You are here

Article Text

Article menu

Download PDFPDF

Paediatric ophthalmology
Continuous oxygen saturation and risk of retinopathy of prematurity in a Japanese cohort
  1. Hiroshi Kubota1,
  2. Yoko Fukushima1,2,
  3. Ryo Kawasaki1,
  4. Takao Endo3,
  5. Yoshikazu Hatsukawa3,
  6. Hiromi Ineyama3,
  7. Katsuya Hirata4,
  8. Shinya Hirano4,
  9. Kazuko Wada4,
  10. Kohji Nishida1,2
  1. 1 Department of Ophthalmology, Osaka University Graduate School of Medicine, Suita, Japan
  2. 2 Integrated Frontier Research for Medical Science Division (iFremed), Institute for Open and Transdisciplinary Research Initiatives (OTRI), Osaka University, Suita, Japan
  3. 3 Department of Ophthalmology, Osaka Women's and Children's Hospital, Izumi, Japan
  4. 4 Department of Neonatal Medicine, Osaka Women's and Children's Hospital, Izumi, Japan
  1. Correspondence to Dr Yoko Fukushima; youko.fukushima{at}ophthal.med.osaka-u.ac.jp

Abstract

Background/aims We assessed the associations between retinopathy of prematurity (ROP) and continuous measurements of oxygen saturation (SpO2), and developed a risk prediction model for severe ROP using birth data and SpO2 data.

Methods This retrospective study included infants who were born before 30 weeks of gestation between August 2009 and January 2019 and who were screened for ROP at a single hospital in Japan. We extracted data on birth weight (BW), birth length, gestational age (GA) and minute-by-minute SpO2 during the first 20 days from the medical records. We defined four SpO2 variables using sequential measurements. Multivariate logistic regression was used to develop a model that combined birth data and SpO2 data to predict treatment-requiring ROP (TR-ROP). The model’s performance was evaluated using the area under the receiver operating characteristic curve (AUC).

Results Among 350 infants, 83 (23.7%) required ROP treatment. The SpO2 variables in infants with TR-ROP differed significantly from those with non-TR-ROP. The average SpO2 and high SpO2 showed strong associations with GA (r=0.73 and r=0.70, respectively). The model incorporating birth data and the four SpO2 variables demonstrated good discriminative ability (AUC=0.83), but it did not outperform the model incorporating BW and GA (AUC=0.82).

Conclusion Data obtained by continuous SpO2 monitoring demonstrated valuable associations with severe ROP, as well as with GA. Differences in the distribution of average SpO2 and high SpO2 between infants with TR-ROP and non-TR-ROP could be used to establish efficient cut-off values for risk determination.

  • Retina
  • Child health (paediatrics)
  • Risk Factors

Data availability statement

Data are available on reasonable request.

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

https://doi.org/10.1136/bjo-2023-324225

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    WHAT IS ALREADY KNOWN ON THIS TOPIC

    • Oxygen supplementation, oxygen saturation (SpO2) and SpO2 targets affect retinopathy of prematurity (ROP). Recent studies have reported that a lower average SpO2 and larger SpO2 fluctuation increase the risk of ROP severity.

    WHAT THIS STUDY ADDS

    • Being outside of the target SpO2 range is associated with ROP severity. Average SpO2 and high SpO2 were both strongly positively associated with gestational age.

    HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

    • Differences in the distribution of average SpO2 and high SpO2 between infants with treatment-requiring ROP (TR-ROP) and non-TR-ROP could help to set efficient cut-off values for risk determination.

