Abstract
Objectives:
Do-not-resuscitate (DNR) orders are used to express patient preferences for cardiopulmonary resuscitation. This study examined whether early DNR orders are associated with differences in treatments and outcomes among patients hospitalized with pneumonia.
Methods:
This is a retrospective cohort study of 768,015 adult patients hospitalized with pneumonia from 2010 to 2015 in 646 US hospitals. The exposure was DNR orders present on admission. Secondary analyses stratified patients by predicted in-hospital mortality. Main outcomes included in-hospital mortality, length of stay, cost, intensive care admission, invasive mechanical ventilation, noninvasive ventilation, vasopressors, and dialysis initiation.
Results:
Of 768,015 patients, 94,155 (12.3%) had an early DNR order. Compared with those without, patients with DNR orders were older (mean age 80.1 ± 10.6 years vs 67.8 ± 16.4 years), with higher comorbidity burden, intensive care use (31.6% vs 30.6%), and in-hospital mortality (28.2% vs 8.5%). After adjustment via propensity score weighting, these patients had higher mortality (odds ratio [OR] 2.39, 95% confidence interval [CI] 2.33–2.45) and lower use of intensive therapies such as vasopressors (OR 0.83, 95% CI 0.81–0.85) and invasive mechanical ventilation (OR 0.68, 95% CI 0.66–0.70). Although there was little relationship between predicted mortality and DNR orders, among those with highest predicted mortality, DNR orders were associated with lower intensive care use compared with those without (66.7% vs 80.8%).
Conclusions:
Patients with early DNR orders have higher in-hospital mortality rates than those without, but often receive intensive care. These orders have the most impact on the care of patients with the highest mortality risk.
Keywords: do-not-resuscitate orders, end-of-life, pneumonia, resuscitation, terminal care
Do-not-resuscitate (DNR) orders are used to express patient preferences and guide management. They specify only whether an individual wants to receive cardiopulmonary resuscitation (CPR) if needed, but they should not limit other treatment. When used appropriately, DNR orders can improve the quality of life near the end of life and reduce patient suffering.1,2 When used inappropriately, however, DNR orders can be discordant with patient wishes.3 In many hospitals, conversations about resuscitation preferences are a routine part of the admission process.4,5
More than 1.5 million adults are hospitalized for pneumonia each year, and the incidence increases with age.6 DNR orders are common among patients with pneumonia, and in-hospital mortality is often attributable to underlying comorbidities, such as cancer and neurological conditions, rather than to pneumonia itself.7,8 In one study, 15% of the patients admitted with pneumonia had DNR orders within 24 hours of admission, and most deaths occurred among these patients.9 Most patients with pneumonia never require CPR, however; in one cohort, <3% of patients admitted with pneumonia experienced cardiac arrest.10
Although DNR orders should apply only to CPR, they may signal a patient’s preference to avoid other intensive treatments such as dialysis or feeding tubes. Such preferences may be real or projected, raising concern that DNR may be misinterpreted as “do-not-treat.”11 Early DNR orders are associated with fewer procedural interventions and higher in-hospital mortality in patients with sepsis and surgical patients.12,13 Patients with these orders may choose to forgo futile care, leading to lower resource utilization. Indeed, for patients who die in the hospital, cost is lower if there is a DNR order on admission.14
It is challenging to study the impact of DNR orders on care because sicker patients are more likely to have these orders and to have poor outcomes. Presumably, DNR orders would most affect the care of those at highest risk of death, but it is not known whether such patients are likely to have them, or whether physicians preferentially approach high-risk patients to obtain advance directives. Because hospitals are usually reimbursed a flat fee for inpatient care via diagnosis-related groups, they should benefit financially from reductions in resource use. Patients, too, may wish to avoid high bills for futile care—costs at the end of life have been identified as a healthcare disparity.15 Using a highly detailed, national hospital discharge administrative database, we examined whether early DNR orders are associated with treatment and outcomes in pneumonia. We also examined whether rates of early DNR orders were associated with predicted mortality risk across hospitals.
