Journal of Perinatology (2014) 34, 130–135 & 2014 Nature America, Inc. All rights reserved 0743-8346/14 www.nature.com/jp
ORIGINAL ARTICLE
Transcribed oxygen saturation vs oximeter recordings in very low birth weight infants TL Ruiz1,2,3, JM Trzaski3,4, DW Sink3,4 and JI Hagadorn3,4 OBJECTIVE: The objective of this study was to compare hand-transcribed oxygen saturation (SpO2) with electronic oximeter data in very low birth weight infants (VLBWI, o1500 g). STUDY DESIGN: Oximeter data were downloaded from birth through 36 weeks postmenstrual age (PMA) for VLBWI before and after interventions to improve neonatal intensive care unit oxygen management. Transcribed SpO2 values were obtained by chart review. Proportions of transcribed and oximetry data in target (85 to 93%), hypoxemic (80 to 84%), and hyperoxemic (X98%) ranges before and after intervention were compared. RESULT: There were 30 441 oximetry hours before intervention and 54 538 oximetry hours after intervention. Transcribed SpO2 values correlated strongly with oximeter overall. However, during hours on supplemental oxygen, transcribed values significantly overdocumented target range and underdocumented values 80 to 84 and X98%. This finding varied by respiratory support and PMA, and increased after intervention. CONCLUSION: Transcribed SpO2 values overdocumented target range and underdocumented hyperoxemic and hypoxemic ranges compared with electronic oximeter data. These results support incorporating electronic oximeter data into medical records. Journal of Perinatology (2014) 34, 130–135; doi:10.1038/jp.2013.157; published online 19 December 2013 Keywords: nursing records; medical records systems; computerized; electronic medical records
INTRODUCTION Pulse oximetry provides an indirect estimate of arterial oxygen saturation (SpO2) using infrared light that is absorbed by hemoglobin and then transmitted to a photodetector.1 Oximeters are used to assess SpO2 to detect and help prevent low or high fluctuations in oxygenation during patient care. This noninvasive, continuous monitoring device is generally initiated in the delivery area for newborn premature infants. Standard practice for subsequent neonatal intensive care requires documentation of the results of noninvasive oxygenation monitoring for very low birth weight (VLBW, o1500 g at birth) infants.2 Accordingly, many neonatal intensive care units (NICUs) direct bedside providers to transcribe SpO2 values serially for infants on supplemental oxygen monitored with continuous pulse oximetry. In the absence of primary electronic oximeter data, these transcribed SpO2 values may be the only permanent record of a premature infant’s SpO2 experience during their NICU stay. The transcribed SpO2 values may be used for clinical decision-making, quality improvement or research.3–5 However, we were unable to locate studies examining the relationship between SpO2 values transcribed by NICU staff and infants’ actual SpO2 experience as reflected by simultaneous oximeter data. The objectives of this study were: first, to explore the relationship between SpO2 values transcribed by bedside providers and infants’ actual SpO2 experience as recorded by pulse oximetry, and second, to examine the relationship between bedside provider and oximeter before and after an intervention to improve the precision of oxygen therapy. Our null hypothesis was that transcribed SpO2 values would not significantly differ from simultaneous oximeter data before or after the intervention.
METHODS This was a single-center observational study linking existing oximeter data with transcribed SpO2 values from medical records. The Institutional Review Board at Connecticut Children’s Medical Center approved this study by expedited review including waiver of informed consent.
The SpO2 practice intervention In March 2009, the Connecticut Children’s level 4 Hartford NICU implemented a clinical practice intervention to improve achievement of SpO2 goals in VLBW infants. Before this intervention, NICU policy set SpO2 targets of 85 to 92% for infants with gestational age o29 weeks and 88 to 92% for 29 to 34 weeks. Oximeter alarm settings were not strictly defined. The intervention revised SpO2 target only slightly, to 85 to 93% for infants with birth weight o1500 g or gestational age o30 weeks. However, the new policy placed a heightened emphasis on achieving SpO2 target while avoiding hyperoxemia and hypoxemia, using several methods. Oximeter alarm settings were standardized. Reminders were posted at the bedside of each VLBW infant displaying the desired saturation target as well as algorithms for responding to oximeter alarms, including non-oxygen interventions and guidelines for titration of FiO2. These practice changes were accompanied by comprehensive NICU staff education with regard to the risks of oxygen therapy in VLBW infants. Computerized oxygen ordering was standardized to reflect the new policy. Adherence to the new policy was monitored using downloaded oximeter data, providing staff with regular feedback about summarized results with regard to achievement of SpO2 goals. The resulting improved achievement of SpO2 goals is described elsewhere.6 As part of the intervention, electronic data were downloaded from bedside oximeters for VLBW infants admitted between January through June 2008 (before intervention) and April through December 2009 (after intervention, after a 1-month washout period). Infants were included in
1 Health Fellows Program, Trinity College, Hartford, CT, USA; 2Clinical Trials Unit, Connecticut Children’s Medical Center, Hartford, CT, USA; 3Division of Neonatology, Connecticut Children’s Medical Center, Hartford, CT, USA and 4Department of Pediatrics, University of Connecticut School of Medicine, Farmington, CT, USA. Correspondence: Dr JI Hagadorn, Division of Neonatology, Connecticut Children’s Medical Center, 282 Washington Street, Hartford, CT 06106, USA. E-mail: [email protected] Received 8 May 2013; revised 10 October 2013; accepted 14 November 2013; published online 19 December 2013
Transcribed oxygen saturation values vs oximeter TL Ruiz et al
131 these samples if they were inborn, had birth weight o1500 g and were alive on day of life four, the time of the first oximeter download. Using Masimo Radical 7 SET oximeters (Masimo Corporation, Irvine, CA, USA), SpO2 was sampled every 2 s during routine bedside care starting from NICU admission through 36 weeks postmenstrual age (PMA). These comprehensive downloads yielded tens of thousands of hours of highly granular oximeter data for subsequent study. For each oximeter, system dates and times were synchronized with NICU clocks and checked periodically. Averaging time for the oximeter display was 8 s. Oximeter data were not collected for care performed away from the bedside (e.g., operating room, radiology, bathing), as during these times monitoring was performed using an oximeter without data storage capability. After intervention, adherence to the new policy was monitored in real time using the downloaded oximeter data, providing staff with regular feedback about of summarized results with regard to achievement of SpO2 goals. In addition, bedside oximeter alarm settings were audited three times per day by respiratory staff while infants received supplemental oxygen.
Derivation of patient care hour groups for analysis Because tens of thousands of hours of continuous oximeter data result in an unwieldy data set, we used more manageably sized groups of patient care hours for analyses. Ten percent of all patient care hours before and after the intervention were randomly selected and assigned to pre- and postintervention Overall groups. During these patient care hours, infants could be in either room air or supplemental oxygen, and could be supported by any mode of respiratory support. The Overall groups thus allowed a summary examination of the relationship between transcribed SpO2 values and simultaneous oximeter data for all VLBW patient care hours before and after the intervention. In addition, the pre- and postintervention Covariate groups consisted of all patient care hours for which oximeter data were available, during which infants were receiving supplemental oxygen (FiO2 421%) from 6-h nurse shifts characterized by a single respiratory support mode and not more than one nurse providing care. Extensive clinical data for this cohort of patient care hours were available from a prior analysis. The pre- and postintervention Covariate groups were used to assess the impact of clinical factors upon the relationship between transcribed SpO2 values and simultaneous oximeter data independently of changes in respiratory support or changes in bedside care provider.