    Introduction

    Retinopathy of prematurity (ROP) is a leading cause of childhood blindness, though it is largely avoidable. The number of infants who develop severe ROP varies considerably depending on the population and the level of neonatal care provided.1 2 Short gestation and low birth weight (BW) (or prematurity) pose the highest risk for developing ROP, regardless of the level of neonatal care. In contrast, neonatal management affects many systemic diseases and related interventions, which have been demonstrated to cause severe ROP, including infection, necrotising enterocolitis, respiratory distress syndrome, late circulatory failure, blood transfusion and oxygen supplementation.3–5 In particular, since oxygen supplementation was first identified as a cause of severe ROP,6 several studies have attempted to achieve appropriate oxygen supplementation for premature infants.5 7–10

    During postnatal early life, premature infants often receive respiratory support. Oxygen supplementation is a pivotal component of intensive neonatal care for preterm infants. Ideally, oxygen administration provides adequate oxygenation to meet the metabolic demands of premature infants while avoiding the consequences of both hypoxaemia and hyperoxia. One of the problems with oxygen treatment is that there is a trade-off between reducing the incidence of severe ROP and chronic lung disease with low oxygenation and achieving low mortality with high oxygenation.4 9 10 Although the associations of ROP with oxygen delivery, monitoring and mechanical ventilation have been evaluated, no key parameters or thresholds to reduce ROP have been established. Differences in the monitoring interval and monitoring period of oxygen saturation (SpO2) and differences in the delivery setting have led to inconsistent results. Therefore, it still remains clinically challenging to define the optimal target levels of oxygen for both low ROP treatment and low motility rate in preterm infants.

    Recently, several studies have demonstrated that the minute-by-minute fluctuations in SpO2 measured by pulse oximetry are strongly associated with ROP in extremely preterm infants.11 12 Non-invasive continuous monitoring of SpO2 for optimal oxygen delivery has become almost universal in the neonatal intensive care unit (NICU).13 The use of non-invasive and continuous monitoring of SpO2 is now common in the NICU. Large numerical datasets from bedside SpO2 monitoring of premature infants have been incorporated into electronic medical records, allowing for objective analysis of continuous SpO2 measurements. In this study, we evaluated the correlation between birth prematurity and early postnatal SpO2, as well as the association between ROP severity and early postnatal SpO2. We also attempted to determine the risk of severe ROP using birth data and continuous early postnatal SpO2 data.

    Materials and methods

    Study population

    This study was conducted at the NICU of a single institution in Japan. We included infants born before 30 weeks of gestation between August 2009 and January 2019 at Osaka Women’s and Children’s Hospital. We excluded infants without a known ROP outcome, those without sequential measurements of SpO2, and those with ocular diseases other than ROP. A flow chart of patient enrolment is shown in figure 1.

    Figure 1
    Figure 1

    Participant flow diagram. GA, gestational age; ROP, retinopathy of prematurity.

    ROP screening

    All infants included in the study underwent ROP examination. We performed the initial ROP examination at 29 or 30 weeks of postmenstrual age (PMA) or 3 weeks after birth, whichever came later unless there was a reason not to conduct this examination. The diagnosis of ROP and the indication for ROP treatment followed the International Classification of ROP Revisited14 and the Early Treatment for ROP (ETROP) study,15 respectively.

    Data collection

    We retrospectively collected birth data, ROP data and SpO2 data from the electronic medical records of the infants. Birth data included sex, BW, birth length (BL) and gestational age (GA). ROP data included the dates of retinal examinations, ROP stages and zone at each fundus examination and treatments for ROP. SpO2 data included all SpO2 measurements during the first 20 days after birth. Specifically, we continuously monitored SpO2 in all infants with a probe and pulse oximeter (TL-535U, Nihon Kohden, Tokyo, Japan) at both NICUs. We averaged the acquired sequential SpO2 measurements over 1 min, and the values were recorded using a biomedical information system (PRM-7400, Nihon Kohden). We extracted all measurements during the period from the system. If measurements were continuously recorded for 20 postnatal days (28 800 min), with no missing measurements, the number of measurements was recorded as 28 800 counts. Subsequently, we removed times with 0% SpO2 owing to artefacts and times with missing SpO2 values. We did not conduct data imputation in this study.

    Definition of SpO2 variables

    We defined four SpO2 variables: SpO2 average, SpO2 fluctuation, high SpO2 and low SpO2. SpO2 average was obtained by dividing the total SpO2 measurements by the total valid recording time. SpO2 fluctuation was obtained by dividing the total difference between two consecutive SpO2 measurements by the total valid time (figure 2).

