Perceptions of Hyperoxemia and Conservative Oxygen Therapy in the Management of Acute Respiratory Failure

Abstract

Rationale: Mechanically ventilated patients in the intensive care unit (ICU) are often managed to maximize oxygenation, yet hyperoxemia may be deleterious to some. Little is known about how ICU providers weigh tradeoffs between hypoxemia and hyperoxemia when managing acute respiratory failure.

Objectives: To define ICU providers’ mental models for managing oxygenation for patients with acute respiratory failure and identify barriers and facilitators to conservative oxygen therapy.

Methods: In two large U.S. tertiary care hospitals, we performed semistructured interviews with a purposive sample of ICU nurses, respiratory therapists, and physicians. We assessed perceptions of oxygenation management, hyperoxemia, and conservative oxygen therapies through interviews, which we audio recorded and transcribed verbatim. We analyzed transcripts for representative themes using an iterative thematic-analysis approach.

Results: We interviewed 10 nurses, 10 respiratory therapists, 4 fellows, and 5 attending physicians before reaching thematic saturation. Major themes included perceptions of hyperoxemia, attitudes toward conservative oxygen therapy, and aspects of titrated-oxygen-therapy implementation. Many providers did not recognize the term “hyperoxemia,” whereas others described a poor understanding; several stated they never encounter hyperoxemia clinically. Concerns about hyperoxemia varied: some providers believed that typical ventilation strategies emphasizing progressive lowering of the fraction of inspired oxygen mitigated worries about excess oxygen administration, whereas others maintained that hyperoxemia is harmful only to patients with chronic lung disease. Almost all interviewees expressed familiarity with lower oxygen saturations in chronic obstructive pulmonary disease. Cited barriers to conservative oxygen therapy included concerns about hypoxemia, particularly among nurses and respiratory therapists; perceptions that hyperoxemia is not harmful; and a lack of clear evidence supporting conservative oxygen therapy. Interviewees suggested that interprofessional education and convincing clinical trial evidence could facilitate uptake of conservative oxygenation.

Conclusions: This study describes attitudes toward hyperoxemia and conservative oxygen therapy. These preferences and uncertain benefits and risks of conservative oxygen therapy should be considered during future implementation efforts. Successful oxygen therapy implementation most likely will require 1) improving awareness of hyperoxemia’s effects, 2) normalizing lower saturations in patients without chronic lung disease, 3) addressing ingrained beliefs regarding oxygen management and oxygen’s safety, and 4) using interprofessional education to obtain buy-in across providers and inform the ICU team.

Supplemental oxygen ranks among the most commonly administered therapies worldwide, particularly among mechanically ventilated patients in the intensive care unit (ICU). Liberal oxygen therapy—variously defined as supplemental oxygen titrated to maintain an oxygen saturation as measured by pulse oximetry (Sp_O2) ≥94–96% (1–4)—is common, possibly because of widespread beliefs that oxygen is benign and that the risks of hypoxia outweigh hyperoxia’s potential harms (5, 6). Indeed, retrospective data suggest that mechanically ventilated patients treated for acute respiratory failure are often exposed to high fractions of inspired oxygen (Fi_O2) and may spend 50–75% of their ICU stay with supraphysiologic partial pressures of arterial oxygen (7, 8).

Most data suggest potential harm from hyperoxemia in critical illness, but recent evidence lacks consistency. Many observational studies and some clinical trials suggest that hyperoxemia during acute respiratory failure is associated with excess morbidity and mortality (1, 9–12). Preclinical data provide biologic plausibility by linking hyperoxemia to numerous sequelae (e.g., lung tissue damage and reactive oxygen species formation) (13). However, a recent clinical trial suggests harm for conservative oxygen therapy in acute respiratory distress syndrome (ARDS) (2). Although ideal oxygenation targets for specific patient populations remain unclear, recent guidelines recommend conservative oxygenation strategies (e.g., targeting Sp_O2 less than 96%) (14–16). Independent of this debate, implementation efforts remain limited because little is known regarding ICU providers’ perceptions of hyperoxemia, the degree to which they consider potential harms of hyperoxemia, and how they weigh tradeoffs between hypoxemia and hyperoxemia in clinical decision-making.

To better understand how providers think about oxygenation during mechanical ventilation and to inform future implementation efforts (17), we used semistructured interviews to investigate provider perceptions regarding oxygen therapy for mechanically ventilated ICU patients. Specifically, we elicited perceptions of hyperoxemia and identified barriers and facilitators to the implementation of conservative oxygenation strategies.

