Bronchiolitis of Infancy Discharge Study (BIDS): a multicentre, parallel group, double-blind, randomised controlled, equivalence trial with economic evaluation

Identifying participants

The study aimed to recruit and randomise only during the historic peaks for bronchiolitis in two northern hemisphere winters. Infants who were eligible were identified by clinicians and research/specialist nurses in the ED/AAA. The clinical decision to admit an infant with bronchiolitis prompted consent for the study.

Settings and locations where the data were collected

The study took place in the ED/AAA and paediatric wards of eight paediatric hospitals in the UK (in Aberdeen, Bristol, Dundee, Edinburgh, Exeter, Glasgow, Kilmarnock and Truro). In season 1, randomisation was open from 3 October 2011 to 30 March 2012 in the five Scottish sites only. In season 2, randomisation was open from 1 October 2012 to 29 March 2013 in all eight sites.

Primary outcome

The primary outcome was equivalent time to no cough (collected by phone calls at follow-up). The primary outcome was assessed by the primary caregiver.

Secondary outcomes (admission questionnaires, clinical data, parental phone calls, home visit)

We were looking for equivalence between treatment groups in the following secondary outcomes:

1. time to re-established feeding (approximately 75% normal)

2. parental perspective of ‘time to back to normal’

3. oxygen saturation at 28 days post randomisation (season 1 only).

We were looking for a difference between treatment groups in the following secondary outcomes:

1. time to fit for discharge (oxygen saturation ≥ 94% for 4 hours including a period of sleep and adequate feeding at ≥ 75% usual)

2. time to discharge

3. health-care reattendance (primary care, ED, readmission)

4. change in anxiety score

5. time to return to work/usual activities for parent(s)/nursery for infants

6. family and societal costs incurred

7. health-care costs related to discharge time and subsequent health-care utilisation.

There were no major changes to trial outcomes following trial commencement. As above, oxygen saturation was measured at 28 days in season 1 with the agreement of the DMC when no significant differences were identified between the groups.

Sequence generation

Randomisation was by central internet-based secure password-protected randomisation database. Patient identifiers and some clinical details were entered to confirm eligibility (inclusion and exclusion criteria) and to prevent re-recruitment. The random allocation sequence was generated by the programmers at the Edinburgh Clinical Trials Unit (ECTU).

Type of randomisation

Randomisation was by blocks of varying length (four and six) without stratification.

Implementation

Infants were enrolled by clinicians and research/specialist nurses in the ED/AAA. Clinicians (mostly nursing staff) attached either a standard or a modified pulse oximeter in accordance with the computer-generated randomisation code. The administering clinician removed non-study oximeters from infants and, with an interval of 1 minute, reapplied the appropriate study oximeter, typically at a time when the infant was preparing to move between departments in the hospital (i.e. ED to a ward area).

Blinding

The monitors were identical in appearance and general function, with the exception of the study number (see Figure 2). All study staff involved in day-to-day running of the trial, hospital staff and parents were blind to study intervention and could not tell what the randomised group was from the study numbers on the machines. To further reduce the opportunity for accidental unblinding, study numbers on oximeters were changed in the period between season 1 and season 2. Those assessing outcomes were blind to the assigned intervention. The blind was not broken for any infant during the study.

General statistical methods

All analyses were by intention to treat (ITT) unless otherwise specified. The ITT population included all infants randomised into the BIDS study. Infants were analysed in the group to which they were randomised, regardless of treatment received. Where a per-protocol analysis was performed, infants were analysed in the group of the treatment they actually received: any use of standard pulse oximeter versus any use of modified pulse oximeter. The group of infants in which no pulse oximeter was used were excluded from the per-protocol population. No infants received both types of oximeter. All applicable statistical tests were two-sided using a 5% significance level. Ninety-five per cent (two-sided) confidence intervals (CIs) were presented. The primary analysis was an unadjusted analysis. Where there were missing data for an outcome variable, in the first instance, those records were removed from any formal statistical analysis, unless otherwise specified.

Primary outcome

The primary outcome was the number of days from randomisation until resolution of cough. The difference in median number of days until cough resolution between the two treatment groups was estimated along with a 95% CI for the difference. The treatment arms were considered equivalent with respect to this outcome if this 95% CI lay entirely within the equivalence limits of ± 2 days. For the primary analysis, the difference between the medians and the 95% CI for this difference were estimated as described by Altman et al. 33 The large-sample formula for derivation of the confidence limits was used. The main primary outcome analysis was conducted on an ITT basis. A per-protocol analysis was also performed. The study would lead to firm conclusions only if the findings from the main analysis and the per-protocol analysis were in agreement.

Secondary outcomes

Secondary outcome measures were tested for a difference between the two arms of the trial.

For the outcome measures, the times, split by treatment group, were presented using a Kaplan–Meier plot. Cox proportional hazards regression was used to estimate the treatment effect: time from randomisation to (1) fit for discharge and (2) actual discharge for all infants admitted with acute viral bronchiolitis; (3) time to no supplemental oxygen; (4) time to readmission to hospital. We considered whether or not the season-1 data for Glasgow (Yorkhill) should be removed from the analysis of time to fit for discharge, as this variable was not recorded consistently at this centre in season 1. However, this made no difference to the results so these data were left in.

The results are presented at multiple time points, and due allowance would be made for this if any of them proved to be statistically significant. The effect of the intervention was estimated using binary logistic regression and reported as an adjusted odds ratio and 95% CI for the proportion of infants with at least one health-care reattendance (primary care, ED, hospital readmission) at days 7, 14 and 28 and at 6 months. The effect of the intervention was estimated using Poisson regression for the number of health-care reattendances (primary care, ED, hospital readmission) at days 7, 14 and 28 and at 6 months. The effect of the intervention was estimated using analysis of covariance models with the baseline score as a covariate for parental anxiety score (anxiety questions from HADS questionnaire at admission, at 7, 14 and 28 days’ follow-up and at 6 months’ follow-up).

For the outcome measures, the mean difference in times between the two trial arms was estimated from a normal linear model, and presented with a 95% CI: heart rate at discharge; respiratory rate at discharge.

We tested for equivalence between the two arms of the trial for the secondary outcome measures. The same method as the primary outcome was used for time in hours from randomisation to re-established feeding (equivalence limits of ± 4 hours), and time in days from randomisation to parental perspective of back to normal (equivalence limits of ± 2 days). For time to re-establish adequate feeding, no imputations for missing data were performed, as the data were recorded only at the end of discharge and they were almost complete. The difference in mean oxygen saturation measurements between the two trial arms was estimated with its corresponding 95% CI for awake oxygen saturation at 28 days after randomisation (equivalence limits of ± 1.0% oxygen saturation – season-1 data collection only).

Subgroup analyses

Overview of economic evaluation

This trial investigated whether or not a 90% oxygen saturation discharge protocol (modified) is cost-effective compared with the standard 94% oxygen saturation discharge procedure. The research question the economic evaluation therefore addressed was: is the modified discharge procedure a cost-effective alternative to the standard discharge procedure? The economic evaluation measured the costs to the NHS and social care and combined this with the main outcome measure, time to cough resolution, within a cost-effectiveness analysis (CEA) framework. CEA is a form of economic evaluation in which both the costs and effects of two or more health interventions are compared, and the results report the incremental difference between the alternatives under consideration as an incremental cost-effectiveness ratio (ICER). The economic evaluation was undertaken alongside the trial, capturing individual resource-use data collected via economic data-collection questions integral to the trial forms. The analysis reported the incremental cost per reduction in time to cough resolution. The economic evaluation also took into consideration the seasonality of the disease by outlining the opportunity cost of hospital bed displacement during peak winter seasons. Individual patient-level data on outcomes and information on resource use were identified and measured during the trial and used in the economic evaluation.