    Figure 2
    Figure 2

    Schematic diagram representing SpO2 fluctuation, high SpO2 and low SpO2. SpO2 fluctuation is defined as the average value of the differences between two consecutive SpO2 measurements during the entire study period. High SpO2 and low SpO2 are cumulative rates; >95% SpO2 and<80% SpO2, respectively. SpO2, oxygen saturation.

    From July 2009 to June 2013, neonatologists used a target SpO2 range of 85%–95% for infants. From July 2013, the target SpO2 range for infants was 90%–95% for infants within the first week of life and 88%–95% for infants through a PMA of 36 weeks. In cases where an infant was over the target range without respiratory support, additional oxygen supplementation was not administrated.12 To evaluate the impact of being outside of the target SpO2 range, we calculated high SpO2 and low SpO2 as the proportion of cumulative time spent over 95% and below 80%, respectively (figure 2).

    Outcome and statistical analysis

    First, we compared the birth data and SpO2 variables of patients with non-TR-ROP with those of patients with TR-ROP. Second, we analysed the correlations between SpO2 variables and GA. Finally, we developed the prediction model using the above variables. After univariate screening to identify potentially important variables, we selected all potential predictive variables with p<0.05 for further analysis. We then analysed the selected variables by multivariate logistic regression. We measured the accuracy, precision and F-measure of the model, and we evaluated the model’s performance in terms of discrimination and calibration using the area under the receiver operating characteristic curve (AUC). In addition to the model with all selected variables, we constructed two other rigorous models by removing the initially included variables while maintaining the predictive power. We compared the performance of the models based on the AUC.

    Continuous variables are presented as the mean and range. For comparisons between the two groups, we used Fisher’s exact test for categorical variables and the Mann-Whitney U test for continuous variables. Values of p<0.05 were considered statistically significant. Correlations were calculated with Pearson’s correlation coefficient. We performed linear regression analysis and tested if the two regression lines of non-TR-ROP and TR-ROP had an equal slope. All analyses were performed by using JMP Pro statistical software, V.16 (SAS Institute).

    Results

    Among 350 infants (mean BW, 839 g; mean GA, 26.5 weeks), 293 infants (83.7%) developed any ROP. Of these infants, 83 (23.7%) required treatment for ROP. The average BW, BL and GA values in TR-ROP infants were lower than those in non-TR-ROP infants (table 1).

    View this table:
    Table 1

    Birth characteristics of the study population

    The average count of SpO2 measurements for all infants was 27 235, which is equivalent to 18.9 days. More than 95% of the maximum of 28 800 counts were valid measurements in 262 infants; values were less than 90% in 20 infants. While the SpO2 average and low SpO2 in TR-ROP infants were significantly lower than in non-TR-ROP infants, SpO2 fluctuation and high SpO2 in TR-ROP infants were significantly higher than in non-TR-ROP infants (table 1).

    We identified significant correlations between all SpO2 variables and GA. The average SpO2 and high SpO2 had strong positive correlations with GA (r=0.73 (0.68–0.78) and r=0.70 (0.65–0.75), respectively), while SpO2 fluctuation and low SpO2 had negative correlations with GA (r=−0.30 (−0.20 to −0.39) and r=−0.52 (−0.44 to −0.59), respectively). Figure 3 shows the distributions of the SpO2 variables by GA for non-TR-ROP and TR-ROP infants. All four SpO2 variables had weaker associations with GA in TR-ROP infants than in non-TR-ROP infants. The regression line slopes of the average SpO2‐GA and the high SpO2‐GA were statistically unequal between non-TR-ROP and TR-ROP infants. Notably, the high SpO2‐GA slope differed by approximately two times between non-TR-ROP and TR-ROP infants.

    Figure 3
    Figure 3

    Associations between SpO2 variables and GA by group. Scatter plots and linear regression lines of four SpO2 variables and GA in non-TR-ROP (blue) and TR-ROP (red) infants. The equation, correlation coefficient and statistical significance are provided for non-TR-ROP (blue) and TR-ROP (red) red and blue in the plots. GA, gestational age; SpO2, oxygen saturation; TR-ROP, treatment-requiring retinopathy of prematurity.