Methods

Study Design, Setting, and Subjects

We conducted a qualitative study within a regional healthcare system and recruited participants who work in a general medical-surgical ICU at a community-based hospital and others from a trauma surgical ICU at a university-based hospital in Pittsburgh, Pennsylvania. We used semistructured interviews rather than focus groups because we were interested in the perceptions of individual providers rather than those of groups, and we elected not to use a survey because interviews allow for flexibility during data collection (18). We opted to interview attending physicians and fellows, respiratory therapists, and nurses because these providers actively manage the oxygenation of mechanically ventilated patients. The research was approved by the University of Pittsburgh Institutional Review Board, and participants provided verbal informed consent. We followed the Consolidated Criteria for Reporting Qualitative Research framework for reporting data (see Appendix E1 in online supplement) (19).

Interview Guide

We developed an initial interview guide after reviewing the literature on hyperoxemia and conservative oxygen therapy, and we piloted this guide with a convenience sample of one physician, three respiratory therapists, and four nurses. During this pilot phase, we iteratively revised the interview guide on the basis of participant feedback and investigators’ reflections (20, 21). The final guide, which was not changed after the pilot phase, consisted of domains exploring practice patterns and others focusing on perceptions of hyperoxemia and provider attitudes toward conservative oxygenation strategies (see Appendix E2). To evaluate whether providers’ views differed in specific clinical scenarios, we iteratively asked about attitudes toward hyperoxemia and oxygenation when managing patients with three archetypal conditions that lead to acute respiratory failure: chronic obstructive pulmonary disease (COPD), ARDS, and sepsis. We defined a conservative oxygenation strategy as one targeting an Sp_O2 of 88–92% (2, 4, 22).

Data Collection (Demographics; Structure of the Interviews)

Two male interviewers (B.R.C. and A.R.) and a female medical anthropologist (K.J.R.) conducted interviews in September and October of 2019. We identified potential interviewees by direct approach and used purposive sampling to interview a range of providers of varying backgrounds and degrees of experience. Several physician participants knew members of the research team; one had collaborated with an investigator (T.D.G.) on a separate project. No other participants knew the interviewers before participation. After providing appropriate study background information and obtaining consent, we conducted interviews in person in or around the ICU when possible or over the phone when necessary. With participants’ consent, we audio-recorded interviews while interviewers took notes. The interviews were subsequently transcribed verbatim. We collected demographic data from and compensated each participant. Between interviews, several investigators (B.R.C., K.J.R., and A.R.) met regularly to discuss initial findings and evaluate data saturation. We considered recruitment complete when data saturation was reached for each provider type, meaning no new themes emerged with additional interviews (23).

Data and Transcript Analysis

We undertook an iterative thematic-content-analysis approach when analyzing contents of interviews. First, we reviewed transcripts and drafted a codebook based on initial perceptions of data. Themes mirrored major sections of the interview guide and emergent topics (e.g., education of providers). We identified “mental models,” or individuals’ representations of their reality and understanding of how actions and effects are related, to aid data collection and analysis (24). This approach helped us analyze individual providers’ and interprofessional ICU teams’ reasoning. Three coders (B.R.C., K.J.R., and A.R.) then met regularly to review transcripts, adding new themes to the codebook and refining existing themes on the basis of observations made in subsequent transcripts. Similar themes were grouped into categories. We specified labels for codes (and subcodes, when appropriate) with a brief definition and inclusion and exclusion criteria when needed. Because we conducted interviews over a 2-month period, we did not expect major shifts in clinical practice or preferences during the study. Participants did not provide feedback on findings.

Once we achieved 90% agreement among codes, two coders (B.R.C. and A.R.) coded all transcripts and then recoded a small sample to ensure reliability was maintained. We used NVivo 12 (QSR International) for data management and analysis. We also organized themes according to the Consolidated Framework for Implementation Research (CFIR) to help us better understand how our results could inform oxygen-therapy implementation (25).

Methods to Ensure Rigor

We enhanced the credibility, or internal validity, of our results by having more than one researcher conduct interviews with individuals of varying professions and experience levels. Interviewers from different backgrounds helped minimize the influence of bias or orientation. Multiple settings and participant demographics allowed us to better identify similarities and variations in providers’ perspectives. When reporting results, we quoted different participants to add transparency and trustworthiness to the results (26, 27).

Results

We reached thematic saturation after interviewing about five providers of each type per unit. In total, we interviewed 29 providers, including 9 physicians (5 attending physicians and 4 critical care medicine fellows), 10 nurses, and 10 respiratory therapists. The duration of interviews ranged from 13 to 50 minutes, with an average of 26 minutes, and none were repeated. Nurse participants tended to be younger and less experienced than other professionals, but age and critical care experience varied for all. We report demographic data on 27 participants (Table 1), as these are missing for 2 participants (1 respiratory therapist and 1 fellow). No interviewee refused participation, and one provider declined compensation. We also identified three primary domains related to provider perceptions toward hyperoxemia and conservative oxygen therapy: 1) perceptions of hyperoxemia, 2) attitudes toward conservative oxygen therapy, and 3) perceived barriers to and facilitators of conservative oxygen therapy implementation.

