Dementia and Early Do-Not-Resuscitate Orders Associated With Less Intensive of End-of-Life Care: A Retrospective Cohort Study

Abstract

Background:

Dementia is a leading cause of death among US older adults. Little is known about end-of-life care intensity and do-not-resuscitate orders (DNRs) among patients with dementia who die in hospital.

Aim:

Examine the relationship between dementia, DNR timing, and end-of-life care intensity.

Design:

Observational cohort study.

Setting/Participants:

Inpatient electronic health record extraction for 2,566 persons age 65 and older who died in 2 New York City hospitals in the United States from 2015 to 2017.

Results:

Multivariable logistic regression analyses modeled associations between dementia diagnosis, DNR timing, and 6 end-of-life care outcomes. 31% of subjects had a dementia diagnosis; 23% had a DNR on day of hospital admission. Patients with dementia were 18%−40% less likely to have received 4 of 6 types of intensive care (mechanical ventilation AOR: 0.82, 95%CI: 0.67 −1.00; intensive care unit admission AOR: 0.60, 95%CI: 0.49–0.83). Having a DNR on file was inversely associated with staying in the intensive care unit (AOR: 0.57, 95%CI: 0.47–0.70) and avoiding other intensive care measures. DNR placement later during the hospitalization and not having a DNR were associated with more intensive care compared to having a DNR upon admission.

Conclusions:

Having dementia and a do-not resuscitate order upon hospital admission are associated with less intensive end-of-life care. Additional research is needed to understand why persons with dementia receive less intensive care. In clinical practice, encouraging advance care planning prior to and at hospital admission may be particularly important for patients wishing to avoid intensive end-of-life care, including patients with dementia.

Introduction

Dementia is the 5th leading cause of death for persons over age 65 in the United States (US).^1,2 The dying trajectory for persons with dementia is often protracted and difficult to prognosticate,^3,4 and end-of-life care intensity in persons with dementia is not well understood. Some studies find dying persons with dementia receive more intensive care,^5,6 others find it is associated with less intensive care⁷ or a mixed picture of care intensity.^8,9 Findings vary with samples, care settings, and care intensity measures studied. No studies that we found focus on care intensity in the 10% of patients with dementia who die in US hospitals,¹⁰ where individuals are likely to receive intensive care.¹¹

Most family caregivers desire comfort care for dying loved ones with dementia.¹² This means avoiding care such as intensive care unit (ICU) admission, mechanical ventilation, intubation, cardiopulmonary resuscitation, and insertion of a feeding tube.^13–26 Among dying older adults and persons with dementia specifically, these procedures can be considered inappropriate,^27,28 do little to extend life span,^25,29,30 and can compromise quality of life,²⁸ functioning,²³ and satisfaction with care.³¹

Advance care planning, including goals of care conversations, advance directives, and physician orders for life-sustaining treatment, allows families and patients to communicate treatment preferences. Persons with dementia can complete advance planning before cognitive impairment advances and before facing end-of-life treatment decisions. Advance planning should be revisited often, including during hospitalization, to capture changes in patient and family preferences. Among dying persons with dementia, advance planning has been linked to fewer hospitalizations^12,32 and reduced ICU utilization.³³

However, advance planning documentation is not always available across care settings, preventing clinicians from having access to this information at critical decision-making points.³⁴ For hospitalized patients, a do-not-resuscitate order (DNR) is a potentially modifiable factor and the most visible marker to clinicians of a patient’s desire for comfort care. For patients with advance planning, DNRs may be recorded in the medical record upon admission. DNRs also may be placed at any point during hospitalization, often following goals of care conversations. In other hospitalized patient populations earlier goals of care conversations and DNR placement were linked to less intensive end-of-life care and better quality of life.^35,36 However, the effect of having a DNR and DNR placement timing on the relationship between dementia and end-of-life care intensity is unknown.

This study is important for two reasons. First, it adds to knowledge about end-of-life care intensity in persons with dementia by focusing on their hospital (as opposed to nursing home) deaths. As the number of persons dying with dementia grows in the coming decades, it is increasingly important to understand the needs and experiences of the 10% of this patient population that dies in the hospital in the US. Second, this study also examines how the presence and DNR placement timing affect end-of-life care intensity among persons with dementia who die in the hospital. DNR placement timing is potentially modifiable (e.g., completion upon hospital admission), and therefore scalable.