Outcome measure for cost-effectiveness analysis

The primary endpoint in the trial is time to cough resolution (measured at 6 months). The economic evaluation used mean time to cough resolution, by measuring the area under the curve. 35,36 The area under the curve method permits comparison of the mean difference between treatment arms based on the entire survival curve, or in this case the entire time to cough resolution curve. 36,37 This is standard practice for an economic evaluation for determining the primary endpoint in time to event/survival analysis, where censored and skewed data are prevalent. The economic analysis considered the joint distribution of costs and effects. Therefore, the economic analysis used the mean time to cough resolution and accompanying standard errors for each arm of the trial to undertake probabilistic analysis, using bootstrapping38 to account for uncertainty around this primary outcome measure and how this affects the cost-effectiveness outcomes. The economic analysis explored the impact that baseline variables have on the mean time to cough resolution. It was hypothesised that variables such as sex, child gestation at birth, length of illness prior to onset, parental smoking status, antibiotics on arrival at accident and emergency (A&E), oxygen saturation and GP visits prior to onset might affect the time to cough resolution. In addition, the quality of life of parents was measured using the anxiety component of the HADS. 29

Resource use data collection

The base-case analysis was undertaken from the perspective of the UK NHS and personal social services. That is, the costs relevant to the economic analysis are those incurred by the NHS and social services. In a sensitivity analysis, the patient perspective was also incorporated, adding the costs incurred by the parent in terms of both financial costs (such as travel to and from hospital) and time lost from normal activities because of the illness.

There are five main resource categories of relevance: initial hospitalisation costs, medication, readmission costs, follow-up care and, in the sensitivity analysis, costs to parents. The base-case total cost (C_T) is a function of the cost of hospital days (or hours) (C_hosp), medication (C_med), readmission (C_readmit) and any follow-up care (C_FUcare). In the sensitivity analysis, the cost to parents is also included (C_parents). Equation 1 illustrates the main components of total cost.

Patient-level resource-use data were collected within trial, for example days (or hours) in an acute paediatric ward, days (or hours) in the HDU, tests and scans undertaken, medication prescribed, hospital readmissions and any additional follow-up care, that is GP appointments, referrals, etc. Parent time was also collected. Resource-use quantities and mean patient cost values are reported separately as per recent recommendations [Consolidated Health Economic Evaluation Reporting Standards (CHEERS) guidelines]. 39

Health-economic analysis methods

Stata version 12 (StataCorp LP, College Station, TX, USA) was used to carry out regression analysis exploring the effect that baseline variables have on the cost of each intervention, such as sex, child gestation at birth, length of illness prior to onset, parental smoking status and antibiotics on arrival at A&E. Potential non-normality in the cost data was explored and, if the cost data were found to be skewed, a generalised linear model (GLM) was employed. It was hypothesised that the cost data were likely to be highly skewed because of the very high costs associated with cases that are referred on to HDU or are readmitted. The GLM for cost regressions, described by Glick et al. ,40 was used in this case, and 95% CIs around the cost estimates were also calculated. 41 If, however, the cost data were found to be normally distributed, then standard parametric tests were used. In line with recent CHEERS guidelines,39 resource-use values, ranges references and probability distributions are reported for all resources. Mean values for the main categories of costs, as well as mean differences between the comparator groups (modified compared with standard), are reported.

Additional analyses were undertaken to explore the effect that baseline variables, such as sex, duration of illness before randomisation, parent smoking, GP visits prior to admission, had on the cost and time to cough resolution of each intervention. Potential non-normality in the cost and time to cough resolution data was explored and a GLM was employed. 40

Outcomes were analysed using multivariate logistic regression, adjusting for child age and sex, parental age left full-time education and heaviness of smoking index. 42 GLM regression was used for analysis of costs if they were skewed. GLM regression for cost was adjusted for treatment group, parental smoking and GP visits prior to admission. The other specified variables in the analysis plan (protocol) had no significant impact. A variety of families and various links for each family were investigated and tested using modified Park tests; gamma was found to be best fit, with a power link of –1. We ran GLM using gamma family and link power –1. GLM regression was used for the time-to-cough-resolution variable, given that the mean is skewed. GLM regression for time to cough resolution was adjusted for treatment group, season, sex, preterm birth, cost (total NHS cost), parental smoking, GP visits prior to admission and antibiotics prior to admission.

All within-trial analyses were performed using Stata version 12.

Unit costs

Unit cost information was combined with the resource-use data collected and the mean cost per patient per arm estimated. Valuing the resource use using the unit costs provided an estimate of the total cost for each resource. These were aggregated to estimate total patient costs within each arm, and the mean cost per patient per arm. The difference in average costs (and significance) between the two trial arms was estimated. All unit costs were collected in UK pounds sterling for the base year 2012/13. Cost information was collected from routine sources such as the British National Formulary (BNF) 63,43 Personal Social Services Research Unit44 and NHS Reference Costs. 45 Some unit costs were collected specifically from the trial/hospital financial records, such as the cost of specific lab tests that are not available on the BNF.

Handling missing cost and outcome data

Missing data were handled through the use of multiple imputation (for both cost and outcome data). The multiple imputation approach has been widely recommended by most experts in biomedical literature. 40 The key dependent variables used to base the imputations on were sex, child gestation at birth, length of illness prior to onset of cough, parental smoking status, antibiotics on arrival at A&E, oxygen saturation and number of GP visits prior to admission. Sensitivity analysis was undertaken to determine the influence of missing data at follow-up (resource-use data and primary outcome) on study conclusions, to assess the strength of association between time to cough resolution and ‘missingness’ (i.e. whether or not data are missing) and to allow for individual, sampling and imputation variation using multiple imputation.

Reporting and presenting of results

Appropriate consideration was given to the distribution of the cost and effect data in the economic evaluation. In order to explore the variation around the costs and effects generated by the trial data, stochastic variance around the cost–effect pairs was estimated using non-parametric bootstrapping methods. The incremental costs and the incremental benefits were reported within an ICER format where appropriate. The 95% CI for each ICER was estimated using Fieller’s theorem, a technique that includes any correlation between cost and outcome. 46 Sample uncertainty was explored using non-parametric bootstrapping to generate 1000 ICER values which are plotted on a cost-effectiveness plane in order to represent graphically uncertainty. 47 The result for the CEA was expressed in terms of positioning on the cost-effectiveness plane as well as translated into cost-effectiveness acceptability curves, indicating the likelihood that the results fall below any given cost-effectiveness ceiling ratio, where appropriate. Cost differences were reported between the arms as standard; however, in a departure from typical ‘treatment minus comparator’ differences for reporting purposes, the cost-effectiveness plane reported the differences as ‘standard minus modified’ to reflect the fact that this trial tested for equivalence in costs and effects.

Adjustment of timing of costs and benefits

Allowance for differential timing of costs and benefits was made using the recommended discount rate of 3.5% (National Institute for Health and Care Excellence Methods Guidance). 48 Data within this trial were collected over two consecutive winter periods, and costs and benefits were estimated for the baseline year, 2012/13.

Sensitivity analyses

A detailed sensitivity analysis of key parameters was undertaken. In addition to this the patient perspective was incorporated to the analysis, adding the costs incurred by the parent (in terms of financial costs such as travel to and from hospital and time lost from normal activities because of the illness) to the NHS and social services costs.

Seasonality impacts

The economic analysis also took into consideration the seasonality of the disease and the economic impact of this with regard to availability of ward space/beds during peak hospital times. Information was collected on the type of ward in which patients were being cared for. It may be that during peak times infants have to be cared for in, say, surgical wards and this will have cost consequences. Further to this, the capacity of wards during such busy times was explored by obtaining hospital attendance information from the business managers at trial centres to allow a more precise costing (saving) in terms of opportunity cost of hospital stay as a result of bronchiolitis. This additional information will allow the economic analysis to provide further relevant policy implications for the potential hospital stay reductions arising through implementation of the 90% oxygen saturation protocol.