Methods
Study Sample
We included patients aged 18 years or older who were hospitalized for pneumonia from 2010 to 2015 at 646 US hospitals that participated in the Premier Healthcare Database. We identified patients with a principal diagnosis of pneumonia present-on-admission or a principal diagnosis of respiratory failure, sepsis, or influenza with a secondary diagnosis of pneumonia present-on-admission (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM]).A For inclusion, patients also must have had a chest x-ray or computed tomography (CT) scan by hospital day 1, as well as antibiotic treatment initiated by hospital day 1 and continued for at least 3 consecutive days, unless they died or were discharged sooner. We excluded patients transferred from an outside hospital and those with cystic fibrosis or ventilator-associated pneumonia. For patients with multiple qualifying admissions, one admission was selected at random for inclusion.
Exposures
The primary exposure was a DNR order present on admission, identified by ICD-9 code V49.86, which has been shown to have high specificity and moderate sensitivity for DNR orders.16 These included orders written before or upon admission to the hospital and that were entered into the electronic health record by the healthcare team. Patients with DNR orders later in the admission were analyzed together with those with no DNR orders.
Outcomes
Outcomes included in-hospital mortality, length of stay, and cost; intensive care admission, length of stay, and cost; and receipt of invasive mechanical ventilation (IMV), noninvasive ventilation (NIV), vasopressors, bronchoscopy, magnetic resonance imaging (MRI)/CT scans, central line placement, and dialysis initiated during hospitalization. These interventions were chosen to represent maximal efforts at treatment. In-hospital mortality also included patients discharged to inpatient hospice. Costs reported by Premier are either the hospital’s report of its actual costs associated with the services provided or estimated costs based on charges and the hospital’s cost-to-charge ratios.
Concomitant Variables
Concomitant variables controlled for as potential confounders included age at admission, sex, race, and insurance payor. Race was defined by the hospital. Comorbidities present on admission also were included as covariates and were defined by ICD-9 codes. Organ failure scores on admission were included to account for severity of illness and were determined using ICD-9 codes.17 We also controlled for several characteristics of admitting hospitals: number of beds (≤200, 201–400, ≥400), teaching (vs nonteaching), US Census Bureau regions (Midwest, Northeast, South, West), and urban (vs rural) location.
Statistical Analysis
We compared baseline demographics, comorbidities, hospital characteristics, and hospitalization outcomes for patients with DNR orders present on admission with those without, using the Pearson χ2 test for categorical and the Student t test for continuous variables.
We created a multiple logistic propensity model regressing DNR orders on age, sex, race, comorbidities, organ failures, and hospital characteristics, including regional location, rural or urban location, teaching status, and size, and assessed model performance using a balance plot18 (this plot, Supplemental Figure 1, identifies the specific predictors;). We then estimated the average effects of DNR orders among patients with such orders, adjusting for these variables, on the following outcomes: in-hospital mortality including those discharged to inpatient hospice, hospital length of stay, intensive care unit (ICU) admission, ICU length of stay, hospital cost, ICU cost, receipt of IMV, vasopressors, bronchoscopy, central or arterial lines, any MRI or CT scan, and initiation of dialysis on day 3 or later of hospitalization. We obtained these adjusted estimates by fitting inverse probability-weighted additive multiple logistic models for the dichotomous outcomes, log-link gamma generalized linear models for costs, and negative binomial regression models for length-of-stays, using the average effect of the treatment on the treated propensity-based weights and without additional covariate adjustment.
To investigate the possible modification of the associations of DNR with outcomes by disease severity at admission, we stratified patients by predicted mortality. The predicted mortality model has been described previously and uses administrative data available in the first two hospital days, with good discrimination for in-hospital mortality prediction.19 Patients were stratified into quartiles by predicted mortality, and we then compared outcomes within groups based on DNR status. The top mortality quartile was further subdivided into tertiles to demonstrate trends among very-high-risk patients.
Finally, to study hospital variation in DNR rates, the ratio of observed DNR orders to expected DNR orders was calculated for each hospital. Expected rates of DNR were determined using the multiple logistic propensity model. Hospitals were stratified into quartiles based on their ratios, with those in the highest quartile having a greater proportion of patients with early DNR orders than expected based on the model. Within each quartile, we reported rates of DNR orders based on quintile of predicted mortality based on a previously published mortality model.19
All of the statistical analyses were performed using SAS 9.4 (SAS Institute, Cary, NC). This study was approved by the Cleveland Clinic institutional review board, procedures were followed in accordance with the ethical standards of the review board and with the 1975 Declaration of Helsinki, and a consent waiver was granted.