Methods

Data

This observational cohort study analyzes data extracted from inpatient electronic health records for all 3,340 individuals age 65 and older who died in two New York City hospitals from January 2015-December 2017. We eliminated 535 individuals who died from January to October 2015 because of data irregularities in ICD procedure codes for mechanical ventilation and resuscitation during this period. We also eliminated 239 individuals for whom race/ethnicity was “unknown” or “refused.” Overall results did not change in sensitivity analyses including individuals with “unknown” or “refused” race/ethnicity. We analyzed information for 2,566 individuals who died from October 2015-December 2017 and for whom race/ethnicity was identified. Because all participants were deceased, this study was deemed exempt from institutional review board approval at participating institutions in accordance with US Code of Federal Regulations 45 CFR §46.102(e).¹

Outcome Measures: Intensive End-of-Life Care

We measured intensive end-of-life care during patients’ terminal hospitalization using 6 dichotomous variables (yes = 1) documenting care commonly identified as intensive. We identified ICU admissions using the locations of patient care within the hospital. We identified mechanical ventilation,^13–17 intubation,¹⁶ resuscitation^16,18–23 and feeding tube insertion^16,24–26 using ICD9 and ICD10 procedure codes. Finally, we classified patients who experienced any of these 5 procedures as receiving any intensive care.

Dementia Diagnosis

Consistent with other studies examining administrative data,^37–39 we classified patients as having a dementia diagnosis documented during any hospitalization at either study hospital, based on ICD9 and ICD10 diagnosis codes.⁴⁰ Our analysis does not include dementia diagnoses in outpatient settings. It allows us to be intentionally inclusive of a range of dementia types, although it does not permit us to identify or control for dementia severity.

DNR and Timing of Placement

We used a dichotomous measure of DNR (yes = 1) based on whether there was a DNR in the electronic record during the terminal hospitalization. We classified timing of the first DNR placement into categories based on the date the order was placed: day of hospital admission (reference), after admission date and prior to day of death, day of death, and no DNR.

Covariates

We controlled for patient demographic and health characteristics that may be confounded with hospital death, intensive care receipt, dementia, or having a DNR: gender, race/ethnicity, age, and insurance status (e.g., Medicaid), which is means tested and a reliable indicator of need. Non-white race and younger age are associated with hospital death.⁴¹ Being male, non-white, younger, and having insurance are associated with intensive end-of-life care.^7,42–44 Being female, non-white, older, and having less education are associated with increased risk of dementia.¹ Being female, white, and older are associated with advance planning and DNR completion, while having Medicaid is associated with not having a DNR.⁴⁵ To account for complexity of patient health that might necessitate complex care or lengthier hospital stays, we constructed a Charlson Comorbidity Index using ICD9 and ICD10 diagnosis codes^46–48 and controlled for length of stay. To account for potential differences in resources and practices over time and by location, we controlled for year of patients’ death and hospital where patients received care.

Statistical Analysis

We calculated descriptive statistics and reported care intensity, DNR status, and covariates by dementia status. We used chi square tests and ANOVAs to identify differences for each. We calculated two sets of multivariable logistic regression models to estimate the association between key independent variables (dementia diagnosis, DNR) and each intensive end-of-life care measure. First we included DNR as a dichotomous variable. Second we included DNR placement timing as a four-category variable. For each set of logistic regression models we tested for multicollinearity and determined if individually including dementia and DNR significantly improved model fit using Wald chi square tests. All analyses were performed using Stata/MP 15.1.

Results

Table 1 presents descriptive statistics for the overall sample and based on dementia status. Care intensity during patients’ terminal hospitalizations differed by type of intervention. Nearly half (47%) of all patients spent time in an ICU, 43% received mechanical ventilation, and 35% were intubated. One in 10 (10%) were resuscitated. Five percent (5%) had a feeding tube initiated during hospitalization. Overall, 61% of patients received one or more of these measures. Approximately one third (31%) of patients had a dementia diagnosis and three-quarters (77%) had a DNR placed during hospitalization. With the exception of feeding tube initiation, patients with dementia received intensive care less often than expected. Patients with and without dementia had a DNR at similar rates, although patients with dementia more often had a DNR in place the day of hospital admission (30% versus 19%) and less often had a DNR put in place for the first time on the day they died (10% versus 15%). On average, patients with dementia were more often female, older, and had a higher score on the Charlson Comorbidity Index.

