86 SECTION II Pediatric Critical Care Tools and Procedures Cardiac Intensive Care Unit Prediction Tools Among the subgroup of infants admitted to the PICU or cardiac ICU following repair or palliation[.]
86 S E C T I O N I I Pediatric Critical Care: Tools and Procedures TABLE Paediatric Index of Mortality (PIM3) Variables 12.2 and Model Coefficients Score calculated based on variables collected from the time of initial patient contact to hour after arrival in the pediatric intensive care unit Variable Pupils fixed to light? (yes/no) Elective admission (yes/no) Coefficient 3.8233 20.5378 Mechanical ventilation in the first hour (yes/no) 0.9763 Absolute value of base excess (mmol/L) 0.0671 SBP at admission (mm Hg) 20.0431 SBP /1000 0.1716 100 Fio2/Pao2 (mm Hg) 0.4214 Recovery postprocedure? Yes, recovery from a bypass cardiac procedure 21.2246 Yes, recovery from a nonbypass cardiac procedure 20.8762 Yes, recovery from a noncardiac procedure 21.5164 Very-high-risk diagnosis (yes/no) 1.6225 Cardiac arrest preceding ICU admission Severe combined immunodeficiency Leukemia or lymphoma after first induction Bone marrow transplant recipient Liver failure is main reason for ICU admission High-risk diagnosis (yes/no) 1.0725 Spontaneous cerebral hemorrhage Cardiomyopathy or myocarditis Additional Algorithms in the Public Domain Hypoplastic left heart syndrome Neurodegenerative disease Necrotizing enterocolitis is main reason for ICU admission Low-risk diagnosis (yes/no) Surgeons and European Association for Cardiothoracic Surgery (STS-EACTS) Congenital Heart Surgery Mortality Score was introduced in 2009 The score was developed on 77,294 procedures entered into the STS and EACTS Congenital Heart Surgery databases and validated on an additional 27,700 procedures.25 Procedure-specific relative risks of in-hospital mortality were estimated for more than 140 congenital heart disease procedures To combine procedure-specific risks with patient-specific factors, the patient’s age, weight, and preoperative hospital length of stay were added to the model The STS-EACTS score has good discrimination for in-hospital mortality (AUC 0.816) and outperformed both the Risk Adjustment for Congenital Heart Surgery (RACHS-1) and the Aristotle Basic Complexity score in the validation sample The Pediatric Cardiac Critical Care Consortium (PC4) has recently leveraged its multi-institutional registry to begin developing risk prediction models, incuding for postoperative mortality, for patients with cardiac disease.26 This case mix model (PC4 Post-Surgical Mortality Model) is based on 8543 patients who had cardiac surgery either during or immediately preceding admission to a participating cardiac ICU in the United States It includes patient-level preprocedural, operative (including STS scoring), and postoperative (within hours) physiologic characteristics and demonstrates good discrimination of mortality (C statistic, 0.92) The inclusion of postoperative characteristics is an attempt to isolate the cardiac ICU performance from the surgical performance Finally, since morbidity and mortality are associated with physiologic instability, the outcomes of postoperative cardiac patients are well predicted with PRISM Simultaneous prediction of morbidity and mortality using PRISM also performs well even without inclusion of surgical complexity scoring.27 22.1766 Asthma is main reason for ICU admission Bronchiolitis is main reason for ICU admission Croup is main reason for ICU admission Obstructive sleep apnea is main reason for ICU admission Diabetic ketoacidosis is main reason for ICU admission It should be clear to any practitioner of critical care that severity of illness and mortality risk are dynamic variables subject to a variety of influences, such as therapies provided, presence of comorbidities, and evolution of the disease process Although not designed to specifically predict mortality risk, there are severity of illness scores obtained over the course of PICU care that correlate with mortality risk The Paediatric Logistic Organ Dysfunction score quantifies degree of organ dysfunction among six different organ systems—neurologic, cardiovascular, respiratory, renal, hematologic, and hepatic.28 The score was developed from 594 patients and subsequently validated on 1806 patients demonstrating good discrimination of mortality (AUC 0.91).29 Increased Multiple Organ Dysfunction Score (MODS) values, as mentioned