Results
We included 768,015 patients admitted for pneumonia from 2010 to 2015; 94,155 (12.3%) had a DNR order present on admission. There were 31,258 (4.1%) patients with DNR orders later in admission; these patients were then included in the unexposed group. Patient demographics and comorbidities stratified by DNR status at admission are shown in Table 1. Patients with DNR orders were older (mean age 80.1 ± 10.6 years vs 67.8 ± 16.4 years), more often female (55.1% vs 51.0%), more often White (79.8% vs 73.7%), and more frequently insured through Medicare (89.9% vs 68.5%). Those with DNR orders had higher rates of some comorbidities, including congestive heart failure (36.2% vs 25.7%), dementia (27.5% vs 11.4%) and other neurological disorders (26.8% vs 14.6%), but had lower rates of others such as chronic pulmonary disease (40.9% vs 46.0%). Cases with DNR orders had more organ failure on admission than those without. DNR orders were more common in the Northeast and West than in the South or Midwest (Supplemental Table 1). All of these comparisons were highly statistically significant (P < 0.001).
Table 1.
Patient demographics, comorbidities, and hospital characteristics stratified by DNR status at admission
| Factor | DNR present on admission (n = 94,155) | No DNR at admission (n = 673,860) |
|---|---|---|
| Age, y, mean ± SD | 80.1±10.6 | 67.8±16.4 |
| Sex, female (%) | 51,864 (55.1) | 343,630 (51.0) |
| Race, no. (%) | ||
| White | 75,147 (79.8) | 496,945 (73.7) |
| Black | 5623 (6.0) | 77,779 (11.5) |
| Other | 13,385 (14.2) | 99,136 (14.7) |
| Insurance, no. (%) | ||
| Medicare | 84,645 (89.9) | 461,370 (68.5) |
| Medicaid | 2975 (3.2) | 64,625 (9.6) |
| Managed care | 3263 (3.5) | 81,501 (12.1) |
| Commercial indemnity | 1375 (1.5) | 20,622 (3.1) |
| Others | 1897 (2.0) | 45,742 (6.8) |
| Admission source, no. (%) | ||
| Emergency department | 75,854 (80.6) | 593,918 (88.1) |
| SNF/ICF | 15,119 (16.1) | 48,005 (7.1) |
| Clinic | 3,150 (3.3) | 31,315 (4.6) |
| Comorbidities | ||
| Anemia, no. (%) | 32,094 (34.1) | 193,921 (28.8) |
| Cancer, no. (%) | 6502 (6.9) | 27,522 (4.1) |
| Chronic pulmonary disease, no. (%) | 38,509 (40.9) | 309,988 (46.0) |
| Coagulopathy, no. (%) | 8181 (8.7) | 50,340 (7.5) |
| Congestive heart failure, no. (%) | 34,088 (36.2) | 173,250 (25.7) |
| Deficiency anemias, no. (%) | 31,396 (33.4) | 176,475 (27.6) |
| Dementia, no. (%) | 25,922 (27.6) | 70,872 (11.1) |
| Depression, no. (%) | 16,102 (17.1) | 91,325 (14.3) |
| Diabetes mellitus, no. (%) | 13,668 (14.5) | 92,956 (13.8) |
| Fluid and electrolyte disorders, no. (%) | 42,494 (45.1) | 240,366 (35.7) |
| Hypertension, no. (%) | 65,376 (69.4) | 428,504 (63.6) |
| Hypothyroidism, no. (%) | 6476 (6.9) | 29,876 (4.4) |
| Immunosuppressed, no. (%) | 11,811 (12.5) | 90,059 (13.4) |
| Obesity, no. (%) | 6736 (7.2) | 94,326 (14.0) |
| Other neurological disorders, no. (%) | 25,209 (26.8) | 98,307 (14.6) |
| Paralysis, no. (%) | 6727 (7.1) | 28,899 (4.3) |
| Peripheral vascular disease, no. (%) | 10,171 (10.8) | 54,663 (8.1) |
| Psychoses, no. (%) | 4823 (5.1) | 38,675 (5.7) |
| Pulmonary circulation disease, no. (%) | 9058 (9.6) | 52,407 (7.8) |
| Valvular disease, no. (%) | 12,150 (12.9) | 62,125 (9.2) |
| Weight loss, no. (%) | 11,720 (12.4) | 50,572 (7.5) |
| Organ failure on admission | ||
| Respiratory failure, no. (%) | 27,475 (29.2) | 149,777 (22.2) |