Table 1.

Descriptive Statistics for 2,566 Persons Age 65 and Older Who Died in Hospital, 2015–2017.

	Sample	Dementia	No dementia
	n = 2,566	n = 801	n = 1,765
	%/Mean (SD)	%/Mean (SD)	%/Mean (SD)	p value
Intensive end-of-life care
ICU admission	47.19	33.96	53.20	<0.001
Mechanical ventilation	42.56	34.71	46.12	<0.001
Intubation	35.00	26.34	38.92	<0.001
Resuscitation	10.09	8.11	10.99	0.025
Feeding tube initiated	4.91	5.24	4.76	0.60
Any intensive care	61.22	51.69	65.55	<0.001
Key independent variables
Dementia diagnosis	31.22
DNR order	77.08	77.53	76.88	0.719
Timing of DNR placement				<0.001
Upon admission (ref)	22.53	30.34	18.98
During hospitalization	41.08	37.20	42.83
Day of death	13.48	9.99	15.07
No DNR	22.92	22.47	23.12
Covariates
Race				0.40
Non-Hispanic white (ref)	52.96	54.93	52.07
Non-Hispanic Black	10.17	10.36	10.08
Asian	18.90	18.35	19.15
Other race	7.05	5.74	7.65
Hispanic	10.91	10.61	11.05
Insurance				0.19
Medicare (ref)	85.89	87.39	85.21
Medicaid	6.47	6.37	6.52
Self/Private pay/Other	7.64	6.24	8.27
Male	49.73	45.57	51.61	0.005
Age at death (r: 65–105)	81.57 (8.97)	85.78 (7.97)	79.66 (8.75)	<0.001
Hospital
Hospital A	41.50	29.46	46.97	<0.001
Hospital B	58.50	70.54	53.03
Length of stay (r: 0–367)	11.43 (15.47)	11.68 (18.29)	11.32 (14.00)	0.58
Charlson Comorbidity Index (r: 2–20)	7.40 (2.52)	7.76 (2.33)	7.24 (2.58)	<0.001
Death year				0.25
2015 (ref)	11.11	9.61	11.78
2016	44.15	44.44	44.02
2017	44.74	45.94	44.19

Notes. ICU = intensive care unit. DNR = do not resuscitate. P value determined using chi square tests for categorical variables, ANOVA test for continuous variables. Bolded values indicate p < 0.05.

Table 2 presents results from multivariable logistic regression models analyzing the association between dementia and having a DNR, and each of the 6 intensive care measures for all patients who died in the hospital. Models control for patient demographic and health characteristics, hospital, length of stay, and year of death. Compared to patients without dementia, patients with dementia were 42% less likely to be admitted to the ICU, 24% less likely to receive mechanical ventilation, 34% less likely to be intubated, and 28% less likely to receive any intensive care (ICU adjusted odds ratio (AOR): 0.58, 95% confidence interval (95%CI): 0.48–0.70; ventilation AOR: 0.76, 95%CI: 0.63–0.92; intubation AOR: 0.66, 95%CI: 0.54–0.81; any intensive care AOR: 0.72, 95%CI: 0.60–0.87). Dementia diagnosis was not significantly associated with resuscitation or feeding tube insertion. Using Wald chi square tests, including dementia significantly improves model fit for ICU admission, mechanical ventilation, intubation, and any intensive care measure, but not for resuscitation or feeding tube insertion. Having a DNR in place during hospitalization is associated with significantly lower odds of intensive care for all interventions except ICU admission. Having a DNR in place was associated with 43% greater odds of ICU admission (AOR: 1.43, 95%CI: 1.15 –1.79). Adjusting for presence of a DNR significantly improves model fit for all outcomes using Wald chi square tests.

Table 2.

Adjusted Odds Ratios and 95% Confidence Intervals for Dementia and DNR on End-of-Life Hospital Care for 2,566 Persons 65 and Older Who Died in Hospital, 2015–2017.