Transcribed SpO2 data Policy at the Connecticut Children’s Hartford NICU required bedside staff to transcribe the displayed oximeter SpO2 value to the patient care flow sheet hourly at the top of the hour. This practice remained unchanged throughout the pre- and postintervention periods for which SpO2 data were collected. For this study, transcribed SpO2 values were obtained retrospectively by review of patient care flow sheets for included pre- and postintervention patient care hours. Patient care hours were excluded if assigned bedside provider or respiratory support modes could not be determined, or if the infant’s flow sheet had more than one SpO2 value or no value recorded for that hour.
Oximeter SpO2 data Oximeter-derived SpO2 values for included patient care hours were obtained retrospectively using archived continuous oximeter saturation data obtained before and after intervention as described above. Less than 1% of SpO2 values were flagged by the oximeter as having poor signal and these were excluded from analysis. Because bedside staff recorded no values o80% on the patient flow sheets, oximeter-recorded SpO2 values o80% were excluded to ensure comparison of similar ranges.
Calculation of proportions of hand-transcribed and oximeter SpO2 values We compared the proportion of SpO2 values recorded by the oximeter at each saturation value between 80 and 100% with the proportion of handtranscribed values at the same saturation value during the same cohort of patient care hours. The proportion of SpO2 data at each saturation value was calculated before and after intervention for both the Overall and Covariate groups. For example, the proportion of transcribed SpO2 values & 2014 Nature America, Inc.
Figure 1. Selection of patient care hours for pre- and postintervention Overall and Covariate group analyses. equal to 90% was calculated as follows: number of patient care hours in group with transcribed SpO2 value equal to 90% total number of patient care hours included in group
Similarly, the proportion of oximeter SpO2 values equal to 90% was calculated as follows: number of oximeter SpO2 values in group equal to 90% total number of included oximeter SpO2 values in patient care hours included in group
In the Covariate groups, patient care hours were further subgrouped by the mode of respiratory support and PMA. For all groups and subgroups, the proportions of SpO2 values within the following ranges were calculated: (1) 85 to 93%, ‘target’ range; (2) 98 to 100%, ‘hyperoxemic’ range; and (3) 80 to 84%, ‘hypoxemic’ range.
Statistical methods Descriptive and bivariate correlation analyses were performed with SPSS 17.0 or Microsoft Excel 2010. Correlations between nurse and oximeter were calculated using concordance correlation coefficients (rc). Proportions of SpO2 values recorded by oximeter vs transcribed values were compared using the w2 test or 2 log-likelihood ratio goodness-of-fit (G).7
RESULTS Before intervention A total of 30 441 h of oximeter data were available from 24 infants before intervention. Infants in the preintervention cohort had median (interquartile range) birth weight 955 (749, 1240) g, gestational age 27 (26, 29) weeks and were mostly males (n ¼ 18, 75%) exposed to antenatal steroids (n ¼ 21, 87.5%). For the preintervention Overall group, 3046 h were selected randomly without regard to age, supplemental oxygen use or respiratory support (Figure 1). After exclusion criteria were applied, 2862 h were analyzed. For the preintervention Covariate group, there were 5027 h on supplemental oxygen for which detailed clinical data were available. After exclusion criteria, 5023 h were analyzed. After intervention After intervention there were 54 538 h of oximeter data available for analysis from 27 infants. Postintervention infants were not significantly different from the preintervention cohort, with median (interquartile range) birth weight 981 (720, 1180) g and gestational age 27 (25.4, 29) weeks. Most were males (n ¼ 17, 63%) and received antenatal steroids (n ¼ 24, 88.9%). For the postintervention Overall group, 5385 h were randomly selected and 4815 h were analyzed after applying exclusion criteria. For the Journal of Perinatology (2014), 130 – 135
Transcribed oxygen saturation values vs oximeter TL Ruiz et al
132
Figure 2. Hourly transcribed SpO2 values compared with oximeter data for Overall Group patient care hours regardless of supplemental oxygen therapy or mode of respiratory support, before and after intervention.
Covariate group, 5484 patient care hours were available, with 4477 h included for analysis.
Overall Proportions of hand-transcribed SpO2 data at each saturation value between 80 and 100% were strongly correlated with the proportions recorded by oximeter in the Overall groups, with rc ¼ 0.988 (Po0.001; Figure 2) in the preintervention group and rc ¼ 0.993 in the postintervention group (Po0.001). Staff transcribed even SpO2 values more frequently than the oximeter displayed (61.1% vs 57.7%, Po0.001, w2).
Table 1. Percentages of oxygen saturation (SpO2) values in target (85 to 93%), hyperoxemic and hypoxemic ranges, transcribed vs oximeter recorded, for patient care hours in the pre- and postintervention Covariate groups (all hours on supplemental oxygen therapy)
In target X98 80–84
Before intervention (5023 h)
After intervention (4477 h)
Transcribed
Oximeter
Transcribed
Oximeter
26.9 34.6 0.5
24.7*** 37.9*** 3.1***
50.8 6.7 3.6
43.2*** 13.8*** 11.4***
***
Po0.001.
Covariate groups ‘Target’ range (85 to 93%). In the Covariate group, SpO2 values in the target range were significantly over-represented in the transcribed group compared with oximeter for both pre- and post-intervention hours on supplemental oxygen (Table 1). Before intervention, SpO2 values in the target range were overrepresented in the transcribed group compared with oximeter for all hours with PMA o32 weeks and for hours on highfrequency oscillator ventilation (HFOV), conventional ventilation or continuous positive airway pressure (CPAP) (Table 2). These findings were most pronounced for hours o28 weeks PMA and for those on HFOV. Transcribed values were not significantly different from oximeter for hours at PMA 32 to 35 weeks or for those on nasal cannula, the two subgroups where the oximeterdocumented target range achievement was lowest (23.5% and 18.6%, respectively). Following the intervention, the tendency to overdocument target values was more prominent. The target range was significantly over-represented among transcribed SpO2 values for all infants regardless of PMA or respiratory support. Differences between transcribed values and oximeter were larger for all subgroups. Journal of Perinatology (2014), 130 – 135
‘Hyperoxemic’ range (X98%). Saturations in the hyperoxemic range were under-represented among transcribed SpO2 values compared with oximeter in both the pre- and postintervention Covariate groups (Table 1). This was true for all PMA subgroups before and after policy revision (Table 3). Transcribed values were not significantly different from oximeter before or after intervention for hours on HFOV, where the oximeter-documented percentage of values in hyperoxemic range was smallest. For hours on all other respiratory support modes (synchronized intermittent mandatory ventilation, continuous positive airway pressure and nasal cannula), percentage of hyperoxemic values in the transcribed group was significantly lower than oximeter. Following policy revision, the percentage of oximeter-documented hyperoxemic SpO2 values decreased for all subgroups except infants on HFOV. All significant differences observed between transcribed values and oximeter persisted. ‘Hypoxemic’ range (80 to 84%). Saturations in the hypoxemic range were under-represented among transcribed SpO2 values compared with oximeter in both the pre- and postintervention Covariate groups (Table 1). Hypoxemic SpO2 values were & 2014 Nature America, Inc.
