When technology creates uncertainty: pulse oximetry and overdiagnosis of hypoxaemia in bronchiolitisBMJ 2017; 358 doi: https://doi.org/10.1136/bmj.j3850 (Published 16 August 2017) Cite this as: BMJ 2017;358:j3850
- Ricardo A Quinonez, associate professor (clinical) and section chief1,
- Eric R Coon, assistant professor (clinical)2,
- Alan R Schroeder, associate professor (clinical) and associate chief for research3,
- Virginia A Moyer, vice president, maintenance of certification and quality4
- 1Section of Pediatric Hospital Medicine, Department of Pediatrics, Baylor College of Medicine, Houston, TX, USA
- 2Department of Pediatrics, University of Utah School of Medicine, Salt Lake City, UT
- 3Department of Pediatrics, Stanford University School of Medicine, Stanford, CA
- 4American Board of Pediatrics, Chapel Hill, NC
- Correspondence to: RA Quinonez
Bronchiolitis admissions have increased substantially over the past 30 years, without a concomitant change in disease severity or mortality
The increase in admission rates for bronchiolitis coincides with the increased widespread use of pulse oximetry in the 1980s
Hypoxaemia may be a main driver for decisions to admit children with bronchiolitis, evidence shows
Overdiagnosis of hypoxaemia may be at least partially responsible for increased admission rates of children with bronchiolitis, recent evidence shows
Lower thresholds for oxygen concentration to determine treatment may be associated with better outcomes such as decreased length of stay, with no demonstrated evidence of adverse outcomes
Karl Matthes, a German physician, is credited with developing the first device to measure oxygen saturation.1 Wide scale use began in the 1980s with the introduction of the first commercial devices.2 Since then, pulse oximetry has permeated healthcare to the point where it has been dubbed the “fifth vital sign” and has become the standard method for non-invasive measurement of oxygenation.34
Pulse oximetry is widely used to help evaluate children with bronchiolitis, a lower respiratory tract viral infection that primarily affects children in their first two years.5 Excluding live births, bronchiolitis is the leading cause of hospital admission in infants under 12 months of age in the United States, accounting for 18% of all admissions in this age group. Each year nearly 3% of all US infants are admitted with bronchiolitis.6
Bronchiolitis is a self limited illness for which most therapeutic interventions have failed to have a positive effect on important outcomes. Treatment, even when children are admitted to hospital, is primarily supportive.57 Different strains of respiratory syncytial virus, the most common agent responsible for bronchiolitis, contribute to differences in disease severity.8
However, the trends in temporal variations for individual strains of respiratory syncytial virus have no direct correlation with a noted significant increase in admission rates for bronchiolitis.910 This increase, which coincides with the growing use of pulse oximetry in the 1980s, has led to a near tripling in the number of admissions for bronchiolitis, which increased from 12.9 to 31.2 per 1000 children under 1 year of age.10 Notably, this impressive rise has not been accompanied by lower disease related mortality,11 leading several investigators to suggest that borderline hypoxaemia is being overdiagnosed.121314
Despite the prevalent use of pulse oximetry for monitoring, particularly in the hospital setting, agreed standard definitions of hypoxaemia are lacking in guidelines and in clinical practice.15 For example, the American Academy of Pediatrics (AAP) clinical practice guideline for bronchiolitis proposes an oxygen saturation threshold of 90%,7 while the UK National Institute for Health and Care Excellence (NICE) guideline recommends a threshold of 92%.16 Additionally, UK guidelines for asthma and pneumonia set a threshold for hypoxaemia at <94% and <92%, respectively.1718
Evidence for overdiagnosis
Overdiagnosis is generally defined as the identification of an abnormality where detection will have no net benefit for the patient.19 Three epidemiological approaches have historically been used to establish overdiagnosis,19 and each of these has been used to demonstrate overdiagnosis of borderline hypoxaemia in bronchiolitis.
Randomised trials affecting the probability of hypoxaemia diagnosis
Three trials have involved randomisation of children with bronchiolitis to varying levels of probability of a hypoxaemia diagnosis.
