Hypoxemia in School-age Children Undergoing One-lung Ventilation: A Retrospective Cohort Study from the Multicenter Perioperative Outcomes Group

Institutional review board approval including a waiver of informed consent for data collection was obtained from individual sites contributing a limited data set to the Multicenter Perioperative Outcomes Group central repository (HUM00024166; University of Michigan, Ann Arbor, Michigan). Further, the institutional review board of Wake Forest University Health Sciences (Winston-Salem, North Carolina) reviewed and approved the protocol for this retrospective observational cohort study (IRB66373). The Strengthening The Reporting of Observational Studies in Epidemiology (STROBE) checklist was used to prepare this article.

The research protocol and statistical plan were approved by the Multicenter Perioperative Outcomes Group Perioperative Clinical Research Committee. The Multicenter Perioperative Outcomes Group central database was queried from July 2004 to January 2022 inclusive, using the current procedural terminology codes for various thoracic procedures. These codes included 32601, 32661, 32662, 32663, 32652, 32650, 32666, 32667, 32100, 32140, and 32141.

Inclusion criteria included age 4 to 17 yr inclusive, noncardiovascular thoracic surgery in which one-lung ventilation was used to facilitate surgery, and operative exposure. Once identified, the patients were divided into two cohorts: 4 to 9 yr of age and 10 to 17 yr of age inclusive. A child who was 9 yr and 6 months old would be part of the 4- to 9-yr-old cohort. The point of division was intended to reflect the anticipated dichotomy in one-lung ventilation management and physical size characteristic of these cohorts to compare like to like and not, for example, to compare a 4-yr-old to a 17-yr-old. Children 4 to 9 yr of age tend to undergo lung isolation via endobronchial intubation or use of a bronchial blocker. Although a 26-Fr double lumen endobronchial tube can technically be used in patients as young as 8 yr, it is roughly equivalent in size to a 6.5-internal-diameter endotracheal tube, which may in some cases be oversized for an 8 or 9 yr old.⁷  The 26-Fr double lumen endobronchial tube is the smallest commercially available double lumen endobronchial tube. Children 10 to 17 yr of age by in large undergo one-lung ventilation using a bronchial blocker or double lumen endobronchial tube.^11,13  Frequency of usage is reported for all approaches to one-lung ventilation in both cohorts; however, 8- and 9-yr-old children in whom a double lumen endobronchial tube was used were included with the older cohort for multivariable analysis. Additionally, patients in the older cohort in whom mainstem intubation was used will be excluded from the primary analysis of risk factors for hypoxemia in this age group.

Children less than 4 yr of age and any child undergoing cardiovascular surgery were excluded. We excluded patients who were American Society of Anesthesiologists (ASA; Schaumburg, Illinois) Physical Status V or VI and patients with cyanotic congenital heart disease. Additionally, we excluded patients with preprocedure room air oxygen saturation of less than 94%. Patients with a period of one-lung ventilation less than 10 min and patients undergoing bronchiolar lavage for alveolar proteinosis were excluded. Finally, patients undergoing a bilateral procedure with sequential episodes of one-lung ventilation were also excluded.

The primary outcome, hypoxemia, was defined as the presence of oxygen saturation measured by pulse oximetry (Spo₂) of less than 90% for 3 min or more continuously during one-lung ventilation. The secondary outcome, severe hypoxemia, was defined as the occurrence of Spo₂ less than 90% for 5 min or more continuously during one-lung ventilation and hypercarbia. While arbitrary, these definitions of hypoxemia were chosen by consensus among members of the Multicenter Perioperative Outcomes Group Perioperative Clinical Research Committee as potentially significant and likely warranting concern or intervention. The other secondary outcome, hypercarbia, was defined as an end-tidal carbon dioxide (ETco₂) of 60 mmHg or more for 5 min or more or a Paco₂ of 60 mmHg or more on arterial blood gas during lung isolation. These outcome definitions mirror those present in a previous research article published by our group looking at risk factors for hypoxemia and hypercarbia in very young children.¹ 