Table 1.

Participant demographics by provider type and ICU

Characteristic	Full Sample	Academic Community General ICU				Academic Medical-Surgical ICU
Provider type	—	All	R.N.	R.T.	M.D.	All	R.N.	R.T.	M.D.
N (%)	29	15 (52)	5 (17)	5 (17)	5 (17)	14 (48)	5 (17)	5 (17)	4 (14)
Demographics
Median age (IQR), yr	34 (29–41)	35 (31–43)	30 (27–31)	50 (35–57)	39 (35–42)	30 (28–34)	25 (24–34)	30 (28–31)	46 (42–49)
Median experience in current ICU (IQR), yr	4 (1–9)	7 (1–9)	3 (1–7)	8 (7–18)	1 (1–4)	2 (1–10)	2 (1–2)	5 (2–8)	10 (9–15)
Median time since obtaining professional license (IQR), yr	8 (4–12)	8 (5–14)	4 (2–8)	8 (8–27)	12 (10–16)	8 (5–12)	3 (2–4)	10 (7–11)	20 (15–21)

Definition of abbreviations: ICU = intensive care unit; IQR = interquartile range; M.D. = Doctor of Medicine; R.N. = registered nurse; R.T. = respiratory therapist.

Perceptions of Hyperoxemia

We identified common themes regarding hyperoxemia among providers. Many reported a lack of recognition of the term “hyperoxemia,” its occurrence in the ICU, and potentially deleterious effects of liberal oxygenation. However, providers were concerned with excess oxygenation when managing chronic lung pathologies (e.g., COPD) because of fears about eliminating patients’ respiratory drive. Alternatively, few providers expressed worry about liberal oxygen therapy when managing patients with ARDS, citing difficulties maintaining adequate oxygenation.

Specific conceptions and definitions of hyperoxemia, beliefs surrounding risks, and reported methods of hyperoxemia assessment appeared to differ by profession. Nurses and respiratory therapists spoke generally of hyperoxemia as “giving too much oxygen” (respiratory therapist) when referring to patient-specific needs. Six nurses and six respiratory therapists defined hyperoxemia precisely as an arterial partial pressure of oxygen cutoff of greater than 100 mm Hg, whereas some respiratory therapists used varying definitions. Few nurses or respiratory therapists described adverse effects of liberal oxygenation, whereas some described hyperoxemia generally.

Some nurses and respiratory therapists stated that they believe hyperoxemia is not a salient concern in critical illness and indicated that although hyperoxemia could occur with prolonged supraphysiologic oxygen pressures, this exposure may not be harmful:

“I’ve never seen someone being over oxygenated (in this unit). . . . with patients diving in (a hyperbaric chamber) for 5 hours you don’t even see any severe risk with that so definitely with a mechanically ventilated patient up here receiving 100% Fi_O2 for 5 hours you don’t see any harm.” —Respiratory Therapist

“You can over oxygenate to the point where their labs show it but . . . there’s not really a thing of oxygen toxicity.” —Nurse

Participants favored titrating Fi_O2 over positive end-expiratory pressure (PEEP) to address hypoxemia because they believed Fi_O2 has a faster, safer, and more reliable effect on oxygenation and that PEEP holds greater risks:

“Risks associated with PEEP are greater than the risks associated with oxygen . . . with oxygen I can safely go up on just about everyone.” —Respiratory Therapist

Other respiratory therapists and nurses believed hyperoxemia could not occur in the ICU, either because of 1) a unit-wide approach to titrate Fi_O2 to progressively lower concentrations or 2) hypoxia remaining an overriding concern:

“I don’t really notice that we over give oxygen because we’re always trying to wean it down to get them out of the ICU.” —Respiratory Therapist

“You don’t hear of too many people being overly oxygenated per se in the ICU . . . It’s more hypoxic.” —Respiratory Therapist

Finally, many respiratory therapists and nurses considered hyperoxemia relatively less concerning in sepsis because such patients’ need for more aggressive treatment overrode concerns about hyperoxemia.