previously, have been linked to relatively poor outcomes Seizure disorder is main reason for ICU admission Constant 21.7928 ICU, intensive care unit; SBP, systolic blood pressure Next Generation: Morbidity and Mortality Prediction—Trichotomous Outcome Morbidity Assessment Cardiac Intensive Care Unit Prediction Tools Among the subgroup of infants admitted to the PICU or cardiac ICU following repair or palliation of congenital heart disease, efforts have been taken to ascribe risk of mortality based on surgical procedure and patient covariates The Society for Thoracic Although mortality adjusted for physiologic status and other case mix factors has been the core methodology of adult, pediatric, and neonatal intensive care assessments for decades, mortality for most pediatric critical illnesses has decreased since these methods were developed More important, therapies are increasingly focused on reducing morbidity in survivors However, most quantitative CHAPTER 12 Prediction of Short-Term Outcomes During Critical Illness in Children outcome assessment methods continue to focus on the dichotomous outcomes of survival or death A major challenge of pediatrics is the development of welldefined morbidity measures that are rapid, reliable, and objective; measure the child’s status at the time of testing; and are applicable to a broad range of ages in a variety of environments In-depth neuropsychologic testing will likely remain the clinical standard, but other methods applicable to all pediatric ages and sufficiently rapid to be used in large samples are necessary The Glasgow Outcome Scale score was adapted to children in the Pediatric Cerebral Performance Category and Pediatric Overall Performance Category scores (PCPC/POPC), but sufficient interrater reliability was achievable only when neighboring categories were combined.30 Therefore, using these scores in outcome studies risks requiring very large sample sizes to detect significant differences Outcome studies in adult medicine have demonstrated the value of scales such as the Activities of Daily Living Scale Thus the foundation of the Functional Status Scale (FSS) was to adapt the concepts of activities of daily living to pediatrics in a manner that met the criteria described earlier.9 The FSS was developed through a formal consensus process involving a variety of pediatric healthcare specialists It is composed of six domains: respiratory status, feeding, motor functioning, communication, sensory functioning, and mental status Each domain is assessed objectively from normal to very severe dysfunction The score was designed to enable assessment from parents’ or caregivers’ reports or from medical records It was validated on more than 800 patients from seven institutions against an adaptive behavior scale, has very good interrater reliability, and has been used in numerous studies, including a study of more than 10,000 pediatric patients Comparisons with the POPC and PCPC demonstrated FSS increases with each higher POPC and PCPC category However, the dispersion of the FSS scores indicated a lack of precision in the POPC/PCPC system when compared with the more objective and granular FSS system.31 A recent multisite study of more than 5000 pediatric ICU patients demonstrated the importance of the development of new functional status morbidities during intensive care.32 The rate of new functional status morbidities (assessed by an increase of 3 points in the FSS score) was 4.8%, twice as high as the rate of hospital deaths On hospital discharge, the good category decreased from a baseline of 72% to 63%, mild abnormality increased from 10% to 15%, moderate abnormality status increased from 13% to 14%, severe status increased from 4% to 5%, and very severe was unchanged at 1% The highest new morbidity rates were seen in patients with neurologic diagnoses (7.3%), acquired cardiovascular disease (5.9%), cancer (5.3%), and congenital cardiovascular disease (4.9%) New morbidities occurred in all ages, especially in those younger than 12 months of age New morbidities involved all FSS domains, with the highest proportions involving respiratory, motor, and feeding dysfunction Comparing recent data with historical data suggests that pediatric critical care may have exchanged mortality for morbidity over the past several decades Although the rates cannot be precisely compared over time because of the different research methods, data from the 1990s