Introduction

Near infrared spectroscopy (NIRS) is routinely used in the noninvasive monitoring of cerebral and somatic regional oxygen saturation (rSO2) in pediatric patients following surgery for congenital heart disease [1,2,3]. Monitoring of cerebral and somatic rSO2 can provide clinicians with insights into regional oxygen delivery and consumption following cardiac surgery [3,4,5,6,7,8]. Multiple studies have demonstrated associations between rSO2 values obtained following cardiac surgery in pediatric patients and postoperative outcomes such as mortality, ventilator free days, and low cardiac output syndrome [9,10,11,12].

For many practitioners, cerebral and somatic rSO2 are viewed as routinely monitored vital signs not unlike mean arterial pressure or arterial oxygen saturation in postoperative cardiac surgical patients. While clear target ranges and treatment approaches to abnormal values are established for several monitored values (e.g., mean arterial pressure) following cardiac surgery, this practice has not fully been extended to rSO2 monitoring [13]. We sought to evaluate the association of a bedside rSO2 thought algorithm with clinical outcomes in a cohort of pediatric patients following cardiac surgery.

Materials and Methods

We conducted a retrospective cohort study of patients admitted following cardiac surgery to the pediatric intensive care unit (PICU) between October 1, 2017, and March 31, 2021, at the University of Virginia Children’s Hospital, an academic, tertiary-care center. The Institutional Review Board for Human Subjects Research at the University of Virginia School of Medicine (Protocol #19035) approved this study. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines [14].

We included patients less than or equal to 18 years of age that underwent cardiac surgery with cardiopulmonary bypass (CPB). Additionally, we included patients that underwent cardiac surgery without CPB for the following procedures: pulmonary artery banding, systemic to pulmonary shunting (e.g., Blalock-Taussig-Thomas shunt), and coarctation of the aorta repair.

Multiple surgeries in an individual patient over the course of the study were considered independent surgical cases if they occurred during separate hospitalizations. In the operating room, cerebral and somatic oximetry sensors were placed on the forehead and right flank below the costovertebral angle overlying the right kidney, respectively. Values of cerebral and somatic rSO2 were continuously monitored in the PICU using the INVOS™ oximeter (Medtronic, Minneapolis, MN). Duration of post-operative rSO2 monitoring was at the discretion of the clinical team.

A multidisciplinary group of PICU providers developed the thought algorithm (Supplementary Fig. 1) through consensus agreement. The primary goal of the implementation was to supply bedside providers with a thought aide to help identify, and guide response to, changes in rSO2 in post-operative cardiac surgical patients. The primary bedside nurse was instructed to complete the thought algorithm upon admission to the PICU following cardiac surgery and after initiation stabilization and resuscitation of the patient within the first hour following admission from the Operating Room. Completion of the thought algorithm included establishing baseline cerebral and somatic rSO2 values based on the individual patient’s physiology as well as thresholds that might suggest significant changes in oxygen delivery or oxygen consumption. Orientation to the thought algorithm as well as background education on rSO2 monitoring was provided to PICU staff (nurses, house staff, respiratory therapists, advanced practice providers, and attending physicians) through a variety of mechanisms (e.g., online education modules, in-person lectures) over a two-month period prior to implementation in June 2019. Our primary objective was to assess differences in clinical outcomes over two 18-month periods before and after thought algorithm implementation. We designated the pre-intervention period as October 2017 to March 2019, and the post-intervention period as October 2019 to March 2021, allowing for a six-month washout period from April 2019 to September 2019. During the period of study, there were no changes in the cardiac surgical or pediatric critical care faculty or in the provision of post-operative care of patients following cardiac surgery.

We collected patient and clinical characteristics including age at time of surgery, CPB and cross clamp times, cardiac diagnosis, surgical procedure, delayed chest closure, use of extracorporeal life support, and postoperative ventilator free days (VFD) and PICU length of stay. Postoperative VFDs was defined as the number of days of invasive mechanical ventilation following surgery to postoperative Day 28, with patients who died before Day 28 assigned zero. For the first 48 h following surgery, peak lactate was collected, and peak vasoactive inotropic score was calculated [15]. The Society of Thoracic Surgeons—European Association for Cardiothoracic Surgery Heart Surgery (STAT) mortality categories were assigned to each surgical procedure [16]. Standardized mortality ratios were calculated for the pre-intervention and post-intervention periods using the Society of Thoracic Surgeon Congenital Heart Surgery Database Mortality Risk Model [17].