Physicians appeared more informed about hyperoxemia, with some defining it more explicitly. Unlike respiratory therapists and nurses, most physician participants expressed familiarity with potentially deleterious consequences of hyperoxemia, particularly oxygen-related free-radical damage or lung injury related to prolonged oxygen exposure. However, they did not further articulate other pathophysiologic mechanisms for adverse effects:

“You’re elevating their blood oxygen, partial pressure to a super physiologic level, and the risk that will come with that includes generation (of) reactive oxygen species and some other stuff that I just can’t think of.” —Fellow

Few physicians defined hyperoxemia with cutoffs like those used by respiratory therapists and nurses, and the majority had a more nuanced approach that physicians believed allowed for more accurate oxygenation assessment. Finally, physician participants indicated more concern about hyperoxemia in sepsis than nurses or respiratory therapists, with physicians stating worry that excess oxygen could exacerbate underlying inflammation.

Attitudes toward Conservative Oxygen Therapy

Providers felt universally receptive to conservative oxygen therapy when managing patients with COPD. Some also cited its use in ARDS and referenced the ARDS Network low-tidal-volume ventilation protocol as justification (28). When speaking about COPD and ARDS, providers, especially nurses and respiratory therapists, referenced preexisting standards or protocols that allowed for “lower-than-normal” oxygen saturations:

“People get really nervous. Which is why it doesn’t happen that often. It’s okay if they’re a COPD patient because everybody understands that relationship. But if it’s not a COPD patient and you say I’m okay with this being 88%. At least with ARDS you have some paper to back you up.” —Fellow

Some providers explicitly stated they would support conservative oxygen therapy if clinical trial data showed benefit.

Few providers, mostly physicians, cited a willingness to use conservative oxygen therapy in other populations and referenced recently published studies. Some remained hesitant to use conservative oxygen therapy in a broad ICU population because of uncertain evidence.

Perceived Barriers to Conservative Oxygen Therapy

Barriers to conservative oxygen therapy are shown according to the CFIR in Figure 1 and Table 2 Providers, especially respiratory therapists and nurses, described concerns about conservative oxygen therapy regarding 1) apprehension about lower oxygen saturations, particularly among less experienced providers, and 2) perceptions that maintaining lower oxygen saturations requires increased work with little patient benefit and increased risk of hypoxia:

Figure 1.

Study results according to the Consolidated Framework for Implementation Research. RCT = randomized controlled trial.

Table 2.

