Oxygen Saturation Targets in Infants Hospitalized With Bronchiolitis: A Multicenter Cohort Study

This study was a secondary analysis of data collected from a multicenter randomized clinical trial of hospitalized infants with bronchiolitis using a cohort study design.^15,16  The sample included all infants who were eligible and enrolled in the randomized clinical trial. Briefly, patients were recruited from general pediatric inpatient units (GPIU) in 3 children’s and 3 community hospitals from November 1, 2016, to May 31, 2019, and were randomized to either intermittent (every 4 hours) or continuous pulse oximetry. Patients were randomized using a computer-generated randomization sequence. Randomization was stratified by center, using block sizes of 4 or 6. Randomization occurred after patients were stable and during daytime hours. None of the study hospitals was at high altitude. The primary outcome for the trial was length of hospital stay from randomization to discharge from hospital. Research ethics boards at all hospital sites approved the trial and parents or guardians provided written informed consent to participate. The institutional research ethics board at the lead site approved the secondary analysis of the trial data.

Infants 4 weeks to 24 months of age admitted to the GPIU with a clinical diagnosis of first-episode bronchiolitis were included in this study if they were generally healthy. Bronchiolitis was defined as signs and symptoms characteristic of a viral lower respiratory tract infection as per American Academy of Pediatrics guidelines.⁹  Infants with and without the need for supplemental oxygen were eligible for inclusion and were randomized to intermittent or continuous monitoring only after meeting prespecified stability criteria (same or decreasing supplemental oxygen requirement and respiratory rate with a respiratory rate less than 70 breaths per minute, heart rate less than 180 beats per minute, or oxygen supplementation less than 40%, or 2 L/min by nasal prongs). Infants treated with heated high-flow oxygen were eligible for randomization only after discontinuation of heated high-flow oxygen. Additionally, infants admitted to the ICU for mechanical or noninvasive ventilatory support, or whose parents were unable to complete questionnaires related to postdischarge health care use for a cost analysis of the primary trial were excluded. Infants with chronic medical conditions, such as congenital heart disease and is cyanotic or requiring diuretics, chronic lung disease with home oxygen, pulmonary hypertension neuromuscular disease, immunodeficiency, and hemoglobinopathy were excluded.

The primary exposure was the hospital oxygen saturation target policy. The oxygen saturation target was determined at the hospital level and guided by local institutional guidelines and practice standards. Hospitals used either an oxygen saturation target policy of 90% or greater in room air while the infant was awake and asleep (90%/90%) (2 hospitals) or 90% or greater in room air while awake and 88% or greater in room air while asleep (90%/88%) (4 hospitals). Along with the clinical status of the child, the target policy guided the use of supplemental oxygen (ie, initiation and discontinuation) and discharge decisions. For 4 sites that used a 90%/88% oxygen saturation threshold, nurses and physicians received education, guidelines were embedded in order sets, monitors were set to alarm at 90%, and clinicians instructed not to intervene if an infant was stable at a saturation between 88% and 90% while asleep. This policy had been implemented before the study: more than 5 years (1 site), 3 to 5 years (2 sites), and 2 years (1 site).

The primary outcome of this study was time to discharge from hospital (hours), measured from admission to the GPIU to discharge from the hospital. Length of hospital stay is an important outcome for patients, clinicians, and the health care system and has been used as the primary outcome in previous bronchiolitis trials.¹²  Secondary outcomes selected a priori included: (1) time to discharge from hospital (hours) among children on supplemental oxygen in the GPIU; (2) initiation of supplemental oxygen in the GPIU (the initiation of supplemental oxygen at any point between GPIU admission and discharge); (3) time to discontinuation of supplemental oxygen among infants who were on supplemental oxygen in the GPIU (duration of oxygen supplementation during the hospitalization in hours); (4) revisits to the hospital after discharge (an ED visit or hospital admission in the first 15 days after discharge from the index hospitalization); (5) parent days missed from work after discharge (number of days missed from work up to the first 15 days after discharge from the index hospitalization); and (6) ICU transfers.

