Associations of Intellectual Disability with Cardiopulmonary Resuscitation and Endotracheal Intubation at End of Life

Abstract

Background:

Little is known about end-of-life intensive care provided to patients with intellectual disabilities (ID).

Objectives:

To identify differences in receipt of end-of-life cardiopulmonary resuscitation (CPR) and endotracheal intubation among adult patients with and without ID and examine whether do-not-resuscitate orders (DNRs) mediate associations between ID and CPR.

Design:

Exploratory matched cohort study using medical records of inpatient decedents treated between 2012 and 2018.

Results:

Patients with ID (n = 37) more frequently received CPR (37.8% vs. 21.6%) and intubation (78.4% vs. 47.8%) than patients without ID (n = 74). In multivariable models, ID was associated with receiving CPR (relative risk [RR] = 2.92, 95% confidence interval = 1.26–6.78, p = 0.012), but not intubation. Patients with ID less frequently had a DNR placed (67.6% vs. 91.9%), mediating associations between ID and CPR.

Conclusions:

In this pilot study, ID was associated with increased likelihood of receiving end-of-life CPR, likely due to lower utilization of DNRs among patients with ID. Further research is needed to confirm these results.

Introduction

People with intellectual disabilities (PwID) face multiple barriers to receiving goal-concordant end-of-life care.¹ Support staff may be hesitant to disclose terminal prognoses to PwID,² and PwID may frequently have limited involvement in making decisions about their own end-of-life care.³ In New York State, some surrogate decision makers for PwID may require legal permission to withdraw or withhold life-sustaining treatments,^4,5 potentially complicating or delaying the enactment of medical directives that limit intensive interventions at the end of life. Conversely, clinicians⁶ and the U.S. Department of Health and Human Services⁷ have expressed significant concern that people with disabilities may be unfairly denied access to life-sustaining care during the coronavirus disease 2019 (COVID-19) pandemic.

Prior research has shown that receiving more intensive interventions at the end of life is associated with decreased quality of life.⁸ Some intensive interventions, such as cardiopulmonary resuscitation (CPR) and endotracheal intubation (intubation), may result in discomfort and painful complications.^9–11 To our knowledge, no study has examined end-of-life care intensity for PwID in the United States. Because PwID may lack capacity for medical decision making,¹ have higher rates of unanticipated death,³ greater difficulty communicating care preferences,¹² and higher medicolegal hurdles to withholding or withdrawing life-sustaining treatment,^4,5 PwID may be more likely to receive intensive interventions and complete fewer do-not-resuscitate orders (DNRs) to prevent unwanted CPR. Alternatively, long-standing concerns by disability advocates, magnified by the COVID-19 pandemic, suggest that PwID may be at risk of receiving less intensive care interventions at the end of life, such as intubation.

In this pilot study, we used a retrospective matched cohort design to identify differences in receipt of CPR and intubation among adults with and without ID who died during their stay at two large urban teaching hospitals. Because our data reflected care in the pre-COVID-19 era, we hypothesized that deceased PwID would be more likely to receive these interventions than a demographically matched cohort of deceased patients without ID. We also explored whether DNR status mediated the relationship between ID and CPR.

Methods

Sample

We reviewed electronic health records of inpatients who died from 2012 to 2018 at New York Presbyterian Queens Hospital (NYP-Queens) and Weill Cornell Medical Center (NYP-WCMC). Records were eligible for review if the patient was 18 years or older and stayed at least one day. Demographic information and ICD-9/ICD-10 diagnoses were collected through automated extraction from health records; ID status and presence or absence of CPR, intubation, and DNRs were manually abstracted. Because all participants were deceased, institutional review boards at both sites determined this study was exempt from review.

Of the 9233 inpatient deaths during the study period, we identified 8494 eligible decedents. We then screened eligible decedents for evidence of ID through ICD codes and manual review, yielding 37 patients (complete screening procedures in Supplementary Data; Supplementary Figure S1). ID status was categorized by etiology of cerebral palsy, Down syndrome, other specified disorder, or unknown etiology.