demonstrated a PICU mortality rate of 4.6% and a PICU morbidity rate of 3.1% (based on a 2 POPC change) while current data reflect a hospital mortality rate of 2.4% and morbidity rate of 4.8% Thus, the “morbidity and mortality rate” decreased only from 7.7% to 7.2% as the mortality rated decreased and the new morbidity rate increased Because these rates are not severity or risk adjusted, the changes in 87 admission criteria and other factors that have occurred in the past several decades could have also significantly influenced this comparison New functional status morbidities associated with PICU stays present on hospital discharge are associated with many of the same factors as mortality, including physiologic status measured by the PRISM III score, age, admission source, and diagnostic factors.33 Importantly, these new morbidities, when measured with the FSS score, can be modeled simultaneously with mortality Critical care mortality is usually associated with physiologic multiple organ system abnormalities It appears that new morbidity significantly affecting functional status is often an event along the path toward mortality, as both outcomes are strongly associated with the degree of physiologic alterations Trichotomous modeling uncovered the phasic association of morbidity risk with physiologic status; morbidity risk increases with physiologic instability but then decreases as patients with potential morbidity die Recent studies indicate that trichotomous logistic regression can produce a well-performing model for simultaneous prediction of both morbidity and mortality suitable for risk adjustment in research, quality, and other studies.33 The addition of morbidity to outcome prediction has wide implications for research trials and quality programs, especially those currently based on internal or external benchmarking of standardized mortality ratios Care assessments that focus on morbidity and mortality will have wide appeal and relevance Potentially, evaluations of, and improvements in, the structure and process of care analogous to those resulting from the investigations of the variability of standardized mortality ratios could result from the inclusion of this important new outcome Initiatives that monitor standardized mortality ratios could find relevance in the inclusion of standardized morbidity ratios as well Application of Prediction Tools in Pediatric Intensive Care Prediction tools can be applied at both the population and individual patient levels At the population level, prediction tools are used in benchmarking, which is a process in which the performance of entities (e.g., individuals, PICUs, institutions) is observed and then compared with internal or external standards Clinical scoring systems (e.g., PRISM III) are used to control for severity of illness and other factors, thus allowing for standardized comparisons The two most common standardized comparisons used in pediatric intensive care are the standardized mortality ratio (observed mortality rate divided by the expected mortality rate) and the standardized length-of-stay ratio (observed length of stay divided by the expected length of stay) Risk adjustment models typically present information regarding performance as standardized mortality ratios (SMR) These are calculated by comparing the number of observed outcomes to the number of expected outcomes based on the clinical characteristics of the patients If the boundaries of this estimate are such that the lower bound is greater than 1, that center has experienced the outcome more than would be expected based on its case mix Alternatively, if the upper bound is less than 1, the outcome is less common than would be expected, and if the bounds cross 1, then the observed values are within the range of what would be expected As researchers move toward evaluating morbidity and other outcomes, the SMR concept can be similarly extended Internal benchmarking relates to the comparison of performance within an entity For example, an individual PICU may want to 88 S E C T I O N I I Pediatric Critical Care: Tools and Procedures compare the impact of a new care protocol on length of stay as compared with the current internal standard of care for a particular illness Calculating a standardized ICU length-of-stay ratio would allow for the comparison of practices while accounting for differences in patients’ severity of illness between the two groups External or competitive