| Cardiovascular failure, no. (%) | 10,680 (11.3) | 54,703 (8.1) |
| Renal failure, no. (%) | 38,739 (41.1) | 211,418 (31.4) |
| Metabolic failure, no. (%) | 10,125 (10.8) | 51,911 (7.7) |
| Neurologic failure, no. (%) | 16,706 (17.7) | 59,449 (8.8) |
| Organ failure score, mean ± SD | 1.1 ± 1.08 | 0.80 ± 0.99 |
Groups were compared for statistical significance using the Pearson χ2 test for categorical variables and the Student t test for age. All of the comparisons were statistically significant, with P < 0.001, except for urban vs rural hospital location (P = 0.57) and the prevalences of AIDS (P = 0.02), lymphoma (P = 0.02), hepatic failure (P = 0.07), and peptic ulcer disease with bleeding (P = 0.07). AIDS, acquired immunodeficiency syndrome; DNR, do not resuscitate; ICF, intermediate care facility; SD, standard deviation; SNF, skilled nursing facility.
Patients with DNR orders had higher in-hospital mortality than those without (28.2% vs 8.5%) (Table 2). They went to the ICU slightly more frequently (31.6% vs. 30.6%) and were treated with NIV more often (21.5% vs 16.0%) but had shorter ICU stays (mean days 4.5 ± 4.4 vs 5.8 ±6.6) and were less likely to receive IMV (11.9% vs 14.8%). They had higher median ($9657 vs $8461) costs, but lower mean ($13,269 vs $14,073) costs (P < 0.001 for all of these comparisons).
Table 2.
In-hospital outcomes stratified by DNR status at admission
| Factor | DNR present on admission (n = 94,155) | No DNR at admission (n = 673,860) |
|---|---|---|
| ICU, no. (%) | 29,762 (31.6) | 206,326 (30.6) |
| Any ventilation, no. (%) | 27,112 (28.8) | 166,223 (24.7) |
| IMV, no. (%) | 11,228 (11.9) | 100,002 (14.8) |
| NIV, no. (%) | 20,208 (21.5) | 107,504 (16.0) |
| Vasopressor use, no. (%) | 12,231 (13.0) | 85,514 (12.7) |
| In-hospital mortality, no. (%) | 26,519 (28.2) | 57,134 (8.5) |
| Length of stay, mean ± SD | 6.7 ± 5.2 | 6.8 ± 6.9 |
| Length of stay, median [Q1, Q3] | 5.0 [3.0, 8.0] | 5.0 [3.0, 8.0] |
| Total cost, mean ± SD | 13,269 ± 13,869 | 14,073 ± 19,573 |
| Total cost, median [Q1, Q3] | 9657 [6123, 15,771] | 8461 [5149, 15,363] |
| Cost per day, mean ± SD | 2021 ± 1071 | 1963 ± 1095 |
| Cost per day, median [Q1, Q3] | 1814 [1427, 2376] | 1754 [1376, 2300] |
| ICU length of stay, mean ± SD | 4.5 ± 4.4 | 5.8 ± 6.6 |
| ICU length of stay, median [Q1, Q3] | 3.0 [2.0, 6.0] | 4.0 [2.0, 7.0] |
| ICU cost, mean ± SD | 13,517 ± 16,379 | 18,055.7 ± 24,336.5 |
| ICU cost, median [Q1, Q3] | 9058 [5172, 16,246] | 10,386 [5400, 21,218] |
| ICU cost per day, mean ± SD | 3098±1585 | 3123±1714 |
| ICU cost per day, median [Q1, Q3] | 2898 [2261, 3666] | 2900 [2233, 3706] |
| Central/arterial lines | 2232 (2.4) | 17,513 (2.6) |
| Bronchoscopy | 1886 (2.0) | 28,635 (4.2) |
| MRI/CT scan | 3538 (3.8) | 35,239 (5.2) |
| Dialysis after day 3, no. (%) | 496 (0.54) | 5917 (0.92) |
Groups were compared for statistical significance using the Pearson χ2 test for categorical variables and the Wilcoxon rank sum test for lengths of stay and costs. All of the comparisons were statistically significant, with P < 0.001, except for vasopressor use (P = 0.01) and ICU cost per day (P = 0.94). CT, computed tomography; DNR, do not resuscitate; ICU, intensive care unit; IMV, invasive mechanical ventilation; MRI, magnetic resonance imaging; NIV, noninvasive ventilation; SD, standard deviation.