    Table 2 summarises the performance of the three models. We first constructed a full prediction model incorporating the following seven variables: BW, BL, GA, SpO2 average, SpO2 fluctuation, high SpO2 and low SpO2. The model had moderate predictive performance, with an AUC of 0.83. Subsequently, we constructed a rigorous model that removed all four SpO2 variables and a minimal model incorporating only BW and GA. Both models demonstrated similar performance to the full model. Ultimately, although the F-measure of the full model was the highest among the three models, its model performance according to the AUC was not superior to the other models (p=0.58).

    View this table:
    Table 2

    Model performance for predicting ROP severity

    Discussion

    Since the first clinical report of ROP in premature infants who received 100% oxygen administration in the 1950s,16 numerous studies have been conducted to determine appropriate oxygen supplementation and respiratory support to reduce ROP severity by analysing factors related to oxygen, such as the fraction of inspired oxygen (FiO2), SpO2 and partial pressure of oxygen.4 17 18 Recent technological advances have enabled the analysis of continuous SpO2 data from infants in the NICU. However, the raw data from continuous SpO2 monitoring, before being compressed and processed, have rarely been used for ROP risk determination and prediction. In this study, we shed light on how the minute-by-minute SpO2 levels in early life influenced ROP severity. Although all four SpO2 variables differed between infants with and without treatment, these differences did not improve the predictive performance for severe ROP. Eventually, GA and BW, which are well-known risk factors for ROP, contributed the most to predicting ROP severity. Moreover, the performance of our model was poorer than several existing models.19–22 One possible explanation could be that we targeted all infants born before 30 weeks of gestation, without setting a lower limit for GA when developing the model to avoid overly optimistic performance with unbalanced data.23

    We demonstrated previously that the SpO2 average and SpO2 fluctuation differed between non-TR-ROP and TR-ROP infants.12 In this study, we further investigated the association between SpO2 variables and GA. Among the four SpO2 variables, both SpO2 average and high SpO2 showed a strong association with GA, despite the target oxygen level remaining constant across GAs. Although no infants had an average SpO2 below the target during the study period, some infants with a longer GA had average SpO2 levels exceeding the target range. This observation may be explained by the fact that premature infants were able to exceed the upper limit of the target SpO2 (SpO2>95%) without receiving supplemental oxygen as they matured postnatally.24 25 Consequently, infants can surpass the target SpO2 range without requiring oxygen supplementation. Although we did not include information on oxygen treatment in this study, the decreasing trend in low SpO2 with a longer GA supports this explanation. Regarding ROP severity, these strong associations did not lead to refinements in the prediction model. However, we observed an increasing gap in the regression line slope of high SpO2 and GA between non-TR-ROP and TR-ROP, suggesting the potential to set an effective cut-off value for high SpO2. In particular, if we had defined infants with high SpO2<0.6 as a high-risk TR-ROP group, the estimated number of infants requiring screening would have decreased to 88 (25.1%), with 100% sensitivity. This promissing cut-off value must be validated in the future studies.

    As for one of the other oxygen-related factors, FiO2 data remain controversial in terms of its ability to determine ROP risk. Recent studies have reported that although incorporating daily FiO2 into the screening criteria proposed in the Postnatal Growth and ROP Study improved both the sensitivity and specificity of type 1 ROP risk determination,26 weekly FiO2 measurement did not improve the predictive performance of GA for TR-ROP.27 With titration of FiO2 to SpO2, FiO2 might demonstrate a correlation with GA. When oxygen-related factors, such as SpO2 and FiO2, are incorporated into the risk determination of ROP, the relationship between GA and oxygen-related factors should be considered. In addition, it would be necessary to properly set the starting point, the time period and the data collection interval of oxygen-related factor analysis when incorporating oxygen-related factors into the predictive model. In view of technological advances in big data analysis, using machine learning would make it possible to explore the best application of continuous data, including SpO2 and FiO2, for ROP risk determination. Using the data-driven approach, we aim to identify novel trends over time or specific points at which SpO2 and FiO2 predict severe ROP. In the future, a real-time analysis system for ROP risk prediction might be implemented using monitoring devices powered by artificial intelligence.

