organization coverage. We identified serious or severe TBI using the ICD Program for Injury Categorization (ICDPIC), version 3.0, a statistical program that can extract injury severity scores based on ICD-9-CM codes. This software has been validated for classifying injury severity of TBI. The ICDPIC tool was therefore used to identify patients with head Abbreviated Injury Scale (AIS) scores for serious (head AIS, 3), severe (head AIS, 4), or critical (head AIS, 5) injury. Patients with maximum injury (head AIS, 6) were not included, as these scores typically represent nonsurvivable injuries. To ensure that we were identifying those with more severe injuries, we also required that patients have an intensive care unit (ICU) stay claim indicator and a procedure code for mechanical ventilation (ICD-9-CM codes 96.04, 96.70, 96.71, or 96.72) during the TBI admission. We excluded patients who were transferred from or to another acute care facility and admissions during which the patient died on the first day of hospitalization, as these latter patients may have met brain death criteria at their initial presentation. For patients with more than 1 eligible TBI admission, each admission was included as a separate event. Because we ranked hospitals based on in-hospital mortality, we followed the “rule of 20s” and excluded patients who were treated at hospitals that had fewer than 20 eligible patients die in the hospital during the study period. This criterion was chosen based on National Center for Health Statistics recommendations on the stability of rates with small numbers of observations.

Mortality

In-hospital mortality was measured as the percentage of patients who died during the TBI admission. The use of in-hospital mortality as a hospital quality benchmark makes it an intrinsically important outcome. Nonetheless, studies have demonstrated high mortality in the month following hospitalization for serious or severe TBI. Therefore, we also measured mortality 30 days or less after discharge as the percentage of patients who were discharged alive and died during the subsequent 30 days to determine differences in this outcome between hospitals with high and low in-hospital mortality. Cumulative mortality was the percentage of patients who died in the hospital or 30 days or less after discharge.

End-of-Life Care

The primary study outcome was end-of-life care measured in 2 domains: (1) intensity of treatment and (2) processes of care. Based on prior studies, hospital length of stay (LOS), ICU LOS (ICU-LOS), and gastrostomy and tracheostomy placement during the TBI admission were included as measures of intensity of treatment. Studies in nontrauma populations have shown that most patients prefer to avoid high-intensity, life-extending interventions at the end of life. Tracheostomy and gastrostomy placement have been previously used to study intensity of end-of-life care and are common procedures in patients with TBI. In a study exploring perspectives of seriously ill hospitalized patients about states of functional impairment, 67% considered requiring a breathing tube and 56% considered relying on a feeding tube to be health states that were equivalent to or worse than death. Therefore, the percentage of patients having gastrostomy placement (ICD-9-CM procedure codes 44.32, 43.1, 43.11, 43.19) or tracheostomy placement (ICD-9-CM procedure codes 31.1, 31.2, 31.21, 31.29) during the TBI admission were considered end-of-life care outcomes representing high-intensity treatment.

End-of-life processes included percentage of patients receiving palliative care specialist consultation and percentage of patients enrolled in hospice before death. These outcomes were selected because they are associated with more favorable perceptions of end-of-life care among seriously ill patients and family members. Palliative care consultations were identified during the TBI admission based on claims codes for palliative care encounters (ICD-9-CM diagnosis code V66.7). Enrollment in hospice was defined as any hospice claim prior to death or discharge location listed as hospice.

Patient Characteristics

Patient characteristics included age at the time of TBI admission, sex, race/ethnicity (white, black, Hispanic, or other), comorbidity, and injury severity. To measure comorbidity, a Charlson Comorbidity Index (CCI) category was computed for each patient, scaled from 0 to 2, with 2 representing greater comorbidity based on ICD-9-CM diagnosis codes at the TBI admission for 17 disease conditions from the Deyo et al adaptation for administrative data. Injury severity was defined using ICDPIC software to assign an Injury Severity Score (ISS) ranging from 1 to 75, which was categorized as mild (<9), moderate (9-15), severe (16-25), or profound (>25). Brain injury severity was defined as serious (head AIS, 3), severe (head AIS, 4), or critical (head AIS, 5).

Low-Mortality Hospitals

The primary independent variable was in-hospital mortality rank. We used observed-to-expected mortality ratios to benchmark hospitals based on in-hospital mortality for patients included in the cohort. To do this, we performed multivariable logistic regression analysis with in-hospital mortality as the outcome to predict the individual probability of mortality for each patient. Covariates included age, sex, race/ethnicity, CCI category, ISS, and head AIS score. The derived probability of in-hospital mortality for each patient in the cohort was summed for each hospital to determine the expected number of deaths. The observed in-hospital mortality was the number of deaths that occurred among patients at each hospital. Hospitals were then ranked into quartiles based on observed-to-expected ratios, and those in the lowest quartile were designated low-mortality hospitals. Patients at all other hospitals composed the comparison group.