Covariate selection was made based on previous work by our group and others in both young and older patients undergoing one-lung ventilation and thoracic surgery.^1,14,15  Records were initially reviewed electronically for the following covariates: age, sex, extremes of weight (lower than 5th percentile or higher than 95th percentile per the U.S. Centers for Disease Control and Prevention growth chart), approach used for one-lung ventilation, procedure type, thoracotomy versus video-assisted thoracoscopic procedure, preoperative Spo₂ less than 98%, ASA Status III or IV, one-lung ventilation time in hours, left- versus right-sided surgical case, and finally low tidal volume ventilation with positive end-expiratory pressure (PEEP) defined as tidal volume of 6 ml/kg or less with PEEP of 4 cm H₂O or greater for more than 80% of the period of lung isolation. Types of procedure in the 4- to 9-yr-old cohort were binned into four types: resection of intrapulmonary lesion, mediastinal surgery, decortication and pleurodesis, and other. Overall, procedures in the 10- to 17-yr-old group were similar to those in the 4- to 9-yr-old cohort, except an additional group, surgery for pneumothorax, was added because of the high prevalence of spontaneous pneumothoraces in these patients and the diversity of surgical treatment options (i.e., blebectomy, chemical or mechanical pleurodesis, and wedge resection).¹⁴  Additionally, an anonymous center identification code was evaluated in a separate sensitivity analysis as a potential covariate to determine whether there might be a center effect. While we were able to extract covariates by various electronic means in many cases, it was necessary to manually review a large number of patient records to attempt to fill in missing covariate data that could not be extracted via electronic review. Cases in which one or more covariates were missing and could not be determined after manual review were excluded from the analysis.

The time of one-lung ventilation was defined as the start of one-lung ventilation to the return to two-lung ventilation. In some records, the beginning or end of one-lung ventilation or both were omitted. In records where both times were omitted, there was typically not enough documentation to support the thesis that one-lung ventilation was used; in such cases, unless it was documented elsewhere in the airway note, the record was excluded. Records in which either the start or stop time were missing were reviewed to see whether a probable beginning or end to one-lung ventilation could be determined looking at the available ventilator data.¹  In most records, the investigator could infer either the beginning or end using changes in either the fractional inspired oxygen tension (Fio₂), ETco₂, or tidal volume. For example, in records in which one-lung ventilation has started, the Fio₂ frequently will be increased to 90% or greater, followed by a fall in tidal volume, and an increase in ETco₂. If an end time was present, but a start time was not evident using a similar approach, incision was used as the start time. In records in which the end time was not documented or could not be determined using the above approach, one-lung ventilation end was considered to be anesthesia end minus 20 min. The number of records in which either start or stop time for one-lung ventilation had to be manually derived was 80 of 228 (35%) in the 4- to 9-yr age group and 184 of 1,012 (18.2%) in the 10- to 17-yr age group. Records in which there were multiple episodes of one-lung ventilation usually required manual review. If the length of time between episodes of one-lung ventilation was 5 min or less, these epochs were merged into a single epoch. In records in which the time between epochs was greater than 5 min, only the first period of one-lung ventilation was used to evaluate for the presence of the primary and secondary outcomes.

All records were screened for the following data: presence of 1-min Spo₂ values and ETco₂ values at a minimum of every 15 min. Records missing 5% or greater of oxygen saturation were excluded. In records missing less than 5% of Spo₂ data, the missing Spo₂ values were imputed using the last value carried forward approach, acknowledging that this approach may overestimate the prevalence of hypoxemia. Following the above exclusions, only three records required imputation of the Spo₂ data. ETco₂ values were also imputed using a last value carried forward approach. A total of 8 records in the 4- to 9-yr-old cohort and 11 records in the 10- to 17-yr-old cohort required imputation of the ETco₂ data. As PEEP values do not vary significantly from moment to moment, a similar imputation approach was used in records in which PEEP was documented at 5-min intervals.