Publication policy

To safeguard the integrity of the trial, data from this study were not presented in public or submitted for publication without requesting comments and receiving agreement from the National Institute for Health Research Health Technology Assessment programme. The success of the trial was dependent on the collaboration of many people. The results were, therefore, presented first to the trial local investigators. A summary of the results of the trial will be made available on the Scottish Children’s Research Network website (www.scotcrn.org) and the research sites can provide these to parents of participating children on request.

Organisation

A Trial Steering Committee (TSC; see Appendix 4) and a DMC were established (see Appendix 5). Day-to-day management of the trial was overseen by a Trial Management Group (see Appendix 6). Each participating centre identified a paediatric consultant as a principal investigator (see Acknowledgements). Each participating centre was reimbursed on a per-patient fee from the core trial grant. The Medicines for Children Research Network and/or local Comprehensive Local Research Networks supported research nursing time, and employed or reallocated a research nurse to support all aspects of the trial at the local centres.

Confidentiality

Patients were identified by their trial number to ensure confidentiality. However, as the main caregivers to the patients in the trial were contacted during the follow-up, the caregivers’ names and contact details were recorded on the data collection forms in addition to the allocated trial number. Stringent precautions were taken to ensure confidentiality of names and addresses at ECTU and the sites. The chief investigator and local investigators ensured conservation of records in areas to which access is restricted.

Audit

No audit of BIDS was carried out. The trial manager monitored all the sites at least once to verify that the site staff had sufficient knowledge of the trial protocol and procedures, that the site file was being properly maintained and that the site adhered to local requirements for consent.

Termination of the study

Before termination of recruitment, ECTU contacted all sites by telephone or e-mail in order to inform sites of the final date for recruitment. Once the recruitment period had expired, the internet-based randomisation database was disabled to prevent further recruitment. After all recruited patients had been followed up until 6 months after randomisation, a declaration of the end of trial form was sent to the Multicentre Research Ethics Committee. The following documents will be archived in each site file and kept for at least 10 years: original consent forms, data forms, trial-related documents and correspondence. The trial master files at ECTU will be archived for at least 10 years.

Funding

The costs for the study itself were covered by a grant from the National Institute for Health Research Health Technology Assessment programme. See the Health Technology Assessment programme website for further project information. Clinical costs were met by the NHS under existing contracts.

Indemnity

If there was negligent harm during the clinical trial, then the NHS body owes a duty of care to the person harmed. NHS indemnity covers NHS staff, medical academic staff with honorary contracts and those conducting the trial. NHS indemnity does not offer no-fault compensation. The cosponsors were responsible for ensuring proper provision was made for insurance or indemnity to cover their liability and the liability of the chief investigator and staff.

Patient disposition and changes post randomisation

Tables 2 and 3 provide information on patient disposition. See Table 2 for the number of infants who reached the end of the study with > 90% of data [N = 584 (95%)]. Thirty-one infants did not reach 6-month follow-up: there were two deaths, one infant was withdrawn by the clinician, 21 infants were lost to follow-up and the parents declined further contact in another seven cases. Breakdown by group is provided in Table 2. Table 3 gives details of those in whom the primary outcome was directly obtained and the number in whom the primary outcome was estimated [standard care 43 (14.0%); modified care 38 (12.3%)].

Protocol deviations and classification are provided in Table 4. There were similar numbers and categories in each group.

Recruitment

Infants were recruited within two winter seasons. For the first season, recruitment was between 3 October 2011 and 30 March 2012 in the five Scottish sites only. For the second season, randomisation was from 1 October 2012 to 29 March 2013 in eight sites. The addition of three sites in south-west England was in response to a quieter RSV season than expected in season 1. An analysis of season-1 peak oximeter use enabled a redistribution of study oximeters across eight sites for season 2.

The trial completed recruitment on schedule and to target on 29 March 2013.

Baseline data

Baseline demographic and clinical characteristics are provided by group in Table 5. The modified care group had slightly fewer males and more preterm infants, though all other demographics and characteristics were similar.

Virology test data

Laboratory virology testing was performed in 81.9% of infants, with near-patient testing in 40.0% of infants (Table 6). The proportion of infants who were RSV positive (either laboratory or near-patient testing) was 69.8% in the standard care group and 76.3% in the modified care group. Laboratory testing identified a positive result for a virus other than RSV in 27.2% of standard care infants and 22.0% of modified care infants.

Numbers analysed

The treatment offered versus allocated is given in Table 7. In eight instances the incorrect treatment was allocated (by staff attaching the wrong oximeter): in seven instances a modified oximeter was provided to an infant randomised to standard care and in one instance a standard oximeter was attached to an infant randomised to modified care. These infants are included in the group to which they were allocated as per ITT. In 44 instances treatment was interrupted, the majority per protocol during an admission to the HDU, with treatment restarted on discharge from the HDU.

Protocol deviation

Seventy-six participants had a protocol deviation, 34 (11.0%) in the standard care group and 42 (13.7%) in the modified care group. The deviation categories are provided in Table 4 and are similar between groups.

Outcomes and estimation

Subgroup analysis of primary outcome

We specified an assessment of subgroups of the primary outcome for the number of days of illness prior to randomisation (would any effect be influenced by the degree of airway inflammation present?), household smoking (would an early discharge to a smoking household prolong symptoms?) and use of antibiotics. Between-subgroup comparisons did not show any evidence of a difference in the magnitude of the treatment effect between subgroups.

Secondary outcomes

Equivalence outcomes

Three further outcomes were assessed for equivalence: (1) time to return to adequate feeding (≥ 75% normal) in hospital, (2) time until parents considered that the infant was back to normal and (3) oxygen saturation measured at 28 days after randomisation.

Time to return to adequate feeding (≥ 75% normal)

Median time to return to adequate feeding was 24.1 hours in the standard care group and 19.5 hours in the modified care group. The median difference was 2.7 hours (95% CI –0.3 to 7.0 hours). This falls outside the prespecified equivalence limits of ± 4.0 hours, so we cannot infer equivalence. The 95% CI overlaps zero, so we cannot conclude that there is a difference either.

Figure 4 provides a survival curve for time to return to adequate feeding. A post-hoc analysis (Table 9) gave a hazard ratio of 1.22 (95% CI 1.04 to 1.44; p-value = 0.015), demonstrating some evidence that the modified care group returned to feeding faster than the modified care group, and the survival curve indicates that any difference between the groups was in the time period following the median time to return to adequate feeding. Thus, we have not proved equivalence. If any difference does exist, it goes in favour of the modified care group.

FIGURE 4.

Time to return to adequate feeding.

TABLE 9 - Patient outcomes: post-hoc analysis. Differences in equivalence outcomes assessed by hazard ratios

IQR, interquartile range.

Parameter	Standard care, median (IQR)	Modified care, median (IQR)	Hazard ratio estimate	Upper and lower 95% CI	p-value
Time feeding returned to ≥ 75% normal	24.1 (6.5–62.1)	19.5 (6.3–47.2)	1.22	1.04 to 1.44	0.015
Time back to normal (days)	12.0 (7.0–25.0)	11.0 (6.0–20.0)	1.19	1.01 to 1.41	0.043

Time to parents considered infant back to normal

Parents were asked in standardised telephone interviews when they considered their child had returned back to normal. This was a median of 12.0 days in the standard care group and 11.0 days in the modified care group. The median difference was 1.0 day (95% CI 0.0 days to 3.0 days). The 95% CI of the median difference fall outside the prespecified limits of ± 2 days and so we cannot infer equivalence. The 95% CI just touches zero, so there is some evidence that there was a small difference in favour of the modified care group.

