In this cohort study, we found that when oxygen is titrated based on SpO2 levels, this results in occult hyperoxemia in a significant proportion of the patients. EMS personnel should especially be reluctant to administer (low flow) oxygen as a standard of care to patients who do not have clear respiratory compromise, as these patients are at a high risk of developing (occult) hyperoxia.

The potential harmful effects of long-term hyperoxia on vascular tone and cellular integrity are well known [4, 8]. Animal studies have indicated that even short-term (1-hour) hyperoxia maybe harmful, inducing functional and morphological changes in rat brains [21], and long-term changes in DNA-repair pathways, even at FiO2 levels as low as 0.3 [22]. In humans, brief exposure of only 15 min to a high FiO2 has been shown to result in an increase in systemic vascular resistance, with potential effects on CO [23, 24]. Although no studies have been published so far on the potential negative effects of (ultra) short exposure to moderate FiO2 levels (as most patients were exposed to in this study), based on these findings, hyperoxia should be avoided if possible.

Therefore several treatment guidelines nowadays focus on the prevention of both hypoxia- and hyperoxia [11, 12]. This study demonstrates however, that awareness of and/or adherence to these guidelines in the prehospital setting is not yet optimal: In 40% of the patients presented to the ED who demonstrated hyperoxia in their ABGA, SpO2 values were above the generally recommended target range of 94–98% (88–92% in severe COPD).

Interestingly, in patients with a history of ischemic heart disease, EMS providers seemed to be more aware of the potential risks of hyperoxia, as in this subgroup, obvious hyperoxia was less often present. This could be explained by the early and explicit emphasis placed by several societies on the potential deleterious effects of hyperoxia in patients with an acute coronary syndrome (ACS) or an out-of-hospital cardiac arrest [25, 26]. Interestingly, for patients with COPD we found the opposite: obvious hyperoxia (with SpO2 values above 92%) was encountered more often in this subgroup of patients, which may be a reflection of the generally higher concerns of developing hypoxia in this group.

This study demonstrates that in the majority of the cases, hyperoxia remains undetected by measuring SpO2 values alone. Over 60% of the patients in our cohort had occult hyperoxia: These patients had normal (or even decreased) SpO2 levels in the presence of an PaO2 > 13,5 Kpa. This may be explained by various reasons. First, the reliability of the SpO2 relies on the quality of the plethysmogram. In patients wearing nail polish, in shocked patients with cold extremities, in sick patients who are shivering or otherwise have movement artefacts, and in patients with dysrhythmia’s it is difficult to obtain a reliable SpO2 trace, and the SpO2 value represented may be an underestimation of actual values, resulting in unnecessary (high amounts of) oxygen administration More importantly, the relation between SpO2 and PaO2 is described by the oxygen dissociation curve. This means that in the upper range of saturations measured, a wide range in PaO2 values may be present: whereas some patients will be normoxemic, others will be hyperoxemic.

Although SpO2 is not a perfect tool to guide oxygen suppletion, often it is the only tool available, as it is important to start oxygen suppletion early to prevent hypoxia in many critically ill patients, and guidance of suppletion by arterial blood gasses (as is done in the intensive care unit (ICU)) is not always possible or desirable in the prehospital setting or the ED. Therefore it is important to establish how occult hyperoxia can be prevented when we only have SpO2 as a guidance, and our findings provide some clues:

In patients with occult hyperoxia, the majority (> 75%) of patients receiving high flow oxygen (by ventimask or NRM) had a P/F ratio < 300, indicative of ARDS [27, 28]. These patients are likely also clinically pulmonary compromised, and hence oxygen therapy is started liberally to prevent hypoxia. For further guidance and prevention of hypoxia, blood gas analysis is warranted. In these patients, a brief period of hyperoxia is likely unavoidable. In contrast, > 75% of the patients receiving low flow oxygen had a P/F ratio > 300. This shows that in patients who are likely clinically less compromised from a pulmonary perspective, and in whom oxygen is started (e.g. to meet increased metabolic demands or for comfort [29, 30]) through a nasal cannula, are at a particular high risk of developing hyperoxia. In these patients, who are generally also only moderately sick (judged by their average triage category), prehospital and ED clinicians should carefully weight potential risks and benefits of oxygen administration, and overall be more reluctant before starting oxygen.

This study had several limitations. First, as this was a proof-of-principle descriptive study, no sample size estimation was performed. Some trends may therefore have remained undetected or may have yielded non-significant results.

Second, patients were selected from the Acutelines data-, image and biobank of a single university hospital in the Netherlands. Only patients with the most-urgent NTS triage categories were included (as only these were included in the biobank). This may have affected the generalizability of our findings, although patients with the lowest triage categories rarely receive oxygen suppletion.

Further, FiO2 levels for calculation of P/F ratio’s were estimated (based on the method of oxygen delivery and the amount of oxygen given) and not measured. Although it is often assumed that the fraction of oxygen that is inspired (above the normal atmospheric level) increases by 4% for every additional liter of oxygen flow administered there may be inter- patient variability in the actual administered FiO2, depending on respiration rate, and pattern (mouth open or closed) [31]. Also, since no FiO2 values were noted in patients receiving non-invasive ventilation (NIV), no conclusion can be drawn about these patients concerning their P/F ratios.

Finally, as the primary outcome in our study was the number of patients with occult hyperoxia, and as we only included patients with confirmed hyperoxia in their ABGA, confounding by indications may be present. Therefore our findings should be interpreted with caution: Although our results demonstrate that hyperoxia remains undetected by SpO2 guided oxygen titration in a significant number of cases, no conclusions can be drawn regarding overall prevalence of hyperoxia in cohorts were SpO2 guided titration is used.