The Bronchiolitis of Infancy Discharge Study (BIDS) randomly allocated 615 infants admitted with bronchiolitis to monitoring with either a standard pulse oximeter or one that had been modified with altered algorithms.20 The oxygen saturation threshold for initiating or weaning oxygen was 94%, representing a saturation of 94% in the true oximetry arm and 90% in the modified oximetry arm. No difference was found in cough resolution, the primary outcome. However, time using oxygen was 22 hours shorter (P=0.002) and time to discharge 10 hours shorter (P=0.003) in the 90% threshold arm. The need for high dependency care and readmission to the hospital was no different between groups.
Schuh et al similarly randomised children presenting to an emergency department with bronchiolitis to having either accurate oximetry readings displayed to the treating physician or values artificially increased by 3 percentage points.21 Independent of clinical appearance, infants with falsely elevated values were significantly less likely to be admitted (25% v 41%; P=0.005), but they showed no difference in complications or unscheduled visit rates.
In a third randomised controlled trial, McCulloh et al assessed the use of continuous pulse oximetry versus intermittent pulse oximetry in children admitted with bronchiolitis, with the hypothesis that the decreased use of pulse oximetry would lead to less detection of hypoxaemia and subsequent oxygen use.22 The authors concluded that intermittent pulse oximetry was safe, in that no differences in escalations of care were seen between groups. Surprisingly, they found no difference in length of stay between groups, perhaps because the intervention was not initiated until children were weaned off oxygen.
These three studies suggest that exposing children to lower probabilities of diagnosis of hypoxaemia seems to be safe and could also lead to improved outcomes.
Delayed or missed diagnosis of hypoxaemia without patient harm
Transient oxygen desaturations often occur in healthy infants particularly during sleep.2324 Recently, Canadian investigators prospectively studied the frequency and degree of hypoxaemia in patients with bronchiolitis managed as outpatients, analysing the association between hypoxaemia and clinical outcomes.25 In this study 118 infants were discharged from the emergency department with pulse oximeters. Oxygen saturations were continuously recorded, but the oxygen threshold alarm and the oxygen saturation display were disabled. After a defined period the information on the oximeters was analysed and compared with the main study outcome of unscheduled medical visits due to bronchiolitis.
The majority (64%) of infants discharged home with bronchiolitis experienced episodes of desaturations, defined as oxygen saturation <90% for at least one minute, and 29 infants had sustained desaturations to 70% or less, levels that certainly would have triggered hospital admission and supplemental oxygen use under most protocols. These episodes of hypoxaemia were effectively missed diagnoses, given that parents and providers were not aware of them. Infants did not receive interventions for these episodes and were no more likely to experience an unscheduled medical visit or readmission than infants without desaturations.
Increased detection of hypoxaemia without a change in bronchiolitis outcomes
Increased detection of an abnormality or a disease accompanied by unchanged rates of important patient outcomes suggests that less severe forms of disease are being detected and, as such, overdiagnosed.19 Bronchiolitis admissions have nearly tripled since the 1980s, an era that coincides with increasing reliance on pulse oximetry.10 Bronchiolitis mortality rates did not change over the same period, however.11 This pattern suggests that children with less severe bronchiolitis are increasingly admitted without measurable benefit, probably driven by overdiagnosis of hypoxaemia.
While no formal meta-analyses have looked at overdiagnosis of hypoxaemia in bronchiolitis, a recently published systematic review,26 evaluating whether pulse oximeter use affects health outcomes in children, stated that pulse oximeter use can reduce mortality rates in children but can also “increase admission of children with previously unrecognized hypoxaemia” and “generally increase resource utilization.” Evidence supporting reduced mortality comes from one study of children with pneumonia in Papua New Guinea, which demonstrated lower mortality after multiple interventions were introduced including pulse oximeters, oxygen concentrators, and personnel training27—findings that may be less relevant to bronchiolitis management in developed countries.
Limitations of the evidence
While no investigations in bronchiolitis, to our knowledge, have demonstrated a measurable benefit to increased pulse oximetry use in short or medium term outcomes—such as preventing escalation of care, re-presentation, or mortality—most studies are underpowered to detect small differences in these uncommon, but important, endpoints.