We performed a sensitivity analysis on the subset of neonates (< 30 days of age) that underwent cardiac surgery with cardiopulmonary bypass to compare several NIRS-derived measures between time periods. Values of cerebral and somatic rSO2 (crSO2 and srSO2, respectively) were continuously captured using the INVOS™ oximeter over the first 48 h following surgery and averaged over 1-min intervals. In addition to average crSO2 and srSO2, we calculated crSO2 and srSO2 desaturation indices defined as the duration of time over the first 48 h following surgery that crSO2 and srSO2 were below 50% [12]. Variability of crSO2 and srSO2 were calculated using the root mean of successive squared differences [18].

Distribution of continuous variables was assessed using the Wilk-Shapiro test for normality. Continuous variables were compared using Student’s T test, Wilcoxon rank sum testing, or linear regression as appropriate. Categorical variables were compared using chi-square test or Fisher’s exact testing as appropriate. Multivariable regression analysis was performed to adjust for confounders. Type I error was set at 0.05. All calculations were performed using STATA/IC 12.1 (STATA Corporation, College Station, TX).

Results

There were 434 surgical cases included in the study with a median patient age at the time of surgery of 3 months (interquartile range 9 days–18 months). The demographic and clinical characteristics of the included patients are listed in Table 1. The distribution of surgical cases over the three time periods was: pre-intervention: 218 cases (50%); wash-out: 91 cases (21%); and post-intervention: 125 cases (29%).

Table 1 Patient and clinical characteristics
Full size table

Patient and clinical characteristics were compared between the pre-intervention and post-intervention periods (Table 2). Patients in the post-intervention period were younger at the time of surgery than patients in the pre-intervention period (median 1 month vs 3 months, p = 0.03). Patients in the post-intervention period were more likely to have delayed sternal closure (20% vs 11%, p = 0.03). We did not observe differences in peak lactate or VIS in the first 48 h following surgery, VFDs, or post-operative PICU length of stay between the study periods.

Table 2 Comparison of patient and clinical characteristics between study periods
Full size table

The observed mortality rate was identical for both study periods (3.2%). Delayed sternal closure, use of extracorporeal life support and peak lactate were all associated with observed mortality (p = 0.003, p < 0.001, and p = 0.025, respectively). Expected mortality was increased in the post-intervention period (p = 0.012). Controlling for significant confounders, study period, delayed sternal closure and peak lactate were all associated with expected mortality (p = 0.037, p < 0.001, and p = 0.009, respectively). The standardized mortality ratio for the pre-intervention period was 0.61 (95% CI 0.16–1.06) and for the post-intervention period was 0.48 (95% CI 0.01–0.95) (p = 0.01). This difference represents a 27% relative risk reduction in standardized mortality in the post-intervention period.

We included 125 neonates (68 pre-intervention, 57 post-intervention) in the NIRS-based measures sensitivity analysis. We did not observe differences in average crSO2, crSO2 or srSO2 variability, or crSO2 or srSO2 desaturation indices. Adjusting for single ventricle physiology, neonates in the post-intervention period had increased average srSO2 values over the first 48 h following surgery (p = 0.045) as compared to the pre-intervention period. Among neonates with single ventricle physiology, the average srSO2 value increased from 70 to 75% following the intervention.

Discussion

Following the implementation of a bedside rSO2 thought algorithm for pediatric patients following cardiac surgery, we observed a 27% decrease in our standardized mortality rate. We did not observe differences in other post-operative clinical outcomes such as peak VIS or lactate, ventilator free days, or post-operative ICU length of stay. In the cohort of neonates with single ventricle physiology, we did observe an increase in the average srSO2 value following the intervention.

