Hi. Thank you for the opportunity to speak today. I'm Kate Steffen. I'm a clinical associate professor at Stanford Children's Health, and I have no financial relationships or potential conflicts of interest to disclose. By way of introduction, in hemodynamically stable children without cardiac disease, there are guidelines that recommend against red cell transfusion if hemoglobin is greater than or equal to 7 grams per deciliter. The objective of this study was to assess the clinical and economic impact of guideline compliance among patients who are enrolled in the age of blood in children in the PICU or ABC PICU trial. As some background, the risk of red blood cell transfusions may outweigh benefits in many critically ill children. Stored red cells may not deliver oxygen as effectively as native red blood cells. They may disrupt oxygen delivery by native cells, and they may have negative procoagulant and immunomodulatory effects. Transfusions increase the risk of infectious and noninfectious seriousness hazards of transfusion, and they've been independently associated with longer durations of mechanical ventilation, reduced risk of multi-organ system dysfunction, and death in multiple studies. The TRI-PICU study was completed in 2007 and compared restrictive versus liberal transfusion thresholds in hemodynamically stable children. In the restrictive group, transfusions were reduced by 44 percent, and there was found to be non-inferior in terms of development of newer progressive multi-organ system dysfunction or NPMODs. Despite this and other studies, liberal transfusion persists within the pediatric population, and the Transfusion and Anemia Expertise Initiative, or TAXI, published a set of evidence-based and expert-informed guidelines in 2018. Among those that was best supported by evidence were a threshold-based guideline to defer transfusion in hemoglobin greater than or equal to seven in children who are hemodynamically stable and did not have cardiac disease. This study is a secondary analysis of patients who were included in the ABC-PICU trial. That trial was initially intended to assess the impact of fresher versus standard issue red cells on newer progressive multi-organ system dysfunction. In the ABC-PICU trial, patients were enrolled between three days and 16 years of age. They received their transfusions within one week of their ICU admission and had an expected length of stay of at least 24 hours beyond the time of their initial transfusion. In our particular secondary analysis, we additionally excluded patients who had cardiac admission types, either medical or surgical, who were deemed to be hemodynamically unstable based on the TRI-PICU definition or had their hemodynamic status be unknown, who had missing data or where we were unable to perform propensity matching. In order to determine study group allocation, we looked at guideline compliance. Patients were placed in the compliant group if their hemoglobin was less than seven based on their pre-transfusion hemoglobin CBC prior to their first transfusion. They were placed in the non-compliant group if the hemoglobin was greater than or equal to seven. Patients who received subsequent transfusions didn't impact their group assignment, as subsequent transfusion could represent a possible outcome in a deteriorating patient or could be a predictor of another outcome. Roughly 37% of our patients in the study received multiple transfusions. Our primary outcome was newer progressive MODS or NP-MODS, and we looked at secondary outcomes including the number of transfusions, incidence of sepsis, ARDS, or nosocomial infection. We looked at worst PLED2 score, ICU free days, ventilator free days, mortality, and estimated cost of ICU stay and physician fees. We performed propensity-weighted regression to reduce confounding from pre-transfusion characteristics. This consisted of three steps. First, we performed a logistic regression to develop a model compliance. This estimated each subject's probability or propensity of compliance with the transfusion guideline. Second, we weighted subjects using stabilized inverse probability of treatment weights. This was designed to ensure that the subjects were balanced for pre-transfusion characteristics but differed around compliance with the transfusion guideline. Our goal was to achieve absolute standardized differences of less than 0.1. Finally, we built regression models using the weighted cohort in order to assess the effect of compliance on outcomes. We controlled for two pre-specified variables, both age and pre-transfusion PLED2. This was designed to further reduce confounding, reduce unexplained variability in models, and increase the overall power. And then finally, we had a pre-specified interaction term between red blood cell storage age and compliance with the transfusion guideline. In the end, this particular interaction term showed no effect and it wasn't included in our final models. The reason we did this was, of course, because ABC-PICU was designed to look at red blood cell storage age, but again, not included in the end. From a cost analysis perspective, unfortunately, we weren't able to obtain individual specific costs and cost to charge ratios, and therefore, we used a mean per diem charge for PICU day from a large tertiary care PICU between 2020 and 21, roughly $13,300 per day. We used a mean per diem physician fee from the same source and period, roughly $2,018 per day. The ABC-PICU study enrolled 1,461 patients. In our study, we excluded an additional, I believe it's 771, and we had 338 were allocated to the non-compliant group and 349 allocated to the complying group. When we looked at our study population prior to creating the weighted cohort, we most notably noticed that in the non-compliant group, roughly 49% of our patients were allocated to the non-compliant group, meaning they'd received a transfusion when the hemoglobin was above 7. The patients were similar in many regards. The exceptions to this were the hemoglobin before the first transfusion with a higher hemoglobin in the non-compliant group, which is obviously expected, and then a slightly higher PLAD2 score in the non-compliant group compared to the compliant group, a PLAD2 score of 5 in the non-compliant group and 4 in the compliant group. When we created our weighted cohort, we found that it was quite well-balanced. All absolute standardized differences were less than 0.1. This indicated minimal differences in our baseline characteristics between the two groups. When we performed our propensity-weighted cohort analysis, comparing compliant and non-compliant transfusions, we found no differences between groups, and our primary outcome, new or progressive multi-organ system dysfunction. The remainder of the secondary outcomes were also not significant between the two different groups, but we did find an increased number of ICU-free days and ventilator-free days in the compliant group. Costs of care were also lower in the compliant group with the reduction of ICU length of stay. We had an estimated decreased cost of PICU care by about $33,000, and estimated physician fees were reduced by about $5,800 in our compliant group. This led to a total reduction of approximately $39,000 per patient in the compliant group. This doesn't account for the cost of transfusion nor the impact on ventilator-free days. In discussing the results, first and foremost, at least in pediatric critical care, we do a fairly abysmal job of integrating clinical, sorry, of integrating evidence from clinical studies into our sort of daily practice. We had 49% of the patients who were actually transfused in a way that was non-compliant with the guidelines, despite the fact that this study, the ABC PICU trial, was conducted seven to 11 years after publication of the Tri-PICU study. While the TAX-C recommendations didn't come out until 2018, after the completion of the ABC PICU trial, the main guideline that was examined here was the evidence is readily available in the Tri-PICU study. Secondly, while this analysis is informative, guidelines based on hemoglobin thresholds don't really reflect individual patient physiology nor anemia tolerance. And in an ideal state, we would have evaluated both of these things along with the patient's trajectory in the ICU in order to make an individual-specific and informed decision about transfusion. And then finally, our NP-MODS criteria didn't show any difference between our compliant and non-compliant patients, however, we did notice improvements in ICU and mechanical ventilation-free days in our compliant group. We think this may be the case because NP-MODS criteria may not be precise nor sensitive enough to capture mild or moderate organ failure that may influence ICU length of stay and need for mechanical ventilation support. There are obvious limitations to this study. First, we were unable to identify subpopulations with different transfusion recommendations as these specific diagnoses were not available in ABC PICU data set. Despite the use of propensity scores, there's still a chance of confounding due to unmeasured variables. To decrease risk of this, we ensured that our comparison groups were well-balanced and we controlled for age and PLOD2 score at randomization. And then finally, or sorry, second to the last, I guess. Differences in our PLOD2 scores were noted between our compliant and non-compliant groups. But when we look at this difference in mortality risk, the 0.87 in the compliant group versus the 1.3 in the non-compliant group, we don't think that this difference is really clinically significant and likely doesn't contribute to our eventual findings in the study. And then, of course, additional transfusions, either given in a compliant or non-compliant manner, may have impacted outcomes in our patients, but we were unable to assess this effect. In conclusion, in hemodynamically stable children without cardiac disease, deferring transfusion until hemoglobin is less than 7 is likely safe. Transfusion with a hemoglobin less than 7 was not associated with increased risk of NPMODs or other adverse outcomes. We noted more ICU-free days and more ventilator-free days in the compliant patients, and we found that deferring transfusion until hemoglobin was less than 7 may reduce ICU costs by more than $30,000 per patient. And then most important to me, as a person who's interested in implementation science, suboptimal compliance of guidelines is pretty common and points to the need for structured formal implementation to integrate guidelines into practice. I wanted to thank all my co-authors and mentors who helped with the study, as well as the blood net subgroup of the Pediatric Acute Lung Injury and Sepsis Investigators Network. Thanks. Thank you.

guideline compliance

transfusion

hemoglobin levels

adverse outcomes

ICU costs