Although some studies suggest improvement in mortality when oxygen is given to maintain oxygen saturation above 90%,28 this finding has been limited to critically ill premature infants treated over a longer period than what would be expected for an acute episode of bronchiolitis. It is certainly possible, however, that some as yet unquantified outcome attributable to the increased use of pulse oximetry in bronchiolitis has led to patient benefit. For example, potential long term cognitive benefits may arise from detecting and treating episodes of mild hypoxaemia, an outcome that has not been rigorously evaluated. Associations between mild desaturations into the 90-94% range in certain conditions and future neurodevelopmental delay have been previously described.2930
Because of this, some investigators have argued that the use of oxygen saturation thresholds in this range in bronchiolitis could lead to adverse cognitive outcomes.3132 However, the desaturations referred to in these studies occurred either in children living at high altitude in South America30 or in children with chronic conditions such as congenital heart disease,29 which are quite distinct from an acute episode of bronchiolitis in an otherwise healthy infant. Additionally, studies on the effects of mild hypoxaemia in other chronic conditions such as obstructive sleep apnoea and asthma have not shown any detrimental effect on long term neurocognitive outcomes.3334
One final limitation is the lack of studies examining parental preferences for monitoring in the bronchiolitis setting. Future work in this area should incorporate fully informed parental preferences and attitudes regarding pulse oximetry monitoring.
Overdiagnosis and harm
The best-documented harm attributable to widespread use of pulse oximetry in bronchiolitis is its association with unnecessary hospital admission and prolongation of hospital stay. While the recent randomised controlled trials by Schuh et al and Cunningham et al provided convincing evidence of the association between pulse oximetry use and hospital admission/length of stay,2021 earlier evidence came from the epidemiological study by Shay et al10 and from a randomised survey of paediatric emergency physicians, which demonstrated that lower oxygen saturations had a strong influence on the decision to admit with bronchiolitis.12 Admission is not only costly but also potentially harmful. A prospective study of children admitted for bronchiolitis found that adverse events while in hospital occur in as many as 10 per 100 admissions.35 These adverse events ranged from simple IV infiltrates to hospital acquired infections such as urinary tract infections or gastroenteritis, even in patients who were classified as “non-critically ill.”
After admission, the continued use of pulse oximetry has been associated with significant increases in length of stay. Schroeder et al demonstrated that a perceived need for oxygen on the basis of pulse oximetry readings led to an increased stay by about 1.6 days in a quarter of their patients,15 while Unger and Cunningham showed that 58% of the patients in their cohort had an increased stay because of this perceived oxygen need.36 Cunningham and McMurray also demonstrated that a pulse oximetry goal of 94% instead of 90% was associated with an increase in stay by as much as 22 hours in children admitted for bronchiolitis.37 And the BIDS trial found significantly shorter stay in the group with the lower oxygen saturation target.20
Aside from prolonging stay, supplemental oxygen may have other direct harms as well. While the harms of prolonged oxygen therapy in other clinical entities in paediatrics are well established,3839 it is unknown whether even low doses of oxygen for short periods are completely safe. In the BIDS trial, Cunningham et al not only demonstrated decreased length of stay with the lower 90% threshold but also reported two other outcomes of “time to feed adequately” (≥75% of usual) and “time to parental perception of back to normal.” In both cases the results favoured the group with the lower threshold for pulse oximetry, as infants in this group returned to adequate feeding 2.7 hours sooner (95% confidence interval –0.3 hours to –7 hours) and were considered back to normal 1 day sooner (0 days to 3 days).20
An additional concern relating to unnecessary pulse oximetry use is alarm fatigue, an issue highlighted as a recent Joint Commission national patient safety goal.40 In a secondary analysis of a prior quality improvement project41 aimed at reducing the time using pulse oximetry in patients with wheezing illnesses, Schondelmeyer et al found no difference in overall alarm frequency despite a significant decrease in pulse oximetry use.42 However, significant reductions in pulse oximetry use occurred in both the intervention and control hospital units in this single centre study. Given that monitors cannot alarm when they are off, future studies examining the effects of pulse oximetry reduction are likely to demonstrate an impact on alarm frequency and fatigue.
Finally, given its frequency, bronchiolitis is a costly disease: at least one study in children estimates the direct cost in the US at $652m (£502m; €556m) a year.43 In an economic analysis of their BIDS trial, Cunningham et al found a decreased cost of £290 (€321; $377) per patient in the group with lower oxygen saturations and £321 per patient when other variables such as travel and time off from work were factored in.44 The cost savings from preventing overdiagnosis could prove significant, both for individual patients and at a societal level.