Statistical Analysis

Patient characteristics and outcome variables were summarized for patients at low-mortality hospitals vs all other hospitals using medians or percentages as directed by the data and were compared using χ² tests for categorical variables and the Mann-Whitney test for continuous variables. Mortality rates in the hospital and after discharge were compared for low-mortality hospitals and all other hospitals using χ² tests. End-of-life care outcomes were analyzed among 2 analytic cohorts: (1) patients who died in the hospital (during the TBI admission) and (2) patients who died after discharge (discharged alive and died within the subsequent 30 days). The association between in-hospital mortality rank and each end-of-life care outcome was analyzed for both analytic cohorts using multivariable negative binomial regression for count outcomes (LOS and ICU-LOS) and multivariable logistic regression for binary outcomes, adjusting for age, sex, race/ethnicity, CCI category, ISS, and head AIS score. Analyses were performed using SAS/STAT, version 9.3 (SAS Institute Inc) and Stata, version 14.0 (StataCorp) where α = .05. A 2-sided P < .05 was considered statistically significant. Data analysis was conducted between August 2016 and February 2017.

Results

The cohort consisted of 34 691 patients at 363 hospitals: median age was 79 years (interquartile range, 72-84 years), 40.8% were women, and 87.9% were white. Based on standardized observed-to-expected ratios for in-hospital mortality, 91 (25.1%) hospitals were designated low-mortality hospitals. Patients at low-mortality hospitals (n = 9983) and all other hospitals (n = 24 708) had similar characteristics (Table 1).

Table 1. Patient Characteristics.

Characteristic	Low-Mortality Hospitals (n = 9983)	All Other Hospitals (n = 24 708)	P Value
Age, median (IQR), y	79 (72-84)	79 (72-84)	.09
Female, %	40.1	41.1	.11
Race/ethnicity, %
White	89.4	87.3	<.01
Black	5.7	6.7
Hispanic	1.7	1.7
Asian	1.2	2.2
Other/missing	2.0	2.1
Charlson Comorbidity Index, %
0	46.8	46.7	.91
1	28.7	28.6
2	24.5	24.7
Injury Severity Score, %
<9 (Mild)	0	0	.19
9-15 (Moderate)	29.5	29.6
16-25 (Severe)	52.2	51.3
>25 (Profound)	18.3	19.1
Head AIS score, %
3 (Serious)	41.1	41.7	<.01
4 (Severe)	48.7	48.0
5 (Critical)	9.5	10.9

Abbreviations: AIS, Abbreviated Injury Scale; IQR, interquartile range.

Mortality

Overall, 49.0% of patients died in the hospital. By definition, in-hospital mortality was lower at low-mortality hospitals: 39.8% compared with 52.7% at all other hospitals (P < .01). Among patients who were discharged alive (n = 17 697), mortality within the subsequent 30 days was 22.8% overall and was higher for low-mortality hospitals than all other hospitals (26.6% vs 20.8%, P < .01). Cumulative mortality, including in-hospital deaths and deaths 30 days or less after discharge, was 60.6% overall and was lower for patients at low-mortality hospitals compared with all other hospitals (55.8% vs 62.5%, P < .01).

End-of-Life Care

Patients Who Died in the Hospital

Among patients who died in the hospital during the TBI admission (n = 16 994), 4.0% underwent gastrostomy placement, 6.1% underwent tracheostomy placement, 17.6% had palliative care consultations, and 2% were enrolled in hospice. Data according to hospital group are presented in Table 2. After adjusting for differences associated with age, sex, race/ethnicity, CCI category, ISS, and head AIS score, patients who died in the hospital at low-mortality hospitals had a clinically small, but statistically significant, reduction in LOS and significantly lower odds of inpatient palliative care consultation than those who died in the hospital at all other hospitals (Table 3).

Table 2. End-of-Life Outcomes for Patients Who Died in the Hospital or After Discharge After Traumatic Brain Injury.

Outcome	Low-Mortality Hospitals	All Other Hospitals	P Value
Died in hospital, No.	3974	13 020	<.01
Length of stay, median (IQR), d	4 (1-9)	4 (1-9)	<.01
ICU length of stay, median (IQR), d	3 (1-7)	3 (1-7)	.26
Gastrostomy tube, %	4.0	4.0	.85
Tracheostomy tube, %	6.7	5.9	.07
Inpatient palliative care, %	14.2	18.6	<.01
Hospice enrollment, %	2.0	1.6	.10
Died ≤30 d after discharge, No.	1597	2430	<.01
Length of stay, median (IQR), d	10 (4-17)	13 (7-22)	<.01
ICU length of stay, median (IQR), d	7 (3-13)	9 (3-16)	<.01
Gastrostomy tube, %	15.9	24.0	<.01
Tracheostomy tube, %	18.2	24.9	<.01
Inpatient palliative care, %	12.2	15.2	.01
Hospice enrollment, %	66.3	52.5	<.01

Abbreviations: ICU, intensive care unit; IQR, interquartile range.

