Introduction

Supplemental oxygen is often administered to preterm infants for hypoxemic episodes during respiratory distress or apnoeas. It is important to prevent hypoxaemia (defined as a decrease in arterial blood saturation (SpO2) of ≤80 % for ≥10 s), as frequent episodes could lead to an increased risk of morbidities, including retinopathy of prematurity (ROP), impaired growth, longer term cardio-respiratory instability and adverse neurodevelopmental outcome [12, 15, 30]. In extreme cases, it can even lead to death [12, 15]. Hyperoxaemia (SpO2 of >95 % for ≥10 s) also needs to be prevented, as administering supplemental oxygen can potentially lead to high oxygen levels. High concentration of oxygen is toxic to living cells and is known to be an important pathogenic factor for bronchopulmonary dysplasia (BPD) and ROP [31] and is correlated with cerebral palsy [3].

Pulse oximetry (PO) is most commonly used for continuous monitoring of oxygen saturation (SpO2) in a non-invasive manner [26]. To prevent hypoxaemia and hyperoxaemia, nurses usually titrate oxygen manually to maintain SpO2 between the prescribed target ranges. However, maintaining the SpO2 within this range can be challenging, and compliance—defined as the nurse’s behaviour that follows the clinical guidelines—[13] is influenced by several factors [40]. This compliance is important, as it can largely influence the effect of a certain SpO2 target range. The optimal range of SpO2 for preterm infants remains undefined, but recent trials have shown that aiming for 91–95 % has decreased mortality but increased incidence of ROP [36]. However, in these trials, oxygen was titrated manually, which caused a large overlap in the distribution of SpO2 between the two groups and may have decreased the observed differences in outcome.

Although comparison of SpO2 target ranges has been subject to systemic review [19], a review in the compliance in SpO2 target ranges is not available but equally important as which target range is optimal. The purpose of this study is to systematically review the available literature in compliance—and the factors influencing this compliance—in targeting SpO2 in preterm infants.

Methods

We performed a systematic review, following PRISMA guidelines where possible (Fig. 1) [28]. The aim of the PRISMA statement is to help authors improve the reporting of systematic reviews and meta-analyses, which made it a particularly useful framework for this report. Eligible studies were identified by searching online databases from January 1965 to January 2015 in PubMed, Embase, Web of Science, Cochrane, CINAHL and ScienceDirect (keywords in Table 1). After selecting the eligible studies, we manually searched the reference lists of the selected studies to identify additional references. The criteria for inclusion limited the selection to articles published in English or Dutch which referred to preterm infants, (nursing) compliance, SpO2 monitoring by PO and targeting oxygen saturation during NICU admission. Both qualitative and quantitative designs were included, but publications that were not primary research studies, i.e. letters, abstracts, reviews and editorials, were not (Fig. 1).

Fig. 1
figure 1

Flow diagram selection studies

Full size image
Table 1 Keywords in different databases
Full size table

Three authors (HvZ, RT, AH) independently graded the selected studies using the QualSyst tool for quantitative and qualitative studies [21]. In case of disagreement, consensus was reached through discussion or consultation of a fourth co-author (AtP). The QualSyst tool for quantitative studies is a validated generic checklist consisting 14 items with scores from zero to two and the possibility to score ‘not applicable’. Items rated not applicable were excluded from the calculation of the summary score. The maximum total score is 28. The summary score was calculated by summing the total score obtained across the relevant items and dividing that by the total possible score.

The QualSyst tool for qualitative studies is a validated generic checklist consisting of ten items with scores from zero to two, with the maximum total score of 20. A summary score was calculated for each study by summing the total score across the ten items and dividing them by the total possible score of 20 [21].

Data from selected studies were extracted using a data extraction form. The following study characteristics were extracted: author, year, design, sample, time points, length of measurement, target range and key results.

Results

Sixteen articles met the inclusion criteria for this review (Fig. 1), detailing studies that included a total of 574 infants and 2935 nurses. Fourteen of these studies used a quantitative design [1, 7–10, 17, 18, 22, 25, 27, 34, 38, 39, 41] while the remaining two used qualitative methods [2, 29]. There was no homogeneity in the study designs, so pooling the data for meta-analysis was not possible. We therefore discuss the studies and their results using a narrative format organized under thematic headings and summarized in tables.

Quality assessment

The studies varied in quality, but none was excluded because of low-quality scores. One observed weakness was the lack of power analysis in four of the studies [10, 17, 27, 39], and all studies were unclear in the reasoning behind the timing and duration of SpO2 data collection [1, 2, 7, 10, 17, 18, 22, 25, 27, 29, 34, 38, 39, 41] (Tables 2 and 3).

