Hospital Variation in Intensive Care Resource Utilization and Mortality in Newly Diagnosed Pediatric Leukemia

Healthcare utilization disparities among children with high-risk neuroblastoma treated on Children's Oncology Group clinical trials

INTRODUCTION

Disparities in relapse and survival from high-risk neuroblastoma (HRNBL) persist among children from historically marginalized groups even in highly standardized clinical trial settings. Research in other cancers has identified differential treatment toxicity as one potential underlying mechanism. Whether racial and ethnic disparities in treatment-associated toxicity exist in HRNBL is poorly understood.

METHODS

This is a retrospective study utilizing a previously assembled merged cohort of children with HRNBL on Children's Oncology Group (COG) post-consolidation immunotherapy trials ANBL0032 and ANBL0931 at Pediatric Health Information System (PHIS) centers from 2005 to 2014. Race and ethnicity were categorized to reflect historically marginalized populations as Hispanic, non-Hispanic Black (NHB), non-Hispanic other (NHO), and non-Hispanic White (NHW). Associations between race-ethnicity and intensive care unit (ICU)-level care utilization as a proxy for treatment-associated toxicity were examined with log binomial regression and summarized as risk ratio (RR) and corresponding 95% confidence interval (CI).

RESULTS

The analytic cohort included 370 children. Overall, 88 (23.8%) patients required ICU-level care for a median of 3.0 days (interquartile range [IQR]: 1.0-6.5 days). Hispanic children had nearly three times the risk of ICU-level care (RR 3.1, 95% CI: 2.1-4.5; fully adjusted RR [aRR] 2.5, 95% CI: 1.6-3.7) compared to NHW children and the highest percentage of children requiring cardiovascular-driven ICU-level care.

CONCLUSION

Children of Hispanic ethnicity with HRNBL receiving clinical trial-delivered therapy were more likely to experience ICU-level care compared to NHW children. These data suggest that further investigation of treatment-related toxicity as a modifiable mechanism underlying outcome disparities is warranted.

Characterization of pediatric beta-adrenergic antagonist ingestions reported to the National Poison Data System from 2000 to 2020

BACKGROUND

When ingested by children, small quantities of beta-adrenergic antagonists (BAA) are described as dangerous and even potentially lethal ("one pill can kill"). We characterize demographics, clinical characteristics, and the rate of serious outcomes among pediatric patients with reported BAA ingestions.

METHODS

This study was a retrospective review of U.S. patients <20 years old with reported single-agent BAA ingestions presenting to a health care facility between January 2000 and February 2020 for whom a poison control center was consulted. Data were abstracted from the National Poison Data System (NPDS). Medical outcomes were assessed by the NPDS scale of no effect, minor effect, moderate effect, major effect, and death. All relevant NPDS fatality narratives were reviewed.

RESULTS

A total of 35,436 reported exposures were identified. A total of 29,155 (82.3%) were <6 years old, of which 29,089 (99.8%) were unintentional. Twenty-five patients (<0.1%) <6 years old had major effects. A total of 2316 (8.8%) of patients with no/mild effects were admitted to a critical care unit. Of all cases, 1460 (4.1%) had hypotension and 1403 (4.0%) had bradycardia. One hundred nineteen (0.3%) developed hypoglycemia. The only four fatalities resulted from intentional ingestions in patients >10 years old who sustained cardiac arrest in the prehospital setting.

CONCLUSIONS

Reported BAA ingestions in this multiyear national pediatric cohort caused infrequent toxicity, and no fatalities resulted from an unintentional ingestion. The frequency of bradycardia, hypotension, and hypoglycemia were low. While severely poisoned patients require aggressive treatment, 8.8% of patients were admitted to a critical care unit despite having no or mild effects, which suggests an opportunity to reduce resource utilization.

Leveraging machine learning to identify acute myeloid leukemia patients and their chemotherapy regimens in an administrative database

BACKGROUND

Administrative datasets are useful for identifying rare disease cohorts such as pediatric acute myeloid leukemia (AML). Previously, cohorts were assembled using labor-intensive, manual reviews of patients' longitudinal chemotherapy data.