In several records, the child’s weight was missing. This was imputed using the 50th percentile for age according to the U.S. Centers for Disease Control and Prevention (Atlanta, Georgia) growth chart. This was necessary in 4 of 228 (1.8%) patients in the younger cohort and 45 of 1,012 (4.4%) patients in the older cohort.

When appropriate, continuous data were examined for extreme values. Records with these values that fall below the 5th percentile or are above 95th percentile were manually reviewed and potentially excluded. Extreme values of Spo₂, ETco₂, and ventilatory data were reviewed to determine their veracity and clinical plausibility. In most cases, the incidence of these values was very low if present at all secondary to preliminary data cleaning algorithms within the Multicenter Perioperative Outcomes Group. Tidal volumes greater than 14 ml/kg were planned to be set to 14 ml/kg. Oxygen saturations measurements greater than 100% were corrected to 100%. ETco₂ values greater than 150 mmHg or less than 15 mmHg during one-lung ventilation were excluded as improbable or not physiologically meaningful. Other data points from the same time point, however, were included if they appeared to be clinically plausible. ETco₂% was converted to mmHg.

Validity of data from the Multicenter Perioperative Outcomes Group limited data set is well recognized.^16,17  Additionally, a random sampling of 10% of the data set from each age group was manually reviewed to validate covariate and outcome data.

Given a lack of standardized practice guidelines and other published standards for one-lung ventilation in either of these age groups, both data sets were evaluated for current practice patterns with regards to approaches to ventilation. The following were evaluated: Fio₂, ETco₂, PEEP, peak inspiratory pressure, and tidal volume.

All possible records meeting inclusion criteria were considered. An initial inquiry performed during project development revealed that there might be as many as 300 records in the 4- to 9-yr age group and as many as 1,000 in the 10- to 17-yr age group. Assuming a 10% event rate for the primary outcome, a sample size of 300 would provide 80% power at the 0.05 significance level to detect a minimal odds ratio of 2.01 in the 4- to 9-yr age group and 1.01 in the 10- to 17-yr age group using a chi-square analysis.

A data analysis and statistical plan was approved by the Multicenter Perioperative Outcomes Group Perioperative Clinical Research Committee before accessing the data. The frequency as well as 95% CI of hypoxemia, severe hypoxemia, and hypercarbia were tabulated for the two age groups (4 to 9 yr of age and 10 to 17 yr of age). All categorical covariates excluding center identification were initially assessed by chi-square analysis to determine their association with the outcome of hypoxemia at a univariate level. In any case where the cell count was less than 5, a Fisher’s exact was used. Normality for continuous variables was initially evaluated using the Shapiro–Wilk test first and, if found to be normally distributed, were compared at the univariate level using unpaired two tailed, t tests. If they were not found to be normally distributed, a Mann–Whitney U test was used. Significance for the univariate analysis was established at P < 0.05. The Wald method was used to compute the 95% CI for proportions throughout.¹⁸ 

The number of events for the primary analysis in both groups using 10 events/covariate to estimate model performance showed that there was an insufficient number of hypoxemic events to use a standard multivariable regression model. Therefore, as dictated by the statistical plan, all covariates, excluding institution, were assessed using a least absolute shrinkage and selection operator regression model with hypoxemia defined as above as the primary outcome.¹⁹  Collinearity was assessed using a variance inflation factor for each covariate. If the variance inflation factor exceeded 10, then one of the covariates was removed. Model overfitting was addressed using the least absolute shrinkage and selection operator approach for regularization. The λ values used for analysis ranged from 0.00001 to 0.05. The model was internally validated using 10-fold cross-validation to ensure that the model performed well across randomly selected validation samples from the data set. Additionally, we performed an analysis of model calibration for the primary outcome in both age groups. Covariates were also evaluated in a separate least absolute shrinkage and selection operator regression model to assess their association with the secondary outcomes of severe hypoxemia and hypercarbia. An additional post hoc least regression model for hypoxemia, severe hypoxemia, and hypercarbia including all covariates was performed for both age groups to assess for potential differences with the model created from covariates remaining after variable selection using the least absolute shrinkage and selection operator approach because approach to lung isolation (bronchial blocker vs. endobronchial intubation) was removed from the multivariable model in both age groups.

