OBJECTIVE. The objective of this study was to determine the rate of compliance with hospital guidelines for alarm limits for pulse oximetry in preterm infants on oxygen therapy.
METHODS. All infants admitted to the nurseries at the Royal Women's Hospital, Melbourne, Australia, with gestational age <32 weeks or birth weight <1500 g between August 2005 and February 2006 were eligible for inclusion. Data on the alarm limits set for infants on oxygen therapy were collected prospectively. The target saturation range recommended in written hospital guidelines was 88% to 92%, with alarm limits set at 85% and 94%.
RESULTS. Data were prospectively collected for 144 subjects with mean (SD) gestational age 29.3 (2.4) weeks and birth weight 1226 (354) g; 1073 alarm limits were collected when infants were on oxygen. The lower alarm limit was set correctly 91.1% of the time. In contrast, the upper alarm limit was set correctly only 23.3% of the time: 76.5% of the time it was too high, and 23.8% of the time it was set at 100%. Infants with an upper alarm limit set correctly on a particular day had a significantly lower birth weight, gestational age, postmenstrual age, and postnatal age than infants who had the upper alarm limit set too high. Use of assisted ventilation, higher inspired oxygen concentrations, and more frequent changes in inspired oxygen concentration were all associated with improved odds of having an appropriately set upper alarm limit.
CONCLUSIONS. This study suggests that current guidelines regarding the upper pulse oximeter alarm limit for infants receiving oxygen might be commonly exceeded, although compliance might be better for infants at higher risk of adverse outcomes. However, there might be less variation from guidelines for the lower alarm limit.
Oxygen, although vital to survival, can be toxic to living cells particularly in high concentrations,1,2 so much so that in preterm infants it is possible to have “too much of a good thing.”3 Oxygen toxicity in preterm infants has been associated with conditions such as bronchopulmonary dysplasia4,5 and retinopathy of prematurity (ROP).1,6 Unfortunately, preventing ROP is not as simple as restricting the amount of oxygen given to these infants: a balance between too much oxygen and too little must be found to reconcile the competing outcomes of ROP and bronchopulmonary dysplasia caused by too much oxygen, and neurologic impairment and death caused by too little.7 Monitoring the level of oxygen in the blood of these infants is, therefore, very important and is part of routine care.
Pulse oximetry is now the most widely used method for monitoring oxygen levels in these infants and is considered the standard noninvasive technique.8,9 There is, however, still no agreement on the optimum saturation range, and considerable variation exists between centers.8 At the Royal Women's Hospital, Melbourne, the guidelines for the use of pulse oximetry in preterm infants receiving supplemental oxygen were changed in June 2005, partly in preparation for a masked randomized, controlled trial comparing 2 different oxygen saturation target ranges. There is little information about compliance with alarms limits for pulse oximetry for infants in oxygen, hence we sought to obtain this information for the most immature and tiny infants and to determine the clinical variables that affect compliance with guidelines. It was hypothesized that compliance with both upper and lower alarm limits would be good, and would be better in infants at highest risk.
We conducted a prospective audit of the use of pulse oximetry in a tertiary neonatal center, the neonatal unit at the Royal Women's Hospital. Infants were recruited from August 29, 2005, to February 28, 2006, and were eligible for participation if they had either a birth weight <1500 g or gestational age <32 completed weeks, and were admitted within the first day of life. Infants who left the unit within 24 hours of admission were excluded.
The guidelines for oxygen target ranges and alarm limits were changed on June 6, 2005 by the issuing of a memorandum to all staff, medical and nursing, involved in pulse oximetry. The new guidelines required target oxygen saturations between 88% to 92%, with alarm limits set at 85% and 94% for infants receiving supplemental oxygen. For infants not on oxygen, the upper alarm limit could be up to 100%. It was expected that all staff would follow the guidelines unless specifically ordered to do otherwise. There was no limit on how many changes were allowed to the inspired oxygen concentration in response to oxygen values outside the target range. Before June 2005, the target range was 90% to 95%, with lower alarm limits of 85% for infants <2 weeks old and 80% for older infants, and upper alarm limits of 96% for infants in oxygen.