					Model fit
	AOR	95% CI		p value	Wald chi square	p value
Outcome 1: ICU Admission
Dementia diagnosis	0.57	0.47	0.70	<0.001	25.38	<0.001
DNR	1.43	1.15	1.79	<0.001	10.22	0.0014
Outcome 2: Mechanical ventilation
Dementia diagnosis	0.76	0.63	0.92	0.01	7.59	0.006
DNR	0.60	0.49	0.75	<0.001	21.61	<0.001
Outcome 3: Intubation
Dementia diagnosis	0.67	0.55	0.82	<0.001	15.46	0.0001
DNR	0.49	0.40	0.61	<0.001	41.9	<0.001
Outcome 4: Resuscitation
Dementia diagnosis	0.80	0.57	1.11	0.19	1.76	0.18
DNR	0.13	0.10	0.18	<0.001	172.63	<0.001
Outcome 5: Feeding tube insertion
Dementia diagnosis	1.23	0.80	1.90	0.35	0.87	0.35
DNR	0.58	0.36	0.92	0.02	5.43	0.02
Outcome 6: Any intensive care
Dementia diagnosis	0.72	0.60	0.87	<0.001	11.4	<0.001
DNR	0.51	0.40	0.64	<0.001	33.23	<0.001

Notes: ICU = intensive care unit. DNR = do-not-resuscitate order. AOR = adjusted odds ratio. CI = confidence interval. Wald chi square tests compare the improvement in model fit when dementia or DNR are included in the model. Bolded values indicate p < 0.05. All models adjust for patient race/ethnicity, insurance status, sex, age at death, hospital of care, length of stay, Charlson Comorbidity Index, and year of death.

Table 3 presents multivariable logistic regression model results analyzing the association between dementia, DNR placement timing, and the 6 intensive care measures for all patients, controlling for covariates. Similar to previous models, dementia continues to be associated with avoiding ICU admission, mechanical ventilation, intubation, and any intensive care. Compared to placing a DNR upon hospital admission, placing a DNR later during hospitalization or not having a DNR is associated with greater odds of ICU admission, mechanical ventilation, intubation, or receiving any intensive care. For example, compared to patients with a DNR upon hospital admission, patients with a DNR placed during hospitalization or on the day they die have 173% and 74% higher odds of ICU admission, respectively (DNR during hospitalization AOR: 2.73, 95%CI: 2.15–3.47; DNR on day of death AOR: 1.74, 95%CI: 1.30–2.34). Patients with no DNR have 33% higher odds of ICU admission than those with a DNR upon admission (AOR: 1.33, 95%CI: 1.01–1.75). Using Wald chi square tests, including DNR placement timing significantly improves model fit for all outcomes.

Table 3.

Adjusted Odds Ratios and 95% Confidence Intervals for Dementia and DNR Timing on End-of-Life Hospital Care for 2,566 Persons 65 and Older Who Died in Hospital, 2015–2017.

					Model fit
	AOR		95% CI	p value	Wald chi square	p value
Model 1: ICU admission
Dementia diagnosis	0.60	0.49	0.73	<0.001	25.38	<0.001
DNR placement (ref: day of admission)					79.42	<0.001
DNR mid-hospitalization	2.73	2.15	3.47	<0.001
DNR on day of death	1.74	1.30	2.34	<0.001
No DNR	1.33	1.01	1.75	0.04
Model 2: Mechanical ventilation
Dementia diagnosis	0.82	0.67	1.00	0.05	3.93	0.05
DNR placement (ref: day of admission)					92.49	<0.001
DNR mid-hospitalization	2.07	1.61	2.65	<0.001
DNR on day of death	3.63	2.69	4.90	<0.001
No DNR	3.14	2.38	4.15	<0.001
Model 3: Intubation
Dementia diagnosis	0.72	0.58	0.88	0.001	10.13	0.002
DNR placement (ref: day of admission)					114.94	<0.001
DNR mid-hospitalization	2.53	1.92	3.33	<0.001
DNR on day of death	4.16	3.02	5.73	<0.001
No DNR	4.53	3.36	6.12	<0.001
Model 4: Resuscitation
Dementia diagnosis	0.82	0.59	1.14	0.24	1.36	0.24
DNR placement (ref: day of admission)					174.69	<0.001
DNR mid-hospitalization	1.42	0.84	2.40	0.19
DNR on day of death	2.28	1.27	4.08	0.006
No DNR	10.91	6.63	17.95	<0.001
Model 5: Feeding tube insertion
Dementia diagnosis	1.27	0.82	1.97	0.28	1.17	0.28
DNR placement (ref: day of admission)					9.29	0.03
DNR mid-hospitalization	2.12	1.04	4.31	0.04
DNR on day of death	1.62	0.67	3.94	0.29
No DNR	3.05	1.44	6.47	0.004
Model 6: Any intensive care
Dementia diagnosis	0.77	0.64	0.94	0.009	6.91	0.01
DNR placement (ref: day of admission)					93.89	<0.001
DNR mid-hospitalization	2.21	1.75	2.79	<0.001
DNR on day of death	2.74	2.04	3.69	<0.001
No DNR	3.52	2.67	4.64	<0.001