Each decedent with ID was matched with two decedents without ID (n = 74) on gender, race, ethnicity, and hospital site, as these characteristics may be associated with intensive end-of-life care.^13–16 This resulted in a total sample of 111 patients.

Outcomes

We examined two dichotomous outcomes: presence or absence of CPR and intubation. To capture the full extent of these interventions performed near death, we included procedures performed while en route to the hospital. We chose to examine these outcomes because they are intensive procedures commonly targeted in advance care planning (i.e., through do-not-intubate orders and DNRs).

Covariates

Charlson Comorbidity Index (CCI) scores were calculated using ICD-9/ICD-10 diagnoses.¹⁷ To account for acute conditions that commonly require intensive care and are not included in the CCI, we created indicator variables of acute kidney injury, sepsis, acute respiratory distress syndrome (ARDS), and shock using ICD-9/ICD-10 diagnoses (codes in Supplementary Data). We also included age and length of stay as covariates in adjusted models.

DNR status

No versus any DNR placed during the terminal hospitalization was measured dichotomously.

Analysis

First, we compared demographic and clinical characteristics of patients with and without ID using descriptive statistics, Wilcoxon signed-rank tests, and Fisher's exact tests. Next, we examined between-group differences in our outcome variables using descriptive statistics and unadjusted risk ratios. Outcomes were then analyzed using multivariable generalized estimating equations with log link, Poisson distribution, and robust standard errors, adjusting for age, length of stay, and CCI scores (measured continuously) and presence of acute kidney injury, sepsis, ARDS, and shock (measured dichotomously).

We also examined whether DNR mediated the association between ID and CPR. First, we compared differences between patients with and without ID on DNR status using descriptive statistics and Fisher's exact tests. Then, we fit multivariable logistic regression models that included the aforementioned covariates for the a path, that is, ID on DNR, and the b path, that is, ID on CPR adjusting for DNR. Unstandardized coefficients from these models were used to calculate a product-method estimation of mediation, or a*b. Because the high frequency of CPR and DNR prevented us from using log-binomial models, this estimate was used to evaluate presence or absence of mediation rather than to provide an effect size estimate or to determine degree of statistical significance.¹⁸ We then confirmed our initial approach using Baron and Kenny's criteria for mediation.¹⁹ Two-sided tests were used with p < 0.05 representing statistical significance. All analyses were completed using SPSS 27.

Results

Patient characteristics

ID etiologies included cerebral palsy (6/37), Down syndrome (6/37), other specified disorders (9/37), and unknown etiology (22/37), allowing multiple diagnoses per patient. Patients with ID had a younger median age of death (53 vs. 81.5 years old, p < 0.001), lower median scores on the CCI (1 vs. 5, p < 0.001), and more frequent diagnoses of ARDS (11/37, 29.7% vs. 9/74, 12.2%, p = 0.035) and shock (10/37, 27% vs. 7/74, 9.5%, p = 0.024) (Table 1).

Table 1.

Demographic and Clinical Characteristics of Intellectual Disability Group and Matched Cohort

	Total sample (N = 111)		Intellectual disability (n = 37)		Matched cohort (n = 74)		Fisher's exact
Matched demographic characteristics	n	%	n	%	n	%	p
Campus							—
NYP-WCMC	48	43.2	16	43.2	32	43.2
NYP-Queens	63	56.8	21	56.8	42	56.8
Gender							—
Male	45	40.5	15	40.5	30	40.5
Female	66	59.5	22	59.5	44	59.5
Race							—
White	39	35.1	13	35.1	26	35.1
Black	15	13.5	5	13.5	10	13.5
Asian	15	13.5	5	13.5	10	13.5
Other not described	24	21.6	8	21.6	16	21.6
Declined	18	16.2	6	16.2	12	16.2
Ethnicity							—
Non-Hispanic	63	56.8	21	56.8	42	56.8
Hispanic	6	5.4	2	5.4	4	5.4
Missing	42	37.8	14	37.8	28	37.8
Unmatched continuous characteristics	Median	IQR	Median	IQR	Median	IQR	pa
Age (years)	74	(60, 86)	53	(28.5, 66.5)	81.5	(70.75, 89)	<0.001
Length of stay (days)	9	(2, 15)	11	(3, 17.5)	7	(2, 15)	0.103
CCI	3	(1, 6)	1	(0, 3)	5	(3, 7)	<0.001
Intensive care diagnoses	n	%	n	%	n	%	p
Acute kidney injury	52	46.8	16	43.2	36	48.6	0.688
Sepsis	44	39.6	18	48.6	26	35.1	0.217
ARDS	20	18	11	29.7	9	12.2	0.035
Shock	17	15.3	10	27	7	9.5	0.024