Figure 5 provides a survival curve for time until parents considered their infant back to normal. A post-hoc analysis (see Table 9) gave a hazard ratio of 1.19 (95% CI 1.00 to 1.41; p-value = 0.043), providing some evidence that, in the parents’ perception, infants in the modified care group returned to normal faster than infants in the standard care group, and the survival curve indicates that any difference between the groups occurred after the median time it took for infants to return to normal. Thus, we have not proved equivalence. If any difference does exist, it is in favour of the modified care group. The benefit to the modified care group was not expected and is a modest overall difference.

FIGURE 5.

Time to return to normal (days).

Oxygen saturation measured at 28 days after randomisation

These data were only collected for season 1, following unblinded review by the DMC at the end of the season. The median peripheral capillary oxygen saturation (SpO₂) value was 99% in both groups. At 28 days there was a mean absolute difference between the treatment groups in oxygen saturation of 0.11% (95% CI –0.35% to 0.57%). We prespecified equivalence limits of 1% in either direction. Oxygen saturation at 28 days was equivalent in the two groups.

Differences outcomes

Time to fit for discharge from hospital

Infants could be considered fit for discharge from hospital once they had achieved adequate feeding (≥ 75% normal) and had been observed to have stable oxygenation in air with continuous oxygen saturation monitoring for a period of hours including a period of sleep. Infants in the standard care group were fit for discharge at 44.2 hours, compared with 30.2 hours in the modified care group (Table 10). The hazard ratio was 1.46 (95% CI 1.23 to 1.73; p-value < 0.001) (Figure 6).

TABLE 10 - Patient outcomes: difference outcomes

IQR, Interquartile range.

Hazard ratio = standard care/modified care (i.e. result < 1 = benefit to standard care, result > 1 = benefit to modified care).

Parameter	Standard care, median (IQR), n	Modified care, median (IQR), n	Hazard ratio estimate	95% CI	All, median (IQR), n	p-value
Time to fit for discharge (hours)	44.2 (18.6–87.5), n = 283	30.2 (15.6–59.7), n = 276	1.46	1.23 to 1.73	38.4 (16.9–69.7), n = 559	< 0.0001
Time to actual discharge (hours)	50.9 (23.1–93.4), n = 303	40.9 (21.8–67.3), n = 301	1.28	1.09 to 1.50	44.5 (22.2–78.4), n = 604	0.003
Time to no further supplemental oxygen (hours)	27.63 (0–68.1), n = 305	5.65 (0–32.4), n = 304	1.37	1.12 to 1.68	14.44 (0–54.5), n = 609	0.002
Time to readmission to hospital (days)	17 (7–22), n = 23	11 (2–21), n = 12	0.925	0.433 to 1.977	15 (3–22), n = 35	0.84

FIGURE 6.

Time to fit for discharge.

Time to actual discharge from hospital

The time at which infants were actually discharged from hospital could be influenced by local practice and logistics. Infants in the standard care group were discharged after a median of 51.0 hours, compared with 40.9 hours in the modified care group (see Table 10). The hazard ratio was 1.28 (95% CI 1.09 to 1.50; p-value = 0.027) (Figure 7).

FIGURE 7.

Time to discharge.

Time to no further supplemental oxygen

We measured the time from randomisation to the time that infants in each group last received supplemental oxygen prior to discharge. In the standard care group this was a median of 27.6 hours, and in the modified care group 5.7 hours (see Table 10). The hazard ratio was 1.37 (95% CI 1.12 to 1.68; p-value = 0.0021) (Figure 8).

FIGURE 8.

Time to no supplemental oxygen.

Time to readmission to hospital

There was no significant difference in time to readmission to hospital in those infants who were subsequently readmitted to hospital following discharge (see Table 10). Median difference was 0.9 days (95% CI 0.4 days to 2.0 days; p-value = 0.8410). Median duration of readmission was 3.0 days in both the modified and standard care groups.

Safety outcomes

It was expected that infants on modified oximeters would be managed on monitors that would report higher oxygen saturation than was actually the case and that reducing oxygen delivery could lead to an increase in the number of adverse events (AEs). It was also expected that infants on modified oximeters would be discharged home sooner than infants on standard monitors and that this might result in increased morbidity.

Deaths

Deaths from bronchiolitis are infrequent, and we did not consider that the intervention would result in more deaths. Two deaths were recorded during the study period; both were in the standard care group and were unrelated to the study intervention.

Admissions to high-dependency care

The study did not recruit infants who were directly admitted to a high-dependency area at admission (n = 56; see Figure 3). There were eight HDU admissions (in eight infants) in the standard care group and 13 (in 12 infants) in the modified care group.

Heart rate and respiratory rate at discharge

Tachycardia and tachypnoea are characteristic clinical features of infants admitted to hospital with bronchiolitis. We wished to determine if infants who were discharged home sooner in the course of their illness (as expected in the modified care group) had significantly higher heart and respiratory rates at discharge. Clinical observations of heart rate and respiratory rate were measured every 8 hours and the final measurement prior to discharge was used for discharge measurement. Contrary to expectation, there were no significant differences between standard and modified care groups in either heart rate (hazard ratio –1.16; p-value = 0.37) or respiratory rate (hazard ratio 0.09; p-value = 0.88) at discharge.

Readmission to hospital

Readmission to hospital was reported as a serious adverse event (SAE) during the first 28 days following randomisation. In the first 7 days after randomisation there were eight readmissions to hospital in six infants in the standard care group and five readmissions in five infants in the modified care group. By 28 days there had been 26 readmissions in 23 infants in the standard care group and 12 readmissions in 12 infants in the modified care group. It was not expected that there would be fewer admissions in the modified care group, as this group was expected to be discharged sooner in the course of their illness than those in the standard care group. The absolute reduction of three readmissions (relative reduction 38%) at 7 days is not a clinically important difference, but an absolute difference of 14 readmissions (relative reduction 54%) at 28 days could be considered a clinically important difference (a 5% absolute reduction in readmission rate for infants in the modified care group).

Reattendance at health care

An earlier discharge from hospital may have been associated with a greater number of contacts with health care after discharge. The numbers of heath-care contacts at 7, 14 and 28 days and at 6 months after randomisation were similar between groups (Table 11). There were no significant differences in the number of infants with a health-care contact at 7 days after randomisation (see Table 11). Infants in the modified care group did not experience a higher number of health-care contacts up to 6 months following randomisation.

TABLE 11 - Outcomes: safety

IQR, interquartile range.

