Article Text
PDF
Abstract
Background Pulse oximeter saturation values are usually obtained by averaging over preceding measurements. This study investigates the dynamics between the averaging time and desaturation level, duration and extent.
Methods and results Prospective observational study of 15 preterm infants. Oxygen saturation was recorded for 168 h using a pulse oximeter. The raw red-to-infrared data were reprocessed using seven different averaging times to determine the number of desaturations below four thresholds and for seven different minimal desaturation durations. The total number of desaturations <80% was 339 with an averaging time of 16 s and 1958 with an averaging time of 3 s (minimal event duration >0 s). There was a significantly lower pulse oximeter saturation nadir with the shorter averaging time, while the maximum duration was significantly longer when using a 16 s averaging time.
Conclusions When using pulse oximeters, more attention should be given to averaging time and duration of desaturations.
- pulse oximetry
- desaturations
- averaging time
Statistics from Altmetric.com
Pulse oximeters average their values over several heart beats. Some studies analysed the influence of averaging time on desaturation rates,1–5 but many details remain unknown. We analysed the influence of averaging time on desaturation levels, durations and extent.
Patients and methods
We studied 15 infants with apnoea of prematurity, who were in room air and otherwise considered healthy. Their gestational age was 24–27 weeks at birth and 32–33 weeks at study. Parents had given written informed consent.
In addition to routine clinical monitoring, infants had a second disposable oximeter sensor (Masimo Radical-7 with LNOP-NeoPt, Irvine, California, USA, software V.4.7, sampling rate 62.5 Hz) attached to a foot for a median of 14 h (range 7–24 h). Study instrument alarms were silenced.
Recordings comprised 226 started hours of data, including those on red-to-infrared absorption, stored on a notebook using PhysioLog software (Masimo). Data were reprocessed offline, with the help of the manufacturer, as if they were from the same patient, but obtained with different averaging times (ie, 3 (2–4), 5 (4–6), 8, 10, 12, 14 and 16 s). For averaging times <8 s, the manufacturer uses a moving average, where averaging time is low (eg, 2 s) if signal quality is high, and long (4 s) if quality is low. The manufacturer claims that all their software validation against measured values uses this approach. The resulting desaturation values were analysed using JMP statistical software (SAS institutes Inc, Cary, NC, USA).
Data were divided into 1-h segments, each being visually checked for data plausibility. Only 1h segments where measurements were undisturbed throughout were included, resulting in the exclusion of 25.7% of segments. The remaining 168 h contained a mean number of 60.9 desaturations per hour (minimal event duration >0 s.) for an averaging time of 3 s (2–4 s) and a threshold value of 90%.
Data were also screened for signal quality using Masimo's Signal-IQ, with values >30 defining good quality. During 99.4% of validated recording time the Signal-IQ was >30.
Results
Time spent at pulse oximeter saturation(SpO2) <80%, 80–85%, 85–90%, 90–95% and 95–100% was not significantly affected by averaging time, while the extent of desaturation, the SpO2 nadir and the desaturation duration were. Thus, the total number of desaturations to <80% was 1958 with an averaging time of 3 s, and 339 with an averaging time of 16 s. With the shortest averaging time (3 s), 96.0% of desaturations were <20 s, and 4.0% ≥20 s. With an averaging time of 16 s, 68.7% of desaturations were <20 s and 31.3% ≥20 s (table 1).
Number and percentage of desaturations to <80% for < or ≥20 s depend on desaturation duration and averaging time
The mean number of desaturation events to <80% for ≥20 s was 4.3 (95% CI 1.6 to 8.3) at 3 s, and 6.0 (2.7 to 10.6) at 16 s averaging. Using an LSMeans Differences Tukey, Honestly Significant Difference with α=0.050, there were significantly more desaturation events with the longer averaging time.