Covariates and baseline characteristics included age, sex, weight (kg), parental cigarette smoking, intravenous (IV) fluids or nasogastric (NG) feeds used before admission, supplemental oxygen in ED before admission to the GPIU, randomization group (intermittent versus continuous monitoring), treatments administered before randomization (antibiotics, salbutamol, nebulized epinephrine, corticosteroids, high-flow oxygen), feeding adequacy scale score (a visual analog scale with scores ranging from 0 to 10, with 0 representing not feeding and 10 indicating feeding similarly to when healthy), respiratory rate (breaths/minute) at randomization on the GPIU, heart rate (beats/minute) at randomization, and hospital site (community versus children’s hospital).

Descriptive baseline characteristics for continuous variables were presented as medians and interquartile ranges, and frequencies and percentages for categorical variables. For the primary outcome, Cox proportional-hazards regression was used to estimate the hazard ratio (HR) for the association between oxygen saturation target and time to discharge from hospital, adjusting for age, sex, randomization group, parental cigarette smoking, feeding adequacy scale score at baseline, use of IV fluids or NG feeds in the ED before admission, respiratory rate, heart rate, hospital type, and oxygen supplementation in the ED before admission. The proportional hazards assumption was assessed using scaled Schoenfeld residuals.¹⁷  To evaluate whether randomization group acted as an effect modifier in the relationship between oxygen saturation target and time to discharge from hospital, we tested an interaction term between the 2 variables.

Multiple logistic regression was used to estimate odds ratios (OR) for the association between oxygen saturation target policy and initiation of oxygen supplementation in the GPIU. Cox proportional-hazards regression was used to estimate the HR for the secondary outcomes of time to hospital discharge among children on supplemental oxygen and time to discontinuation of supplemental oxygen. These models were adjusted for age, sex, randomization group, parental cigarette smoking, respiratory rate, heart rate, use of IV fluids or NG feeds in the ED before admission, feeding adequacy scores at baseline, and hospital type. A zero-inflated generalized linear mixed model with a log link and γ distribution was used to assess the association between oxygen saturation target and parent days missed from work.¹⁸  Using this model, we presented the expected number of workdays missed for each group.¹⁹  Logistic regression was used to estimate the OR for the relationship between oxygen saturation target policy and revisits after discharge. These models were adjusted for age, sex, randomization group, and type of hospital. Because oxygen saturation targets differed according to hospital site rather than individual patients, we used mixed-effects models with hospital site as a random effect to account for clustering at the site level. All effect estimates are reported with 95% confidence intervals (CI). Multicollinearity for all models was assessed using the variance inflation factor.²⁰  Because of a small number of events, only descriptive statistics were reported for the secondary outcome of ICU transfers. A 2-sided P value < .05 was used as the threshold for statistical significance. Given that this study sample size was fixed as a secondary analysis of a trial, post hoc power calculations were not conducted as they are not informative and can be misleading. Rather, the CI around the primary outcome estimate was assessed as is recommended.^21,22  All analyses were performed using R, version 4.0.2 (R Foundation for Statistical Computing; Vienna, Austria).

A participant flowchart is included in the Supplemental Information and Supplemental Figure 2. A total of 229 infants were included in this study. Characteristics of the infants are shown in Table 1. A total of 162 infants at 4 hospitals (2 community and 2 children’s) were managed with an oxygen saturation target policy of 90%/88%, whereas 67 infants at 2 hospitals (1 community and 1 children’s) were managed at 90%/90%. Age and sex were similar between the 2 groups. The median (interquartile range [IQR]) age of infants was 4.0 months (2.3-8.7) in the 90%/88% group and 4.0 months (2.8-8.3) in the 90%/90% group. Sixty-eight of 162 (42.0%) infants in the 90%/88% group and 25 of 67 (37.3%) infants in the 90%/90% group were female.

Some baseline characteristics differed between the 2 groups. Specifically, 58 of 162 (35.8%) of infants in the 90%/88% group versus 31 of 67 (46.3%) of infants in the 90%/90% group received IV fluids or NG feeds in the ED before admission. A total of 57 of 162 (35.2%) infants in the 90%/88% group compared with 36 of 67 (53.7%) in the 90%/90% group received salbutamol treatment in the ED. Similarly, a smaller proportion of infants in the 90%/88% group received corticosteroid treatments compared with those in the 90%/90% group (13.6% and 20.9%, respectively).