Table 3. End-of-Life Care Outcomes for Patients Who Died in the Hospital, Comparing Patients at Low-Mortality Hospitals vs All Other Hospitals^a.

Outcome	Unadjusted Ratio (95% CI)b	P Value	Adjusted Ratio (95% CI)b,c	P Value
Length of stay	0.93 (0.89-0.96)	<.01	0.94 (0.90-0.97)	<.01
ICU length of stay	0.99 (0.95-1.03)	.48	0.99 (0.95-1.03)	.66
Gastrostomy tube	0.98 (0.82-1.18)	.85	0.99 (0.82-1.18)	.88
Tracheostomy	1.14 (0.99-1.32)	.07	1.14 (0.98-1.31)	.09
Inpatient palliative care	0.72 (0.65-0.80)	<.01	0.72 (0.65-0.79)	<.01
Hospice enrollment	1.25 (0.96-1.62)	.10	1.25 (0.97-1.62)	.09

Abbreviation: ICU, intensive care unit.

^aLow-mortality hospitals, 3974 patients; all other hospitals, 13 020 patients.

^bIncidence rate ratios are reported for length of stay and ICU length of stay and represent change in number of hospitalized days associated with low-mortality hospitals. Odds ratios are reported for binary outcomes (gastrostomy tube, tracheostomy, inpatient palliative care, and hospice enrollment).

^cAdjusted for age, sex, race/ethnicity, Charlson Comorbidity Index category, Injury Severity Score, and head Abbreviated Injury Scale score.

Patients Who Died After Discharge

Among patients who survived the TBI admission and then died 30 days or less after discharge (n = 4027), 20.8% underwent gastrostomy placement, 22.2% underwent tracheostomy placement, 14.0% had inpatient palliative care consultations during the TBI admission, and 58.0% enrolled in hospice. Data according to hospital group are presented in Table 2. After adjusting for age, sex, race/ethnicity, CCI category, ISS, and head AIS score, patients who died after discharge from low-mortality hospitals had shorter LOS; shorter ICU-LOS; lower odds of gastrostomy (15.9% vs 24.0%; adjusted OR, 0.61, 95% CI, 0.52-0.72), tracheostomy (18.2% vs 24.9%; adjusted OR, 0.71; 95% CI, 0.60-0.83), and inpatient palliative care consultation and higher odds of enrolling in hospice (66.3% vs 52.5%; adjusted OR, 1.72; 95% CI, 1.50-1.96) (Table 4).

Table 4. End-of-Life Care Outcomes for Patients Who Died After Discharge, Comparing Patients at Low-Mortality Hospitals vs All Other Hospitals^a.

Outcome	Unadjusted Ratio (95% CI)b	P Value	Adjusted Ratio (95% CI)b,c	P Value
Length of stay	0.78 (0.74-0.82)	<.01	0.81 (0.77-0.85)	<.01
ICU length of stay	0.81 (0.76-0.86)	<.01	0.83 (0.78-0.89)	<.01
Gastrostomy tube	0.60 (0.51-0.71)	<.01	0.61 (0.52-0.72)	<.01
Tracheostomy	0.67 (0.57-0.78)	<.01	0.71 (0.60-0.83)	<.01
Inpatient palliative care	0.77 (0.64-0.93)	<.01	0.73 (0.60-0.88)	<.01
Hospice enrollment	1.78 (1.56-2.03)	<.01	1.72 (1.50-1.96)	<.01

Abbreviation: ICU, intensive care unit.

^aLow-mortality hospitals, 1597 patients; all other hospitals, 2430 patients.

^bIncidence rate ratios are reported for length of stay and ICU length of stay and represent change in number of hospitalized days associated with low-mortality hospitals. Odds ratios are reported for binary outcomes (gastrostomy tube, tracheostomy, inpatient palliative care, and hospice enrollment).

^cAdjusted for age, sex, race/ethnicity, Charlson Comorbidity Index category, Injury Severity Score, and head Abbreviated Injury Scale score.

Discussion

In this national retrospective study of older patients with serious or severe TBI, over 60.6% died in the hospital or within 30 days after discharge. Among patients who died within 30 days after discharge, those at low-mortality hospitals spent fewer days in the hospital or ICU during the TBI admission, underwent fewer high-intensity treatments, and received more hospice care without increasing cumulative mortality.