Additionally, we performed a post hoc analysis of the risk of hypoxemia and age across the entire cohort to determine whether the risk of hypoxemia decreased with increasing age. A simple linear regression was performed using the proportion of patients in whom hypoxemia occurred grouped by 2-yr age cohorts inclusive increasing from 4 and 5, to 6 and 7, and ultimately up to 16 and 17. We also performed a sensitivity analysis to determine whether there might be an institutional effect by repeating a least absolute shrinkage and selection operator regression analysis on the primary outcome with the addition of the institutional ID as a covariate to see whether there was any impact on the results. At the request of the peer reviewers, we performed a post hoc sensitivity analysis excluding cases in which one-lung ventilation start or stop times were manually derived to determine whether this might affect the primary or secondary analyses. Finally, we also performed a peer reviewer–requested ordinary multivariable regression analysis of all covariates for all three outcomes including hypoxemia, severe hypoxemia, and hypercarbia without using the least absolute shrinkage and selection operator approach to multivariable regression model development. Statistical analyses were performed in R version 4.0.2 (R Foundation for Statistical Computing, Austria) using RStudio environment version 1.3.1056 (RStudio, USA).

A query of the Multicenter Perioperative Outcomes Group database yielded 320 records in the 4- to 9-yr age group and 1,480 records in the 10- to 17-yr age group. These records hailed from 42 different institutions. After an initial review of the associated case postings for all records to determine eligibility for inclusion, 56 records were excluded from the 4- to 9-yr age group, and 277 records were excluded from the 10- to 17-yr age group. After exclusions for missing covariates and physiologic data, the final cohorts included 228 records in the 4- to 9-yr-old cohort and 1,012 records in the 10- to 17-yr old cohort. The number of missing covariates for each cohort, as well as exclusions and the flow of patients, are summarized in figure 1.

The prevalence of hypoxemia in the 4- to 9-yr-old cohort was 24 of 228 (10.5%; 95% CI, 6.5 to 14.5%). The prevalence of severe hypoxemia and hypercarbia were 14 of 228 (6.1%; 95% CI, 3.0 to 9.3%) and 23 of 228 (10.1%; 95% CI, 6.2 to 14.5%), respectively. In the 10- to 17-yr-old cohort, the prevalence of hypoxemia, severe hypoxemia, and hypercarbia were 76 of 1,012 (7.5%; 95% CI, 5.9 to 9.1%), 47 of 1,012 (4.6%; 95% CI, 3.3 to 5.8%), and 40 of 1,012 (4.0%; 95% CI, 2.3 to 5.2%), respectively. Overall, the prevalence of hypoxemia appeared to decrease with increasing age when looking at 2-yr increments across both cohorts (fig. 2).

In the 4- to 9-yr-old cohort, the use of endobronchial intubation, and an initial room air Spo₂ less than 98% were associated with an increased risk of hypoxemia at the univariate level. Conversely, age, sex, extremes of weight, type of surgery, ASA III or IV status, right-sided surgery, use of video-assisted thoracoscopic surgery, use of lower tidal volume ventilation during lung isolation, and duration of one-lung ventilation were not associated with an increased risk of hypoxemia (table 1).

In the older and larger 10- to 17-yr-old cohort, younger age, extremes of weight, ASA Physical Status III or IV, right-sided cases, and an initial room air Spo₂ less than 98% were associated with an increased risk of hypoxemia. Interestingly, female sex was also found to be significant. This may be due to selection bias arising from a higher rate of spontaneous pneumothorax in males and the observation that procedures for this diagnosis have a lower risk of hypoxemia.^20,21  No other covariates were found to be statistically significant (table 1).