^a Wilcoxon signed-rank test used for continuous variables.

ARDS, acute respiratory distress syndrome; CCI, Charlson Comorbidity Index; IQR, interquartile range; NYP-Queens, New York Presbyterian Queens Hospital, NYP-WCMC, New York Presbyterian Weill Cornell Medical Center.

Unadjusted analyses

14/37 (38%) of patients with ID received CPR, compared with 16/74 (22%) of patients without ID (RR = 1.75, 95% confidence interval [CI] = 0.96–3.19, p = 0.067). Patients with ID more frequently received intubation (78.4% vs. 47.8%, RR = 1.64, 95% CI = 1.22–2.21, p = 0.001) (Table 2).

Table 2.

End-of-Life Care Outcomes among Patients with Intellectual Disabilities and Matched Cohort

	Total sample (N = 111)		Intellectual disability (n = 37)		Matched cohort (n = 74)		Relative risk
Outcome	n	%	n	%	n	%	RR	95% CI	p
CPR	30	27.0	14	37.8	16	21.6	1.75	0.96–3.19	0.067
Endotracheal intubationa	62	58.5	29	78.4	33	47.8	1.64	1.22–2.21	0.001

^a Excludes participants with tracheostomy tube at admission for a total N of 106, including 37 patients with ID and 69 patients without ID.

CI, confidence interval; CPR, cardiopulmonary resuscitation; ID, intellectual disabilities; RR, relative risk.

Adjusted analyses

In multivariable models, ID was associated with increased likelihood of CPR (RR = 2.92, 95% CI = 1.26–6.78, p = 0.012), but not intubation (RR = 0.96, 95% CI = 0.60–1.53, p = 0.858) (Table 3).

Table 3.

Generalized Estimating Equation Models Adjusting for Clinical Factors (N = 111)

	CPR			Endotracheal intubationa
Variable	RR	95% CI	p	RR	95% CI	p
Intellectual disability	2.92	1.26–6.78	0.012	0.96	0.60–1.53	0.858
Age	1.00	0.99–1.02	0.645	0.99	0.98–1.00	0.047
CCI	1.06	0.93–1.20	0.384	0.98	0.92–1.05	0.505
Length of stay	1.00	0.97–1.03	0.941	1.01	1.00–1.02	0.018
Acute kidney injury	0.89	0.43–1.85	0.756	1.21	0.87–1.70	0.258
Sepsis	0.55	0.27–1.12	0.099	1.42	0.98–2.05	0.063
ARDS	0.61	0.27–1.41	0.251	1.60	1.12–2.29	0.009
Shock	0.62	0.26–1.47	0.277	1.18	0.86–1.62	0.301

^a Excluded patients with tracheostomy tube at admission, n = 106.

Mediation analyses

Patients with ID had fewer DNRs (25/37, 67.6% vs. 68/74, 91.9%, p = 0.002). The product-method estimation of the indirect effect of DNRs on the association between ID and CPR (10.2) suggested that DNR status may mediate the association between ID and CPR (Supplementary Tables S1 and S2). These results conformed with Baron and Kenny's criteria for full mediation.