After adjusting for demographics, comorbidities, organ failure scores, and hospital characteristics, those with DNR orders had higher mortality (odds ratio [OR] 2.39, 95% confidence interval [CI] 2.33–2.45) and were more likely to receive noninvasive ventilatory support (OR 1.39, 95% CI 1.36–1.43) than those without a DNR order (Fig. 1). They were slightly less likely to be treated in an ICU (OR 0.95, 95% CI 0.93–0.97) and much less likely to receive IMV (OR 0.68, 95% CI 0.66–0.70), vasopressors (OR 0.83, 95% CI 0.81–0.85), bronchoscopy (OR 0.56, 95% CI 0.53–0.60), MRI or CT scans (OR 0.68, 95% CI 0.65–0.71), and central lines (OR 0.76, 95% CI 0.71–0.81), and to initiate dialysis (OR 0.64, 95% CI 0.57–0.71). Patients with DNR orders had shorter lengths of stay (OR 0.91, 95% CI 0.91–0.92), particularly in the ICU (OR 0.78, 95% CI 0.78–0.79). Their total hospitalization and ICU costs were lower, but the cost per day for hospital stay and ICU stay was similar between the groups. All of the associations were highly statistically significant (P < 0.001).
Fig. 1.
Adjusted associations of outcomes with DNR status on admission. CT, computed tomography; DNR, do not resuscitate; ICU, intensive care unit; IMV, invasive mechanical ventilation; MRI, magnetic resonance imaging; NIV, noninvasive mechanical ventilation; OR, odds ratio.
There was little consistent relationship between predicted mortality and having a DNR order (Supplemental Figure 2). Within each quartile of predicted mortality, patients with DNR orders had higher mortality than those without, but resource utilization was generally lower for those with a DNR order (Fig. 2). As mortality rose, patients with DNR orders went from using more resources than other patients to using less than others. In the lowest predicted mortality group, patients with DNR orders stayed longer in the hospital (6.0 ± 4.9 days vs 5.1 ± 4.8 days, P < 0.001) and had higher costs than those without DNR orders, whereas in the highest predicted mortality group, patients with DNR orders had shorter lengths of stay than those without (7.7 ± 6.9 days vs 10.9 ± 10.0 days, P < 0.001) (Supplemental Table 2). They also went to the ICU less frequently (66.7% vs 80.8%, P < 0.001) and had much lower costs.
Fig. 2.
Difference in outcomes between those with DNR and those without, for the patients in the lower three quartiles and upper three-twelfths of the distribution of predicted mortality. DNR, do not resuscitate; ICU, intensive care unit; ref, reference.
The distribution of observed DNR rates across hospitals is shown in Supplemental Figure 3. For most hospitals, fewer than 20% of admitted patients had early DNR orders. Figure 3 shows DNR rates grouped by hospital-level quartiles of DNR rates, further stratified by quintiles of predicted mortality. Within each hospital quartile, patients with high predicted mortality were more likely to have a DNR order present on admission; however, this trend became much more pronounced at hospitals with the highest rates of DNR orders.
Fig. 3.
Rates of early DNR orders grouped by hospital-level quartiles of DNR rates, further stratified by quintiles of predicted mortality. DNR, do not resuscitate.