In the multivariable analysis of the 4- to 9-yr-old cohort, only having an initial room air Spo₂ less than 98% was associated with an increased risk of hypoxemia and severe hypoxemia. No other covariates were found to be significant. Finally, none of the covariates appear to be associated with an increased risk of hypercarbia (table 2).

The multivariable analysis in the 10- to 17-yr-old cohort revealed that younger age, extremes of weight, right-sided cases, and an initial Spo₂ less than 98% were associated with increased risk of hypoxemia. Looking more closely at those with extremes of weight, obesity was more common among those who developed hypoxemia compared to those that had weights less than the 5th percentile. In fact, 17 of 26 (65%) of those who were at the extremes of weight who developed hypoxia had weights higher than the 95th percentile. Approach to one-lung ventilation, double lumen endobronchial tube or bronchial blocker, was not associated with an increased risk of hypoxemia and was in fact removed from the model. Extremes of weight, right-sided cases, and an initial oxygen saturation less than 98% were also associated with an increased risk of severe hypoxemia. Only younger age was associated with an increased risk of hypercarbia (table 3). A post hoc sensitivity analysis using all covariates in an ordinary least regression model for both age groups did not reveal any significant differences from the model created using the least absolute shrinkage and selection operator (LASSO) approach to variable reduction (Supplemental Table 1, https://links.lww.com/ALN/D315; Supplemental Table 2, https://links.lww.com/ALN/D316).

Sensitivity analysis including facility ID did not reveal a significant site effect in either cohort (Supplemental Table 3, https://links.lww.com/ALN/D317; Table 4, https://links.lww.com/ALN/D318). Additionally, an analysis of model calibration across all institutions for hypoxemia demonstrated reasonable predictive accuracy (Supplemental Figure 1, https://links.lww.com/ALN/D313; Supplemental Figure 2, https://links.lww.com/ALN/D314). A further sensitivity analysis excluding cases in which the one-lung ventilation start or stop time was manually derived did not change the primary and secondary analyses significantly in either cohort (Supplemental Table 5, https://links.lww.com/ALN/D319; Supplemental Table 6, https://links.lww.com/ALN/D320).

A summary of ventilatory parameters including tidal volume, ventilatory pressures, PEEP, and Fio₂ during one-lung ventilation for each cohort are summarized in table 4. The median Fio₂ used in the older cohort appeared to be fairly comparable to that used in younger cohort 75% versus 78%. A histogram of approaches to one-lung ventilation by age is presented in figure 3.

The primary finding of this observational retrospective cohort study was that the risk of hypoxemia in both cohorts exceeds that seen in adult patients undergoing one-lung ventilation and thoracic surgery.²  Additionally, the risk of hypoxemia in school-age children undergoing one-lung ventilation appears to decrease with increasing age. This risk likely approaches that of an adult after early adolescence. The approach to one-lung ventilation did not appear to increase the risk of hypoxemia in either cohort. This is important in that there does not seem to be an approach to one-lung ventilation in either of these cohorts that confers an increased risk of hypoxemia. In the older cohort, right-sided surgical cases appear to be associated with an increased risk of hypoxemia and severe hypoxemia. Extremes of weight and, perhaps unsurprisingly, a starting room air Spo₂ less than 98% were also associated with an increased risk of hypoxemia. Additionally, the risk of hypercarbia in both age cohorts appears to be significantly reduced when compared to the risk of hypercarbia in younger patients and also likely continues to decrease with increasing age and size.¹  While somewhat anticipated, risk factors for hypoxemia in these patients primarily represent underlying surgical or patient factors and therefore may offer little in terms of material intervention at first glance. Awareness of these factors, however, may allow the clinician to potentially implement preventive strategies, for example use of an increased Fio₂, or at a minimum be better prepared to manage hypoxemic events in higher-risk scenarios.