Transcribed oxygen saturation values vs oximeter TL Ruiz et al
133 Table 2. Percentage of SpO2 values in the ‘target’ range, 85 to 93%, before and after intervention, transcribed vs oximeter recorded, for patient care hours in the pre- and postintervention Covariate groups, subgrouped by postmenstrual age (weeks) and by respiratory support Before intervention
o28 28–31 32–35 HFOV SIMV CPAP NC
After intervention
Hours
Transcribed
Oximeter
Hours
Transcribed
Oximeter
660 2151 2212 374 917 2088 1644
51.5 39.6 23.0 64.2 39.9 38.5 17.7
42.8** 35.0** 23.5 56.0** 35.2** 34.3*** 18.7
937 2190 1350 481 414 2598 984
69.8 58.5 47.7 68.0 63.3 57.9 49.4
54.9** 42.2** 37.2** 54.0*** 50.7*** 41.7*** 39.1***
Abbreviations: CPAP, continuous positive airway pressure; HFOV, high-frequency oscillator ventilator; NC, nasal cannula; SIMV, synchronized intermittent mandatory ventilation; SpO2, oxygen saturation. ** Po0.01; ***Po0.001.
Table 3. Percentage of SpO2 values in the ‘hyperoxemic’ range, 98 to 100%, before and after intervention, transcribed vs oximeter recorded, for patient care hours in the pre- and postintervention Covariate groups, subgrouped by postmenstrual age (weeks) and by respiratory support Before intervention
o28 28–31 32–35 HFOV SIMV CPAP NC
After intervention
Hours
Transcribed
Oximeter
Hours
Transcribed
Oximeter
660 2151 2212 374 917 2088 1644
11.5 16.0 28.3 2.7 16.6 16.9 32.4
14.8* 19.9*** 32.0*** 3.8 20.0* 20.9*** 36.4***
937 2190 1350 481 414 2598 984
3.1 8.8 13.3 2.9 4.3 8.7 14.3
5.1** 14.2*** 18.7*** 4.9 8.0** 14.7*** 17.7**
Abbreviations: CPAP, continuous positive airway pressure; HFOV, high-frequency oscillator ventilator; NC, nasal cannula; SIMV, synchronized intermittent mandatory ventilation; SpO2, oxygen saturation. *Po0.05; **Po0.01; ***Po0.001.
Table 4. Percentage of SpO2 values in the ‘hypoxemic’ range, 80–84%, transcribed vs oximeter recorded, for patient care hours in the pre- and postintervention Covariate groups, subgrouped by postmenstrual age (weeks) and by respiratory support Before intervention
o28 28–31 32–35 HFOV SIMV CPAP NC
After intervention
Hours
Transcribed
Oximeter
Hours
Transcribed
Oximeter
660 2151 2212 374 917 2088 1644
1.4 0.7 0.2 1.3 1.6 0.2 0.2
5.4*** 4.8*** 2.4*** 5.9*** 5.6*** 4.4*** 1.6***
937 2190 1350 481 414 2598 984
8.0 2.7 1.9 6.0 8.7 3.0 1.8
16.0*** 11.6*** 8.2*** 14.4*** 16.3*** 11.7*** 7.5***
Abbreviations: CPAP, continuous positive airway pressure; HFOV, high-frequency oscillator ventilator; NC, nasal cannula; SIMV, synchronized intermittent mandatory ventilation; SpO2, oxygen saturation. ***Po0.001.
significantly undertranscribed when compared with oximeter for all PMA and respiratory support subgroups before and after the oxygen practice intervention (Table 4). Following the intervention, differences between transcribed values and oximeter were larger for all subgroups regardless of PMA or supplemental oxygen. DISCUSSION In this study, we report that SpO2 values transcribed hourly from the oximeter display by bedside staff generally correlated well with the overall oximeter-recorded SpO2 experience of premature infants during their NICU stay. However, for high-risk VLBW infants receiving supplemental oxygen, bedside providers over-recorded & 2014 Nature America, Inc.
‘target’ SpO2 values (85 to 93%) and under-recorded SpO2 values in ‘hyperoxemic’ (X98%) and ‘hypoxemic’ (80 to 84%) ranges. This bias was present before an intervention designed to improve the achievement of SpO2 goals in VLBW infants, and varied by PMA and mode of respiratory support. Following the intervention, the discrepancy became more pronounced, and was most evident for infants supported with high frequency oscillator ventilation and those at the most premature PMAs. These differences were both statistically and clinically significant. Recent research has resulted in an increased emphasis on achieving SpO2 goals and improving precision of oxygen management in premature infants.8–12 Our results suggest that transcribing hourly SpO2 values may not be sufficiently precise Journal of Perinatology (2014), 130 – 135
Transcribed oxygen saturation values vs oximeter TL Ruiz et al
134 to determine adequately the extent of achievement of SpO2 target or departures from target. In this study, primary oximeter data provided a significantly more accurate reflection of premature infants’ SpO2 experience than SpO2 values hand transcribed from oximeter into the medical record by bedside providers. We were unable to locate other studies comparing hand-transcribed SpO2 values to source oximeter data. However, others have shown that a continuous electronic record is superior to hand-transcription of spot values with respect to data integrity during cardiopulmonary bypass,13 suggesting that use of primary electronic data provides the best opportunity to minimize transcription error and bias. Although our unit policy directed bedside staff to transcribe the oximeter value at the top of each hour, this study did not assume that staff did so with rigorous precision with respect to either time or SpO2 value. Staff may have been preoccupied with other responsibilities, so that the transcribed SpO2 value may not have been from exactly at the top of the hour. Moreover, staff may have chosen to transcribe a more ‘typical’ or ‘desirable’ value than that displayed by the oximeter. Thus, comparing the transcribed SpO2 value with the value that the oximeter recorded precisely at the top of the hour or at any specific time would not have been meaningful or useful. Instead, we compared the proportion of hand-transcribed SpO2 values in the saturation ranges of interest (target, hyperoxemic, hypoxemic) with the proportion of values in those ranges recorded by oximeter during the same patient care hours. This approach allowed us to obtain a valid answer to our research question while avoiding highly problematic attempts to align a specific hand-transcribed SpO2 value to a specific value recorded by the oximeter. This study included several measures to ensure that the comparison of transcribed SpO2 values with the primary oximeter data was valid. Because downloaded oximeter data were not available for care provided away from the NICU bedside (e.g., in radiology, during surgery, etc.), patient care hours and flow sheet SpO2 values corresponding to such times were not considered for inclusion. Each oximeter’s electronic date and time were synchronized with NICU clocks and were checked periodically. Also, because bedside staff recorded no SpO2 values o80% on the patient flow sheets, only oximeter values 80% or higher were included for analysis. These measures maximized the similarity between the cohorts of transcribed and oximeter SpO2 values with respect to time and range, thus optimizing internal validity. We surmise that values o80% were not recorded on bedside flow sheets because these values were interpreted by bedside staff as transient, unrepresentative desaturation episodes or were recorded on a separate bedside log of episodic apnea/bradycardia/desaturation. The effect of emphasizing precision of oxygen management on documentation of discrete episodes below 80% saturation warrants further investigation. Although this was a single-center study, with only 51 infants included—24 before intervention and 27 after intervention—it was not a small study. The unit of analysis was not infants, but patient care hours. Thousands of patient care hours in each group were analyzed, reflecting a broad variety of chronological ages, PMAs, and modes of respiratory support. Because patient care hours were clustered within infants, analyses were performed with hierarchical regression to adjust our results appropriately for nonindependent sampling. Oximeter SpO2 values flagged by the oximeter as having poor signal were excluded from analysis. This represented much o1% of all oximeter SpO2 values. In contrast, 18% of otherwise-eligible patient care hours were excluded from the postintervention Covariate cohort because of hand-transcribed SpO2 values not meeting the criteria for clarity. The great majority of these were due to transcribing multiple saturation values or a range of values rather than a single SpO2 value. As these patient care hours were excluded from analysis, oximeter data for these hours were excluded as well. The need to exclude patient care hours due to Journal of Perinatology (2014), 130 – 135