Notes: ICU = intensive care unit. DNR = do-not-resuscitate order. AOR = adjusted odds ratio. CI = confidence interval. Wald chi square tests compare the improvement in model fit when dementia or timing of DNR placement are included in the model. Bolded values indicate p < 0.05. All models adjust for patient race/ethnicity, insurance status, sex, age at death, hospital of care, length of stay, Charlson Comorbidity Index, and year of death.

Discussion

Main Findings

This study examined the relationship between dementia and intensive end-of-life care receipt among patients who die in the hospital. Consistent with prior research that patients with dementia receive less intensive end-of-life care, we found patients with dementia were less likely to be admitted to the ICU or receive mechanical ventilation, intubation, or any intensive care. This association persists, regardless of whether patients have a DNR in place, the point during hospitalization at which the DNR is placed, and controlling for patient and care-related factors. At the individual patient level, there could be differences in reasons for hospital admissions that result in less intensive care for patients with dementia. For example, prior meta-analysis showed patients with dementia were admitted to hospital more often for falls, infections, or behavioral disturbances⁴⁹ which can often be controlled without ICU admission or other intensive care. The trajectory of decline during terminal hospitalization among patients with dementia, although difficult to predict, may also require less intensive care than for patients without dementia. At the family level, prior to hospitalization and regardless of advance planning completion, most family members express a desire for comfort care for dying loved ones with dementia.¹² As this preference is known by family members before critical care decisions need to be made, it may be more consistently communicated by family members to hospital clinicians, resulting in less intensive care overall, and regardless of whether and when a DNR is placed. At the broader systemic level, for decades research has shown that patients who hold greater perceived social value receive more interventions and attempts to save their lives than patients with less perceived social value.^50,51 Older age and serious illness are strong indicators of low social value⁵¹; cognitive impairment and the change in personhood that accompanies it may be an additional indicator of low social value that results in less intensive care for dying patients with dementia. Additional research is needed at the patient, family, and systemic levels to understand why patients with dementia who die in the hospital receive less intensive care than patients without dementia. From a clinical perspective, it is important to understand why patients with dementia receive less intensive care during terminal hospitalizations, and whether this is a reflection of patient and family member wishes or an unintentional difference in clinical practice.

