If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Clinicians prescribe and administer oxygen in response to reports of dyspnea, in the face of dropping oxygen saturation, as a “routine” comfort intervention, or to support anxious family members. Oxygen may produce nasal irritation and increase the cost of care.
To determine the benefit of administering oxygen to patients who are near death.
Patients were 66% female, 34% white, and 66% African American, and ages 56–97 years. Patients had heart failure (25%), chronic obstructive pulmonary disease (34%), pneumonia (41%), or lung cancer (9%). Most (91%) patients tolerated the protocol with no change in respiratory comfort. Three patients (9%) displayed distress and were restored to baseline oxygen; one patient died during the protocol while displaying no distress. Repeated-measure analysis of variance revealed no differences in the Respiratory Distress Observation Scale under changing gas and flow conditions.
The routine application of oxygen to patients who are near death is not supported. The n-of-1 trial of oxygen in clinical practice is appropriate in the face of hypoxemic respiratory distress.
leading to clinician uncertainty about symptoms experienced in the absence of a reliable patient report. Does dyspnea increase or abate during the near death stage?
Oxygen administered via face mask or nasal cannula is indicated for patients with chronic obstructive pulmonary disease (COPD) with severe hypoxemia, resulting in improved oxygenation, reduced dyspnea, and improved functionality and quality of life.
reported no additional symptomatic benefit for relief of refractory dyspnea comparing oxygen with room air in a sample of nonhypoxemic patients with advanced illness. Of interest, terminally ill patients who lived alone were less likely to have oxygen prescribed than those with a caregiver, suggesting that reasons other than patient distress often prompt the initiation of therapy.
Patients who were near death or unable to provide a self-report about distress were not included in previous studies. Thus, whether oxygen confers a patient benefit when applied during the last days to hours of life remains unknown and was the basis for this study.
Palliative oxygen is routinely prescribed at life's end in response to patient reports of dyspnea, declining oxygen saturation, and signs of imminent death regardless of the patient's ability to experience or report distress and/or in the absence of behaviors that signify respiratory distress. Oxygen may be a wasteful treatment if the patient is not experiencing distress; additionally, oxygen administration has potential adverse effects.
Oxygen delivery by nasal cannula leads to nasal dryness, nasal irritation, and nosebleed, particularly at high flow rates in excess of 4 L/minute. Oxygen may also prolong the dying process without conferring benefit if the patient is experiencing no respiratory distress or is unable to experience distress. Prolonging death without conferring comfort also may burden the patient's family by extending caregiving days and the anticipatory grief phase. Additionally, oxygen delivery carries a financial cost that may be wasteful if there is no patient benefit.
We hypothesized that the routine application of oxygen to patients who were near death would not be beneficial. We expected that the ability to experience distress when death is imminent differs among individuals, and the benefits of breathing oxygen will vary among individual patients.
This double-blind, repeated-measure study with the patient as his/her own control was conducted from August 2010 through September 2011 after obtaining approval from the Wayne State University Institutional Review Board. After obtaining patient or family written consent, a trained observer, blinded to the nasal cannula flow (oxygen or medical air), collected baseline data about patients' respiratory status via the Respiratory Distress Observation Scale® (RDOS; ML Campbell, Detroit, MI), end-tidal carbon dioxide (et-CO2), and peripheral oxygen saturation (SpO2) on whichever gas was flowing at enrollment. Patients who were not in respiratory distress at baseline (RDOS ≤4) were randomly rotated through a trial of low flow oxygen alternating with medical air by nasal cannula or room air (no flow) on a 10-minute washout rotation. Demographic characteristics and administration of benzodiazepines and opioids were noted.
Study participants were recruited from a population of patients who were referred for palliative care consultation or were enrolled in hospice, were near death (Palliative Performance Scale [PPS] ≤30%), at risk for respiratory distress, and institutionalized at a university hospital or a free-standing hospice facility in the midwestern U.S. (Fig. 1). Patients were included if they had one or more of the following diagnoses putting them at risk for dyspnea: lung cancer, heart failure, COPD, or pneumonia. Patients were excluded if they were mechanically ventilated or on high-flow oxygen by face mask, had a tracheostomy, or were experiencing respiratory distress at the time of potential study enrollment. The required sample size of 32 patients for this study was based on an effect size of 0.25, a correlation coefficient of at least 0.3 between a six time-point repeated measures, a statistical power of 0.80, and a 0.05 level of significance. G*Power computer software version 3 (F. Faul, Universität Kiel, Germany) was used to calculate the required sample size.