Discussion
In this retrospective cohort study, we found that approximately 12% of patients admitted for pneumonia had DNR orders present on admission and that these were associated with markedly higher in-hospital mortality even after adjusting for differences in patient characteristics and measures of comorbidity and severity of illness. We found that those with early DNR orders were older and had higher rates of neurological disease than patients without these orders, but lower rates of chronic obstructive pulmonary disease. Early DNR orders were associated with less intensive treatment overall, but a high proportion of patients still received intensive therapies. Patients with early DNR orders were only slightly less likely to be treated in the ICU, with almost one-third admitted to ICU during their admission. Costs per ICU-day suggested that they received similar intensity of treatment while there. The high rate of NIV also suggests an attempt to treat respiratory failure aggressively, at least initially. Their ICU stays, however, were shorter than those of patients without DNR orders, suggesting that intensive care was abandoned more quickly when the prognosis was poor. Patients with early DNR orders did have lower rates of some intensive therapies and procedures such as IMV, dialysis, bronchoscopy, and central lines. Even imaging tests were used less frequently. This may reflect patient preferences that are not captured by administrative data, such as do-not-intubate orders or other advance directives. It also may indicate that early DNR orders are markers of more frequent goals of care conversations, leading to less intensive therapy. Although potential biases from care teams may play a role in this association, early DNR orders were not associated with systematic denial of intensive care, because a high proportion of patients still received each treatment.
Although hospital characteristics such as size and teaching status had a minimal relationship to DNR rates in our sample, regional effects were present. Patients in the Northeast and West regions had higher rates of early DNR orders than those in the South and Midwest. This is slightly different than geographic patterns described in a study from Kabaria and colleagues, who observed higher rates in the Midwest and West regions among patients with cirrhosis.20 This difference may be due to the patient population studied or hospitals sampled, because others have found significant variation even at the state, county, and hospital levels.21 Local and cultural beliefs, individual and community values, and physician practice patterns likely contribute to this variation. Similar to others, we found that Black and Hispanic patients had lower rates of DNR compared to White patients.20 Although patient preference and cultural beliefs may account for some of this difference, lack of access to advance care planning, communication quality, and mistrust of the healthcare system also likely contribute to this disparity.20,21 Disparities in DNR rates may be a significant contributor to markedly higher healthcare bills for Black and Hispanic patients at the end of life, which may have lasting consequences for families.22
We found that although patients with DNR orders had a high rate of mortality, most survived to discharge. Even among patients with the highest predicted mortality and most severe disease, almost 50% left the hospital alive. Lower resource utilization associated with DNR orders were the most pronounced in the setting of high expected mortality, where the relative infrequency of high-cost outliers suggests drawn-out care was forgone. There was only modest correlation between predicted mortality and having a DNR order, however. Even among patients with the highest predicted mortality, only 17% had such orders. As hospitals attempt to improve patient-centered care through better understanding of end-of-life preferences, it makes sense to focus these efforts on patients with the most severe presentation and comorbidity burden at admission.23 These patients most frequently have preference-based changes in their care, and conversations about their values can ensure that their care aligns with these goals. Hospitals may have difficulty identifying such patients. Mortality prediction tools such as CURB65 and PSIB are available but are usually used for determining treatment location and are not appropriate for prognostication.24 Having better tools could help physicians initiate these discussions with the patients for whom they are the most relevant. In our sample, hospitals with the most DNR orders appeared to do the best job of differentiating among patients based on mortality risk, but more study is needed to understand how these hospitals did so.
Approaching patients early is important, because patients with DNR orders placed late in admission often have difficult hospitalizations with long lengths of stay and highly invasive, aggressive care with poor quality of life at the end.25 These late orders are frequently signed by surrogates, adding stress to loved ones.26 Better communication between outpatient and inpatient records also may help with this approach because patients may be more likely to have conversations about their values in the outpatient setting.