Parameter	Standard care	Modified care	All	Mean difference (95% CI)	Odds ratio (95% Cl)	p-value
Deaths	2	0	2	–	–	–
High-dependency care (episodes)	8	13	21	–	–	–
Heart rate at discharge, median	133 (IQR 122–144)	135 (IQR 123–147)	134 (IQR 122–145)	–1.16 (–3.70 to 1.37)	–	0.37
Respiratory rate at discharge, median	38 (IQR 34–41)	38 (IQR 34–42)	38 (IQR 34–42)	0.09 (–1.05 to 1.23)	–	0.88
Readmission to hospital within 7 days (episodes)	8	5	13	–	–	–
Readmission to hospital within 7 days (infants)	6	5	11	–	–	–
Readmission to hospital within 28 days (episodes)	26	12	38	–	–	–
Readmission to hospital within 28 days (infants)	23	12	35	–	–	–
Reattendance at health care within 7 days (episodes)	41	43	84	–	–	–
Reattendance at health care within 7 days (infants)	39 (14.4%) (n = 270)	34 (12.7%) (n = 267)	73 (13.6%) (n = 537)	–	0.98 (0.65 to 1.49)	0.94
Reattendance at health care within 14 days (episodes)	92 (n = 270)	88 (n = 267)	180 (n = 537)	–	–	–
Reattendance at health care within 14 days (infants)	76 (28.5%) (n = 267)	70 (27.1%) (n = 258)	146 (27.8%) (n = 525)	–	1.07 (0.73 to 1.57)	0.73
Reattendance at health care within 28 days (episodes), n	199 (n = 290)	188 (n = 288)	387 (n = 578)	–	–	–
Reattendance at health care within 28 days (infants)	127 (46.4%) (n = 274)	128 (48.9%) (n = 262)	256 (47.6%) (n = 536)	–	0.90 (0.64 to 1.27)	0.56
Reattendance at health care within 6 months (episodes)	802 (n = 295)	774 (n = 293)	1576 (n = 588)	–	–	–
Reattendance at health care within 6 months (infants)	214 (84.6%) (n = 253)	209 (80.1%) (n = 261)	423 (82.3%) (n = 514)	–	1.37 (0.87 to 2.16)	0.18
Antibiotics after discharge	24 (7.8%) (n = 305)	10 (3.3%) (n = 304)	34 (5.5%) (n = 609)	–	–	–
SpO2 measured at 28 days (season 1 only)	99 (IQR 97–100) (n = 94)	99 (IQR 97–100) (n = 101)	99 (IQR 97–100) (n = 195)	0.111 (–0.350 to 0.572)	–	0.64

Antibiotics after discharge

An earlier discharge from hospital and attendance at primary care in an earlier part of disease recovery might be associated with a higher level of antibiotic prescribing after discharge in the modified care group. Infants in the modified care group received fewer courses of antibiotics in the 28 days after discharge than the standard care group (see Table 11; a 58% relative reduction).

Hospital treatments

The hospital treatments received by both treatment groups are shown in Table 12. Use of intravenous fluids, antibiotics and bronchodilator (salbutamol) was similar in both groups.

TABLE 12 - Patient therapies in hospital

Parameter	Standard, n (%)	Modified, n (%)	All, N (%)
Need for supplemental oxygen	223 (73.1)	169 (55.6)	392 (64.4)
Use of nasogastric tube feeding	141 (46.2)	125 (41.3)	266 (43.8)
Use of intravenous fluids	29 (9.5)	28 (9.2)	57 (9.4)
Use of antibiotics	44 (14.4)	39 (12.8)	83 (13.6)
Use of salbutamol	25 (8.2)	21 (6.9)	46 (7.6)

Supplemental oxygen

Supplemental oxygen was provided to 73.1% of infants in the standard care group and 55.6% of those in the modified care group. It was understood that there could be a difference between the groups, as oxygen saturation measured in the ED may decline following admission to hospital (in association with disease progression and continuous monitoring). The difference (17.5%) should approximate to those infants in the modified care group in whom oxygen saturation fell below 94% but remained above 90%, so the monitor would display at ≥ 94% and they would not have been commenced on supplemental oxygen. In other words, approximately one in six infants admitted to hospital who would receive oxygen at a target oxygen saturation of 94% would not do so at a target oxygen saturation of 90%.

Use of nasogastric tube feeding

The proportion of infants receiving nasogastric tube feeding was lower in the modified care group (41.3%) than in the standard care group (46.2%).

Anxiety scores

We wished to understand whether or not earlier discharge from hospital at an earlier stage of disease recovery might be associated with greater anxiety for parents looking after their child at home.

Parents’ levels of anxiety were similar at the time of their child’s admission to hospital, but the intervention did not result in parents experiencing greater levels of anxiety and there were no significant differences in anxiety scores at 7, 14 and 28 days or at 6 months (Table 13).

TABLE 13 - Parental/family outcomes

IQR, interquartile range.

Parameter	Standard care, median (IQR)	Modified care, median (IQR)	All, median (IQR)	Mean difference (95% CI)	p-value
Anxiety score at admission	7 (4–10)	7 (4–11)	7 (4–11)	–	–
Anxiety score at 7 days	4 (2–8)	4 (2–7)	4 (2–7)	–0.18 (–0.75 to 0.39)	0.53
Anxiety score at 14 days	3 (1–6)	3 (1–5)	3 (1–5)	0.00 (–0.57 to 0.56)	0.99
Anxiety score at 28 days	3 (1–7)	3 (1–6)	3 (1–6)	–0.27 (–0.88 to 0.34)	0.39
Anxiety score at 6 months	4 (1–7)	3 (1–6)	4 (1–7)	–0.16 (–0.82 to 0.49)	0.62

Family life

Discharging a child from hospital earlier in the course of their illness could have an impact on family life, with lead carers having to give up time to care for the child, secondary carers taking additional time off work and children returning later to paid and unpaid child care than they otherwise would have done (Table 14). Lead carers were generally mothers (94.4% overall), and mothers of infants in the modified care group lost fewer hours to usual activities, with 23% fewer hours lost at 7 days, 28% at 14 days, 25% at 28 days and 21% at 6 months. In contrast, levels of activity lost by secondary carers (typically fathers) were similar regardless of intervention group, and most of this time was lost in the first 7 days. Relatively few children had childcare placements and the time to return to child care was similar between groups.

TABLE 14 - Carer hours missed from usual activities

IQR, interquartile range.

Parameter	Standard	Modified	All
Lead carer (= mother), n (%)	288 (95.4)	283 (93.4)	571 (94.4)
Lead carer work status (= employed), n (%)	43 (14.2)	39 (12.9)	82 (13.5)
Lead carer hours missed, 0–7 days, median (IQR)	58.3 (25.1–96.5)	44.8 (24.6–72.0)	49.3 (24.8–85.5)
Lead carer hours missed, 0–14 days, median (IQR)	62.3 (25.7–97.3)	45.0 (25.3–72.7)	50.9 (25.6–85.5)
Lead carer hours missed, 0–28 days, median (IQR)	63.4 (27.5–101.2)	47.3 (25.7–76.4)	53.4 (26.3–89.3)
Lead carer hours missed, 0–6 months, median (IQR)	67.6 (28.1–114.8)	53.2 (30.4–86.2)	59.0 (29.0–97.8)
Secondary carer work status (= employed), n (%)	233 (83.3)	251 (89.6)	484 (86.5)
Secondary carer hours missed, 0–7 days, median (IQR)	15.0 (8.0–24.0)	16.0 (8.0–27.0)	15.0 (8.0–24.0)
Secondary carer hours missed 0–14 days, median (IQR)	15.0 (8.0–27.5)	16.0 (8.3–30.5)	16.0 (8.0–30.0)
Secondary carer hours missed 0–28 days, median (IQR)	16.0 (8.0–33.8)	17.3 (8.5–34.0)	16.5 (8.0–34.0)
Secondary carer hours missed 0–6 months, median (IQR)	22.0 (10.0–43.0)	18.0 (9.0–36.0)	20.0 (9.5–40.0)
Child well enough to return to child care by 7 days, n (%)	24 (48.0)	18 (52.9)	42 (50.0)
Child well enough to return to child care by 14 days, n (%)	28 (54.9)	22 (62.9)	50 (58.1)

Prespecified subgroup analyses of secondary outcomes

Time to fit for discharge by oxygen supplementation and by treatment allocation

Time to fit for discharge by oxygen requirement is demonstrated in Figure 9. We have already shown that, overall, infants in the modified care group were fit for discharge earlier than those in the standard care arm (hazard ratio 1.46, 95% CI 1.23 to 1.73). There was a significant difference (p-value = 0.04) in the magnitude of this treatment effect between those who received oxygen supplementation (hazard ratio 1.37, 95% CI 1.11 to 1.69) and those who did not (hazard ratio 0.95, 95% CI 0.71 to 1.26). Among those who received oxygen supplementation, infants in the modified care group seemed to do better than those in the standard care group, whereas among those who did not have oxygen supplementation, the groups were similar. These results are difficult to interpret, as some of this oxygen supplementation occurred after randomisation, and will therefore have been affected by the allocated treatment: more infants in the standard care group were given oxygen supplementation, as we would expect.