Provider perceived barriers to conservative oxygen therapy

Domain or Construct	Respective Theme	Supporting Quotations
Intervention characteristics
Evidence strength and quality	Evidence base uncertain on overall benefit for patients.	“We don’t really understand if there would be long-term neurocognitive sequelae to lower levels of oxygen in critical illness or not.” [Physician[
Relative advantage	Providers may perceive conservative oxygen therapy to be of marginal benefit for patient with increased work.	“I think the concern from the team would be that if it’s viewed as potentially marginal or if they have an added work or burden at some point during their patient day that they then would be at risk of having a hypoxic event and satting [having oxygen saturations] below 88. Which would then result in potential bad outcomes.” [Physician]
	Providers may also perceive increased risk of hypoxia or less time to react before hypoxia onset with conservative oxygen therapy.
Complexity	Narrow window for oxygen saturation target could make oxygen titration difficult.	Q: Any other challenges you could see . . . just using that within your patient population?
		A: “Probably just keeping it in between that 4%. I would think. Like it’s rare . . . like I said how something reads, how a patient is moving, if they’re not perfusing correctly, what kind of drugs are they on. All that that could decrease their SpO2 and it would probably be hard to keep it in like a 4% range like that, 88–92%. It might go over; it might go under.” [Nurse]
Outer setting
Patient needs and resources	Family members become concerned about patients when they see lower numbers and hear alarms on monitors, and providers may feel the need to adjust oxygen to prevent this situation.	“That really, for lack of a better word, alarms family members that are in the room and they’ll ask about it. So, not only would you need to reset the alarm limits, so things aren’t going off when they hit 87, give them a little bit of buffer, but it would need to be consistent education of families too. Like, ‘This blue line is the oxygen level, and our goal is for it to be [at] 88–92. Yep 88 is just fine.’ Because they ask a lot, you know? ‘Why is the blood pressure this? What are you going to do about that?’ And I think that would definitely be a barrier because it’s a lot easier to just turn up the oxygen instead of staying there and having a 5-minute conversation about why it’s there.” [Physician]
	Patient discomfort related to lower oxygen saturations can impede conservative oxygen-therapy implementation.	“When we extubate people and stuff, we have a lot of patients like that who are tight in the chest and feel like they can’t breathe, even though you tell them ‘You’re at 100%,’ they’re like ‘No no, turn my oxygen up.. Between a lot of anxiety and, two, the tightness of all their injuries and everything.” [Nurse]
External policy and initiatives	The process of updating protocols and guidelines so that they reflect changes in the evidence base is slow.	“I think as with anything new there would be a knowledge translation problem. There would be a need for multiple randomized trials suggesting that it’s safe for patients. [If it] doesn’t result in new hypoxic events or worse outcomes, [then] there is the system issue of having protocols changed, which probably don’t say target that low amount but target a higher amount.” [Physician]
Inner setting
Structural characteristics	Shift work and rotating staff make instituting a permanent change in unit practice difficult.	“When things go to hell in the middle of the night, then you start trying new and exciting things just to get your feet under you. And that may leave you in the end with something that’s like a pseudo-stable patient but also maybe not optimal settings. And so, then the next day someone will come and like tinker with it until its back to something that we all recognize.” [Fellow]
Networks & communications	Discordance between electronic and verbal orders can sow confusion regarding oxygen management.	“I think making sure that everyone’s on the same page about it. So, like if someone tells me the goal sats [saturations] are 88–92, but it’s not ordered, then the next person might think like ‘oh, that’s too low’ and then would go up on their oxygenation. So, things often get lost in like the handoff reports and the kinda telephone of everything. And oftentimes, orders are not always, ya know, sometimes the notes say one thing the orders say another thing, the doctor that’s there says another thing. So, I think just like communication and making sure that everybody’s on the same page.” [Nurse]
Culture	Institutional bias and patterns in oxygen management can hinder conservative oxygen-therapy implementation.	"I think that there are years and years of institutional bias. When I started critical care medicine, I remember being a resident in the intensive care unit and having my attending tell me that the three rules of critical care were: number 1, believe no one; number 2, trust nothing; and number 3, give oxygen. There’s been this sort of long-standing belief that the application of oxygen is always beneficial and never harmful, and [I] think that perception has gone for a long, long time, and so there’s probably a large sort of degree of inertia that would be relate[d] to trying to change those attitudes." [Physician]
Implementation climate	Institutional belief that oxygen is always beneficial and never harmful	"I think partly it would be just overcoming long-held, possibly erroneous beliefs about the benefits of oxygen. I think that would be a big institutional hurdle. There’s just a lot of old-school belief that oxygen is always beneficial and never harmful. And then doing anything like making a systematic change in the way we provide care is harder than it seems. You just can’t tell people ‘well now we’re go for sats [saturations] of 88–92’ and expect it to happen. There will need to be sort of all the implementation science behind rolling that out and making it happen in various intensive care units." [Physician]
Available resources	Faulty SpO2 monitoring equipment makes monitoring oxygenation difficult.	Q: What sorts of things would impede or limit the use of a conservative oxygenation strategy in ICU patients?
		A: “I think if like our equipment wasn’t picking up right. We get kind of erroneous numbers on the monitors or blood pressures. If somebody’s blood pressure is tanking and they’re on blood pressure meds. Cause the oxygen parameters and the waveforms kind of fluctuate with those blood pressures. So that would be hard for us. Cause a lot of times they’ll call us and be like ‘oh my patient’s desatting [having a decrease in saturation],’ but their pressures are like 60s over 40s, for instance.” [Respiratory Therapist]
Characteristics of individuals
Knowledge and beliefs about intervention	Providers are uncomfortable with lower oxygen saturations because of concern for hypoxia.	"I think that’s why you see some issues with the wanting to run people on lower FiO2 where they’re right where they should be is because they’re more likely to occasionally drop below, which I don’t think is a dangerous thing to happen, but it makes staff concerned, and so, for that reason, they’ll be cranked back up. If they’re on 40, they’ll get cranked back up to 70 to make sure that we make the numbers look good." [Fellow]
Other personal attributes	Provider patterns and preferences instituted are ingrained starting in professional schools.	“I think a lot of people just want to see 100. They want to see 99; they want to see 100. 98. They see 94 and then get nervous. They want to hit the button. They want the oxygen turned up to 50, or you know they think somebody needs [to be] suctioned. When all it is . . . it’s okay. You know, I think people get real nervous. They see numbers; they don’t look at patients. You know, sometimes the numbers aren’t even real, you know. But we’re not talking like that, but you know people just want to see a certain number.”
		Q: Who gets nervous the most?
		A: “The nurses, I would think. Nurses, newer therapists, you know, people maybe fresh out of school. But now it’s nurses a lot. They literally want to see numbers. They want a heart [rate] of 80. Blood pressure of 120/80. You know they want a textbook patient, and it doesn’t work. You know, a textbook is what it is, you know. Ninety-two and above is okay for anybody. Like, you know, my sat [saturation] right now might not be . . it might 94 its fine.” [Respiratory Therapist]

Definition of abbreviations: A = answer; Fi_O2 = fraction of inspired oxygen; ICU = intensive care unit; Q = question; Sp_O2 = supplemental oxygen titrated to maintain a saturation.