Data for the current study were collected from August 29, 2005, until March 31, 2006. Data were collected at approximately the same time daily during weekdays for each infant during this period until the infant was discharged from the unit (and did not return within 24 hours), or until the end of the data collection period, whichever occurred first. Data collected included the length of time (hours) the infant was on an oximeter in the preceding 24-hour period, the targeted oxygen saturation range, and pulse oximeter alarm limits. If there was no specific saturation range ordered by the doctor in the patient's treatment notes, it was assumed that the target range was 88% to 92% and that alarm limits should be set at 85% and 94% for all infants on supplemental oxygen, as per the hospital guidelines.
We recorded whether the infant was breathing air or supplemental oxygen. Only days when the infant was on supplemental oxygen are included in this study. The fractional inspired oxygen (Fio2) (percent) or flow rate (mL/min) of oxygen at that time, the range of the Fio2 or flow rate of oxygen (lowest and highest values) in the preceding 24-hour period, the number of times the amount of oxygen was changed, and whether the infant was on assisted ventilation at the time of data collection (any of continuous positive airway pressure, conventional ventilation, or high-frequency ventilation) were recorded. After all the data were collected, the oxygen requirements on any given day were categorized into approximate tertiles as low (Fio2: 22%–23%), moderate (Fio2: 24%–29%), or high (Fio2 > 29%); Fio2 values were derived for infants on low-flow oxygen before this categorization.
Ethics approval for this audit was obtained from the Royal Women's Hospital Research and Ethics Committees.
Data were analyzed by using SPSS 14.010 and Stata 9.1.11 Continuous data were expressed as mean (SD) or median (interquartile range [IQR]) if the data were skewed. Dichotomous data were presented as percentages.
All comparative analyses were adjusted for clustering, because of the repeated observations on the same subjects. For continuous variables (birth weight, gestational age, postmenstrual age, and postnatal age), linear regression with robust variance estimation (to adjust for clustering) was used to estimate the difference in mean values between days when the upper alarm limit was set at 94% and days when the limit was set at >94%. Multivariable logistic regression with robust variance estimation was then used to determine the independent effects of gestational age and postnatal age on the odds that the upper alarm limit would be set correctly; we did not want to test birth weight and postmenstrual age simultaneously in this analysis because they were too highly correlated with gestational age and postnatal age. χ2 with adjustment for clustering was used to investigate whether categorical variables (gender, whether receiving assisted ventilation, Fio2 tertile, and grouped number of changes in oxygen in the preceding 24-hour period) influenced the upper alarm-limit setting (94%, >94%) on each day. Multivariable logistic regression with robust variance estimation was used to estimate the independent effects of the last 3 variables on the upper alarm setting; results are presented as adjusted odds ratios (ORs) with 95% confidence intervals (CIs) on the basis of the robust variance estimate.
The sample size of this study was determined by the number of infants admitted to the unit within the study period who met the inclusion criteria, and by the number of days that those infants received oxygen therapy during their admission.
The study sample comprised 144 consecutive infants who met the inclusion criteria and who were born over the 6-month period between August 28, 2005, and February 28, 2006. This comprised all eligible infants admitted over this time. The subjects had a mean gestational age of 29.3 weeks (SD: 2.4) and mean birth weight of 1226 g (SD: 354). Seventy-seven (53.5%) infants were male. Infants were followed for a median of 31 days (IQR: 13.2–48.8). Of the 144 infants, 129 were followed until they were discharged from the unit or died, whereas 15 were still in the unit at the end of the data collection period. Some data for 11 of 144 infants were missed during the prospective data collection, having initially been excluded incorrectly on the basis of birth weight. Of 144 infants, 64 never received any oxygen. Of 80 who received oxygen, 55.6% (n = 44) were male with a mean gestational age of 28.4 weeks (SD: 2.4) and mean birth weight of 1123 g (SD: 383); they were in hospital for a median of 37 days (IQR: 17.0–61.5), during which time they had oximetry for a median of 32.5 days (IQR: 11.5–67.0) and were in oxygen for a median of 5 days (IQR: 2.0–34.5).