FIGURE 9.

Time to fit for discharge by oxygen required.

Time to discharge by oxygen supplementation and by treatment allocation

Time to discharge by oxygen requirement is demonstrated in Figure 10. We have already shown that, overall, infants in the modified care group were discharged earlier than those in the standard care arm (hazard ratio 1.28, 95% CI 1.09 to 1.50). There was a significant difference (p-value = 0.002) in the magnitude of this treatment effect between those who received oxygen supplementation (hazard ratio 1.22, 95% CI 0.99 to 1.49) and those who did not (hazard ratio 0.71, 95% CI 0.54 to 0.94). Among those who received oxygen supplementation, infants in the modified care group seemed to do better than those in the standard care group, whereas among those who did not received oxygen supplementation, infants in the standard care group seemed to do better than those in the modified care group. These results are difficult to interpret, as some of this oxygen supplementation occurred after randomisation and will therefore have been affected by the allocated treatment: more infants in the standard care group were given oxygen supplementation, as we would expect.

FIGURE 10.

Time to discharge by oxygen requirement.

Serious adverse events

Serious adverse events were recorded up to 28 days and were recorded in 56 participants (Table 15a and 15b). There were 35 SAEs in 32 infants in the standard care group and 25 SAEs in 24 infants in the modified care group. The standard care group had 8 infants with an HDU transfer, 23 with a readmission and one with prolonged hospitalisation. The modified care group had 12 infants with a HDU admission and 12 infants with a readmission to hospital.

TABLE 15a - Serious adverse events: number of patients

Parameter	Category	Allocated intervention		Overall, N = 615
		Standard pulse oximeter, n = 308	Modified pulse oximeter, n = 307
Number of patients with at least one SAE	Overall	32	24	56
Number of patients with SAEs: HDU transfer	Overall	8	12	20
Number of patients with SAEs: readmission	Overall	23	12	35
Number of patients with SAEs: hospital prolongation for ‘other’	Overall	1	0	1

TABLE 15b - Serious adverse events: number of events

Parameter	Category	Allocated intervention		Overall, N = 615
		Standard pulse oximeter, n = 308	Modified pulse oximeter, n = 307
Number of SAEs	Overall	35	25	60
Number of SAEs by classification	HDU transfer	8	13	21
	Hospital prolongation for ‘other’	1	0	1
	Readmission	26	12	38
Number of SAEs by severity	Mild	11	1	12
	Moderate	17	15	32
	Severe	7	9	16
Number of SAEs by outcome	Recovered	34	25	59
	Death	1	0	1

Adverse events

Adverse events were recorded to 28 days and were recorded in 144 infants (Table 16a and 16b). An AE was recorded at least once for each SAE. Overall, there were 173 AEs recorded in 144 infants. There were 89 AEs in 75 infants in the standard care group and 84 AEs in 69 infants in the modified care group. No important differences between groups were noted in terms of subcategory (respiratory, gastrointestinal, other) or severity of AEs.

TABLE 16a - Adverse events: number of patients

Parameter	Category	Allocated intervention		Overall (N = 615)
		Standard pulse oximeter (n = 308)	Modified pulse oximeter (n = 307)
Number of patients with at least one AE	Overall	75	69	144
Number of patients with respiratory AEs by severity	Mild	26	23	49
	Moderate	17	18	35
	Severe	6	9	15
Number of patients with gastrointestinal AEs by severity	Mild	7	7	14
	Moderate	0	5	5
Number of patients with other AEs by severity	Mild	19	12	31
	Moderate	6	2	8
	Severe	2	0	2

TABLE 16b - Adverse events: number of events

Parameter	Category	Allocated intervention		Overall (N = 615)
		Standard pulse oximeter (n = 308)	Modified pulse oximeter (n = 307)
Number of AEs	Overall	89	84	173
Number of respiratory AEs by severity	Mild	28	25	53
	Moderate	17	20	37
	Severe	7	9	16
Number of gastrointestinal AEs by severity	Mild	7	8	15
	Moderate	0	5	5
Number of other AEs by severity	Mild	19	12	31
	Moderate	8	2	10
	Severe	2	0	2

Exclusion of infants under 6 weeks of age

Our exclusion criteria were kept to a minimum, as we wished, if successful, for this study to be widely applicable to acute bronchiolitis admissions; however, we did exclude infants under 6 weeks of age. The reasons for this were twofold. First, in our recruitment feasibility assessment in the season prior to the study, parents of children in this age group indicated that they would decline consent to the study because of concerns about the age of their child and their first acute illness. Second, infants in this age group frequently present with apnoea and, although we accommodated this within the protocol, the clinical and parental anxiety associated with infant apnoea may have provided an undue skew to greater length of stay in this age range. We consider that infants under 6 weeks of age will require a higher degree of personalisation of oxygen saturation targets depending on their disease course. Although there are some who may be stable and able to be managed at an oxygen saturation target of ≥ 90%, there will be others who will require a higher target for management and discharge (particularly those with apnoea) and this will be as is clinically appropriate.

Measures of concurrent symptom relief

Supplemental oxygen is provided in hypoxaemia for both tissue oxygenation and perceived symptom relief. In adults, there is no demonstrable effect of supplemental oxygen on relief of respiratory symptoms. 4,5 We used proxy measures of symptom relief, assessing heart and respiratory rate as indicators of comfort in acute respiratory disease. At time of discharge, heart rate and respiratory rate were similar in the two groups, suggesting that stopping supplemental oxygen sooner in the modified care group was not associated with an increase in discomfort reflected by an increase in respiratory or heart rate.

Measurement of clinical scores

The study did not use a bronchiolitis clinical score. A range of clinical scores are reported in studies, some specific for bronchiolitis, others adapted from asthma scores. Our reason for not using a bronchiolitis score was threefold. First, bronchiolitis scores are not used clinically in the majority of UK hospitals, in particular in our study site partners, as they have not been demonstrated to be of greater value than routine clinical decision-making. Second, there is no agreed best clinical score. Third, agreement between observers tends to be poor unless the number of trained observers is limited. 54 To have study staff available 24 hours per day for scoring would have been expensive for measurement of single outcome. The alternative approach of training clinical staff across all sites to clinical score accurately and precisely for bronchiolitis may have changed behaviour with regard to routine care (which we wished to observe) and still have been associated with unacceptable variance in scoring with corresponding concerns for data surety.

Measures of neurocognitive development

Neurocognitive delay has been associated with hypoxaemia in children, with most anxieties stemming from observation of lower school attainment in children with obstructive sleep apnoea. 54 Such children experience recurrent, variable, hypoxaemia with obstructed breathing while asleep. Hypoxaemia in acute respiratory illness tends to be less variable, and often less severe, and with a pattern of resolution over a significantly shorter period of time. Longer periods of hypoxaemia than that experienced by infants with bronchiolitis may have no associated neurocognitive impact. Preterm infants maintained at an oxygen saturation target of 91–94% had no neurocognitive deficit at 2 years compared with those maintained at an oxygen saturation target of 95–98%. 55 Neurocognitive scores in children with mild/moderate obstructive sleep apnoea observed for a period of 6 months were no different from scores in children who had undergone immediate tonsillectomy. 56 These studies support the perspective that children may be neurocognitively tolerant of short-term borderline hypoxia.