“The concern from the team would be that if it’s viewed as potentially marginal or if they have an added work or burden at some point during their patient day that they then would be at risk of having a hypoxic event and satting [having an oxygen saturation] below 88.” —Physician

Providers of all types indicated that poor interprofessional communication in the ICU would impede implementing conservative oxygen therapy:

“Things often get lost in like the handoff reports and the kind of telephone of everything . . . sometimes the notes say one thing the orders say another thing, the doctor . . . says another.” —Nurse

Less commonly cited barriers centered around individual provider attitudes, ICU-specific factors, and elements from outside the ICU. Some felt that nurses and respiratory therapists would find it difficult to maintain a narrow oxygenation target in critically ill patients; this impediment could foil implementing lower oxygen saturations. Others described ingrained patterns of oxygenation management and faulty equipment as barriers to implementation. Finally, one physician mentioned that protocols and guidelines from professional societies, which may be slow to adapt to changing evidence, could impede employing evidence-based strategies (Table 2).

Perceived Facilitators of Conservative Oxygen Therapy

Participants also discussed various facilitators for the uptake of conservative oxygen therapy (listed in Figure 1 and Table 3), including 1) strong evidence showing that conservative oxygen therapy benefits a range of ICU patients, 2) interprofessional buy-in, and 3) education provided to the team:

Table 3.

Provider-perceived facilitators of conservative oxygen therapy

Domain or Construct	Respective Theme	Supporting Quotations
Intervention characteristics
Intervention source	Providers are more likely to trust studies that were conducted internally (in same settings as they practice).	“You know, I think if we see a benefit for our patients, we will try to do our best to do that intervention.” [Nurse]
Evidence strength and quality	Physicians feel that conservative oxygen therapy helps to give FiO2 that is more physiologic, rather than standard 40%.	“I think the pros would be it’s just more physiologic. I mean, I know I maybe oversimplify a little bit, but, you know, God put us on this earth to tolerate an FiO2 . . . you know we’re used to breathing 20 room percent oxygen [20% oxygen in room air], so anything that we can provide that more closely approximates normal physiology is, one would assume, better for patients. But I don’t have any more scientific of an explanation than that. It just makes intuitive sense to me.” [Physician]
	Lower oxygen requirements allow for lower mechanical ventilator parameters, decreasing risk of iatrogenic injury from PEEP.	“I would think that the two tools we have to improve oxygenation, PEEP, and FiO2 both potentially have deleterious effects at high levels. By minimizing PaO2 goals you would expose the patients to lower levels of airway-pressure FiO2, which might potentially have long-term benefits. [Physician]
	Lower oxygen targets avoid deleterious effects of hyperoxemia.	“Again, so I would say the two major pros are one, avoiding hyperoxygenation, which we have pretty good evidence is harmful, and then two, I think you have to subject people to less potentially physiologically damaging effects of the mechanical ventilator in an attempt to maintain this really high oxygen level that really isn’t necessary. So again, just doing more harm to the patient with the ventilator than good by trying to get their sat [oxygen saturation] to something that they don’t need it to be.” [Physician]
	Evidence in certain patient populations indicates that conservative oxygen therapy may be beneficial.	“I think that some of the initial or early cardiac-arrest literature is reasonably compelling. I think using a relatively lower oxygenation target, there is potentially beneficial.” [Physician]
Relative advantage	Aiming for lower oxygen requirements avoids struggles, and effort may be associated with maintaining higher saturations, especially in critically ill patients.	“Usually if I’m doing it, it’s because it’s really hard to do it any other way. Like I’m not like ‘oh I have an ARDS patient; we can aim for 88%; just wean down.’ It’s because getting them to 95% requires so much extra PEEP and FiO2 and maneuvers by the RT [respiratory therapist]. Maneuvers with the vent. That it’s probably not worth it, cause I think you have diminishing returns in those patients. And so, 88% is fine. It’s more that I think it’s not harmful than [that] I think it wouldn’t be necessarily better to be at 95.” [Fellow]
Trialability	Providers exposed to studies that use alternative practices become more open to using those practices.	“And when they did the ARDSnet trial or the ROSE trial. Like a year or two ago, they had two separate PEEP and FiO2 strategy ladders for that, and so the units where that was done the most are more comfortable with the high-PEEP strategy, so they will wean sometimes differently. If you tell them that you are using that strategy, they will wean according to that as opposed to the standard one.” [Fellow]
		Q: Interesting.
		A: “Like, TICU will do it that way. Like, I can go up to somebody in TICU and be like ‘I would like you to use the high-PEEP–low-FiO2 ladder,’ and they will say ‘okay, we know what that is.’” [Fellow]
Outer setting
External policy & initiatives	Ongoing clinical trials and studies can provide evidence for changing practice patterns.	Q: What do you think might prompt your ICU to change how they approach oxygenation for acute respiratory failure?
		A: “An adequately powered well-done randomized control trial.” [Physician]
Inner setting
Leadership engagement	Unit leadership that is widely respected by all provider types can improve uptake of intervention.	“In this ICU . . . we have pretty strong leadership that's well respected by the multidisciplinary team, so I think a top-down approach.” [Physician]
Access to knowledge & info	Knowledge delivered to interprofessional team through physicians appears to be effective for respiratory therapists and nurses.	“Well usually, well, I like to round with the doctors. When I have time, like if it’s not too busy of a morning, like when I’m worried, I like to round . . . cause I learn, I mean I learn a lot, pick up a lot of things I didn’t know, like I said. That’s one of the ways I know.” [Respiratory Therapist]
Process
Engaging	Structured education is critical to implementation of conservative oxygen therapy.	“I would start with education. Education seems to be the key component across the board. Okay, I would include all patient contact personnel. I think that no one segment of clinical practice is more important than the other. . . . Doctors, nurses, respiratory [therapists]. Everybody has to be on board with the same concept.” [Respiratory Therapist]
	Interprofessional buy-in helps preempt conflict that can occur during implementation.	“I think probably provider comfort is a big one in sort of getting everyone acclimated to the new targets and what peoples’ preconceived notions about that would be both from the nursing staff to the physician staff. But I mean, that’s probably the major limitation.” [Fellow]