Importantly, the prevalence of hypoxemia during one-lung ventilation in both cohorts appears to be significantly reduced when compared to the prevalence of hypoxemia in very young children (10.5% and 7.5% vs. 26%).¹  This is consistent with the study by Byon et al.,¹³  which found that younger school-age patients were at increased risk of hypoxemia compared to older children. This smaller risk in older children is likely multifactorial in origin as older children have reduced rates of oxygen consumption, a more rigid chest wall to maintain resting lung volumes, an increased hydrostatic gradient as a result of their increased size that augments perfusion of the ventilated lung.⁷  Interestingly, however, the risk of hypoxemia in both younger and older cohorts still exceeds that commonly seen in adults (10.5% and 7.5% vs. 4%) despite being generally “younger and healthier.”²  While there is not a readily available explanation for this, it may be that in some cases adults with underlying lung disease have become significantly less dependent on the diseased segments of the lung that necessarily are undergoing surgery, and therefore, they are less subject to physiologic perturbation when these diseased segments are excluded from gas exchange during one-lung ventilation.

In contrast to findings in a very young cohort (2 months to 3 yr of age), use of a bronchial blocker does not appear to be associated with a reduced risk of hypoxemia in the 4- to 9-yr-old cohort.¹  In this previous multicenter retrospective study, the risk of hypoxemia was substantially higher in patients in whom mainstem intubation was used for lung isolation. Interestingly, the frequency of bronchial blocker use for one-lung ventilation in the 4- to 9-yr age cohort appears to greatly exceed the use of endobronchial intubation, with a prevalence of 72% compared to just 20%, respectively. We hypothesize that this is the result of increased technical ease in placing a bronchial blocker in 4- to 9-yr-old children compared to placing a blocker in a young infant. This stands in significant contrast to current trends in very young children, in which endobronchial intubation was preferred 2:1 over the use of a bronchial blocker.¹  Clinicians appear to be significantly more comfortable using a bronchial blocker approach to one-lung ventilation in this intermediate age cohort. In the older cohort of patients, utilization of a double lumen endobronchial tube for lung isolation far exceeded the use of a bronchial blocker, which parallels current trends seen in the adult population (87% vs. 11%).²² 

Mirroring experience in both very young children and adults, right-sided surgical procedures were associated with an increased risk of hypoxemia during one-lung ventilation in the 10- to 17-yr-old cohort.²  This is likely the result of the exclusion of more lung tissue from gas exchange in right-sided cases versus left-sided cases because there is innately more lung tissue on the right when compared to the left. Intuitively, it is reasonable to expect a similar finding in the 4- to 9-yr-old cohort, but the smaller number of records available for analysis in this cohort may have ultimately obscured this finding.

Overall, there seem to be few material differences in practice patterns related to the ventilatory management in both cohorts. Interestingly, in the older cohort of patient, the Fio₂ does appear to be higher during one-lung ventilation in patients in whom a bronchial blocker was used compared to a double lumen endobronchial tube or endobronchial intubation. Median tidal volumes are only slightly lower in the older cohort. Peak inspiratory pressures and median ETco₂ levels during one-lung ventilation were not different between the groups.

The limitations of this study importantly include the retrospective nature of the inquiry and the small sample size in the younger cohort. While it is possible in some cases to observe trends in this younger cohort, the relatively small number of hypoxemic events necessarily reduced the available statistical power. The Fio₂ before a hypoxemic event also could have been confounding, as a lower Fio₂ in proximity to a desaturation event may have been associated with increased rates of desaturation. In the majority of cases, however, as reflected in table 4, the Fio₂ during one-lung ventilation tended to be fairly high, likely to prevent hypoxemia and encourage atelectasis in the operative lung. Further, it seems unlikely that clinicians would have intentionally maintained a lower Fio₂ in the face of ongoing hypoxemia once it occurred and given that the desaturation event had to last a minimum of 3 min continuously to be counted as an event, it seems less likely that this event would simply have been a result of a reduced initial Fio₂. We did not specifically control for cases in which the same patient may have undergone a repeat procedure on the same side or the contralateral side in close temporal proximity, although patients getting bilateral procedures on the same day were excluded. Additionally, children from different age groups may be referred to specialty centers depending on their age. In such cases, institutional preferences may have introduced a selection bias. However, a sensitivity analysis looking at the impact of a given site in combination with the other covariates failed to detect a “site effect” on the primary outcome. It is also possible that the surgical context may have affected the primary outcome with certain events requiring a surgical intervention in preference to a recruitment breath or other more definitive airway or ventilatory interventions leading to a hypoxemic event. Unfortunately, it was not possible to correlate this with the available data set. Finally, the statistical findings of this study imply only association and should not be interpreted as causation.

