Titration of oxygen therapy in critically ill emergency department patients: a feasibility study

This prospective cohort study was performed at the ED of the University Medical Center Groningen, a tertiary care teaching hospital with 33.000 ED visits/year. During a 14 week period from February until May 2016, consecutive patients > 18 years arriving at the ED, registered for cardiology, internal medicine, pulmonary medicine and emergency medicine, requiring oxygen therapy (according to the judgement of the ambulance nurse, ED nurse or ED physician) were included from 8.00 h a.m. until 23.00 h p.m. The institutional review board of our hospital waived informed consent. Exclusion criteria were: intubated patients or patients requiring immediate intubation, patients with unreliable pulse oximetry recording [8] and patients using bleomycin, because high concentrations of inspired oxygen (FiO2) increases lung toxicity in these patients.

Also, patients with a known specific target saturation due to other causes at presentation were excluded, such as patients with congenital heart disease in whom saturations < 85% were normal.

A protocol was developed aiming for normoxia, defined as PaO₂ 9,5–13,5 kPa (70–100 mmHg) or a corresponding oxygen saturation of 94–98%. Hyperoxia was defined - according to normal values used in our hospital - as PaO₂ > 13.5 kPa or SaO₂ > 98%, and hypoxia as PaO₂ < 9.5 kPa or SaO₂ < 94%. In consultation with our pulmonologists saturations between 90 and 92% were used to aim for normoxia in patients with known severe COPD (GOLD III or IV), in order to avoid hypercapnia, while further using the same study protocol (Fig. 1). The study protocol is shown in Fig. 1. For example, if a patient was admitted with a non-rebreathing mask (NRM) (15 L O₂/minute, FiO₂ = ± 0.8) and peripheral oxygen saturation (SaO₂) immediately after arrival was ≥98%, the nurse would apply a Ventimask (VM) 40% with 10 L O₂/minute (FiO₂ = ± 0.4). If after 5 min, in arterial blood gas analysis PaO₂ was still > 13.5 kPa or (peripheral) SaO₂ was ≥98%, oxygen was titrated towards an oxygen saturation (SaO₂) between 94 and 98%, preferably administered by a nasal oxygen cannula. If the patient was admitted with a nasal oxygen cannula and SaO₂ was ≥94% immediately after arrival, oxygen was titrated towards a SaO₂ between 94 and 98%. If SaO₂ was < 94%, a VM 40% with 10 L O₂/minute (FiO₂ = ± 0.4) would be applied and the same procedure as described above would be followed. Patients with known Chronic Obstructive Pulmonary Disease (COPD) GOLD III of IV dropped out of the study if they became hypercapnic (PaCO₂ > 6.0 kPa or > 45 mmHg) in the first hour after arrival. This threshold was chosen in consultation with our pulmonologists to prevent respiratory acidosis caused by relative hyperoxia lowering the respiratory rate. Oxygen saturation values were measured by non-invasive pulse oximetry [Masimo Corporation, CA, USA] and if the treating physician requested an arterial blood gas, PaO₂ as well as oxygen saturation were used [ABL800 flex, Radiometer, Copenhagen, DK].

All ED nurses and physicians were educated about the purpose of the study and the protocol. Data was collected using a case report form. The values of SpO₂ and PaO₂ (if available) were measured at ED presentation (T₀), after 5–10 min (T₁), after 1 h (T₂), and after 3 h (T₃) together with the dose of oxygen and the method used to administer oxygen. Vital signs were measured at the same time points. When the patient was admitted to a nursing ward or ICU, the same parameters were also collected 24 h after arrival at the ED (T₄). No advice was given about aiming for normoxia at the nursing ward or ICU. The Modified Early Warning Score (MEWS) was calculated on arrival and after 3 h.

The primary outcome parameter was whether the protocol was followed and whether normoxia was obtained. This was based on the measurements made by the pulse oxymeter, to enchance clinical applicability. The protocol was followed when during the time the patient was admitted at the ED (between arrival at the ED and 3 h after arrival at the ED) normoxia was achieved. If only at the final measurement (3 h after arrival at the ED) the SaO₂ was too high or too low, but the first three measurements were in the normal range, it was also concluded that the protocol was followed.

In the study period 201 patients were screened for inclusion; 162 patients were included. The flowchart of patient inclusion and exclusion is shown in Fig. 2. In 140 patients the protocol was followed and normoxia successfully achieved (86,4%) within 1 h after arrival at the ED. In 7 normoxic patients multiple blood gases were additionally sampled during ED admission, of these 6 were normoxic measuring both SaO2 and PaO2. In 22 patients the protocol was not followed, of which there were 5 known COPD patients who dropped out due the development of hypercapnia in the first hour at the ED. There were different reasons why the protocol was not followed in the other 17 patients. The most important observation was that most of these patients suffered from COPD. It appeared that in some COPD patients the threating physician seemed reluctant to follow the protocol, resulting in not reducing oxygen therapy in patients with severe COPD (n = 11) or too low oxygen saturations in patients with mild COPD (n = 3). For other patients we could not find a specific reason for not following the protocol.

The baseline characteristics of the 162 study patients are shown in Table 1. Significantly more patients had severe COPD (GOLD III or IV) in the group in which the protocol was not followed (P < 0.001). This group also received significantly more often oxygen before ED arrival.

An overview of the diagnosis at discharge from the ED for the total study population is shown in Table 2. Some patients were discharged with more than one diagnosis. In the group in which the protocol was not followed more patients were discharged with a COPD exacerbation (P = 0.03) or with other pulmonary conditions (P = 0.04) compared to the group in which the protocol was followed. Other diagnoses at discharge were not significantly different between the patient groups.

The median PaO₂ of patients arriving with prehospital oxygen was 9.5 (8.1–11.8) kPa. There is an effect of the method of oxygen delivery on causing hyperoxia at T_0. In patients arriving with a NRM, 50% was hyperoxic at T₀ compared to 18% of the patients with oxygen administered via a nasal cannula (P < 0.001).

Figure 3 shows the course of the oxygen saturation at the ED and 24 h after arrival at the ED for all included patients. When patients stopped receiving oxygen therapy and their saturation was > 94% or > 90% in severe COPD patients, they were considered to be normoxic. In some patients their oxygen saturation fluctuated during ED admission, resulting in shifting between hypoxia, normoxia and hyperoxia at the different time points. From the figure it becomes clear that most patients were in the normoxia range after 1 h at the ED. In the first 10 min (T₀-T₁) it was more difficult to attain normoxia, maybe due to the amount of oxygen treatment in the ambulance or due to the number of actions that have to be performed when a patient arrives at the ED. Some patients were normoxic at first but at T₃ they became hypoxic or hyperoxic, probably because they were checked less frequently. When patients were admitted to the ICU or nursing ward, most patients stayed in the normoxia range.

In patients with severe COPD the protocol was more often not followed. For this reason, we made a separate graph for this subgroup, shown in Fig. 4.

Figure 4 shows that almost half of the severe COPD patients were hyperoxic during the first 3 h after arrival at the ED. Almost none were hypoxic and at T₂ most patients were in the normal range. After being admitted to the nursing ward or ICU, more than half of the severe COPD patients (69%) stayed hyperoxic.