Our study has several important limitations. We could not account for other patient preferences that may affect the level of care they receive, such as do-not-intubate orders or other advance directives. We also could not account for potential discordance between advanced care documentation and patient wishes. We relied on the ICD code to identify DNR orders, which has high specificity and moderate sensitivity in identifying patients with DNR.16 We also relied on present-on-admission indicators, which are not always accurate, to identify early DNR orders. This may have led us to include some patients with late DNR orders in our exposed cohort, which would likely increase our estimates for mortality and intensive care among this population. Conversely, there may be a group of patients with DNR orders that were not recorded as present on admission. We would have misclassified these patients into the no-DNR group. Although we included a large number of patient and hospital characteristics in our model, the possibility of unmeasured confounding, which would affect our outcome estimates, cannot be excluded. The strengths of our study include the use of a large, nationally representative sample and the inclusion of a highly accurate mortality prediction model in our analyses.
In summary, among patients with pneumonia, DNR orders present on admission are associated with much higher in-hospital mortality and less intensive treatment, although patients with these orders still receive intensive care and are most often discharged from the hospital alive. Differences in cost based on DNR status were not observed in patients with low predicted mortality, but among those with high probability of death, DNR orders were associated with lower costs and shorter hospital and intensive care stays. To have the greatest impact, hospitals should focus efforts to elicit patient preferences for care on those sickest at admission.
Supplementary Material
Key Points.
Do-not-resuscitate (DNR) orders were associated with less intensive care, higher mortality, and markedly lower costs.
DNR orders had the most impact on patients at the highest risk of mortality.
Only 17% of the highest risk patients older than 65 years who were hospitalized for pneumonia had a DNR order on admission.
Acknowledgments
Funding for this study was obtained through the Agency for Healthcare Research and Quality (AHRQ) R01 HS0242777-01A1 and AHRQ 1K08 HS025026-01.
Footnotes
Please note that only the information in the disclosures made by each coauthor is included in the COI statement.
Is “work for hire” in error; if not, the entity must be disclosed.
From this point, you mention ICD-9, but without CM. Are all of them ICD-9-CM? Please clarify.
Please define CURB65 and PSI.
Supplemental digital content is available for this article. Direct URL citations appear in the printed text, and links to the digital files are provided in the HTML text of this article on the journal’s Web site (http://sma.org/smj).
BM.M.S. did not report any financial relationships or conflicts of interest. M.D.Z. has received compensation from Cleveland Clinic, Spero/The Medicines Company, scPharma, and Melinta.
P.K.L. has received compensation from the National Institutes of Health. T.L.H. has received compensation from Cerner Corporation (now Oracle). P.B.I. has received compensation from AHRQ, Colgate-Palmolive, and the California Attorney General’s Office. A.D. has received compensation from AHRQ and Merck. S.D.H. and M.B.R. have received compensation from AHRQ.
N.G. lists this as a “Work for Hire”C
References
- 1.Garrido MM, Balboni TA, Maciejewski PK, et al. Quality of life and cost of care at the end of life: the role of advance directives. J Pain Symptom Manage 2015;49:828–835. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Ouyang DJ, Lief L, Russell D, et al. Timing is everything: early do-not-resuscitate orders in the intensive care unit and patient outcomes. PLOS ONE 2020;15:e0227971. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Mirarchi FL, Juhasz K, Cooney TE, et al. TRIAD XII: are patients aware of and agree with DNR or POLST orders in their medical records. J Patient Saf 2019;15:230–237. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Binder AF, Huang GC, Buss MK. Uninformed consent: do medicine residents lack the proper framework for code status discussions? J Hosp Med 2016;11:111–116. [DOI] [PubMed] [Google Scholar]
- 5.Patient self determination act of 1990. H.R. 4449, 101st Cong (1990). https://www.congress.gov/bill/101st-congress/house-bill/4449. Accessed December 8, 2021.