Presence and intensity of respiratory distress was measured using the RDOS, an eight-variable ordinal scale. The RDOS is the only known tool for assessing respiratory distress when the adult patient cannot self-report dyspnea. The RDOS has undergone rigorous clinical testing to establish scale reliability, inter-rater reliability, convergent validity, construct validity, and discriminant validity.
Each item is scored from zero to two points and the points are summed; higher scores suggest higher-intensity respiratory distress. We proposed that a change of two points or more represented a clinically significant change. Internal consistency with Cronbach's alpha was established before data analysis in this application.
Nearness to death was quantified using the PPS. This scale grades a patient's general condition as 0 (dead) to 100 (normal) in increments of 10 points. The scale incorporates five observer-rated parameters: ambulation, activity, self-care, intake, and level of consciousness. The scale has been validated with patients with cancer (all types), patients in an acute tertiary hospital setting and a home care setting, and in patients with heterogeneous diagnoses.
Scores range from 1 to 8, with lower scores signifying higher levels of consciousness; scores from 4 to 8 signify unconsciousness.
Hypoxemia and hypercarbia were assessed by measuring SpO2, and et-CO2 with a Nellcor® N-85 capnograph/oximeter (Covidien, Mansfield, MA). This device provides exceptional pulse oximetry performance via a finger probe even during low perfusion states that typify the study population. A capnoline oral/nasal circuitry is attached to the patient similarly to a nasal oxygen cannula.
The research assistant (RA) placed a nasal capnoline on the patient's nose and an oximetry probe on a finger. The capnoline tubing connected to a Y connector with tubing to each of an oxygen and medical air flow meter. A hand towel covered the flow meters masking them from both the patient and RA; flow could be heard at the bedside but whether it was oxygen or air could not be determined. Blinding was lost when the patient was crossed over to the no flow encounter.
The patient's baseline flow was either oxygen (O2) or medical air, or no flow, whichever modality the patient was receiving before study enrollment. The flow rate used in the study protocol ranged from 1 to 4 L/minute to correspond to the flow rate the patient was experiencing at enrollment, if any. Patients who were not receiving oxygen at study enrollment received flow rates at 2 L/minute during the protocol. After study monitors were placed, the patient rested for five minutes before any subsequent data collection. Baseline data (RDOS, et-CO2, and SpO2) were collected after the five-minute rest period. The research nurse left the room and the patient's staff nurse changed the flow to correspond to the randomization pattern previously established. After the flow meters were re-masked, the RA resumed observation and collected data after 10 minutes. This crossover pattern was repeated every 10 minutes for six randomized gas/flow encounters. After the sixth encounter, the patient was restored to his/her baseline modality. Additional data were recorded from the patient record, including age, gender, race, diagnoses, and benzodiazepine and opioid administration for the 24-hour interval before enrollment.
Patients who displayed distress (RDOS >4) during the protocol were restored to their baseline modality and subsequent flow rotations ceased. The patient's nurse was alerted to the onset of patient distress by the RA.
Analyses were done with SAS version 9.2 (SAS Institute Inc., Cary, NC). Descriptive statistics were used to characterize the sample. Internal consistency of the RDOS was confirmed with Cronbach's α before we proceeded with analysis; an internal consistency coefficient (α=0.87) for RDOS was achieved in this analysis. Procedure mixed in SAS was used to test the potential change in repeated measures of RDOS, SpO2, and et-CO2 over time.
Table 1 illustrates sample characteristics; the frequency of diagnoses is more than 100% because some patients had more than one. Most (84%) patients had oxygen flowing at baseline. Three patients did not tolerate switching from oxygen to no flow as evidenced by an increased RDOS; they were restored to baseline oxygen, were restored to baseline comfort, and dropped from further rotations. All three patients who experienced distress were conscious and had PPS 20–30; two of three patients experienced a rapid and severe hypoxemia after the crossover (Table 2). A fourth patient died during the fourth gas/flow encounter without any increase in RDOS from baseline, in other words, without displaying distress. No oxygen adverse effects, such as nosebleed, were detected.