Definition of abbreviations: A = answer; ARDS = acute respiratory distress syndrome; ARDSnet = ARDS Network; Fi_O2 = fraction of inspired oxygen; ICU = intensive care unit; Pa_O2 = arterial partial pressure of oxygen; PEEP = positive end-expiratory pressure; Q = question; ROSE = Reevaluation of Systemic Early Neuromuscular Blockade; TICU = trauma ICU.

“If we got anything that’s approaching level 1 data that shows a good benefit then I think we would take that and probably roll it into a meeting or discussion that got everybody on board or on the same page.” —Physician

“Education . . . there’s no replacement for it. Doctors, nurses, respiratory. Everybody has to be on board with the same concept.” —Respiratory Therapist

Suggested modalities of education and buy-in included 1) traditional continuing education, 2) practice champions, 3) conventional quality-improvement strategies (e.g., signage), and 4) policy changes on a unit level. Nurses and respiratory therapists often endorsed physician-led teaching during interprofessional rounds as an effective way to learn about evidence-based practice:

“Like the ICU as a whole? I think it would have to be a push from the more administrative level to get it at that degree of change.” —Fellow

“Most of the critical care team, especially our attendings, are very good at during rounds, they educate us. . . .” —Nurse

“I think a lot of people probably don’t know about hyperoxia and so, educating on that, educating on the safety of lower sats [oxygen saturations]. And then, posting signs in every room, is what they did in this (omitted) study, at the head of the bed, ‘Goal oxygen saturation, 88–92.’” —Physician

Discussion

In this qualitative study, we found that interprofessional ICU providers have variable perceptions regarding oxygen therapy in the management of critically ill patients. Many have little or no concern for hyperoxemia because of overriding concerns about hypoxia, beliefs that oxygen is harmless, and a lack of clinical recognition of hyperoxemia. Most providers felt conservative oxygen therapy was acceptable only when managing patients with COPD and has limited efficacy for other ICU patients. Although physicians did discuss risks of PEEP, none described potential indirect effects of strategies that promote hyperoxemia, such as lung injury due to increased driving pressure or delayed extubation due to slow titration of ventilator settings. Finally, we identified multiple potential barriers and facilitators to the implementation of conservative oxygen therapy in the ICU.

Our findings substantiate existing literature while offering novel insights. We corroborate previously reported provider attitudes toward oxygen therapy, including a belief that oxygen is risk-free and apprehensions regarding hypoxia risks (5, 6). Other results indicate that some acknowledge potential adverse effects of prolonged exposure to excessive amounts of oxygen (5), including misattributing risks of hyperoxemia in COPD to the eliminating respiratory drive rather than exacerbating ventilation–perfusion mismatch (29, 30). However, participants in our study did not seem to act on concerns regarding hyperoxemia and favored giving excess oxygen to avoid hypoxia. Overall, our findings suggest that hyperoxemia is not well known as a clinical entity among ICU providers (apart from that recognized in patients with COPD), and its potential risks are even less known. Several providers acknowledged a theoretical existence of hyperoxemia but indicated that harm from hyperoxemia is impossible, suggesting that beliefs about oxygen therapy are strongly entrenched and difficult to change.