benchmarking allows for direct comparisons between individual hospitals or PICUs by controlling for differences in case mix At the individual patient level, the uses of prediction tools for short-term outcomes are varied Clinical scoring systems are often employed in clinical trials and outcome analyses to control for patient severity of illness or in measuring changes in physiologic status after a novel therapy has been initiated, but this has not yet achieved clinical relevance for individual patients.10 Future Directions: Predictive Analytics and Tools for Decision Support The use of computerized decision support in adult, pediatric, and neonatal ICUs has grown considerably over the past several years The integration of alerts, reminders, and protocols into computer order entry systems can help in guiding therapy and the reduction of medication errors.34 Children in the ICU have their physiologic parameters, including vital signs and laboratory results, monitored frequently The dynamic and relatively dense nature of these variables in the electronic health record (EHR) allows for opportunities for real-time prediction and intervention in ways that are not possible in care locations with sparser data, such as the general care ward There is, however, still a paucity of effective and validated decision support tools to guide the critical care practitioner in making real-time decisions that affect patient outcomes The use of continuous monitoring data in the ICU to predict clinical deterioration is a recent subject of great interest, as it ideally allows for intervention to occur in a timely fashion and potentially prevent the deterioration Moorman et al.35 developed a predictive algorithm for clinical deterioration in very-low-birth-weight infants employing heart rate characteristics monitoring and demonstrated a 20% relative reduction of mortality in a randomized clinical trial of 3003 patients Efforts to develop predictive monitoring algorithms in older children and adults are underway Predictive monitoring algorithms have been used in adult patients for a variety of conditions, both in and outside of the ICU Perhaps the most prominent area of work for predictive algorithms relates to the detection of sepsis Sepsis is relatively common and presents a large burden to the healthcare system With regulatory and organizational mandates for appropriate sepsis care, the hope is that these models will preemptively identify patients before clinical decompensation Adult critical care researchers have commonly used the Medical Information Mart for Intensive Care (MIMIC) database, a publicly available database sourced from Beth Israel Deaconess Medical Center, or institutional EHR data to predict sepsis hours to days before clinical diagnosis in adults In neonates, Masino et al recently demonstrated the use of a variety of machine-learning models to predict sepsis hours prior to clinical diagnosis with EHR data from a single institution.36 AUCs for the different models ranging from 0.80 to 0.82 when considering only those with positive cultures and 0.85 to 0.87 when children treated for sepsis despite negative cultures were also included The increasing availability of large pediatric critical care datasets potentially provides the foundation for the creation of decision support tools The various pediatric critical care data sources in the United States and worldwide vary greatly as they relate to accessibility/cost, clinical detail, and represented population.37 The expectation of the National Institutes of Health that federally funded investigators make their data widely and freely available to the public will only increase the amount of available data Currently, datasets generated from research conducted by the CPCCRN and the Pediatric Emergency Care Applied Research Network are freely available to researchers The Pediatric Data Science and Analytics group of the Pediatric Acute Lung Injury and Sepsis Investigators maintains a regularly updated list of pediatric critical care data sources (https://www palisi.org/palisi-pedal) Key References Graciano AL, Balko JA, Rahn DS, Ahmad N, Giroir BP The pediatric multiple organ dysfunction score (P-MODS): development and validation of an objective scale to measure the severity of multiple organ dysfunction in critically ill children Crit Care Med 2005;33:14841491 Leteurtre S, Duhamel A, Salleron J, et al PELOD-2: an update of the pediatric