The average RDOS baseline score was 1.47 (range 0–4) signifying respiratory comfort. The primary outcome, respiratory distress, did not differ over time during the study period (F=0.55, P=0.74). The average SpO2 baseline level was 93.6 (range 69–100). SpO2 decreased significantly for some patients but the change over time for the sample was statistically nonsignificant (F=1.97, P=0.089). The baseline SpO2 and RDOS were inversely correlated (r=−0.41, P<0.01). Controlling for baseline, we found no change in et-CO2 during the study period (F=0.15, P=0.98) (Fig. 2). A few patients (n=12) received an average 7.3 mg of morphine in the eight hours prior to the protocol but there was no association with the baseline RDOS (χ2=0.78, P=0.94). Only two patients received a low dose (1 mg) of lorazepam prior to the study protocol; the small number precluded statistical analysis.
Neither consciousness (RLS85) nor nearness to death (PPS) correlated with baseline RDOS. Consciousness was inversely and significantly correlated with nearness to death (r=−0.87, P<0.01).
This adequately powered investigation demonstrated that most patients near death receive no palliative benefit from oxygen or medical air. We have added to the evidence about palliative oxygen because previous studies have not included patients who were near death and/or unable to provide a dyspnea report.
All the patients were comfortable at baseline and most were receiving oxygen. The reason oxygen was initiated in this sample is an unanswered question. With no significant changes in respiratory comfort over time under changing gas and flow conditions, it is apparent that the oxygen was an unnecessary intervention. In addition, there was no benefit to medical air flowing. Previous investigators have identified a palliative benefit to air flowing on the face
Our findings depart from other investigations, however; some of our patients were profoundly hypoxemic yet they did not display respiratory distress. This finding suggests that the hemodynamic and metabolic derangements that characterize imminent death may suppress distress experience. Does an endogenous opioid release characterize imminent death? Further investigation is indicated.
Oxygen is not a benign intervention. Adverse nasal effects and decreased mobility are patient experiences.
These results suggest that oxygen need not be prescribed to patients who are near death and not exhibiting respiratory distress regardless of oxygen saturation. Oxygen can often be withdrawn when the patient makes a transition from terminal illness to imminent death, particularly as consciousness decreases. Whereas most patients in our study tolerated a crossover from oxygen to air or no flow, three patients experienced distress that was relieved by a return to oxygen. An inability to reliably predict which patient will experience distress requires close clinical observation when withdrawing oxygen.
Baseline et-CO2 and SpO2 levels on room air were not obtained because the aim of the study was to determine if distress occurs during a crossover from oxygen. We wanted to observe the naturally occurring changes, distress and physiological, with withdrawal of oxygen as may happen during usual care.
We were limited by missing SpO2 data for six patients. Six patients were hypothermic and/or hypotensive and had insufficient peripheral circulation to support SpO2 measurement. We noted a trend toward significance with the SpO2 changes over time. We were interested in identifying whether respiratory comfort was maintained without oxygen even in the face of naturally occurring hypoxemia.
We had incomplete blinding during the no flow arms of the protocol. We attempted disconnecting the nasal cannula at its distal end from the oxygen tubing and keeping the flow under a towel but the different sound was detectable to the RA. Thus, the patient and RA were blinded to oxygen and medical air but the no flow condition was not masked.
Palliative oxygen is routinely prescribed for a number of reasons, including to “do something.” A growing body of evidence supported by these findings suggests that oxygen is a useful palliative intervention when the patient is experiencing distress and is hypoxemic. There is no support from these findings for the initiation or continuation of oxygen therapy when the patient is comfortable and near death.
Disclosures and Acknowledgments
This investigation was funded by the Blue Cross Blue Shield of Michigan Foundation. The sponsor had no role in the study design or in the collection, analysis, and interpretation of the data. The authors declare no conflicts of interest.
The authors thank Judy Wheeler and Julia Walch (Detroit Receiving Hospital—Palliative Care Service) and Denise Grabowski (Angela Hospice) for their assistance in identifying and enrolling participants. They also thank the patients and their families who generously participated in the study.