Can a predictive model for extubation readiness in preterm infants improve rates of successful extubation?

December 28, 2020

MANUSCRIPT CITATION

Gupta D, Greenberg R, Sharma A, Natarajan G, Cotton M, Thomas R, Chawla S. A predictive Model for Extubation Readiness in Extremely Preterm Infants. Journal of Perinatology 2019; 39:1663-9. PMID: 31455825

REVIEWED BY

Rebecca Shay MD
Neonatology Fellow
Department of Pediatrics
University of Colorado Hospital and
Children’s Hospital Colorado

Clyde J. Wright, MD
Associate Professor
Section of Neonatology
Department of Pediatrics
Children’s Hospital Colorado and
University of Colorado School of Medicine

TYPE OF INVESTIGATION

Clinical prediction guides

QUESTION

In a population of preterm infants requiring mechanical ventilation, could available demographic and clinical data be used to create an extubation readiness estimator tool that reliably predicts extubation success?

METHODS

  • Design: Retrospective Cohort Study
  • Allocation: Not applicable
  • Blinding: Not applicable
  • Follow-up period: Premature infants born in the study setting were followed through their hospitalization and assessed for mechanical ventilation requirement as well as first extubation attempt during the first 60 days of life. Patients were then monitored for extubation success or failure. Assessments at different time intervals were not performed.
  • Setting: This study was performed at the Hutzel Women’s Hospital, a single center level III neonatal intensive care unit (NICU)
  • Patients:
    • Eligible infants included preterm neonates born between January 2009 and December 2015. Infants were included if they had a birth weight ≤1250 grams, received mechanical ventilation via endotracheal tube, and underwent a first elective extubation attempt within 60 days of life.
    • Infants were excluded if they died before an extubation attempt was made, if they had an unplanned extubation, if they had missing data, or if they were part of an ongoing multicenter study evaluating extubation readiness.
  • Intervention: Creation of a predictive model for extubation readiness based on the results from the retrospective cohort study.
  • Outcomes: The primary outcome for this study was to develop an estimator for predicting successful extubation for an individual preterm infant. Successful extubation was defined as survival for five or more days without the need for respiratory support from an endotracheal tube. Secondary outcomes included time to reintubation and characteristics of infants with successful or failed extubation outcomes.
  • Analysis and Sample Size: The sample size of infants born in the study period with a birth weight ≤1250 grams were 621 eligible patients; after exclusion criteria, 312 infants were included in the final analysis.
  • Patient follow-up: All 312 patients that were included in the final analysis were followed up with 228 infants (73%) meeting study criteria of successful extubation and 84 infants (27%) failing extubation.

MAIN RESULTS

As summarized in Table 1, successfully extubated infants demonstrated statistically significant demographic and clinical differences. Initial intubation criteria were not provided. These infants had higher gestational age, increased birth weight, lower incidence of histologic chorioamnionitis, and lower peak respiratory severity score (RSS) in the first six hours of life. The RSS was calculated as the product of the mean airway pressure and fraction of inspired oxygen (FiO2) and used as the primary measure of respiratory health in this study. At the time of successful extubation, infants in the successful extubation group had higher postmenstrual age, higher weight at extubation, higher pre-extubation blood gas pH, lower pre-extubation FiO2 requirement, and lower pre-extubation RSS. Re-intubation criteria was not standardized. Clinical data on the ductus arteriosus, sepsis/infection, nutritional status, brain ultrasound, of hemoglobin were provided.

Variable Extubation success (n=228) Extubation failure (n=84) P value
Gestational age (weeks) 27 (26 – 28) 26 (25 – 27) <0.001
Birth weight (g) 931 (790 – 1090) 815 (690 – 914) <0.001
Histologic chorioamnionitis 100/213 (47%) 46/76 (61%) 0.04
Peak RSS in first 6h 3.6 (2.5 – 4.6) 3.8 (2.7 – 6.3) 0.03
Postmenstrual age at extubation (weeks) 28 (27 – 29) 27 (26 – 28) <0.001
Weight at extubation (g) 970 (810 – 1100) 810 (700 – 980) <0.001
Pre-extubation pH 7.36 (7.31 – 7.41) 7.33 (7.29 – 7.38) 0.002
Pre-extubation FiO2 (%) 23 (21 – 28) 25 (21 – 30) <0.001
Pre-extubation RSS 1.4 (1.2 – 1.8) 1.6 (1.4 – 1.8) 0.005

Table 1. Statistically significant differences in patient demographics and clinical variables between infants successfully extubated and those who failed extubation.