Table 2 Quality appraisal of included quantitative studies
Full size table
Table 3 Quality appraisal of included qualitative studies
Full size table

Study designs

The designs of the quantitative studies varied and were composed of the following: one efficacy study [9], two pilot clinical trials [6, 41], three randomized clinical trials [7, 18, 38] and eight observational studies, of which six had a prospective design [1, 10, 17, 22, 25, 27] and two were retrospective [34, 39] (Table 4). Both qualitative studies employed a descriptive design [2, 29] (Table 4).

Table 4 Summery of included studies
Full size table

TRs of SpO2

The lower limit of the target ranges (TRs) varied between studies from 80 to 92 % [17, 18], and upper limits of TR varied from 92 to 96 %, respectively [1, 9, 10, 17, 25, 27, 34] (Table 4).

Time points and length of measurements

All studies were conducted in the period that the infants needed supplemental oxygen, but the starting time points and duration of data collection differed between studies. The starting time point varied between the first day of life [1, 2] and 33 days [8] (Table 4). In one study, the postnatal age was not described [27]. The duration of data collection also varied widely, the shortest covering only 4 h [9] and the longest the entire period between admission and discharge [10]. The data were collected continuously in eight studies [1, 7–9, 18, 34, 38, 39, 41] and intermittently in the remaining studies [10, 17, 22, 27] (Table 4).

Compliance in TR

Twelve studies investigated how often SpO2 values were in or outside the TR, expressed as the percentage of monitored time [1, 7–9, 17, 18, 22, 25, 34, 38, 39, 41]. In a multicentre study, Hagadorn et al. observed that SpO2 was below, within or above TR in 16 (0–47 %), 48 (6–75 %) and 36 (5–90 %), respectively, of the monitored time [17]. Van der Eijk et al. reported similar values, finding that SpO2 was below TR for 16 % of the time and above it for 30 % [39]. In contrast, Lim et al. only studied infants receiving supplemental oxygen during CPAP and SpO2 was below TR for 9 % and above it for 58 % of the time [25].

Education and training

Two studies demonstrated the impact of an educational program in targeting SpO2. Laptook et al. observed that training did not change the time that SpO2 was below (26.9 vs. 26.6 %; not significant (ns)) or above TR (15.4 vs. 14.0 %; ns) [22]. Interestingly, Arawiran et al. even observed that training had an adverse effect and that the time that SpO2 was within TR decreased after training (44.5 ± 14.4 vs 40.4 ± 12.8 %) with an increase in time above TR (from 36.9 ± 17.2 vs 41.9 ± 15.6 %) [1].

Nurse/patient ratio

Sink et al. studied the influence of the nurse/patient ratio on compliance in SpO2 targeting. They observed that the proportion of time that SpO2 was below TR decreased from 0.06 to 0.03 and time above TR increased from 0.56 to 0.82 when a third or fourth patient was added to the nurse’s workload [34]. The high percentage of time above TR was probably due to the use of a lower upper limit (92 %) in comparison with other studies [7–9, 22, 38, 39]. Lim et al. also confirmed that more than one infant per nurse was associated with an increase in the time when SpO2 was above TR (Table 4) [25].

Automated regulation of inspired oxygen

Six recent studies reported that, when compared to manual titration, the use of automated regulation of inspired oxygen increased the time that SpO2 spent within TR [7–9, 18, 38, 41]. In a multicenter crossover study of ventilated preterm infants, Claure et al. (2011) observed that the time that SpO2 was within TR increased significantly during the automated period compared with the manual period (40 % (14) vs 32 % (13) (mean (SD) p < 0.001). The time periods with SpO2 >93 % or >98 % were thus significantly reduced during the automated period [7]. Although most studies observed that the time that SpO2 was above TR decreased [7–9, 38, 41] while the time below TR increased [7, 8, 38, 41], Hallenberger et al. found different results. They observed no change in time above TR (16 (0.0–60) vs 15.9 (1.9–34.8) p = 0.108) during automatic control of inspired oxygen and, therefore, no difference with manual control [18] (Table 4).

Compliance in alarm limit setting

Two studies investigated nursing compliance in setting the appropriate alarm limits for PO in preterm infants [10, 27]. The actual SpO2 values were not reported, but Clucas et al. observed that the lower and upper alarm limit was set correctly in 91 and 23 % of monitored time, respectively [10]. Mills et al. compared compliance in alarm settings of SpO2 according to whether or not infants participated in a trial. When infants were participating in the BOOST II trial, the lower and upper alarm limit for SpO2 was set correctly in 94 % (88–100 %) and 80 % (71–88 %) of the monitored time period. However, this decreased to 87 % (75–99 %) and 29 % (17–40 %) when infants were not participating in the trial [27] (Table 4).

Nurses’ perception and awareness

Armbruster et al. interviewed nurses who stated that the following would improve their compliance: further education, prompt response to alarm limits, a favourable patient to staff ratio, root cause analyses at the bedside and high priority given to control oxygen therapy [2]. Nghiem et al. reported that 63 % of the nurses were aware of the local oxygen saturation guidelines and 57 % of them correctly identified the target limits specified by their NICU guidelines (Table 4) [29].