The results of this study confirm those of previous work demonstrating high mortality in the hospital and 30 days or less after discharge in older adults with TBI and expand the current literature by describing use of high-intensity treatments and hospice enrollment for patients who do not survive their injuries. Our finding that a greater proportion of patients died shortly after discharge from low-mortality hospitals could raise concerns that these hospitals are undertreating or prematurely discharging patients who might otherwise survive their injuries. However, cumulative mortality within 30 days or less was lower for low-mortality hospitals than all other hospitals (55.8% vs 62.5%). These data corroborate work in other seriously ill patient populations, showing that hospice does not compromise survival and underscores the need for palliative care education for surgeons, many of whom remain resistant to palliative care.

In a single-institution retrospective study, Toevs and Philp examined palliative medicine consultation in geriatric trauma patients, finding that palliative care consultation was associated with fewer gastrostomy and tracheostomy procedures for geriatric trauma patients. In contrast, we found that decedents treated at low-mortality hospitals received fewer palliative care consultations than those in other hospitals; nonetheless, they also received fewer gastrostomy and tracheostomy procedures during the TBI admission and more enrolled in hospice. A report from the Improving Palliative Care in the Intensive Care Unit Project Advisory Board and the Center to Advance Palliative Care described 2 strategies for delivering palliative care in the surgical and trauma ICU: a consultative model in which palliative care specialists are sought for patients at highest risk for poor outcomes and an integrative model in which palliative care principles are embedded in routine care for all critically ill and injured surgical patients. A study by Mosenthal et al evaluated the integrative approach in the trauma ICU and reported no change in mortality during the intervention but decreased intensity of end-of-life care for those who died. Unfortunately, administrative data used in our study did not allow us to determine whether low-mortality hospitals differed from other hospitals in terms of their processes and structure for delivering palliative care; however, these data suggest that primary palliative care, which includes discussions about prognosis, goals of care, and code status as well as basic symptom management, is delivered in low-mortality hospitals without palliative care consultation. Further research is needed to delineate optimal delivery of palliative and end-of-life care for trauma patients, as are initiatives by health systems and professional societies to broadly educate trauma surgeons in providing primary palliative care. Doing so may help to meet the public’s demand to reduce burdensome, unwanted treatments that do not restore quality of life.

Given the high short-term mortality among older patients with serious or severe TBI who can be expected to die in the hospital or within 30 days after discharge, measuring the intensity and processes of end-of-life care for those who die is important for assessing quality of care. In light of ongoing efforts to improve end-of-life care for trauma patients, including forthcoming best practice guidelines from the American College of Surgeons Trauma Quality Improvement Program, measures for end-of-life care, such as treatment intensity and hospice enrollment, are needed. Current trauma quality benchmarks based on mortality do not represent the outcome of treatment for most patients in this high-mortality cohort nor do they reflect the outcomes most meaningful to patients at the end of life. However, common metrics to assess quality in palliative care, including management of pain and dyspnea, do not account for the complexity of acute injury-related symptoms. Development and validation of end-of-life care quality metrics specific to trauma care will allow monitoring of practice patterns and trends, analysis of quality improvement initiatives, and development of measurable quality standards.

Limitations and Strengths

Limitations of this study include its retrospective design and the inability to measure other important information that may influence patient prognosis and clinical decision making using secondary data (ie, frailty and response to treatment). Nonetheless, the end-of-life outcomes were examined among patients who died during or shortly after their TBI admissions. By limiting our cohort to patients with the same outcome (death after TBI), we were able to compare variation in health care utilization in the last days of life, which has been used previously as a proxy for value. Treatment for patients with TBI is managed by a variety of clinicians, including trauma surgeons, neurosurgeons, and surgical intensivists. The primary responsibility for treatment decisions may vary between specialists and across hospitals but could not be examined in this study.

Using secondary data, we were not able to identify patients’ preferences for treatment intensity or hospice enrollment and so cannot determine whether the end-of-life care that they received was concordant with their preferences. Gastrostomy or tracheostomy may be performed for comfort or to facilitate hospital discharge; however, ICD-9-CM codes do not indicate whether procedures were performed with palliative intent. In addition, the occurrence and timing of family meetings to delineate goals of care is an important care process that we cannot identify in these data. Finally, this study is limited by the lack of validated outcomes that reflect high-quality end-of-life care in trauma and other surgical subspecialties.

Strengths of this study include its use of a large, national data set to compare patterns of end-of-life care across multiple hospitals. In addition, by using Medicare data, we were able to identify patients who died after discharge from the TBI admission, which cannot be accomplished using existing trauma registries.

Conclusions

In this study of older patients with serious or severe TBI, those who died shortly after discharge from low-mortality hospitals were less likely to receive high-intensity treatments and were more likely to enroll in hospice. Although low-mortality hospitals had greater mortality after discharge, cumulative mortality remained lower for 30 days or less after discharge. These results suggest that low-mortality hospitals may be better able to identify patients with the worst prognoses and provide less aggressive end-of-life care without increasing cumulative mortality. Further work is needed to understand the processes of care at these hospitals that allow them to achieve better outcomes.