The minimal (nadir) SpO2 recorded with 16 vs 3 s averaging was significantly lower with shorter averaging. When analysing the longest desaturation event found for each averaging time, we found a mean of 57.8 s (95% CI 37.3 to 89.4) for 16 s versus 38.3 s (95% CI 24.8 to 59.3) for 3 s averaging time (Tukey Honestly Significant Difference, α=0.050). Finally, we determined the relationship between the number of desaturations per hour and averaging time for the three desaturation thresholds, seven minimal desaturation durations and seven averaging times and found a linear relationship between the logarithms of the desaturation rate (number/hour) and averaging time, where log refers to base 10. The slope of the linear regression depended on the desaturation duration and was negative for a minimal event duration of >0s or 5 s, and positive for a minimal duration of 20s or 30 s (figure 1).
Linear relationships between the desaturation rate logarithms (N des) and the averaging time logarithms (T des) at a desaturation threshold of 80% for seven different minimal desaturation durations (>0 s, 5 s, 10 s, 15 s, 20 s, 25 s, 30 s) are shown.
Discussion
This study shows that desaturation rates below certain thresholds on simulated data can vary up to fivefold depending on the averaging time used. We confirmed that significantly lower saturation values are measured with shorter than with longer averaging times, but that the averaging time does not affect the time spent in different SpO2 ranges.
In contrast to Ahmed,1 however, we found that the averaging time had a statistically significant influence on the maximum desaturation duration (3s vs 16 s). This difference may be due to the fact that our SpO2 data were obtained from the same sensor site.
Furthermore, our results suggest that if the influence of the averaging time is analysed, the duration of desaturation must be taken into account: the number of desaturations <20 s decreased with increasing averaging time, while the reverse was true for desaturations longer than 20 s.
With longer averaging (eg, 16 s), several short desaturations are likely lumped together as one, potentially increasing the number of prolonged desaturations. Conversely, the number of short desaturations will be reduced. Similar findings have been reported before, but without taking the effect of different desaturation durations into account.
The reason for this phenomenon is that most desaturations occurring in our patients were brief events. Prolonged desaturations, in which a longer averaging time has the reverse effect, were rare. Had the respective duration of desaturation been taken into account, this effect would have been obvious. In our study, we recorded 69% of desaturations as short events (<20 s) with an averaging time of 16 s, but 96% at an averaging time of 3 s.
Another parameter that can be changed on some instruments is the alarm delay time. We could show that averaging time has little effect on events of 10–15 s duration, that is, it has little effect on an alarm delay of 10–15 s (the default with Radical 7 is 10 s). In contrast, the averaging time has a significant influence on shorter or longer desaturation durations, as discussed above. To our knowledge, it has not yet been described that the slope of the regression line depends on the desaturation duration. It was negative for a minimal desaturation event duration of >0 or 5 s. The negative slope might be explained by the fact that two or even three short desaturations are summed up to one long desaturation by smoothing the SpO2 curve as soon as a long averaging time is used. Some authors included only desaturations ≥20 s. As 96% of desaturations <80% are shorter than 20 s, the influence of the averaging time has to be considered.
The slope was positive for a minimal duration of >20 s. This means that the number of such desaturations, described as severe events (even though they represent only 4% of all events) was also influenced by the averaging time. Again, the underlying mechanism is based on the smoothing effect resulting from longer averaging times. As yet, it remains unknown which minimal desaturation durations as well as which threshold have to be considered relevant.
Limitations of our study are that results are based on reprocessed data. Although the company claims that all their software validation applies the approach also used here, an external validation would have been preferable. Another limitation is that our analyses refer to desaturations in preterm infants, thus our results should be verified in children and adults.
Acknowledgments
We are grateful to the parents for allowing us to study their infants and to Friederike Hanel, MD, Cornelia Wichers, MD, and Anja Bialkowski, MD for their help in obtaining the recordings, Troy Vine, PhD and Jeff Martin for editorial work.
Footnotes
Contributors JV designed the study, helped to analyse the data and wrote the first draft of the manuscript. KD was responsible for data analysis. CP supervised the study and was involved in writing the manuscript.
Competing interests Dr Poets has served on an advisory board for Masimo Inc. and has received a speaker honorarium from Philips. Masimo Inc. also provides oximeters free of charge for studies conducted in his department.
Patient consent Obtained.
Ethics approval Ethics Committee of Tübingen University Hospital.
Provenance and peer review Not commissioned; externally peer reviewed.