unclear flow sheet documentation of SpO2 reinforces that hand transcription is more ambiguous and less precise than primary electronic oximeter data and supports incorporating electronic oximeter data into the medical record. Methods for incorporating primary oximeter data into the medical record are available and deserve further attention. Some oximeters are equipped with a histogram function that summarizes distribution of SpO2 in the preceding hours for printout or export. It is feasible to print such data for inclusion in a written medical record, or to import directly into an electronic record. Adopting such methods may eliminate the need for busy bedside caretakers to transcribe oximeter displays by hand, and would enhance quality improvement and research by improving access to reliable data reflecting VLBW infants’ SpO2 experience. However, the need for clinical judgment and interpretation of electronic oximeter data remains necessary for optimal patient care. Under our new policy, the SpO2 target was changed only minimally, and the intervention did not involve any change in policy-driven practice with respect to transcription of oximeter SpO2 by bedside staff. Thus, this study reports the effects upon documentation practice of a newly increased emphasis on precise oxygen management, largely separated from the effect of a change in saturation target. In theory, the intervention should have provided no reason for staff documentation practices to change. Instead, our results demonstrate that staff were more likely to transcribe desired values and less likely to transcribe undesired values following the intervention. We speculate that this reflects the new policy’s heightened emphasis on achieving SpO2 target while avoiding hyperoxemia and hypoxemia. The SpO2 data collected from oximeters in this study were recorded every 2 s; however, averaging time was 8 s for the oximeter displays used during care before and after intervention. This difference is unlikely to have influenced the results of our study, as averaging time has been demonstrated not to affect the time spent within various target SpO2 ranges.14 Moreover, any postulated effect of averaging time would not have differential effects before and after intervention or on different subgroups as seen in our results. Also, it has been reported that the Masimo Radical’s internal calibration algorithm had a artifact that caused the oximeter to under-record values from 87 to 90%, while higher values were recorded in slight excess.15 As staff transcribed SpO2 values from the oximeter display regardless of the underlying algorithm, this artifact could not affect the results we report. The single-center design of this study precludes extrapolating its results to NICUs generally. Others have documented discrepancies between transcribed SpO2 values and those witnessed by an observer over brief periods or in simulated scenarios; however, real-patient electronic oximeter data were not used.16,17 Nonetheless, it is plausible that biased transcription of oximeter SpO2 values may be occurring to greater or lesser extent in other NICUs. This possibility should give pause to any temptation to use hand-transcribed SpO2 data for any rigorous purpose. In summary, this study suggests that transcribed SpO2 values may overestimate achievement of saturation target and may underestimate hyperoxemic and hypoxemic exposure for high-risk VLBW infants receiving intensive care. These results provide evidence for clinicians, investigators and those engaged in quality improvement with regard to the possible impact on handtranscribed SpO2 values of NICU practice changes intended to improve precision of oxygen management. These results support developing methods and practices to introduce primary electronic oximeter data directly into the permanent medical record.
CONFLICT OF INTEREST The authors declare no conflict of interest.
& 2014 Nature America, Inc.
Transcribed oxygen saturation values vs oximeter TL Ruiz et al
135 ACKNOWLEDGEMENTS No external funding was secured for this study.
REFERENCES 1 Fouzas S, Priftis KN, Anthracopoulos MB. Pulse oximetry in pediatric practice. Pediatrics 2011; 128(4): 740–752. 2 American Academy of Pediatric and American College of Obstetricians and Gynecologists. Guidelines for Perinatal Care. American Academy of Pediatric and American College of Obstetricians and Gynecologists 2007. 3 Adhami N, Arabi Y, Raees A, Al-Shimemeri A, Ur-Rahman M, Memish ZA. Effect of corticosteroids on adult varicella pneumonia: Cohort study and literature review. Respirology 2006; 11: 437–441. 4 Boyle M, Wong J. Prescribing oxygen therapy. An audit of oxygen prescribing practices on medical wards at North Shore Hospital, Auckland, New Zealand. N Z Med J 2006; 119: 1–5. 5 Taheri P, Abbasi E, Abdeyazdan Z, Fathizadeh N. The effects of a designed program on oxygen saturation and heart rate of premature infants hospitalized in neonatal intensive care unit of Al-Zahra Hospital in Isfahan in 2008–2009. Iran J Nurs Midwifery Res 2010; 15: 66–70. 6 Sink DW, Thomas P, Bober B, Hagadorn JI. Forty Percent Achievement of the Oxygen Saturation (SpO2) Target Range 85–93% is Associated with Improved Retinopathy of Prematurity Outcomes. Pediatric Academic Societies: Vancouver, Canada 2010, E-PAS20104420.502; E-PAS20104420.502. 7 Zar J. Biostatistical Analysis. 3rd edn (Prentice-Hall: Upper Saddle River, NJ, USA, 1996).
& 2014 Nature America, Inc.
8 Carlo WA, Finer NN, Walsh MC, Rich W, Gantz MG, Laptook AR et al. Target ranges of oxygen saturation in extremely preterm infants. N Engl J Med 2010; 362(21): 1959–1969. 9 Stenson BJ, Tarnow-Mordi WO, Darlow BA, Simes J, Juszczak E, Askie L et al. Oxygen saturation and outcomes in preterm infants. N Engl J Med 2013; 368(22): 2094–2104. 10 Schmidt B, Whyte RK, Asztalos EV, Moddemann D, Poets C, Rabi Y et al. Effects of targeting higher vs lower arterial oxygen saturations on death or disability in extremely preterm infants: a randomized clinical trial. JAMA 2013; 309(20): 2111–2120. 11 Polin RA, Bateman D. Oxygen-saturation targets in preterm infants. N Engl J Med 2013; 368(22): 2141–2142. 12 Bancalari E, Claure N. Oxygenation targets and outcomes in premature infants. JAMA 2013; 309(20): 2161–2162. 13 Ottens J, Baker RA, Newland RF, Mazzone A. The future of the perfusion record: automated data collection vs. manual recording. J Extra Corpor Technol 2005; 37(4): 355–359. 14 Ahmed SR, Finer NN. The effect of averaging time on oximetry values in the premature infant. Pediatrics 2010; 125: e115–e121. 15 Johnston E, Boyle B, Juszczak E, King A, Brocklehurst P, Stenson B. Oxygen targeting in preterm infants using the Masimo SET Radical pulse oximeter. Arch Dis Child Fetal Neonatal Ed 2011; 96: F429–F433. 16 Eastwood GM, O’Connell B, Considine J. Low-flow oxygen therapy in intensive care: an observational study. Austr Crit Care 2011; 24: 269–278. 17 Devitt JH, Rapanos T, Kurrek M, Cohen MM, Shaw M. The anesthetic record: accuracy and complete-ness. Canad J Anaesth 1999; 46(2): 122–128.