Compliance with study protocol

It could be considered a limitation of the study that we did not collect and download the time-matched ‘true’ oxygen saturation of the modified oximeters; this was predominantly for logistic and cost reasons, and we considered that the additional study resource would not have proportionally added to the study outcome or understanding of the effects of oxygen in borderline hypoxia. The differences noted for use of oxygen and time to stopping supplemental oxygen in each of the groups suggest good compliance to the study protocol in those who received the intervention. We were not aware of any episodes of unblinding associated with the study protocol; hospital oximeters were removed and a study oximeter applied after an interval of 1 minute. The average difference between the hospital and study oximeter display was 2% SpO₂ and, as a consequence, to our knowledge, there were no instances of accidental unblinding.

Reducing variance in clinical practice

This is the first study of oxygen saturation targets for acute respiratory infection in children in a developed health-care setting. The recommendation of the AAP that an oxygen saturation of ≥ 90% is acceptable in acute bronchiolitis has not been widely adopted, with varying practice and continued debate. 19,57 The AAP recommendation has, however, led to a drift in clinical practice without sufficient evidence, with clinicians now managing bronchiolitis to a range of oxygen saturation targets even within the same hospital. The risks of practice drift are well demonstrated by recent studies in preterm infants, in which a progressive clinical acceptance of lower oxygen saturation targets58 was only subsequently shown to be associated with increased risk of death in exposed infants. 7 The results of this study therefore enable the range of oxygen saturation targets currently in use to be coalesced into a clear oxygen saturation target of ≥ 90%.

Unifying oxygen management strategy for acute bronchiolitis

Debate in oxygen management in acute respiratory disease focuses on target oxygen saturations, whether or not oxygen saturation monitoring should be continuous or intermittent and for how long oxygen saturation should be observed to be stable prior to discharge. The debate on oxygen saturation targets is presented in the rationale for the study in Chapter 1, Controversies in approach to hypoxaemia in bronchiolitis, but important within this is the differential approach to starting and stopping oxygen supplementation. In guidelines for respiratory disease, target oxygen saturation for commencing supplemental oxygen is typically lower than that for stopping. In children with respiratory infection, lower oxygen saturation early in the course of the illness often represents clinical instability and the clinical logic for a higher threshold at these times is unclear. A strength of our study was the use of a single target oxygen saturation for starting and stopping supplemental oxygen.

There are concerns that continuous oxygen saturation monitoring leads to clinical overinterpretation of minor physiological and artefactual brief and self-correcting desaturation. Such minor desaturation is considered to delay patient progress to management in air and, consequently, discharge. Unfortunately, there is no agreement on the frequency or duration of intermittent monitoring, and, as a consequence, we considered it safest to provide continuous monitoring over a short period of time.

There is also no agreement on the length of time that infants should be observed to have stable oxygen saturation in room air prior to discharge, with 8–24 hours considered appropriate by over two-thirds of clinicians and 24–48 hours considered appropriate by nearly one-fifth (Clare van Miert, University of Liverpool, 2012, personal communication).

In developing our protocol, while acknowledging these concerns, we hoped to provide a pragmatic, sensible and safe approach that would not miss important desaturation events in acutely infected infants but also would not inappropriately prolong admissions. We therefore elected to have a shorter period of continuous monitoring. The results of this study, by combining these three issues of debate, provide an evidence-based structure for clinical decision-making at discharge.

Applicability within UK hospitals and beyond

In general, the care in each of the paediatric units within the BIDS study team was similar and we would consider that the results are generalisable to the care of infants with bronchiolitis in UK paediatric hospitals. We have no reason to believe that the study results would not be applicable to other similar health-care settings across the world. Infants living at higher altitudes have lower oxygen saturation in health and may be managed at lower oxygen saturation limits during disease. 1 This study was performed at sea level and, therefore, is widely applicable to most urban areas, in line with recent requests for evidence at sea level. 59

Although the study protocol was for infants admitted to hospital, we deliberately chose a more intense (continuous monitoring) but shorter period of observation (4 hours) than most current practice. The debate of continuous or intermittent oxygen saturation monitoring is discussed above; however, our shorter period of observation was intended to capture infants with important prolonged episodes of desaturation during a short period of observation, such that the results could be applicable to acute paediatric observation areas (short stay, acute assessment units, etc.) and negate the need for admission if the criteria for safe discharge were fulfilled. Such acute observation areas typically have fewer time pressures than EDs and therefore provide a better safety net for sick infants.

Use of oxygen saturation target ≥ 90% in emergency departments and primary care

The study did not address appropriate oxygen saturation targets for primary care and EDs. Infants in the first year of life with an acute respiratory infection are a vulnerable patient group. Our study boundary when considering the ethical approach to this research question was the knowledge of a greater risk of death in children with acute respiratory infection managed below an oxygen saturation of 90%,9 a finding recently also seen in infants born preterm. 7 Bronchiolitis typically has good outcomes, with few deaths, under current management strategies. Although there is some evidence that infants recovering from bronchiolitis have temporary dips below 90% at home,60 we did not consider it ethical or appropriate to devise a protocol that would lead to a sustained period below the 90% oxygen saturation threshold, and therefore we restricted the trial to those who would be observed in hospital.

In infants with bronchiolitis presenting to an ED with an oxygen saturation of ≤ 92% there is a high probability that oxygen saturation will fall below 90% during the observation period. 61 Among ED physicians, the threshold for admitting an infant with an oxygen saturation of 92% is much lower than that for admitting those with an oxygen saturation of 94%. 2 These studies, together with the wish to provide a sufficient safety net to young infants, persuaded us that a study of a 90% oxygen saturation target in EDs may not be in the best interests of infants with acute viral bronchiolitis, and a better understanding of the safety and clinical impact of target oxygen saturations during a typically longer period of observation would be most appropriate. The same reasoning would be applicable to primary care.

Applicability of oxygen saturation target ≥ 90% in other acute respiratory disease in childhood

The generalisability of targeting oxygen saturation ≥ 90% in other acute respiratory conditions has not been tested in this study. Our population was infants under 1 year of age with acute respiratory infection, who could be considered potentially more vulnerable than many older children with acute respiratory infection. In children recovering from acute pneumonia (viral or bacterial) or acute virus-induced wheeze, oxygen saturation often follows a pattern of a long tail of recovery (particularly during sleep), during which time the child’s clinical status may have significantly improved, and with no or stable chest signs on auscultation, and with hospitalisation needed only for the provision of supplemental oxygen. Clinicians caring for children who are recovering from such illnesses and who are cardiovascularly stable may consider targeting to a lower threshold of ≥ 90% oxygen saturation for hospital care and discharge. Until appropriately tested, children with acute pneumonia and acute asthma/wheeze at presentation should continue to receive supplemental oxygen according to current guideline recommendations (typically a target of ≥ 92%) because of the risk of acute change in symptoms and tissue hypoxia in these conditions during the acute phase of the illness.

Potential risks not explored within this study

The principal risk not explored by this study methodology was that children who present with clinical bronchiolitis may have an alternative diagnosis. Many conditions masquerading as bronchiolitis would typically be picked up as current (e.g. congenital heart disease). Possible exceptions are rare lung diseases in children, which may present a similar clinical picture similar to bronchiolitis. Children with such conditions (e.g. neuroendocrine hyperplasia of infancy or bronchiolitis obliterans) often struggle to maintain oxygen saturation in air ≥ 90%, and in some cases the presentation will resemble that of recurrent bronchiolitis with oxygen saturation maintained at ≥ 90% but < 94%. Clinicians should be educated that any infant presenting with a second episode of bronchiolitis should be assessed for lack of chest signs and normal oxygen saturation at discharge or follow-up.

Is 90% oxygen saturation an appropriate target threshold for supplementing oxygen in acute respiratory infection in children?