In total, 1073 lower and upper alarm limits were collected when infants were on oxygen. Of the 1073 lower alarm-limit values, 977 (91.1%) were set correctly at 85%, 6.3% were set lower than 85%, and 2.7% were set higher (Fig 1). Of the 1073 upper alarm-limit values, 250 (23.3%) were set correctly at 94%; 76.5% were set above 94%, and 23.8% were set at the maximum value of 100%. Only 2 of the upper alarm limits collected were set lower than 94% (0.2% of the values collected) (Fig 2). Of the 1073 pairs of lower and upper alarm limits collected when infants were on supplemental oxygen, there were 236 cases (22.0%) in which both the lower and upper alarm limits were set correctly. In 7.6% of cases both alarm limits were incorrect, and in 70.4% only 1 was correct.
The upper alarm limit was more likely to be set correctly on a particular day in infants of significantly lower birth weight, gestational age, postmenstrual age, and postnatal age (Table 1). Gestational age and postnatal age had independent effects; every week's decrease in gestational age increased the odds that the upper alarm limit would be set correctly by 36% (adjusted OR: 1.36; 95% CI: 1.20 to 1.54; P < .001), and for every week's decrease in postnatal age, the odds that the upper alarm limit would be correct increased by 11% (adjusted OR: 1.11; 95% CI: 1.06 to 1.16, P < .001).
Of the observations taken for infants requiring high levels of supplemental oxygen, the upper alarm limit was set correctly in 35.7% of cases, compared with 23.6% in the moderate oxygen group and 6.2% in the low oxygen group, a statistically significant difference between the groups (χ2: 20.8; P < .001 adjusted for clustering).
The proportions of alarm limits set correctly was also related to the number of changes to an infant's Fio2 on a given day (adjusted χ2: 21.4; P < .001); 37.2% of infants with the highest number of changes (>12 changes) had correct upper alarm limit, compared with 23.1% of infants who had 6 to 12 changes, and 7.3% who had 0 to 5 changes.
A greater proportion of infants who were on assisted ventilation had their upper alarm limit set correctly compared with infants who were not on assisted ventilation (31.1% compared with 6.3%, adjusted χ2: 19.1; P < .001). Gender did not have an effect (adjusted χ2: 0.03; P = .87).
Being on assisted ventilation, receiving moderate or high levels of oxygen, and having a moderate (6–12) or high (>12) number of changes to an infants' Fio2 all independently increased the odds of having the upper alarm limit set correctly on a given day (Table 2).
This study found that the lower alarm limit for very preterm infants on supplemental oxygen was usually set correctly; however, the upper alarm limit was set too high the majority of the time. The percentage of times when both the upper and lower alarm limits were set correctly was only 22.0%, leaving at least 1 alarm limit set incorrectly 78.0% of the time. This was not what we expected to find and compares poorly with a study conducted by Laptook et al12 in the United States in which, unannounced, intermittent surveys found pulse oximeter alarm limits to be set incorrectly only 26% (SD: 15%) of the time when lower and upper alarm limits were to be set at 80% and 96%, respectively, and 23% (SD: 16%) of the time when alarm limits were to be set at 80% and 94%. As in our study, they found that it was predominantly the upper alarm limit that was set incorrectly. They related this particularly to infants whose oxygen requirement varied from air to low levels of supplemental oxygen. This is similar to the results in our study; infants receiving lower concentrations of supplemental oxygen (Fio2: 22% or 23%) had the lowest proportion of upper alarm limits set correctly.
Although infants with a lower gestational age and postnatal age and those on ventilation or receiving higher levels of oxygen were more likely to have a correctly set upper alarm limit, as we hypothesized, the compliance was still poor. Moreover, although lower risk infants of increased maturity and postnatal age and those not on assisted ventilation may be considered less of a concern for staff than other infants, they are still at risk of developing ROP.
The results of our study suggest that improvement in compliance with the guidelines for the upper pulse oximeter alarm limit for preterm infants requiring supplemental oxygen at our hospital is needed. The possibility that setting, and keeping, the upper alarm limit at 94% may not always be realistic may also need to be considered. However, the results of the study by Laptook et al12 suggest that better compliance is possible. Because of a lack of other published studies on this topic of compliance with alarm limits, it is unclear whether our results or those of Laptook et al12 are more typical.