    There are several limitations to consider in this study. First, the study included a small cohort of Japanese patients in a single NICU. Because the rates of infants treated for ROP in our study were comparable to those in the Neonatal Research Network of Japan database, our cohort can be considered a group of standard Japanese infants. However, our results may not be generalisable to other populations.2 Second, we have concerns about the SpO2 data, such as SpO2 measurement. For example, oximeter inaccuracy and failure owing to motion artefact may have decreased the valid recording duration of SpO2.28 An increase in the invalid recording duration or low accuracy may have affected the SpO2 variables, resulting in difficulties in setting the cut-off values. Another concern is the target SpO2 range, which has been changed as a result of some clinical trials over the years.7–10 With further optimisation of the target range, it may be necessary to revise the definitions of SpO2 variables, such as high SpO2 and low SpO2. Finally, we did not compare our model with other predictive models.

    Conclusion

    In conclusion, this study demonstrated that SpO2 variables, which were obtained by continuous data monitoring, were associated with severe ROP and GA. Of note, different distributions of average SpO2 and high SpO2 between non-TR-ROP and TR-ROP could help to set efficient cut-off values for risk determination. Future ROP research should be directed at data-driven analysis of continuous data monitoring using artificial intelligence, which could help to discover novel risk determination models for ROP.

    Data availability statement

    Data are available on reasonable request.

    Ethics statements

    Patient consent for publication

    Ethics approval

    This study involves human participants and was approved by the ethics review board of Osaka Women’s and Children’s Hospital (the approved number, 888-2). The study protocol did not require that each patient provide written informed consent (based on the Ethical Guidelines for Medical and Health Research Involving Human Subjects issued by the Japanese government). We instead posted the protocol on the hospital organisation website to notify study guardians of all infants.

    Acknowledgments

    We thank Emily Woodhouse, PhD, and Analisa Avila, MPH, ELS, of Edanz (https://jp.edanz.com/ac) for editing a draft of this manuscript.