Our analyses revealed a nuanced picture with regard to the relationship between DNRs and intensive end-of-life care. In initial analysis, having a DNR was associated with greater likelihood of ICU admissions and decreased likelihood of all other types of intensive care. The negative association between DNRs and resuscitation suggests DNRs work as intended: patients with a DNR are less likely to be resuscitated. ICU admission, mechanical ventilation, and intubation are often provided in the aftermath of resuscitation. However, supplementary analyses revealed that resuscitation did not explain the association between DNRs and these other intensive care measures. Negative associations between DNRs and mechanical ventilation, intubation, and feeding tube insertion warrant further investigation. DNRs, which are often easily located or visible in the electronic record, may simply be the most visible marker of comprehensive goals of care conversations in which patients and families express a tendency to avoid invasive or intensive care. Unfortunately, other planning documents and details from goals of care conversations are usually not easily identified or accessible to clinicians in inpatient records.⁵² Accessing planning documents may require clicking through several screens; obtaining details on goals of care, which are often dynamic and evolving, requires reading through notes from multiple providers which may be located in separate sections of the electronic record. Clinicians faced with medically urgent situations may not have time to scour the electronic record before needing to make a recommendation or decision about patient care. However, presence of a DNR should not necessarily be related to other types of intensive care. Although hospitals use more specific goals of care notes in their electronic records, these can be time consuming to document and review, placing perceived burdens on busy clinicians. Patients and families may wish to avoid resuscitation but be amenable to ICU admission, mechanical ventilation, or feeding tube use. Anecdotally, hospitalized patients and family members hesitate to implement DNRs because they think it will result in less treatment, and prior research suggests clinicians provide less intensive care in general to patients who have a DNR in their electronic record.⁵³ Regardless, from a clinical perspective, our findings underscore the importance of comprehensive and multiple goals of care conversations with hospitalized patients and their families that address multiple aspects of intensive care to ensure patients receive care consistent with their preferences and are not unintentionally denied desired care other than resuscitation. More easily accessible information on detailed patient preferences in the inpatient electronic record or on a bracelet worn by the patient may help to ensure that procedures offered and received are concordant with dying patients’ and families’ preferences.

Finally, we found DNR placement timing was also important for avoiding intensive care during terminal hospitalization. In initial analyses, having a DNR was associated with greater likelihood of ICU admission, suggesting that, for some patients, DNRs are placed after intensive care has been administered. In these cases, a DNR may reflect a desire to discontinue, rather than avoid, intensive care. This possibility was supported in further analyses examining the DNR placement timing, finding early DNR placement was related to avoiding ICU admission. Compared to having a DNR upon hospital admission, patients with DNRs placed later in the hospital stay and with no DNRs had higher odds of ICU admission. Compared to having a DNR upon hospital admission, later and no DNRs were also associated with greater odds of mechanical ventilation, intubation, resuscitation, feeding tube insertion, and any intensive care measure. Early DNR placement has also been linked to decreased suffering and better quality of life.³⁶ This finding has two implications for clinical practice. First, these findings underscore the importance of early planning for avoiding intensive care and potentially maintaining quality of life among dying hospitalized patients. Discussing patient wishes in advance is particularly important for persons with dementia, for whom cognitive decline may prevent patients from being involved in care decisions when the time comes. Second, our finding emphasizes the importance of incorporating goals of care conversations into the hospital admission process, as not doing so may represent a missed opportunity for patients wishing to avoid intensive care.

This study has limitations. While avoiding intensive end-of-life care is desired by many patients and family members, it is not a universal preference. We were unable to identify patient preferences for care from the inpatient records, limiting our ability to determine if care received was concordant with patient and family preferences. Second, our analysis was limited to information that could be extracted from the inpatient medical record in a standardized manner and was therefore limited in the amount of detail it provided. For example, we were unable to assess dementia severity, which likely varied among the patients identified as having dementia in this sample. We relied on ICD-9 and ICD-10 diagnostic codes, which is a recognized method of analyzing administrative data.⁵⁴ Thorough review of individual patient inpatient and outpatient records may provide more information and a more detailed picture of dementia severity, patient and family preferences, and care decision-making, but is beyond the scope of this study. However, our analysis provides a conservative estimate of end-of-life care of the associations between dementia, DNRs, and commonly accepted measures of intensive care. Third, we were not able to reliably control for patient nursing home residence prior to hospitalization, markers of goals of care conversations, palliative or spiritual care consultations, or referrals to hospice, which might provide additional insight about the associations we observed.

Implications

Dementia and early DNR placement were independently associated with less intensive end-of-life care among patients who died in the hospital. Additional research is needed to better understand what patient, family, and systemic factors contribute to hospitalized dementia patients receiving less intensive end-of-life care. Research is also needed to understand patient, family, and clinician perceptions of DNRs and how they relate to acceptance or avoidance of intensive care other than resuscitation. Early planning may facilitate DNR placement upon hospital admission, leading to less intensive care for patients and families preferring conservative or comfort care. Early planning also provides an opportunity for persons with dementia to express their care preferences before advanced cognitive decline precludes them from doing so. DNRs placed at later points during the hospital stays do not help avoid intensive care, but may reflect changing patient and family preferences toward comfort care in response to already-administered intensive procedures.