Journal of Perinatology (2014), 130 – 135
ORIGINAL ARTICLE
Transcribed oxygen saturation vs oximeter recordings in very low birth weight infants TL Ruiz1,2,3, JM Trzaski3,4, DW Sink3,4 and JI Hagadorn3,4 OBJECTIVE: The objective of this study was to compare hand-transcribed oxygen saturation (SpO2) with electronic oximeter data in very low birth weight infants (VLBWI, o1500 g). STUDY DESIGN: Oximeter data were downloaded from birth through 36 weeks postmenstrual age (PMA) for VLBWI before and after interventions to improve neonatal intensive care unit oxygen management. Transcribed SpO2 values were obtained by chart review. Proportions of transcribed and oximetry data in target (85 to 93%), hypoxemic (80 to 84%), and hyperoxemic (X98%) ranges before and after intervention were compared. RESULT: There were 30 441 oximetry hours before intervention and 54 538 oximetry hours after intervention. Transcribed SpO2 values correlated strongly with oximeter overall. However, during hours on supplemental oxygen, transcribed values significantly overdocumented target range and underdocumented values 80 to 84 and X98%. This finding varied by respiratory support and PMA, and increased after intervention. CONCLUSION: Transcribed SpO2 values overdocumented target range and underdocumented hyperoxemic and hypoxemic ranges compared with electronic oximeter data. These results support incorporating electronic oximeter data into medical records. Journal of Perinatology (2014) 34, 130–135; doi:10.1038/jp.2013.157; published online 19 December 2013 Keywords: nursing records; medical records systems; computerized; electronic medical records
INTRODUCTION Pulse oximetry provides an indirect estimate of arterial oxygen saturation (SpO2) using infrared light that is absorbed by hemoglobin and then transmitted to a photodetector.1 Oximeters are used to assess SpO2 to detect and help prevent low or high fluctuations in oxygenation during patient care. This noninvasive, continuous monitoring device is generally initiated in the delivery area for newborn premature infants. Standard practice for subsequent neonatal intensive care requires documentation of the results of noninvasive oxygenation monitoring for very low birth weight (VLBW, o1500 g at birth) infants.2 Accordingly, many neonatal intensive care units (NICUs) direct bedside providers to transcribe SpO2 values serially for infants on supplemental oxygen monitored with continuous pulse oximetry. In the absence of primary electronic oximeter data, these transcribed SpO2 values may be the only permanent record of a premature infant’s SpO2 experience during their NICU stay. The transcribed SpO2 values may be used for clinical decision-making, quality improvement or research.3–5 However, we were unable to locate studies examining the relationship between SpO2 values transcribed by NICU staff and infants’ actual SpO2 experience as reflected by simultaneous oximeter data. The objectives of this study were: first, to explore the relationship between SpO2 values transcribed by bedside providers and infants’ actual SpO2 experience as recorded by pulse oximetry, and second, to examine the relationship between bedside provider and oximeter before and after an intervention to improve the precision of oxygen therapy. Our null hypothesis was that transcribed SpO2 values would not significantly differ from simultaneous oximeter data before or after the intervention.
METHODS This was a single-center observational study linking existing oximeter data with transcribed SpO2 values from medical records. The Institutional Review Board at Connecticut Children’s Medical Center approved this study by expedited review including waiver of informed consent.
The SpO2 practice intervention In March 2009, the Connecticut Children’s level 4 Hartford NICU implemented a clinical practice intervention to improve achievement of SpO2 goals in VLBW infants. Before this intervention, NICU policy set SpO2 targets of 85 to 92% for infants with gestational age o29 weeks and 88 to 92% for 29 to 34 weeks. Oximeter alarm settings were not strictly defined. The intervention revised SpO2 target only slightly, to 85 to 93% for infants with birth weight o1500 g or gestational age o30 weeks. However, the new policy placed a heightened emphasis on achieving SpO2 target while avoiding hyperoxemia and hypoxemia, using several methods. Oximeter alarm settings were standardized. Reminders were posted at the bedside of each VLBW infant displaying the desired saturation target as well as algorithms for responding to oximeter alarms, including non-oxygen interventions and guidelines for titration of FiO2. These practice changes were accompanied by comprehensive NICU staff education with regard to the risks of oxygen therapy in VLBW infants. Computerized oxygen ordering was standardized to reflect the new policy. Adherence to the new policy was monitored using downloaded oximeter data, providing staff with regular feedback about summarized results with regard to achievement of SpO2 goals. The resulting improved achievement of SpO2 goals is described elsewhere.6 As part of the intervention, electronic data were downloaded from bedside oximeters for VLBW infants admitted between January through June 2008 (before intervention) and April through December 2009 (after intervention, after a 1-month washout period). Infants were included in
1 Health Fellows Program, Trinity College, Hartford, CT, USA; 2Clinical Trials Unit, Connecticut Children’s Medical Center, Hartford, CT, USA; 3Division of Neonatology, Connecticut Children’s Medical Center, Hartford, CT, USA and 4Department of Pediatrics, University of Connecticut School of Medicine, Farmington, CT, USA. Correspondence: Dr JI Hagadorn, Division of Neonatology, Connecticut Children’s Medical Center, 282 Washington Street, Hartford, CT 06106, USA. E-mail: [email protected] Received 8 May 2013; revised 10 October 2013; accepted 14 November 2013; published online 19 December 2013
Transcribed oxygen saturation values vs oximeter TL Ruiz et al
131 these samples if they were inborn, had birth weight o1500 g and were alive on day of life four, the time of the first oximeter download. Using Masimo Radical 7 SET oximeters (Masimo Corporation, Irvine, CA, USA), SpO2 was sampled every 2 s during routine bedside care starting from NICU admission through 36 weeks postmenstrual age (PMA). These comprehensive downloads yielded tens of thousands of hours of highly granular oximeter data for subsequent study. For each oximeter, system dates and times were synchronized with NICU clocks and checked periodically. Averaging time for the oximeter display was 8 s. Oximeter data were not collected for care performed away from the bedside (e.g., operating room, radiology, bathing), as during these times monitoring was performed using an oximeter without data storage capability. After intervention, adherence to the new policy was monitored in real time using the downloaded oximeter data, providing staff with regular feedback about of summarized results with regard to achievement of SpO2 goals. In addition, bedside oximeter alarm settings were audited three times per day by respiratory staff while infants received supplemental oxygen.
Derivation of patient care hour groups for analysis Because tens of thousands of hours of continuous oximeter data result in an unwieldy data set, we used more manageably sized groups of patient care hours for analyses. Ten percent of all patient care hours before and after the intervention were randomly selected and assigned to pre- and postintervention Overall groups. During these patient care hours, infants could be in either room air or supplemental oxygen, and could be supported by any mode of respiratory support. The Overall groups thus allowed a summary examination of the relationship between transcribed SpO2 values and simultaneous oximeter data for all VLBW patient care hours before and after the intervention. In addition, the pre- and postintervention Covariate groups consisted of all patient care hours for which oximeter data were available, during which infants were receiving supplemental oxygen (FiO2 421%) from 6-h nurse shifts characterized by a single respiratory support mode and not more than one nurse providing care. Extensive clinical data for this cohort of patient care hours were available from a prior analysis. The pre- and postintervention Covariate groups were used to assess the impact of clinical factors upon the relationship between transcribed SpO2 values and simultaneous oximeter data independently of changes in respiratory support or changes in bedside care provider.