The median (IQR) time from hospital admission in the GPIU to discharge was 48 hours (34-86 hours) in the 90%/88% group and 46 hours (34-74 hours) in the 90%/90% group (Table 2). After adjusting for covariates, the association between use of an oxygen saturation target policy of 90%/88% compared with 90%/90% and time to discharge from hospital was not statistically significant (HR, 0.83; 95% CI, 0.61–1.14; P = .25). A Kaplan-Meier plot illustrating time to hospital discharge by oxygen saturation target is shown in Fig 1. The interaction term between randomization group and oxygen saturation target was not statistically significant (P = .26).

Secondary outcomes are shown in Table 3. All secondary outcomes used the 90%/90% group as the comparator.

Among infants on oxygen supplementation in the GPIU (N = 103), the median (IQR) time to discharge was 69 hours in the 90%/88% (43–110 hours) and 65 hours (42–89 hours) in the 90%/90% group. The use of an oxygen saturation target policy of 90%/88% was not associated with time to discharge among infants on oxygen supplementation (HR, 0.72; 95% CI, 0.44–1.16; P = .18) after adjusting for covariates.

A total of 38 of 162 (23.5%) infants in the 90%/88% group versus 16 of 67 (23.9%) in the 90%/90% group had supplemental oxygen initiated in the GPIU. After adjusting for covariates, the use of an oxygen saturation target policy of 90%/88% was not associated with initiation of oxygen supplementation in the GPIU (OR, 0.98; 95% CI, 0.47–2.02; P = .95).

For those infants on oxygen in the GPIU (N = 103), the median (IQR) time to discontinuation of oxygen supplementation was 21 hours (11–48 hours) in the 90%/88% group and 20 hours (9.4–49 hours) in the 90%/90% group. The use of an oxygen saturation target policy of 90%/88% was not associated with time to discontinuation of oxygen supplementation (HR, 0.75; 95% CI, 0.44–1.27; P = .28) after adjusting for covariates.

Eighteen of 162 (11.1%) infants in the 90%/88% group and 6 of 67 (9.0%) infants in the 90%/90% group revisited the hospital within 15 days of initial discharge from the index hospitalization. After adjusting for covariates, the use of an oxygen saturation target policy of 90%/88% was not associated with revisiting the hospital after discharge (OR, 1.38; 95% CI, 0.52–3.71; P = .52).

The median (IQR) parent days missed from work up to the first 15 days after discharge from the index hospitalization was 2.0 days (0.25–3.0 days) in the 90%/88% group and 2.0 days (1.0–4.0 days) in the 90%/90% group. The association between oxygen saturation target policy and parent days missed from work was not statistically significant for either the logit component of the model or the γ component (OR from logit model, 2.41; 95% CI, 0.90–6.41; P = .08, relative risk from γ model = 0.95; 95% CI, 0.70–1.30; P = .77).

Three of 162 (1.9%) infants in the 90%/88% group and no infants in the 90%/90% group were transferred to the ICU. In the 3 infants who were transferred to the ICU, there was no concern about delayed recognition of hypoxia or delayed intervention.

Bronchiolitis clinical practice guidelines recommend managing infants using an oxygen saturation target of 90%.^9,10  However, some hospitals use a lower oxygen saturation target policy of 88% while infants are asleep to reduce potential unnecessary interventions, escalation of care, and/or prolongation of hospital stay associated with oxygen desaturation that commonly occur during sleep. In this multicenter, prospective cohort study, we examined the use of a 90% while awake and 88% while asleep oxygen saturation target with a 90% while awake and asleep target policy for managing hospitalized infants with bronchiolitis. Clinical outcomes of time to discharge from hospital, initiation of and time to discontinuation of oxygen supplementation, revisits to the hospital after discharge, and parent days missed from work were similar between the 2 oxygen saturation target policies. However, the CI around the HR for the primary outcome of length of stay was not narrow enough to exclude a meaningful difference between groups.