While there are several studies demonstrating the association of rSO2 measures with both invasively measured parameters (e.g., central venous oxygen saturation) and clinical states (e.g., low cardiac output syndrome), there is a paucity of data to support the role rSO2 monitoring plays in impacting clinical outcomes [3,4,5,6, 10]. A recently published systematic review and meta-analysis by Hansen et al. included 25 randomized clinical trials involving 2606 adult and pediatric patients randomized to either cerebral rSO2 monitoring or no monitoring [19]. The Authors noted a high risk of bias in reported trials and concluded that the effects on clinical care with access to cerebral rSO2 monitoring versus clinical care without access to cerebral rSO2 monitoring remains uncertain [19].

From a practical standpoint, there are well-defined assessment and treatment algorithms for abnormalities in routine vital signs such as hypotension or hypoxemia. Inherent in this process is the care team’s ability to understand the physiologic property that is being measured by any given vital sign and possess a shared understanding of what that value represents. With rSO2 the interpretation of values may require a more nuanced approach. This characteristic may help explain why the use of NIRS monitoring is still viewed primarily as a trending tool by many practitioners [13]. Irrespective of how the end-user incorporates rSO2 measures into clinical decision making, NIRS is routinely used in the PICU following cardiac surgery, contributing to a lack of equipoise for the conduct of a randomized controlled trial in this population.

Our approach in this study was modeled on the work of Woods-Hill et al. who designed a novel framework to improve blood culture utilization in the PICU [20]. As one of the participating PICU’s in the Woods-Hill study, our team learned the importance of stakeholder involvement in the creation of the algorithm and practical education of the end-users which we integrated into our study approach [20]. We recognized that several processes would be necessary for successful adoption of the rSO2 thought algorithm, including understanding the physiology monitored by rSO2 and the science behind the thought algorithm and suggested clinical pathways [21]. Anecdotally, following our educational efforts and implementation of the thought algorithm, we observed increased attention to, and reporting of, rSO2 values by bedside providers in patients following cardiac surgery.

Our most interesting finding was the decrease in our standardized mortality ratio for the post-intervention period. In a cohort of healthy children who underwent acute normovolemic hemodilution, Fontana et al. demonstrated that significant drops in rSO2 occurred well before changes in lactate were observed [22]. We hypothesize that the monitoring of rSO2 and informed responses to changes to rSO2 contributed to the earlier detection of imbalance of oxygen homeostasis allowing for intervention and improved outcomes. While we did not observe a statistically significant difference in surgical complexity between periods, we did note a 21% increase in the rates of STAT 4 and STAT 5 cases in the post-intervention period which may contribute to the lack of differences observed in post-operative ventilator free days and post-operative PICU length of stay.

We did observe that patients in the post-intervention period were younger and more likely to have delayed sternal closure as compared to the pre-intervention period. We surmise that these differences were highly influenced by the COVID-19 pandemic, the early stages of which occurred during the post-intervention period and was characterized by routine postponement of non-urgent and elective cardiac surgical procedures. Additionally, the pandemic effect likely drove the increase in expected mortality in the post-intervention period. While we did not see an increase in observed mortality with the increase in expected mortality, caution is warranted when evaluating the impact of our intervention on mortality outcomes.

Our study has several limitations, most notably the single-center design and the semi-passive nature of our intervention. While end-users were provided with education on both the principles of rSO2 monitoring and the thought algorithm, we did not mandate or protocolize clinical responses to changes in rSO2 values nor did we attempt to assess knowledge retention over time. It should be acknowledged that our intervention may simply have led to an overall increased level of clinical engagement at the bedside in the post-intervention period, irrespective of rSO2 monitoring. A potential next step would be to randomize patients to a protocolized response versus standard of care. Randomizing to rSO2 monitoring or no monitoring seems impractical in the current practice environment for children following cardiac surgery.

Monitoring of rSO2 in children following cardiac surgery has become routine. Providing frontline clinical staff with education and tools, such as a bedside rSO2 thought algorithm, may aide in the earlier detection of imbalance between oxygen delivery and consumption and may contribute to improved patient outcomes. Future studies aimed at evaluating specific therapeutic response strategies are needed.