Discussion

The wide variation in study methodologies made it necessary to use narrative reporting when discussing the results of this systematic review. Although the power of some of the studies was limited and the quality varied, all were considered eligible for inclusion. Moreover, they focused on different aspects of compliance in targeting SpO2. The design, TR of SpO2, time points and duration of each study differed.

The central finding is that compliance in targeting SpO2 was low, as were the alarm settings. All studies in compliance in oxygen targeting reported that maintaining the SpO2 below the upper limit was the most difficult to adhere to [1, 6, 7, 10, 17, 18, 25, 27, 34, 39, 41]. The analysis of the large clinical trials comparing lower- vs higher-oxygen-saturation TR was based on the intention to treat principle. However, the larger proportion of the SpO2 was either below or above the intended TR and there was also an overlap between the two TRs [4, 32, 36]. Although compliance was audited [27], it is possible that this has influenced the outcome of the trials. This underlines the importance in improving compliance in SpO2 targeting, as improved compliance could have influenced the results.

According to the studies

Several factors may play a role in low compliance in targeting oxygen saturation: lack of awareness of the TR settings, limited knowledge of the effects of hypoxaemia and hyperoxaemia and an increased nurse/patient ratio [2, 23, 25, 29, 34]. Many caregivers were unaware of the appropriate SpO2 limits [29]. In addition, nurses tend to rely on subjective observations for oxygen titration, such as skin colour and chest excursions, as well as PO and blood gases [35]. So far, studies indicate that the effects of education and training in improving the compliance targeting SpO2 are disappointing [1, 23].

On the other hand, the use of automated FiO2 regulation, which eliminates the need for the nurses’ compliance, has been shown to improve the time that SpO2 remains within TR [7–9, 38, 41]. The increase in time within TR was small, but it is possible that the effect of automated FiO2 regulation has been underestimated. A Hawthorne effect could have increased the nurses’ compliance during the short study period, thus decreasing the difference between the manual and automated periods. The effectiveness of automated regulation on oxygenation variability, and whether this results in an improved outcome, remains to be investigated [5].

It has been suggested that the absence of a FiO2 titration protocol would lead to saturations which would frequently exceed or fall below the TR [24]. Manual adjustments of FiO2 can vary widely in frequency and step size, so standardization of these adjustments could decrease large fluctuations in SpO2 [39]. After implementing an oxygen titration protocol for reducing the incidence of severe ROP, Lau et al. observed that the period during which SpO2 was above TR decreased significantly [24].

Although fewer studies investigated this, compliance with alarm settings appeared to be low as well, especially the upper alarm limit [10, 27]. In addition, even when alarm limits are appropriately set, caregivers seem to have a preference for SpO2 close to the upper alarm limit [4, 20]. This was also demonstrated in the large trials comparing TR of SpO2 [36]. It is possible that caregivers are more accustomed to preventing hypoxaemia than hyperoxaemia. It is also possible that infants are more stable in SpO2 when kept at the higher end of the TR. A regular check of alarm limit settings each shift could increase awareness of this issue.

Educational programs on hyperoxaemia improved knowledge levels [11, 16] but did not lead to better compliance. Earlier research has shown that after education in risks related to hyperoxaemia, the nurses’ performance was still variable and only 51 % of nurses were successful in minimizing exposure of their infants to hyperoxaemia [37]. Nurses usually take care of more than one patient and perform multitasking [14], and an increased workload decreases their compliance in TR [25, 34]. Also, nurses frequently have to deal with alarms, but a large proportion of the alarms are false [33]. The common occurrence of false alarms or “cry wolf” phenomenon could lead to no or delayed response of caregivers.

The decision not to limit inclusion criteria in terms of study design and methodology led to a high level of variety within the chosen studies, necessitating a narrative review. The advantage of this method, however, is that it enabled us to have a complete overview of a range of different aspects related to compliance in SpO2 targeting. However, the review was restricted to recent studies published in English and Dutch; similar studies published in other languages may have been missed. In addition, the selection process was conducted by the first author only and selection bias could have occurred. However, in any case of doubt of including a publication, peers were approachable for discussion and were resolved by consensus to avoid bias in the selection process.

In conclusion, the main finding of this literature review is that there is a low compliance in SpO2 targeting and alarm settings during oxygen therapy in preterm infants, especially in maintaining the SpO2 below the upper limit and in setting the upper alarm limit. Although there is little data available, it is likely that training, titration protocols and decreasing the nurses’ workload could improve awareness and compliance. Automated oxygen regulations have been shown to increase the time that SpO2 remains within the TR. Improving the compliance in SpO2 targeting and automated control has the potential to improve the outcome in preterm infants. The effect of training, implementing protocols and automated oxygen regulators needs further investigation.