logistic organ dysfunction score Crit Care Med 2013: 41(7):1761-1773 Moorman JR, Carlo WA, Kattwinkel J, et al Mortality reduction by heart rate characteristic monitoring in very low birth weight neonates: a randomized trial J Pediatr 2011;159:900-906 Pollack MM, Holubkov R, Funai T, et al Simultaneous prediction of new morbidity, mortality, and survival without new morbidity from pediatric intensive care: a new paradigm for outcomes assessment Crit Care Med 2015;43:1699-1709 Pollack MM, Holubkov R, Funai T, et al The Pediatric risk of mortality score: update 2015 Pediatr Crit Care Med 2016;17(1):2-9 Pollack MM, Holubkov R, Glass P, et al Functional status scale: new pediatric outcome measure Pediatrics 2009;124(1):e18-e28 Reid S, Bajuk B, Lui K, et al Comparing CRIB-II and SNAPPE-II as mortality predictors for very preterm infants J Paediatr Child Health 2015;51(5):524-528 Straney L, Clements A, Parslow RC, et al Paediatric index of mortality 3: an updated model for predicting mortality in pediatric intensive care Pediatr Crit Care Med 2013;14:673-681 Tabbutt S, Schuette J, Zhang W, et al A novel model demonstrates variation in risk-adjusted mortality across pediatric cardiac ICUs after surgery Pediatr Crit Care Med 2019;20(2):136-142 The full reference list for this chapter is available at ExpertConsult.com e1 References Cullen DJ, Civetta JM, Briggs BA, Ferrara LC Therapeutic intervention scoring system: a method for quantitative comparison of patient care Crit Care Med 1974;2(2):57-60 Lefering R, Zart M, Neugebauer EA Retrospective evaluation of the simplified therapeutic intervention scoring system (TISS-28) in a surgical intensive care unit Intensive Care Med 2000;26(12): 1794-1802 Wilkinson JD, Pollack MM, Ruttimann UE, et al Outcome of pediatric patients with multiple organ system failure Crit Care Med 1986;14(4):271-274 Agus MS, Wypij D, Hirshberg EL, et al Tight glycemic control in critically ill children N Engl J Med 2017;376(8):729-741 Tucci M, Lacroix J, Fergusson D, et al The age of blood in pediatric intensive care units (ABC PICU): study protocol for a randomized controlled trial Trials 2018;19(1):404 Graciano AL, Balko JA, Rahn DS, Ahmad N, Giroir BP The Pediatric Multiple Organ Dysfunction Score (P-MODS): development and validation of an objective scale to measure the severity of multiple organ dysfunction in critically ill children Crit Care Med 2005;33:1484-1491 Lin JC, Spinella PC, Fitzgerald JC et al for the Sepsis Prevalence, Outcomes, and Therapy Study Investigators New or progressive multiple organ dysfunction syndrome in pediatric severe sepsis: a sepsis phenotype with higher morbidity and mortality Pediatr Crit Care Med 2017;18(1):8-16 Lamas A, Otheo E, Ros P, et al Prognosis of child recipients of hematopoietic stem cell transplantation requiring intensive care Intensive Care Med 2003;29(1):91-96 Pollack MM, Holubkov R, Glass P, et al Functional status scale: new pediatric outcome measure Pediatrics 2009;124(1):e18-e28 10 Marcin JP, Pollack MM Review of the methodologies and applications of scoring systems in neonatal and pediatric intensive care Pediatr Crit Care Med 2000;1:20-27 11 Parry G, Tucker J, Tarnow-Mordi W CRIB II: an update of the clinical risk index for babies score Lancet 2003;361:1789-1791 12 Richardson DK, Corcoran JD, Escobar GJ, et al SNAP-II and SNAPPE-II: simplified newborn illness severity and mortality risk scores J Pediatr 2001;138:92-100 13 Horbar JD, Soll RF, Edwards WH The Vermont Oxford Network: a community of practice Clin Perinatol 2010;37:29-47 14 Cockburn F, Cooke RWI, Gamsu HR, et al The CRIB (clinical risk index for babies) score: a tool for assessing initial neonatal risk and comparing performance of neonatal intensive care units Lancet 1993;342:193-198 15 Richardson DK, Gray JE, McCormick MC, et al Score for neonatal acute physiology: a physiologic severity index for neonatal intensive care Pediatrics 1993;91:617-623 16 Zupancic JAF, Richardson DK, Horbar JD, et al Revalidation of the score for neonatal acute physiology in the Vermont Oxford Network Pediatrics 2007;119:e156-e163 17 Reid S, Bajuk B, Lui K, et al Comparing CRIB-II and SNAPPE-II as mortality predictors for very preterm infants J Paediatr Child Health 2015;51(5):524-528 18 Pollack MM, Patel KM, Ruttimann UE PRISM III: an updated Pediatric Risk of Mortality score Crit Care Med 1996;24(5): 743-752 19 Straney L, Clements A, Parslow RC, et al Paediatric