Transcribed SpO2 data Policy at the Connecticut Children’s Hartford NICU required bedside staff to transcribe the displayed oximeter SpO2 value to the patient care flow sheet hourly at the top of the hour. This practice remained unchanged throughout the pre- and postintervention periods for which SpO2 data were collected. For this study, transcribed SpO2 values were obtained retrospectively by review of patient care flow sheets for included pre- and postintervention patient care hours. Patient care hours were excluded if assigned bedside provider or respiratory support modes could not be determined, or if the infant’s flow sheet had more than one SpO2 value or no value recorded for that hour.
Oximeter SpO2 data Oximeter-derived SpO2 values for included patient care hours were obtained retrospectively using archived continuous oximeter saturation data obtained before and after intervention as described above. Less than 1% of SpO2 values were flagged by the oximeter as having poor signal and these were excluded from analysis. Because bedside staff recorded no values o80% on the patient flow sheets, oximeter-recorded SpO2 values o80% were excluded to ensure comparison of similar ranges.
Calculation of proportions of hand-transcribed and oximeter SpO2 values We compared the proportion of SpO2 values recorded by the oximeter at each saturation value between 80 and 100% with the proportion of handtranscribed values at the same saturation value during the same cohort of patient care hours. The proportion of SpO2 data at each saturation value was calculated before and after intervention for both the Overall and Covariate groups. For example, the proportion of transcribed SpO2 values & 2014 Nature America, Inc.
Figure 1. Selection of patient care hours for pre- and postintervention Overall and Covariate group analyses. equal to 90% was calculated as follows: number of patient care hours in group with transcribed SpO2 value equal to 90% total number of patient care hours included in group
Similarly, the proportion of oximeter SpO2 values equal to 90% was calculated as follows: number of oximeter SpO2 values in group equal to 90% total number of included oximeter SpO2 values in patient care hours included in group
In the Covariate groups, patient care hours were further subgrouped by the mode of respiratory support and PMA. For all groups and subgroups, the proportions of SpO2 values within the following ranges were calculated: (1) 85 to 93%, ‘target’ range; (2) 98 to 100%, ‘hyperoxemic’ range; and (3) 80 to 84%, ‘hypoxemic’ range.
Statistical methods Descriptive and bivariate correlation analyses were performed with SPSS 17.0 or Microsoft Excel 2010. Correlations between nurse and oximeter were calculated using concordance correlation coefficients (rc). Proportions of SpO2 values recorded by oximeter vs transcribed values were compared using the w2 test or 2 log-likelihood ratio goodness-of-fit (G).7
RESULTS Before intervention A total of 30 441 h of oximeter data were available from 24 infants before intervention. Infants in the preintervention cohort had median (interquartile range) birth weight 955 (749, 1240) g, gestational age 27 (26, 29) weeks and were mostly males (n ¼ 18, 75%) exposed to antenatal steroids (n ¼ 21, 87.5%). For the preintervention Overall group, 3046 h were selected randomly without regard to age, supplemental oxygen use or respiratory support (Figure 1). After exclusion criteria were applied, 2862 h were analyzed. For the preintervention Covariate group, there were 5027 h on supplemental oxygen for which detailed clinical data were available. After exclusion criteria, 5023 h were analyzed. After intervention After intervention there were 54 538 h of oximeter data available for analysis from 27 infants. Postintervention infants were not significantly different from the preintervention cohort, with median (interquartile range) birth weight 981 (720, 1180) g and gestational age 27 (25.4, 29) weeks. Most were males (n ¼ 17, 63%) and received antenatal steroids (n ¼ 24, 88.9%). For the postintervention Overall group, 5385 h were randomly selected and 4815 h were analyzed after applying exclusion criteria. For the Journal of Perinatology (2014), 130 – 135
Transcribed oxygen saturation values vs oximeter TL Ruiz et al
132
Figure 2. Hourly transcribed SpO2 values compared with oximeter data for Overall Group patient care hours regardless of supplemental oxygen therapy or mode of respiratory support, before and after intervention.
Covariate group, 5484 patient care hours were available, with 4477 h included for analysis.
Overall Proportions of hand-transcribed SpO2 data at each saturation value between 80 and 100% were strongly correlated with the proportions recorded by oximeter in the Overall groups, with rc ¼ 0.988 (Po0.001; Figure 2) in the preintervention group and rc ¼ 0.993 in the postintervention group (Po0.001). Staff transcribed even SpO2 values more frequently than the oximeter displayed (61.1% vs 57.7%, Po0.001, w2).
Table 1. Percentages of oxygen saturation (SpO2) values in target (85 to 93%), hyperoxemic and hypoxemic ranges, transcribed vs oximeter recorded, for patient care hours in the pre- and postintervention Covariate groups (all hours on supplemental oxygen therapy)
In target X98 80–84
Before intervention (5023 h)
After intervention (4477 h)
Transcribed
Oximeter
Transcribed
Oximeter
26.9 34.6 0.5
24.7*** 37.9*** 3.1***
50.8 6.7 3.6
43.2*** 13.8*** 11.4***
***
Po0.001.
Covariate groups ‘Target’ range (85 to 93%). In the Covariate group, SpO2 values in the target range were significantly over-represented in the transcribed group compared with oximeter for both pre- and post-intervention hours on supplemental oxygen (Table 1). Before intervention, SpO2 values in the target range were overrepresented in the transcribed group compared with oximeter for all hours with PMA o32 weeks and for hours on highfrequency oscillator ventilation (HFOV), conventional ventilation or continuous positive airway pressure (CPAP) (Table 2). These findings were most pronounced for hours o28 weeks PMA and for those on HFOV. Transcribed values were not significantly different from oximeter for hours at PMA 32 to 35 weeks or for those on nasal cannula, the two subgroups where the oximeterdocumented target range achievement was lowest (23.5% and 18.6%, respectively). Following the intervention, the tendency to overdocument target values was more prominent. The target range was significantly over-represented among transcribed SpO2 values for all infants regardless of PMA or respiratory support. Differences between transcribed values and oximeter were larger for all subgroups. Journal of Perinatology (2014), 130 – 135
‘Hyperoxemic’ range (X98%). Saturations in the hyperoxemic range were under-represented among transcribed SpO2 values compared with oximeter in both the pre- and postintervention Covariate groups (Table 1). This was true for all PMA subgroups before and after policy revision (Table 3). Transcribed values were not significantly different from oximeter before or after intervention for hours on HFOV, where the oximeter-documented percentage of values in hyperoxemic range was smallest. For hours on all other respiratory support modes (synchronized intermittent mandatory ventilation, continuous positive airway pressure and nasal cannula), percentage of hyperoxemic values in the transcribed group was significantly lower than oximeter. Following policy revision, the percentage of oximeter-documented hyperoxemic SpO2 values decreased for all subgroups except infants on HFOV. All significant differences observed between transcribed values and oximeter persisted. ‘Hypoxemic’ range (80 to 84%). Saturations in the hypoxemic range were under-represented among transcribed SpO2 values compared with oximeter in both the pre- and postintervention Covariate groups (Table 1). Hypoxemic SpO2 values were & 2014 Nature America, Inc.