Discussion

In this pilot study, we found that patients with ID were more likely to receive potentially burdensome end-of-life care in the form of CPR. This may have been, in part, due to fewer DNRs among patients with ID. These results may be explained by the barriers in end-of-life decision making experienced by PwID¹ resulting in less utilization of DNRs and more frequent provision of CPR as the “default” treatment. Patients in our matched cohort may also have been more likely to have previously had capacity to appoint a surrogate decision maker and share their preferences about end-of-life care, as well as greater forewarning to do so due to higher comorbidity burden and older age. Future research is needed to confirm these results and understand their relationship to goal concordance of care and quality of life at the end of life.

We also found that ID was associated with increased likelihood of intubation in unadjusted, but not adjusted, analyses. These results are consistent with higher mortality rates from respiratory illnesses among PwID,^20,21 suggesting intubation for mechanical ventilation may be more frequently indicated in this population. After accounting for this by including ARDS as a covariate in our models, the magnitude of the association between ID and intubation decreased. This may also explain why differences were found for CPR, but not intubation, in multivariable models; differences in CPR may have been due to disparities in advance care planning, whereas the initial observed differences in intubation may have been the result of clinical differences between groups.

Overall, our results align with previous research documenting disparities in end-of-life care for PwID. In the United Kingdom, problems in advance care planning more frequently contributed to avoidable death among PwID than those without ID.²² In the Netherlands, terminally ill cancer patients with ID were less likely to access specialist care than those without ID.²³ Our results contrast with a more recent study from the United Kingdom, which found that PwID admitted to hospitals due to COVID-19 in the first months of the pandemic were less likely to be intubated than a matched cohort despite having more frequent respiratory symptoms.²⁴

Advocacy organizations²⁵ and clinicians⁶ have also expressed serious concerns that PwID may be at risk for unfairly being denied life-sustaining care; our results should not delegitimize these concerns. Viewed together, however, this may represent how disparities in care for PwID may vary based on the health system and context in which they occur. End-of-life care provided for PwID must prioritize dignity and autonomy, be clinically appropriate, and align with patient values.

Our exploratory study has limitations. Our sample was small and restricted to one city and hospital system, although differences between our groups were large enough to generate statistically significant results after matching. Our sample size also precluded adequately powered subanalyses based on severity of ID. In addition, we did not match cohorts based on age or chronic/acute illnesses. As such, our ID group had a markedly lower median age of death, reflective of general trends among people with and without ID,²² as well as greater prevalence of ARDS and shock and lower CCI scores, corresponding with higher rates of unanticipated death among this population.³

We chose not to match on these characteristics as doing so would have likely resulted in a comparator group unrepresentative of most inpatient decedents without ID and prevented us from investigating how the unique clinical context in which PwID in our sample died (i.e., younger, with more acute illnesses and different comorbidity burden) may have contributed to more intensive end-of-life care and less advance care planning. Using these characteristics as covariates rather than matching variables also allowed us to demonstrate how age and illness may have influenced some, but not all, of our outcomes by comparing adjusted and unadjusted results.

Finally, our results may not be representative of care provided to patients whom clinicians expected to die but survived the study period,²⁶ nor do they represent end-of-life care provided in nonhospital settings such as homes or inpatient hospice facilities, where less intensive end-of-life care may be delivered, and which may be underutilized by PwID.²⁷

Conclusion

Among a small inpatient sample, we found that PwID more frequently received CPR at the end of life, potentially due to lower utilization of DNRs. Further studies should examine the quality and intensity of end-of-life care provided to PwID.

Authors' Contributions

Conception or design of the study by M.V., E.A.L., and H.G.P. Data collection by M.V., J.G., and Z.A. Data analysis and interpretation by M.V., R.T.B., E.A.L., and H.G.P. Drafting the article by M.V., R.T.B., and H.G.P. Critical revision of the article by E.A.L., C.X.P., L.L., J.G., Z.A., and P.D. All authors read and approved the final report.

Funding Information

This study was supported by the WITH Foundation, the National Cancer Institute (CA197730), and the National Institute on Aging (AG065624, 1K01AG075246).

Author Disclosure Statement

No competing financial interests exist.

Supplementary Material

Supplementary Data

Supplementary Figure S1

Supplementary Table S1

Supplementary Table S2