In summary, the risk of hypoxemia in school-age children during one-lung ventilation and surgery appears to significantly exceed that seen in adult patients and generally appears higher in younger children. Children 10 to 17 yr of age undergoing right-sided procedures or with extremes of weight or a lower starting Spo₂ (less than 98%) appear to be at higher risk of hypoxemia. There does not appear to be a difference in the risk of hypoxemia when using a bronchial blocker or double lumen endobronchial tube for one-lung ventilation in the 10- to 17-yr-old cohort. An evaluation of a larger cohort of children 4 to 9 yr of age may provide additional insights into risk factors for hypoxemia during thoracic surgery requiring one-lung ventilation.

The authors thank the many Multicenter Perioperative Outcomes Group investigators for their many recommendations and insights during the development and completion of this project.

Supported by departmental and institutional resources at each contributing site. In addition, partial funding to support underlying electronic health record data collection into the Multicenter Perioperative Outcomes Group registry was provided by Blue Cross Blue Shield of Michigan/Blue Care Network (Detroit, Michigan) as part of the Blue Cross Blue Shield of Michigan/Blue Care Network Value Partnerships program. Although Blue Cross Blue Shield of Michigan/Blue Care Network and Multicenter Perioperative Outcomes Group work collaboratively, the opinions, beliefs and viewpoints expressed by the authors do not necessarily reflect the opinions, beliefs, and viewpoints of Blue Cross Blue Shield of Michigan/Blue Care Network or any of its employees.

Dr. Mathis has received U.S. National Institutes of Health (Bethesda, Maryland) NIDDK grant No. R01DK133226. The other authors declare no competing interests.

The data sets involved in this study are defined as limited data sets per U.S. federal regulations and require execution of a data use agreement for transfer or use of the data. They are derived from data shared within the Multicenter Perioperative Outcomes Group. The investigative team is able to share data securely and transparently conditional on the following: (1) the team receives of a detailed written request identifying the requestor, purpose, and proposed use of the shared data; (2) the requestor uses of a secure enclave for the sharing of personally identifiable information; and (3) the request is permissible within the confines of existing data use agreements executed between Multicenter Perioperative Outcomes Group members.

Supplemental Figure 1. Composite Calibration Curve LASSO Model: 4 to 9 yr, https://links.lww.com/ALN/D313

Supplemental Figure 2. Composite Calibration Curve LASSO Model: 10 to 17 yr, https://links.lww.com/ALN/D314

Supplemental Table 1. Ordinary Regression Model all Covariates 4 to 9 yr, https://links.lww.com/ALN/D315

Supplemental Table 2. Ordinary Regression Model all Covariates 10 to 17 yr, https://links.lww.com/ALN/D316

Supplemental Table 3. LASSO Including Site ID 4 to 9 yr, https://links.lww.com/ALN/D317

Supplemental Table 4. LASSO Including Site ID 10 to 17 yr, https://links.lww.com/ALN/D318

Supplemental Table 5. LASSO Excluding Patients with Derived OLV time 4 to 9 yr, https://links.lww.com/ALN/D319

Supplemental Table 6. LASSO Excluding Patients with Derived OLV time 10 to 17 yr, https://links.lww.com/ALN/D320

The following collaborators attest to their substantial role in research protocol revisions or data collection/validation efforts as part of the Multicenter Perioperative Outcomes Group Perioperative Clinical Research Committee: Robert B. Schonberger, M.D., M.H.S., Yale University, New Haven, Connecticut.