index of mortality 3: an updated model for predicting mortality in pediatric intensive care Pediatr Crit Care Med 2013;14:673-681 20 Ruttimann UE, Pollack MM Variability in duration of stay in pediatric intensive care units: a multi-institutional study J Pediatr 1996; 128(1):35-44 21 Tibby SM, Taylor D, Festa M, et al A comparison of three scoring system for mortality risk among retrieved intensive care patients Arch Dis Child 2002;87:421-425 22 Pollack MM, Holubkov R, Funai T, et al The pediatric risk of mortality score: update 2015 Pediatr Crit Care Med 2016;17(1):2-9 23 Pollack MM, Dean JM, Butler J, et al The ideal time interval for critical care severity-of-illness assessment Pediatr Crit Care Med 2013;14(5):448-453 24 Czaja AS, Scanlon MC, Kuhn EM, Jeffries HE Performance of the Pediatric Index of Mortality for pediatric cardiac surgery patients Pediatr Crit Care Med 2011;12(2):184-189 25 O’Brien SM, Clarke DR, Jacobs JP, et al An empirically based tool for analyzing mortality associated with congenital heart surgery J Thorac Cardiovasc Surg 2009;138:1139-1153 26 Tabbutt S, Schuette J, Zhang W, et al A novel model demonstrates variation in risk-adjusted mortality across pediatric cardiac ICUs after surgery Pediatr Crit Care Med 2019;20(2):136-142 27 Berger JT, Holubkov R, Reeder R, et al Morbidity and mortality prediction in pediatric heart surgery: physiological profiles and surgical complexity J Thorac Cardiovasc Surg 2017;154(2):620-628.e6 28 Leteurtre S, Duhamel A, Salleron J, et al PELOD-2: an update of the pediatric logistic organ dysfunction score Crit Care Med 2013:41(7):1761-1773 29 Leteurtre S, Martinot A, Duhamel A, et al Validation of the paediatric logistic organ dysfunction (PELOD) score: prospective, observational, multicentre study Lancet 2003;362:192-197 30 Fiser DH Assessing the outcome of pediatric intensive care J Pediatric 1992;121:69-74 31 Pollack MM, Holubkov R, Funai T, et al Relationship between the functional status scale and the pediatric overall performance category and pediatric cerebral performance category scales JAMA Pediatrics 2014;168(7):671-676 32 Pollack MM, Holubkov R, Funai T, et al Pediatric intensive care outcomes: development of new morbidities during pediatric critical care Pediatr Crit Care Med 2014;15(9):821-827 33 Pollack MM, Holubkov R, Funai T, et al Simultaneous prediction of new morbidity, mortality, and survival without new morbidity from pediatric intensive care: a new paradigm for outcomes assessment Crit Care Med 2015;43:1699-1709 34 Williams CN, Bratton SL, Hirshberg EL Computerized decision support in adult and pediatric critical care World J Crit Care Med 2013;2:21-28 35 Moorman JR, Carlo WA, Kattwinkel J, et al Mortality reduction by heart rate characteristic monitoring in very low birth weight neonates: a randomized trial J Pediatr 2011;159:900-906 36 Masino AJ, Harris MC, Forsyth D, et al Machine learning models for early sepsis recognition in the neonatal intensive care unit using readily available electronic health record data PLoS One 2019; 14(2):e0212665 37 Bennett TD, Spaeder MC, Matos RI, et al Existing data analysis in pediatric critical care research Front Pediatr 2014;2:79 e2 Abstract: Outcomes research in pediatric critical care is an important aspect of providing high-quality, error-free medical care Scoring systems and predictive models can add objectivity to these assessments The historical basis for these scoring systems was the therapeutic intensity required by the patient; mortality was the outcome of focus As the data available from critically ill patients increase over time and critical care shifts from solely mortality prevention to include morbidity prevention, additional outcomes and analytic techniques have become relevant Key Words: outcomes, morbidity, mortality, predictive analytics, PRISM score ... factors that have occurred in the past several decades could have also significantly influenced this comparison New functional status morbidities associated with PICU stays present on hospital... investigations of the variability of standardized mortality ratios could result from the inclusion of this important new outcome Initiatives that monitor standardized mortality ratios could find relevance... of expected outcomes based on the clinical characteristics of the patients If the boundaries of this estimate are such that the lower bound is greater than 1, that center has experienced the outcome