Transcribed oxygen saturation values vs oximeter TL Ruiz et al
133 Table 2. Percentage of SpO2 values in the ‘target’ range, 85 to 93%, before and after intervention, transcribed vs oximeter recorded, for patient care hours in the pre- and postintervention Covariate groups, subgrouped by postmenstrual age (weeks) and by respiratory support Before intervention
o28 28–31 32–35 HFOV SIMV CPAP NC
After intervention
Hours
Transcribed
Oximeter
Hours
Transcribed
Oximeter
660 2151 2212 374 917 2088 1644
51.5 39.6 23.0 64.2 39.9 38.5 17.7
42.8** 35.0** 23.5 56.0** 35.2** 34.3*** 18.7
937 2190 1350 481 414 2598 984
69.8 58.5 47.7 68.0 63.3 57.9 49.4
54.9** 42.2** 37.2** 54.0*** 50.7*** 41.7*** 39.1***
Abbreviations: CPAP, continuous positive airway pressure; HFOV, high-frequency oscillator ventilator; NC, nasal cannula; SIMV, synchronized intermittent mandatory ventilation; SpO2, oxygen saturation. ** Po0.01; ***Po0.001.
Table 3. Percentage of SpO2 values in the ‘hyperoxemic’ range, 98 to 100%, before and after intervention, transcribed vs oximeter recorded, for patient care hours in the pre- and postintervention Covariate groups, subgrouped by postmenstrual age (weeks) and by respiratory support Before intervention
o28 28–31 32–35 HFOV SIMV CPAP NC
After intervention
Hours
Transcribed
Oximeter
Hours
Transcribed
Oximeter
660 2151 2212 374 917 2088 1644
11.5 16.0 28.3 2.7 16.6 16.9 32.4
14.8* 19.9*** 32.0*** 3.8 20.0* 20.9*** 36.4***
937 2190 1350 481 414 2598 984
3.1 8.8 13.3 2.9 4.3 8.7 14.3
5.1** 14.2*** 18.7*** 4.9 8.0** 14.7*** 17.7**
Abbreviations: CPAP, continuous positive airway pressure; HFOV, high-frequency oscillator ventilator; NC, nasal cannula; SIMV, synchronized intermittent mandatory ventilation; SpO2, oxygen saturation. *Po0.05; **Po0.01; ***Po0.001.
Table 4. Percentage of SpO2 values in the ‘hypoxemic’ range, 80–84%, transcribed vs oximeter recorded, for patient care hours in the pre- and postintervention Covariate groups, subgrouped by postmenstrual age (weeks) and by respiratory support Before intervention
o28 28–31 32–35 HFOV SIMV CPAP NC
After intervention
Hours
Transcribed
Oximeter
Hours
Transcribed
Oximeter
660 2151 2212 374 917 2088 1644
1.4 0.7 0.2 1.3 1.6 0.2 0.2
5.4*** 4.8*** 2.4*** 5.9*** 5.6*** 4.4*** 1.6***
937 2190 1350 481 414 2598 984
8.0 2.7 1.9 6.0 8.7 3.0 1.8
16.0*** 11.6*** 8.2*** 14.4*** 16.3*** 11.7*** 7.5***
Abbreviations: CPAP, continuous positive airway pressure; HFOV, high-frequency oscillator ventilator; NC, nasal cannula; SIMV, synchronized intermittent mandatory ventilation; SpO2, oxygen saturation. ***Po0.001.
significantly undertranscribed when compared with oximeter for all PMA and respiratory support subgroups before and after the oxygen practice intervention (Table 4). Following the intervention, differences between transcribed values and oximeter were larger for all subgroups regardless of PMA or supplemental oxygen. DISCUSSION In this study, we report that SpO2 values transcribed hourly from the oximeter display by bedside staff generally correlated well with the overall oximeter-recorded SpO2 experience of premature infants during their NICU stay. However, for high-risk VLBW infants receiving supplemental oxygen, bedside providers over-recorded & 2014 Nature America, Inc.
‘target’ SpO2 values (85 to 93%) and under-recorded SpO2 values in ‘hyperoxemic’ (X98%) and ‘hypoxemic’ (80 to 84%) ranges. This bias was present before an intervention designed to improve the achievement of SpO2 goals in VLBW infants, and varied by PMA and mode of respiratory support. Following the intervention, the discrepancy became more pronounced, and was most evident for infants supported with high frequency oscillator ventilation and those at the most premature PMAs. These differences were both statistically and clinically significant. Recent research has resulted in an increased emphasis on achieving SpO2 goals and improving precision of oxygen management in premature infants.8–12 Our results suggest that transcribing hourly SpO2 values may not be sufficiently precise Journal of Perinatology (2014), 130 – 135
Transcribed oxygen saturation values vs oximeter TL Ruiz et al
134 to determine adequately the extent of achievement of SpO2 target or departures from target. In this study, primary oximeter data provided a significantly more accurate reflection of premature infants’ SpO2 experience than SpO2 values hand transcribed from oximeter into the medical record by bedside providers. We were unable to locate other studies comparing hand-transcribed SpO2 values to source oximeter data. However, others have shown that a continuous electronic record is superior to hand-transcription of spot values with respect to data integrity during cardiopulmonary bypass,13 suggesting that use of primary electronic data provides the best opportunity to minimize transcription error and bias. Although our unit policy directed bedside staff to transcribe the oximeter value at the top of each hour, this study did not assume that staff did so with rigorous precision with respect to either time or SpO2 value. Staff may have been preoccupied with other responsibilities, so that the transcribed SpO2 value may not have been from exactly at the top of the hour. Moreover, staff may have chosen to transcribe a more ‘typical’ or ‘desirable’ value than that displayed by the oximeter. Thus, comparing the transcribed SpO2 value with the value that the oximeter recorded precisely at the top of the hour or at any specific time would not have been meaningful or useful. Instead, we compared the proportion of hand-transcribed SpO2 values in the saturation ranges of interest (target, hyperoxemic, hypoxemic) with the proportion of values in those ranges recorded by oximeter during the same patient care hours. This approach allowed us to obtain a valid answer to our research question while avoiding highly problematic attempts to align a specific hand-transcribed SpO2 value to a specific value recorded by the oximeter. This study included several measures to ensure that the comparison of transcribed SpO2 values with the primary oximeter data was valid. Because downloaded oximeter data were not available for care provided away from the NICU bedside (e.g., in radiology, during surgery, etc.), patient care hours and flow sheet SpO2 values corresponding to such times were not considered for inclusion. Each oximeter’s electronic date and time were synchronized with NICU clocks and were checked periodically. Also, because bedside staff recorded no SpO2 values o80% on the patient flow sheets, only oximeter values 80% or higher were included for analysis. These measures maximized the similarity between the cohorts of transcribed and oximeter SpO2 values with respect to time and range, thus optimizing internal validity. We surmise that values o80% were not recorded on bedside flow sheets because these values were interpreted by bedside staff as transient, unrepresentative desaturation episodes or were recorded on a separate bedside log of episodic apnea/bradycardia/desaturation. The effect of emphasizing precision of oxygen management on documentation of discrete episodes below 80% saturation warrants further investigation. Although this was a single-center study, with only 51 infants included—24 before intervention and 27 after intervention—it was not a small study. The unit of analysis was not infants, but patient care hours. Thousands of patient care hours in each group were analyzed, reflecting a broad variety of chronological ages, PMAs, and modes of respiratory support. Because patient care hours were clustered within infants, analyses were performed with hierarchical regression to adjust our results appropriately for nonindependent sampling. Oximeter SpO2 values flagged by the oximeter as having poor signal were excluded from analysis. This represented much o1% of all oximeter SpO2 values. In contrast, 18% of otherwise-eligible patient care hours were excluded from the postintervention Covariate cohort because of hand-transcribed SpO2 values not meeting the criteria for clarity. The great majority of these were due to transcribing multiple saturation values or a range of values rather than a single SpO2 value. As these patient care hours were excluded from analysis, oximeter data for these hours were excluded as well. The need to exclude patient care hours due to Journal of Perinatology (2014), 130 – 135