Our results should be interpreted in the context of recent clinical trials. Two recent trials comparing liberal versus conservative oxygen therapy in critical illness offer mixed conclusions, possibly because they included different patient populations. The ICU-ROX (Intensive Care Unit Randomized Trial Comparing Two Approaches to Oxygen Therapy) trial demonstrated the safety of a conservative oxygen approach in a heterogenous group of mechanically ventilated ICU patients, whereas the LOCO₂ (Liberal Oxygenation versus Conservative Oxygenation in Acute Respiratory Distress Syndrome) trial, which specifically enrolled patients with ARDS, was halted early because of concerns about increased mortality in the conservative oxygen arm (2, 31). Though these findings suggest that effects of conservative oxygen therapy are modified by the etiology of respiratory failure, they may foster doubts about the safety of widespread conservative oxygen therapy by reinforcing concerns about risks related to hypoxia (1, 2, 31, 32). Therefore, implementation of conservative oxygen therapy in critically ill populations may be difficult until clear and consistent evidence supports its use in certain patient populations. Though expert opinion and recent guidelines (16) call for a reduction in oxygen exposure for ICU patients on the basis of evidence suggesting current practice may lead to excess morbidity and mortality, barriers to conservative oxygen implementation likely dwarf supporting evidence. Stronger data are required to clarify if conservative oxygen approaches are beneficial for critically ill populations. This would allow ICU providers to feel more confident implementing an oxygen strategy, conservative or otherwise, that is not harmful to patients so implementation can proceed more effectively (33, 34).

When probing potential barriers and facilitators to conservative oxygen therapy implementation (and potentially other implementation efforts), we identified factors in three domains—individual-specific factors, ICU-level features, and influences from outside the ICU—which corresponded to domains found in the CFIR (25), facilitating the creation of a conceptual model that can guide broader implementation efforts of oxygen therapy (Figure 1) (35).

Our findings indicate that successful implementation of oxygenation strategies (and potentially other implementation efforts) in the ICU should include interprofessional education to obtain buy-in from all members of the ICU team, given the interdisciplinary nature of caring for patients’ oxygenation, and preempt conflict during implementation. However, this process must begin with high-quality evidence showing that a chosen oxygenation strategy benefits mechanically ventilated patients in the ICU, a standard that remains lacking for conservative oxygen therapy in light of conflicting clinical trials (1, 2, 31). For example, supporting evidence for conservative oxygenation would help to normalize lower oxygen saturations in patients without COPD when disseminated through interprofessional education. Individual ICUs and/or hospital systems can complement interprofessional education with “adaptable” components, such as changing ventilator alarm settings, targeting provider types differentially with educational initiatives and use of electronic reminders.

On the basis of the conceptual model derived from our findings, we suggest several strategies for oxygen-therapy implementation. The first would use physicians, respected as educators by the interdisciplinary team, who would teach about oxygenation strategies primarily during rounds. Respiratory therapists and nurses viewed such teaching as an effective method to disseminate knowledge and change practice patterns of ICU providers. A second would involve respiratory therapists in an education initiative, led by respiratory-therapist and physician champions, to educate providers about the benefits and safety of a chosen oxygenation strategy. These educational strategies should be supported by changes in ICU-wide or hospital-wide protocols and policies regarding patient selection and oxygenation targets.

Limitations

This study has several limitations. First, the evidence regarding conservative oxygen therapy in the ICU continues to evolve; two relevant trials, for example, were published shortly after we completed our study (2, 31). Our interview guide was therefore influenced by older literature and may have reflected a bias toward the acceptability and safety of conservative oxygen strategies. Our findings, however, are relevant to populations for whom conservative oxygen therapy is appropriate (e.g., those with COPD) and are likely to have implications for any implementation efforts involving oxygen therapy in the ICU. Second, generalizability may be limited because providers came from a single healthcare system where many practice in academic medical centers involved in clinical research. This may have influenced providers’ knowledge of hyperoxemia, although our results suggest relatively low awareness. Third, our findings may not be generalized to settings where a dedicated respiratory-therapist role does not exist. Fourth, our definition of a conservative oxygen strategy (Sp_O2 of 88–92%) may be considered too low. However, this definition has precedent in clinical trials and evidence-based approaches to oxygenation (28, 30, 36, 37). Finally, several physician participants were familiar with some investigators, which may have biased their responses. This possibility, however, likely had little effect on the overall results because the known investigators did not conduct interviews, and only a few participants knew the investigators.

Conclusions

Providers’ attitudes toward hyperoxemia and conservative oxygen therapy in the ICU should inform future implementation efforts surrounding oxygen therapy in mechanical ventilation. Future research should address barriers and implement facilitators to oxygen therapy to improve uptake of safer oxygen therapy in critical illness.