This study demonstrates that, in infants with bronchiolitis, management to a target oxygen saturation of 90% confers some modest benefits compared with management to 94%. A reasonable question is whether or not this benefit would be further extended at a target SpO₂ of < 90%. This inflection point for the oxyhaemoglobin dissociation curve would require careful ethical consideration, as it has been associated with increased deaths in preterm infants managed below this target. Exploring the clinical effectiveness of oxygen supplementation at an SpO₂ target of < 90% may be considered ethically more appropriate in health-care settings at altitude and/or with limited resources to supply supplemental oxygen in acute lower respiratory tract infection.

What is the clinical effectiveness and cost-effectiveness of community-based supplemental oxygen provision at home for infants with acute bronchiolitis who have been discharged home from hospitals that have adopted a target oxygen saturation of 90%?

In the past 10 years, with increasing pressure on paediatric inpatient beds, there has been a move to provide supplemental oxygen to infants with bronchiolitis at home once they are clinically stable. Although most reports come from the USA and Australia, there are some reports in the UK. The infrastructure to provide home care oxygen has involved primary care physicians and dedicated nursing teams. This has required financial expenditure due to the associated costs of home-based oxygen delivery and monitoring. To date there has been no published comprehensive cost-effectiveness analysis of this. Assuming that hospitals will adopt a 90% target for administration of supplemental oxygen in bronchiolitis (with fewer infants receiving supplemental oxygen for a shorter duration), a reasonable research question would be to gauge the clinical effectiveness and cost-effectiveness of home care oxygen-delivery systems in acute bronchiolitis.

What is the clinical effectiveness and cost-effectiveness of continuous or intermittent oxygen saturation monitoring on outcomes in acute bronchiolitis managed to a target oxygen saturation of 90%?

Measured oxygen saturation varies over time and with movement and sleep state. Minor dips in oxygen saturation are not considered to be of clinical significance but are thought to influence clinician behaviour unduly with respect to starting and stopping supplemental oxygen. Intermittent monitoring is understood to identify only important changes in baseline oxygen saturation. Some clinicians monitor oxygen saturation intermittently as infants improve, but typically to a higher baseline target than the 90% SpO₂ recommended by this study. There are no agreed standards for frequency or duration of intermittent monitoring. A reasonable question, therefore, could be to identify whether or not intermittent monitoring to a target oxygen saturation of 90% could be clinically effective and cost-effective.

What are safe oxygen saturation targets for management of infants with acute bronchiolitis in primary care and emergency department discharge?

This report provides evidence for patients observed as inpatients. The majority of patients with bronchiolitis, however, are managed in primary care. Infants with bronchiolitis in the community may be maintaining adequate feeding but have an oxygen saturation of < 94% but > 90%. Currently, they would be referred to hospital and may be admitted (in most, SpO₂ is at or below 92%), but a reasonable question arising from our study is whether or not such infants could safely remain in the community under primary care review.

Other contributors

Mrs Kay Riding* (Lead Paediatric Research Nurse, Children’s Clinical Research Facility, Royal Hospital for Sick Children, Edinburgh) Involved in the set-up, conduct and acquisition of data of this research and assisted with TSC meetings.

Mrs Fiona Wee* (née Sloan) (Trial Manager for this trial until January 2012, ECTU, University of Edinburgh, Edinburgh) and Dr Morag MacLean* (Trial Manager for this trial from January 2012, ECTU, University of Edinburgh, Edinburgh) Involved in the day-to-day trial management, data management and were members of the Trial Management Group.

Bronchiolitis of Infancy Discharge Study (BIDS)

We would like to invite you to take part in our research study. Before you decide we would like you to understand why the research is being done and what it would involve for you and your child. One of our team will go through the information sheet with you and answer any questions you have.

Quick summary

Your child has been diagnosed with a common chest infection, ‘bronchiolitis’, and needs to come into hospital for help with feeding and/or breathing. Children admitted to hospital with bronchiolitis go home once they improve and are able to breathe and feed OK. At present we keep them in hospital until their blood oxygen has reached a normal level (more than 93% oxygen saturation), even though otherwise they would be fit to go home.

Children’s doctors in the USA have been advised that infants with bronchiolitis who have improved feeding and breathing can be managed without being given extra oxygen once blood oxygen levels are nearly normal (90% oxygen saturation or higher). Many doctors in the UK also think this is sensible as children recovering from bronchiolitis often have mildly low blood oxygen levels despite looking much better and feeding well. At present these children would need to stay in hospital, but we think that this time in hospital does not help recovery and could be time spent at home with family. This study is to investigate whether children going home from hospital, once breathing and feeding have improved, continue their recovery just as quickly as those who stay a little longer.

Our study doesn’t involve any additional tests. Apart from filling in a questionnaire at the start of the study, there will be no need for further contact with you by the study team until a week from now.

Part 1 tells you the purpose of this study and what will happen to you if you take part. Part 2 gives you more detailed information. Ask if there is anything that is not clear. The study diagram summarises what will happen to your child in the study.

Part 1

Part 2

Bronchiolitis of Infancy Discharge Study (BIDS)

We would like to invite you to take part in our research study. Before you decide we would like you to understand why the research is being done and what it would involve for you and your child. One of our team will go through the information sheet with you and answer any questions you have.

Quick summary

Your child has been diagnosed with a common chest infection, ‘bronchiolitis’, and needs to come into hospital for help with feeding and/or breathing. Children admitted to hospital with bronchiolitis go home once they improve and are able to breathe and feed OK. At present we keep them in hospital until their blood oxygen has reached a normal level (more than 93% oxygen saturation), even though otherwise they would be fit to go home.

Children’s doctors in the USA have been advised that infants with bronchiolitis who have improved feeding and breathing can be managed without being given extra oxygen once blood oxygen levels are nearly normal (90% oxygen saturation or higher). Many doctors in the UK also think this is sensible as children recovering from bronchiolitis often have mildly low blood oxygen levels despite looking much better and feeding well. At present these children would need to stay in hospital, but we think that this time in hospital does not help recovery and could be time spent at home with family. This study is to investigate whether children going home from hospital once breathing and feeding have improved continue their recovery just as quickly as those who stay a little longer.

Our study doesn’t involve any additional tests. Apart from filling in a questionnaire at the start of the study, there will be no need for further contact with you by the study team until a week from now.

Part 1 tells you the purpose of this study and what will happen to you if you take part. Part 2 gives you more detailed information. Ask if there is anything that is not clear. The study diagram summarises what will happen to your child in the study.

Part 1

Part 2

Bronchiolitis of Infancy Discharge Study (BIDS)

We would like to invite you to take part in our research study. Before you decide we would like you to understand why the research is being done and what it would involve for you and your child. One of our team will go through the information sheet with you and answer any questions you have.

Quick summary

Your child has been diagnosed with a common chest infection ‘bronchiolitis’, and needs to come into hospital for help with feeding and/or breathing. Children admitted to hospital with bronchiolitis go home once they improve and are able to breath and feed OK. At present we keep them in hospital until their blood oxygen has reached a normal level (more than 93% oxygen saturation), even though otherwise they would be fit to go home.

Children’s doctors in the USA have been advised that infants with bronchiolitis who have improved feeding and breathing can be managed without being given extra oxygen once blood oxygen levels are nearly normal (90% oxygen saturation or higher). Many doctors in the UK also think this is sensible as children recovering from bronchiolitis often have mildly low blood oxygen levels despite looking much better and feeding well. At present these children would need to stay in hospital, but we think that this time in hospital does not help recovery and could be time spent at home with family. This study is to investigate whether children going home from hospital once breathing and feeding have improved continue their recovery just as quickly as those who stay a little longer.

Our study doesn’t involve any additional tests. Apart from filling in a questionnaire at the start of the study, there will be no need for further contact with you by the study team until a week from now.

Part 1 tells you the purpose of this study and what will happen to you if you take part. Part 2 gives you more detailed information. Ask if there is anything that is not clear. The study diagram summarises what will happen to your child in the study.

Part 1

Part 2