    References

    1. ↵
      2. Blencowe H ,
      3. Lawn JE ,
      4. Vazquez T , et al
      . Preterm-associated visual impairment and estimates of retinopathy of Prematurity at regional and global levels for 2010. Pediatr Res 2013;74 Suppl 1(Suppl 1):35–49. doi:10.1038/pr.2013.205
      OpenUrlPubMed
    2. ↵
      2. Darlow BA ,
      3. Lui K ,
      4. Kusuda S , et al
      . International variations and trends in the treatment for retinopathy of Prematurity. Br J Ophthalmol 2017;101:1399–404. doi:10.1136/bjophthalmol-2016-310041
      OpenUrlAbstract/FREE Full Text
    3. ↵
      2. Hellström A ,
      3. Smith LEH ,
      4. Dammann O
      . Retinopathy of Prematurity. Lancet 2013;382:1445–57.:S0140-6736(13)60178-6. doi:10.1016/S0140-6736(13)60178-6
      OpenUrlCrossRefPubMedWeb of Science
    4. ↵
      2. Tsai AS ,
      3. Chou H-D ,
      4. Ling XC , et al
      . Assessment and management of retinopathy of Prematurity in the era of anti-vascular endothelial growth factor (VEGF). Prog Retin Eye Res 2022;88:101018.:S1350-9462(21)00079-3. doi:10.1016/j.preteyeres.2021.101018
      OpenUrl
    5. ↵
      2. Darlow BA ,
      3. Vento M ,
      4. Beltempo M , et al
      . Variations in oxygen saturation targeting, and retinopathy of Prematurity screening and treatment criteria in neonatal intensive care units: an international survey. Neonatology 2018;114:323–31. doi:10.1159/000490372
      OpenUrl
    6. ↵
      2. CAMPBELL K
      . Intensive oxygen therapy as a possible cause of Retrolental Fibroplasia; a clinical approach. Med J Aust 1951;2:48–50.
      OpenUrlPubMed
    7. ↵
      1. Supplemental Therapeutic Oxygen for Prethreshold Retinopathy Of Prematurity (STOP-ROP)
      . A randomized, controlled trial I: primary outcomes. Pediatrics 2000;105:295–310. doi:10.1542/peds.105.2.295
      OpenUrlCrossRefPubMedWeb of Science
    8. ↵
      2. Askie LM ,
      3. Henderson-Smart DJ ,
      4. Irwig L , et al
      . Oxygen-saturation targets and outcomes in extremely Preterm infants. N Engl J Med 2003;349:959–67. doi:10.1056/NEJMoa023080
      OpenUrlCrossRefPubMedWeb of Science
    9. ↵
      2. Carlo WA ,
      3. Finer NN ,
      4. Walsh MC , et al
      . Target ranges of oxygen saturation in extremely Preterm infants. N Engl J Med 2010;362:1959–69. doi:10.1056/NEJMoa0911781
      OpenUrlCrossRefPubMedWeb of Science
    10. ↵
      2. Askie LM ,
      3. Darlow BA ,
      4. Finer N , et al
      . Association between oxygen saturation targeting and death or disability in extremely Preterm infants in the neonatal oxygenation prospective meta-analysis collaboration. JAMA 2018;319:2190–201. doi:10.1001/jama.2018.5725
      OpenUrlCrossRefPubMed
    11. ↵
      2. Das A ,
      3. Mhanna M ,
      4. Sears J , et al
      . Effect of fluctuation of oxygenation and time spent in the target range on retinopathy of Prematurity in extremely low birth weight infants. J Neonatal Perinatal Med 2018;11:257–63. doi:10.3233/NPM-1757
      OpenUrl
    12. ↵
      2. Imanishi Y ,
      3. Hirata K ,
      4. Nozaki M , et al
      . Effect of fluctuation of oxygenation on the development of severe retinopathy of Prematurity in extremely Preterm infants. J Perinatol 2020;40:515–21. doi:10.1038/s41372-019-0571-y
      OpenUrlPubMed
    13. ↵
      2. Kumar N ,
      3. Akangire G ,
      4. Sullivan B , et al
      . Continuous vital sign analysis for predicting and preventing neonatal diseases in the twenty-first century: big data to the forefront. Pediatr Res 2020;87:210–20. doi:10.1038/s41390-019-0527-0
      OpenUrlPubMed
    14. ↵
      1. International Committee for the Classification of Retinopathy of Prematurity
      . The International classification of retinopathy of Prematurity Revisited. Arch Ophthalmol 2005;123:991–9. doi:10.1001/archopht.123.7.991
      OpenUrlCrossRefPubMedWeb of Science
    15. ↵
      1. Early Treatment For Retinopathy Of Prematurity Cooperative Group
      . Revised indications for the treatment of retinopathy of Prematurity: results of the early treatment for retinopathy of Prematurity randomized trial. Arch Ophthalmol 2003;121:1684–94. doi:10.1001/archopht.121.12.1684
      OpenUrlCrossRefPubMedWeb of Science
    16. ↵
      2. KINSEY VE