The current body of evidence suggests that we should challenge assumptions regarding the detection and aggressive management of borderline hypoxaemia in non-critically ill infants with bronchiolitis (see box).
In 2013, as part of its participation in the American Board of Internal Medicine Foundation’s Choosing Wisely campaign, the Society of Hospital Medicine published a list of five therapies or tests to avoid, which can lead to overuse of healthcare resources in children in hospital.45 One of the five recommendations was to avoid continuous pulse oximetry in children admitted for respiratory illness who are not using supplemental oxygen. A key driver behind this recommendation was to avoid overdiagnosis of hypoxaemia.
Since then, the updated AAP clinical practice guideline on bronchiolitis has also advised against the use of continuous pulse oximetry in bronchiolitis.7 Separately, a single centre quality improvement project inspired by the Choosing Wisely recommendation used a standardised protocol to discontinue pulse oximetry in children admitted for bronchiolitis who met defined criteria.41 Their intervention appeared safe and resulted in significant reductions in the time using pulse oximetry in patients (10.3 hours pre-intervention v 3.1 hours post-intervention). Similar to the randomised controlled trial by McCulloh et al, the intervention did not occur until after participants were weaned off oxygen and was not associated with a reduction in stay.
Because overdiagnosis of hypoxaemia can also occur in infants taking oxygen, future investigations evaluating intermittent pulse oximetry use over the duration of admission will be informative. At least one such trial is currently under way (https://clinicaltrials.gov/ct2/show/NCT01646606).
The most impactful intervention may be to decrease overdependence on pulse oximetry as a major decision point in admission of children with bronchiolitis. Schuh et al showed the unequivocal influence of pulse oximetry values on admission and demonstrated the ubiquitous but benign nature of hypoxaemia in children with bronchiolitis who are treated as outpatients.2125 The AAP clinical practice guideline for bronchiolitis has set thresholds for hypoxaemia at the lower target of 90%.7 Cunningham et al’s work has demonstrated that this threshold is safe and suggests better outcomes than higher targets.20 However, given that oxygen desaturations to well below 90% are common in infants with25 and without23 bronchiolitis, even this lower target may result in substantial overdiagnosis of hypoxaemia.
Finally, as previously mentioned, parental perspectives in this discussion are lacking. Future research and guidelines should include parental perspectives of the acceptable risks and benefits of oxygen saturation monitoring.
Pulse oximetry as a technology represents a major and significant advance in medicine. Its prevalent use in the care of the critically ill patient with respiratory illness, and in the operating room as a monitoring tool for the anaesthetised patient, has contributed to safer care. However, its increasing and widespread use in stable infants and young children with bronchiolitis, a self limited disease with a generally benign course, has led to technology driven overdiagnosis of hypoxaemia—fueling uncertainty, increased use of resources, and patient harm.
Challenging assumptions for the detection of hypoxaemia in bronchiolitis
Pulse oximetry is the fifth vital sign, and all infants with bronchiolitis should have their oxygen saturation assessed.
A rigid cut-off threshold for treating hypoxaemia is needed to guide medical providers.
There are no strict thresholds for other vital signs in bronchiolitis.7
Continuous oximetry is safer than intermittent oximetry measurements.
The only trial to investigate this statement found equivalent risk for need of escalation of care between infants randomly allocated to intermittent versus continuous pulse oximetry.22
A quality improvement project similarly demonstrated that a reduction in continuous pulse oximetry use did not lead to any adverse events.41
Contributors and sources: RAQ conceptualised the manuscript, developed the initial outline and literature search strategy, drafted the initial manuscript, reviewed and critically revised the manuscript, and approved the final manuscript as presented. ERC and ARS reviewed and critically revised the manuscript and approved the final manuscript as presented. VAM conceptualised the manuscript, reviewed and critically revised the manuscript, and approved the final manuscript as presented.
Competing interests: We have read and understood BMJ policy on declaration of interests and declare that we have no competing interests.
Provenance and peer review: commissioned; externally peer reviewed.