A multiple logistic regression analysis demonstrated that successful extubation was associated with higher gestational age, higher chronologic age at the time of extubation, higher blood gas pH, lower pre-extubation FiO2, and lower RSS in the first six hours of life. Based on this information, the authors created a prediction model with area under the curve of 0.77. The model had sensitivity/specificity of 87%/53%, 76%/66%, and 54%/81% for cutoffs of 60%, 70%, and 80% probability of extubation success, respectively.

CONCLUSION

The authors concluded that they were able to create an extubation readiness estimator based on demographic and clinical data collected in their retrospective cohort study. They report that their calculator can provide the probability of extubation success for an individual preterm infant.

COMMENTARY

Extubation readiness is a complex, multifactorial topic with important clinical implications. In the premature neonatal population, there are no proven predictors of extubation readiness, success, or failure; decisions surrounding extubation are left to clinician judgment resulting in practice variability and high rates of failure (1). Frequency of failed elective extubation in this patient population ranges from 23-42% (2-4). Extubation failure has been demonstrated to be associated with increased rates of death, bronchopulmonary dysplasia, and retinopathy of prematurity as well as prolonged respiratory support requirement and hospitalization (3-4).

In the current study, Gupta and colleagues created an extubation readiness estimator with available clinical and demographic data. The authors have made their calculator available to the public (http://elasticbeanstalk-us-east-2-676799334712.s3-website.us-east-2.amazonaws.com). After entering the gestational age, extubation day of life, pre-extubation percent oxygen, highest RSS in first 6 hours, weight at extubation and pre-extubation pH, the clinician will receive the probability of successful extubation. Importantly, in this predictive model, the sensitivity and specificity of the prediction change with the results of the calculated probability of successful extubation. If the calculator predicts a 60% chance of success, the authors report an 87% sensitivity and 53% specificity. This indicates that 87% of patients with similar values inserted into the calculator who succeed extubation will be predicted to succeed, and 53% of patients who fail extubation will be predicted to fail. If the calculator predicts an 80% chance of success, the sensitivity falls considerably to 54%; this implies that the ability to predict who will succeed is decreased at this level. Conversely, the sensitivity increases to 81% suggesting that the ability to predict who will fail improves.

Though difficult to conceptualize, this change in “test performance” is intuitive and analogous to the way a clinician might make their own prediction. For example, a provider who feels an infant has a 95% chance of successful extubation likely has a high degree of certainty about the estimation. A clinical who feels an infant has only a 60% chance of extubation success may have uncertainty regarding the exact percentage. In this way, the predictive model performs similarly to clinical judgment, performing less well as uncertainty increases. A predictive model is most useful when it provides added certainty in the latter scenario; it is unlikely to be applied or improve outcomes if it only performs well when extubation success is either highly probable or very unlikely.

It is unknown how the tool derived retrospectively will perform prospectively, and this uncertainty has potential to result in unanticipated harm. The authors comment that use of the estimator may help facilitate earlier extubation in infants estimated to have a high chance of success – if this hypothesis is true, rates of successful extubation should increase. Alternatively, a poor score could prevent an extubation attempt in some babies, with the unanticipated consequence of increasing ventilation days and lung injury. The balance of these potential harms and benefits should be prospectively assessed with broad implementation of this tool.

REFERENCES

  1. Shalish W, Latremouille S, Papenburg J, Sant’Anna GM. Predictors of extubation readiness in preterm infants: a systematic review and meta-analysis. Arch Dis Child Fetal Neonatal Ed. 2019;104:F89-F97.
  2. Hermeto F, Martins BM, Ramos JR, Bhering CA, Sant’Anna GM. Incidence and main risk factors associated with extubation failure in newborns with birth weight <1250 grams. J pediatr. 2009;85:397-402.
  3. Chawla S, Natarajan G, Shankaran S, Carper B, Brion LP, Keszler M, et al. Markers of successful extubation in extremely preterm infants, and morbidity after failed extubation. J Pediatr. 2017;189:113-e.112.
  4. Manley BJ, Doyle LW, Owen LS, Davis PG. Extubating extremely preterm infants: predictors of success and outcomes following failure. J Pediatr. 2016;173:4509.
css.php