      . Retrolental Fibroplasia; cooperative study of Retrolental Fibroplasia and the use of oxygen. AMA Arch Ophthalmol 1956;56:481–543.
      OpenUrlCrossRefPubMedWeb of Science
    17. ↵
      2. Srivatsa B ,
      3. Hagan JL ,
      4. Clark RH , et al
      . Oxygenation factors associated with retinopathy of Prematurity in infants of extremely low birth weight. J Pediatr 2022;247:46–52.S0022-3476(22)00318-3. doi:10.1016/j.jpeds.2022.03.057
      OpenUrl
    18. ↵
      2. Hartnett ME ,
      3. Lane RH
      . Effects of oxygen on the development and severity of retinopathy of Prematurity. J AAPOS 2013;17:229–34.:S1091-8531(13)00133-X. doi:10.1016/j.jaapos.2012.12.155
      OpenUrlCrossRefPubMedWeb of Science
    19. ↵
      2. Pivodic A ,
      3. Hård A-L ,
      4. Löfqvist C , et al
      . Individual risk prediction for sight-threatening retinopathy of Prematurity using birth characteristics. JAMA Ophthalmol 2020;138:21–9. doi:10.1001/jamaophthalmol.2019.4502
      OpenUrl
    20. ↵
      2. Binenbaum G ,
      3. Bell EF ,
      4. Donohue P , et al
      . Development of modified screening criteria for retinopathy of Prematurity: primary results from the postnatal growth and retinopathy of Prematurity study. JAMA Ophthalmol 2018;136:1034–40. doi:10.1001/jamaophthalmol.2018.2753
      OpenUrl
    21. ↵
      2. Cao JH ,
      3. Wagner BD ,
      4. McCourt EA , et al
      . The Colorado-retinopathy of Prematurity model (CO-ROP): postnatal weight gain screening algorithm. J AAPOS 2016;20:19–24.:S1091-8531(16)00020-3. doi:10.1016/j.jaapos.2015.10.017
      OpenUrlPubMed
    22. ↵
      2. Shiraki A ,
      3. Fukushima Y ,
      4. Kawasaki R , et al
      . Retrospective validation of the postnatal growth and retinopathy of Prematurity (G-ROP) criteria in a Japanese cohort. Am J Ophthalmol 2019;205:50–3.:S0002-9394(19)30138-2. doi:10.1016/j.ajo.2019.03.027
      OpenUrl
    23. ↵
      The relationship between precision-recall and ROC curves. Proceedings of the 23rd international conference on Machine learning; 2006
    24. ↵
      2. McGregor ML ,
      3. Bremer DL ,
      4. Cole C , et al
      . Retinopathy of Prematurity outcome in infants with Prethreshold retinopathy of Prematurity and oxygen saturation >94% in room air: the high oxygen percentage in retinopathy of Prematurity study. Pediatrics 2002;110:540–4. doi:10.1542/peds.110.3.540
      OpenUrlCrossRefPubMedWeb of Science
    25. ↵
      2. Di Fiore JM ,
      3. Bloom JN ,
      4. Orge F , et al
      . A higher incidence of intermittent Hypoxemic episodes is associated with severe retinopathy of Prematurity. J Pediatr 2010;157:69–73. doi:10.1016/j.jpeds.2010.01.046
      OpenUrlCrossRefPubMedWeb of Science
    26. ↵
      2. Estrada MM ,
      3. Tomlinson LA ,
      4. Yu Y , et al
      . Daily oxygen supplementation and risk of retinopathy of Prematurity. Ophthalmic Epidemiol 2023;30:317–25. doi:10.1080/09286586.2022.2111687
      OpenUrl
    27. ↵
      2. Chen JS ,
      3. Anderson JE ,
      4. Coyner AS , et al
      . Quantification of early neonatal oxygen exposure as a risk factor for retinopathy of Prematurity requiring treatment. Ophthalmol Sci 2021;1:100070. doi:10.1016/j.xops.2021.100070
    28. ↵
      2. Petterson MT ,
      3. Begnoche VL ,
      4. Graybeal JM
      . The effect of motion on pulse Oximetry and its clinical significance. Anesth Analg 2007;105(6 Suppl):S78–84. doi:10.1213/01.ane.0000278134.47777.a5
      OpenUrlCrossRefPubMedWeb of Science

    Footnotes

    • X @Hatsuhi

    • Contributors Conception and design: YF. Data acquisition and/or research execution: HK, YF, TE, YH and HI. Analysis and interpretation: HK, YF, RK, YH, KH, SH, KW and KN. Manuscript preparation: HK and YF. Guarantor: YF.

    • Funding The authors have no proprietary or commercial interest in any materials discussed in this article. This work was supported by the Japan Society for the Promotion of Science KAKENHI (grant 21K09717), with additional funding from the Takeda Science Foundation.

    • Disclaimer The funding organisation had no role in the design or conduct of this research.

    • Competing interests None declared. The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

    • Provenance and peer review Not commissioned; externally peer reviewed.