- 6.Ramirez JA, Wiemken TL, Peyrani P, et al. Adults hospitalized with pneumonia in the United States: incidence, epidemiology, and mortality. Clin Infect Dis 2017;65:1806–1812. [DOI] [PubMed] [Google Scholar]
- 7.Stefan MS, Jaber R, Lindenauer PK, et al. Death among patients hospitalized with pneumonia: implications for hospital outcome measures. JAMA Intern Med 2015;175:851. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Mortensen EM, Coley CM, Singer DE, et al. Causes of death for patients with community-acquired pneumonia: results from the Pneumonia Patient Outcomes Research Team cohort study. Arch Intern Med 2002;162:1059–1064. [DOI] [PubMed] [Google Scholar]
- 9.Marrie TJ, Fine MJ, Kapoor WN, et al. Community-acquired pneumonia and do not resuscitate orders. J Am Geriatr Soc 2002;50:290–299. [DOI] [PubMed] [Google Scholar]
- 10.Marrie TJ, Shariatzadeh MR. Community-acquired pneumonia requiring admission to an intensive care unit: a descriptive study. Medicine 2007;86:103–111. [DOI] [PubMed] [Google Scholar]
- 11.Beach MC, Morrison RS. The effect of do-not-resuscitate orders on physician decision-making. J Am Geriatr Soc 2002;50:2057–2061. [DOI] [PubMed] [Google Scholar]
- 12.Sarkari NN, Perman SM, Ginde AA. Impact of early do-not-attempt-resuscitation orders on procedures and outcomes of severe sepsis. J Crit Care 2016;36:134–139. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Scarborough J, Pappas T, Bennett K, et al. Failure-to-pursue rescue: explaining excess mortality in elderly emergency general surgical patients with preexisting “do-not-resuscitate” orders. Ann. Surg 2012;256:453–461. [DOI] [PubMed] [Google Scholar]
- 14.Maksoud A, Jahnigen DW, Skibinski CI. Do not resuscitate orders and the cost of death. Arch Intern Med 1993;153:1249–1253. [PubMed] [Google Scholar]
- 15.Wachterman MW, Sommers BD: Dying poor in the US—disparities in end-of-life care. JAMA 2020;325:423–424. [DOI] [PubMed] [Google Scholar]
- 16.Fonseca L, Walkey AJ, Ma X, et al. Validation of the V49.86 code for do-not-resuscitate status in hospitalized patients at a single academic medical center. Ann Am Thorac Soc 2018;15:1234–1237. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Lagu T, Rothberg MB, Shieh M-S, et al. Hospitalizations, costs, and outcomes of severe sepsis in the United States 2003 to 2007. Crit Care Med 2012;40:754–761. [DOI] [PubMed] [Google Scholar]
- 18.Zhang Z, Kim HJ, Lonjon G, et al. Balance diagnostics after propensity score matching. Ann Transl Med 2019;7:16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Rothberg MB, Pekow PS, Priya A, et al. Using highly detailed administrative data to predict pneumonia mortality. PLOS ONE 2014;9:e87382. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Kabaria S, Gupta K, Bhurwal A, et al. Predictors of do-not-resuscitate order utilization in decompensated cirrhosis hospitalized patients: a nationwide inpatient cohort study. Ann Hepatol 2021;22:100284. [DOI] [PubMed] [Google Scholar]
- 21.Zingmond DS, Wenger NS. Regional and institutional variation in the initiation of early do-not-resuscitate orders. Arch Intern Med 2005;165:1705–1712. [DOI] [PubMed] [Google Scholar]
- 22.Hanchate A, Kronman AC, Young-Xu Y, et al. Racial and ethnic differences in end-of-life costs: why do minorities cost more than whites? Arch Intern Med 2009;169:493–501. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Swamy L, Cooper AZ. Did you confirm code status? Code status discussions as a checkbox on hospital admission. Chest 2021;160:2335–2337. [DOI] [PubMed] [Google Scholar]
- 24.Chalmers JD, Singanayagam A, Akram AR, et al. Severity assessment tools for predicting mortality in hospitalised patients with community-acquired pneumonia. Systematic review and meta-analysis. Thorax 2010;65:878–883. [DOI] [PubMed] [Google Scholar]
- 25.Wright AA, Zhang B, Ray A, et al. Associations between end-of-life discussions, patient mental health, medical care near death, and caregiver bereavement adjustment. JAMA 2008;300:1665–1673. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Levin TT, Li Y, Weiner JS, et al. How do-not-resuscitate orders are utilized in cancer patients: timing relative to death and communication-training implications. Palliat Support Care 2008;6:341–348. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.