unclear flow sheet documentation of SpO2 reinforces that hand transcription is more ambiguous and less precise than primary electronic oximeter data and supports incorporating electronic oximeter data into the medical record. Methods for incorporating primary oximeter data into the medical record are available and deserve further attention. Some oximeters are equipped with a histogram function that summarizes distribution of SpO2 in the preceding hours for printout or export. It is feasible to print such data for inclusion in a written medical record, or to import directly into an electronic record. Adopting such methods may eliminate the need for busy bedside caretakers to transcribe oximeter displays by hand, and would enhance quality improvement and research by improving access to reliable data reflecting VLBW infants’ SpO2 experience. However, the need for clinical judgment and interpretation of electronic oximeter data remains necessary for optimal patient care. Under our new policy, the SpO2 target was changed only minimally, and the intervention did not involve any change in policy-driven practice with respect to transcription of oximeter SpO2 by bedside staff. Thus, this study reports the effects upon documentation practice of a newly increased emphasis on precise oxygen management, largely separated from the effect of a change in saturation target. In theory, the intervention should have provided no reason for staff documentation practices to change. Instead, our results demonstrate that staff were more likely to transcribe desired values and less likely to transcribe undesired values following the intervention. We speculate that this reflects the new policy’s heightened emphasis on achieving SpO2 target while avoiding hyperoxemia and hypoxemia. The SpO2 data collected from oximeters in this study were recorded every 2 s; however, averaging time was 8 s for the oximeter displays used during care before and after intervention. This difference is unlikely to have influenced the results of our study, as averaging time has been demonstrated not to affect the time spent within various target SpO2 ranges.14 Moreover, any postulated effect of averaging time would not have differential effects before and after intervention or on different subgroups as seen in our results. Also, it has been reported that the Masimo Radical’s internal calibration algorithm had a artifact that caused the oximeter to under-record values from 87 to 90%, while higher values were recorded in slight excess.15 As staff transcribed SpO2 values from the oximeter display regardless of the underlying algorithm, this artifact could not affect the results we report. The single-center design of this study precludes extrapolating its results to NICUs generally. Others have documented discrepancies between transcribed SpO2 values and those witnessed by an observer over brief periods or in simulated scenarios; however, real-patient electronic oximeter data were not used.16,17 Nonetheless, it is plausible that biased transcription of oximeter SpO2 values may be occurring to greater or lesser extent in other NICUs. This possibility should give pause to any temptation to use hand-transcribed SpO2 data for any rigorous purpose. In summary, this study suggests that transcribed SpO2 values may overestimate achievement of saturation target and may underestimate hyperoxemic and hypoxemic exposure for high-risk VLBW infants receiving intensive care. These results provide evidence for clinicians, investigators and those engaged in quality improvement with regard to the possible impact on handtranscribed SpO2 values of NICU practice changes intended to improve precision of oxygen management. These results support developing methods and practices to introduce primary electronic oximeter data directly into the permanent medical record.
CONFLICT OF INTEREST The authors declare no conflict of interest.
& 2014 Nature America, Inc.
Transcribed oxygen saturation values vs oximeter TL Ruiz et al
135 ACKNOWLEDGEMENTS No external funding was secured for this study.
REFERENCES 1 Fouzas S, Priftis KN, Anthracopoulos MB. Pulse oximetry in pediatric practice. Pediatrics 2011; 128(4): 740–752. 2 American Academy of Pediatric and American College of Obstetricians and Gynecologists. Guidelines for Perinatal Care. American Academy of Pediatric and American College of Obstetricians and Gynecologists 2007. 3 Adhami N, Arabi Y, Raees A, Al-Shimemeri A, Ur-Rahman M, Memish ZA. Effect of corticosteroids on adult varicella pneumonia: Cohort study and literature review. Respirology 2006; 11: 437–441. 4 Boyle M, Wong J. Prescribing oxygen therapy. An audit of oxygen prescribing practices on medical wards at North Shore Hospital, Auckland, New Zealand. N Z Med J 2006; 119: 1–5. 5 Taheri P, Abbasi E, Abdeyazdan Z, Fathizadeh N. The effects of a designed program on oxygen saturation and heart rate of premature infants hospitalized in neonatal intensive care unit of Al-Zahra Hospital in Isfahan in 2008–2009. Iran J Nurs Midwifery Res 2010; 15: 66–70. 6 Sink DW, Thomas P, Bober B, Hagadorn JI. Forty Percent Achievement of the Oxygen Saturation (SpO2) Target Range 85–93% is Associated with Improved Retinopathy of Prematurity Outcomes. Pediatric Academic Societies: Vancouver, Canada 2010, E-PAS20104420.502; E-PAS20104420.502. 7 Zar J. Biostatistical Analysis. 3rd edn (Prentice-Hall: Upper Saddle River, NJ, USA, 1996).
& 2014 Nature America, Inc.
8 Carlo WA, Finer NN, Walsh MC, Rich W, Gantz MG, Laptook AR et al. Target ranges of oxygen saturation in extremely preterm infants. N Engl J Med 2010; 362(21): 1959–1969. 9 Stenson BJ, Tarnow-Mordi WO, Darlow BA, Simes J, Juszczak E, Askie L et al. Oxygen saturation and outcomes in preterm infants. N Engl J Med 2013; 368(22): 2094–2104. 10 Schmidt B, Whyte RK, Asztalos EV, Moddemann D, Poets C, Rabi Y et al. Effects of targeting higher vs lower arterial oxygen saturations on death or disability in extremely preterm infants: a randomized clinical trial. JAMA 2013; 309(20): 2111–2120. 11 Polin RA, Bateman D. Oxygen-saturation targets in preterm infants. N Engl J Med 2013; 368(22): 2141–2142. 12 Bancalari E, Claure N. Oxygenation targets and outcomes in premature infants. JAMA 2013; 309(20): 2161–2162. 13 Ottens J, Baker RA, Newland RF, Mazzone A. The future of the perfusion record: automated data collection vs. manual recording. J Extra Corpor Technol 2005; 37(4): 355–359. 14 Ahmed SR, Finer NN. The effect of averaging time on oximetry values in the premature infant. Pediatrics 2010; 125: e115–e121. 15 Johnston E, Boyle B, Juszczak E, King A, Brocklehurst P, Stenson B. Oxygen targeting in preterm infants using the Masimo SET Radical pulse oximeter. Arch Dis Child Fetal Neonatal Ed 2011; 96: F429–F433. 16 Eastwood GM, O’Connell B, Considine J. Low-flow oxygen therapy in intensive care: an observational study. Austr Crit Care 2011; 24: 269–278. 17 Devitt JH, Rapanos T, Kurrek M, Cohen MM, Shaw M. The anesthetic record: accuracy and complete-ness. Canad J Anaesth 1999; 46(2): 122–128.
Journal of Perinatology (2014), 130 – 135