No previous studies have examined the use of an 88% oxygen saturation target while asleep in bronchiolitis hospital management. In fact, few studies have evaluated oxygen saturation targets in bronchiolitis hospital management. Cunningham et al conducted a randomized clinical trial comparing oxygen saturation targets of 90% vs. 94% and found that time to discharge was significantly lower in the 90% oxygen saturation target group, without an increase in adverse events, supporting the use of the lower 90% oxygen saturation target.¹²  In our study, the association between an oxygen saturation target policy of 90%/88% and time to discharge or other outcomes, compared with 90%/90%, was not statistically significant. There are several reasons that might explain these results. First, an oxygen saturation target difference of 2% between groups and only while the infant is asleep may not be large enough to impact clinical outcomes. Second, clinicians at hospitals that use a 90%/90% target may not sometimes intervene when mild and transient oxygen desaturations (eg, between 86% and 90%) occur during sleep, recognizing that transient desaturations are common in recovering infants and the importance of focusing on the clinical state.²³  Third, some providers may not consistently adhere to the 88% while asleep target policy. These individual practice variations may result in an underestimation of the true effect of managing infants at the lower oxygen saturation target and outcome differences between groups. In this cohort study, unmeasured confounding of variables at the patient or hospital level could bias the results to the null hypothesis.²⁴  Last, length of stay may also be associated with hospital level factors, such as oxygen weaning protocols or guidance on observation after oxygen weaning. Participating hospitals did not have such protocols. We included hospital site as a random effect in the statistical models to account for unmeasured factors related to hospital specific practice.

Given the similar outcomes observed with either oxygen saturation target policy, our study results suggest there may not be a difference between the use of the lower 88% while asleep strategy currently used by some hospitals and 90% while awake and asleep. Clinicians and hospitals may prefer the lower target strategy because it may reduce sleep disruption and nursing workload associated with transient oxygen desaturation and monitor alarms. Alternatively, other hospitals may choose a 90% while awake and asleep target policy because it may be simpler to implement in practice and/or better aligned with other local practice factors (eg, consistency with broader hospital practice or with other condition-specific local practice guidelines).

Strengths of this study include its multicenter design that included community hospitals, where the majority of infants hospitalized with bronchiolitis are cared for.³  Data were collected prospectively, allowing for the inclusion of relevant and important covariates and both postdischarge outcomes and patient-centered outcomes. This study has several limitations. First, although this cohort study used statistical models that accounted for important patient and hospital-level covariates associated with outcomes, a randomized clinical trial would be a better study design to address unmeasured confounding and establish a causal relationship between oxygen saturation target policy and outcomes. A cluster randomized design would be preferred because patient-level randomization would be at high risk of contamination of the oxygen saturation target strategies across patients. Second, we did not have data on adherence to the hospital oxygen saturation target policy. This study was designed as a pragmatic study of real-world practice. If there was a lack of adherence to the policies, this would result in an underestimation of any group differences in outcomes. Third, the sample size was not adequate to detect differences in safety outcomes such as ICU transfer, precluding any statistical analysis. However, given the infrequency of safety outcomes, a very large sample size would be required. In addition, although we used all available data from the previously conducted randomized clinical trial, a substantially larger sample size would be needed to increase the precision of our estimate and to draw a more definitive conclusion. Fourth, we did not measure outcomes such as infant sleep disruptions or nursing workload. Fifth, this study only included otherwise healthy infants. Results should not be generalized to children with medical comorbidities or complex chronic conditions who have longer hospital length of stay. Sixth, although supplemental oxygen was a secondary proximal outcome measure, we did not have other proximal outcome measures, such as desaturation alarm frequency or documentation of desaturation events that would have provided mechanistic information. Seventh, the original trial excluded families with non-English language preference, which may limit the generalizability of study results to this population. Last, we did not have data on darker skin pigmentation, which is important because pulse oximetry overestimates arterial oxygen saturation in those with darker skin pigmentation.

Among hospitalized infants with bronchiolitis, clinical outcomes, including length of hospital stay, supplemental oxygen use, revisits, and parent days missed from work, were similar between infants managed with an oxygen saturation hospital policy of 90% while awake and 88% while asleep compared with 90% while awake and asleep. Future trials generalizable to diverse populations, and comparing different oxygen saturation target strategies, such as specification of a minimum and maximum target, are needed.

The authors thank non-author collaborating group members of the Canadian Paediatric Inpatient Research Network (PIRN). PIRN Executive Committee Members: Sanjay Mahant, MD, Hospital for Sick Children, University of Toronto; Peter Gill, MD, DPhil, Hospital for Sick Children, University of Toronto; Gita Wahi, MD PhD, McMaster Children’s Hospital, McMaster University; Patricia Li, Montreal Children’s Hospital, McGill University, Evelyn Constantin, MD, Montreal Children’s Hospital, McGill University; Olivier Drouin, MD, CHU Sainte-Justine, University of Montreal; Mahmoud Sakran, MD, Lakeridge Health, Queen’s University; Karen Forbes, MD, Stollery Children’s Hospital, University of Alberta.