Our study did not attempt to provide any evidence of a link between alarm-limit compliance and ROP rates. It was a descriptive study aiming to provide information about the use of oximeters in the unit. Our study was limited by the inability to download data from the oximeters; instead, we relied on snapshots of alarm settings rather than continuous recording of these variables. In a recent study, Hagadorn et al13 collected oximetry data continuously for 72 hours each week for the first 4 weeks of life in 84 infants <28 weeks' gestational age from 14 different centers with varying upper and lower target ranges for oximetry. They reported that infants receiving modifiable oxygen had median SPo2 values within the target ranges in 12 of 14 centers. Overall, infants spent 16% below, 48% within, and 36% above their NICU's intended range. The studies of Laptook et al12 and Hagadorn et al13 and our own study indicate that there are problems with compliance with both target ranges and alarm settings. Clinicians need to be aware of these problems. Interpretation of the results of randomized, controlled trials targeting different saturation ranges will be problematic if these problems are not addressed; indeed, if there is not a clear separation of targeted ranges between groups in such trials, then is it possible that no benefit of either range may be found.
Data were collected once daily in our study, thus only the alarm limits at that time are known. The proportion of the time, or number of times per day, the alarm limits were correct or incorrect is unknown; only the number of times that the alarm limits were correct or incorrect at the time of data collection could be determined. Nursery staff were aware of the purpose of the study and the approximate time the observations were going to be made and had the opportunity to modify the alarm limits if they so desired; that the upper alarm limit was too high on the majority of days that infants were in oxygen suggests that they did not avail themselves of this opportunity. We speculate that the total number of times alarm limits are incorrect in a day might be worse than we have described. Time spent within the target saturation range or alarm-limit range also could not be determined. Correct alarm limits may not mean the infant's saturations are being kept in this range, whereas infants with incorrect alarm limits may in fact be saturating correctly. Additional research looking at whether the alarm limits accurately reflect the oxygen saturation range in which the infant is being nursed, and if better compliance with alarm limits improves oxygen targeting, would be of interest.
The current guidelines regarding the upper pulse oximeter alarm limit for very preterm infants in oxygen at the Royal Women's Hospital are rarely followed. The majority of the time the upper alarm limits are being set too high, with a concerning proportion set at the maximum value of 100%. Compliance with the guidelines for the lower alarm limit is better, with little deviation from the guideline of 85%. Infants at higher risk of adverse outcomes from oxygen toxicity had an increased likelihood of having the alarm limit set correctly on a given day. Although this trend is somewhat reassuring, compliance rates for the upper alarm limits for these infants were still poor.
- Accepted January 18, 2007.
- Address correspondence to Lex W. Doyle, MD, FRACP, Department of Obstetrics and Gynaecology, Royal Women's Hospital, 132 Grattan St, Carlton 3053, Australia. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
- ↵Duc G, Sinclair J. Oxygen administration. In: Sinclair J, Bracken M, eds. Effective Care of the Newborn Infant. Oxford, England: Oxford University Press; 1992:178– 199
- ↵Haugaard N. Cellular mechanisms of oxygen toxicity. Physiol Rev.1968;48 :311– 373
- ↵Tin W. Oxygen therapy: 50 years of uncertainty. Pediatrics.2002;110 :615– 616
- ↵Tin W, Milligan DW, Pennefather P, Hey E. Pulse oximetry, severe retinopathy, and outcome at one year in babies of less than 28 weeks gestation. Arch Dis Child Fetal Neonatal Ed.2001;84 :F106– F110
- ↵Silverman WA. A cautionary tale about supplemental oxygen: the albatross of neonatal medicine. Pediatrics.2004;113 :394– 396
- ↵SPSS for Windows [computer program]. Version 14.0. Chicago, IL: SPSS Inc; 2005
- ↵Intercooled Stata 9.1 for Windows [computer program]. College Station, TX: Stata Corporation; 2005
- ↵Hagadorn JI, Furey AM, Nghiem TH, et al. Achieved versus intended pulse oximeter saturation in infants born less than 28 weeks' gestation: the AVIOx Study. Pediatrics.2006;118 :1574– 1582
- Copyright © 2007 by the American Academy of Pediatrics