METHODS

We utilized a two-step machine learning (ML) method to (i) identify pediatric patients with newly diagnosed AML, and (ii) among the identified AML patients, their chemotherapy courses, in an administrative/billing database. Using 2558 patients previously manually reviewed, multiple ML algorithms were derived from 75% of the study sample, and the selected model was tested in the remaining hold-out sample. The selected model was also applied to assemble a new pediatric AML cohort and further assessed in an external validation, using a standalone cohort established by manual chart abstraction.

RESULTS

For patient identification, the selected Support Vector Machine model yielded a sensitivity of 0.97 and a positive predictive value (PPV) of 0.97 in the hold-out test sample. For course-specific chemotherapy regimen and start date identification, the selected Random Forest model yielded overall PPV greater than or equal to 0.88 and sensitivity greater than or equal to 0.86 across all courses in the test sample. When applied to new cohort assembly, ML identified 3016 AML patients with 10,588 treatment courses. In the external validation subset, PPV was greater than or equal to 0.75 and sensitivity was greater than or equal to 0.82 for patient identification, and PPV was greater than or equal to 0.93 and sensitivity was greater than or equal to 0.94 for regimen identifications.

CONCLUSION

A carefully designed ML model can accurately identify pediatric AML patients and their chemotherapy courses from administrative databases. This approach may be generalizable to other diseases and databases.

Multiple Organ Dysfunction and Critically Ill Children With Acute Myeloid Leukemia: Single-Center Retrospective Cohort Study

OBJECTIVES

To describe the prevalence of multiple organ dysfunction syndrome (MODS) and critical care utilization in children and young adults with acute myeloid leukemia (AML) who have not undergone hematopoietic cell transplantation (HCT).

DESIGN

Retrospective cohort study of MODS (defined as dysfunction of two or more organ systems) occurring any day within the first 72 hours of PICU admission.

SETTING

Large, quaternary-care children's hospital.

PATIENTS

Patients 1 month through 26 years old who were treated for AML from 2011-2019.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

Eighty patients with AML were included. These 80 patients had a total of 409 total non-HCT-related hospital and 71 PICU admissions. The majority 53 of 71 of PICU admissions (75%) were associated with MODS within the first 72 hours. MODS was present in 49 of 71 of PICU admissions (69%) on day 1, 29 of 52 (56%) on day 2, and 25 of 32 (78%) on day 3. The organ systems most often involved were hematologic, respiratory, and cardiovascular. There was an increasing proportion of renal failure (8/71 [11%] on day 1 to 8/32 [25%] on day 3; p = 0.02) and respiratory failure (33/71 [47%] to 24/32 [75%]; p = 0.001) as PICU stay progressed. The presence of MODS on day 1 was associated with a longer PICU length of stay (LOS) (β = 5.4 [95% CI, 0.7-10.2]; p = 0.024) and over a six-fold increased risk of an LOS over 2 days (odds ratio, 6.08 [95% CI, 1.59-23.23]; p = 0.008). Respiratory failure on admission was associated with higher risk of increased LOS.

CONCLUSIONS

AML patients frequently require intensive care. In this cohort, MODS occurred in over half of PICU admissions and was associated with longer PICU LOS. Respiratory failure was associated with the development of MODS and progressive MODS, as well as prolonged LOS.

Indirect Standardization as a Case Mix Adjustment Method to Improve Comparison of Children's Hospitals' Antimicrobial Use

Antimicrobial use (AU) in days of therapy per 1000 patient-days (DOT/1000 pd) varies widely among children's hospitals. We evaluated indirect standardization to adjust AU for case mix, a source of variation inadequately addressed by current measurements. Hospitalizations from the Pediatric Health Information System were grouped into 85 clinical strata. Observed to expected (O:E) ratios were calculated by indirect standardization and compared with DOT/1000 pd. Outliers were defined by O:E z-scores. Antibacterial DOT/1000 pd ranged from 345 to 776 (2.2-fold variation; interquartile range [IQR] 552-679), whereas O:E ratios ranged from 0.8 to 1.14 (1.4-fold variation; IQR 0.93-1.05). O:E ratios were moderately correlated with DOT/1000 pd (correlation estimate 0.44; 95% confidence interval, 0.19-0.64; P = .0009). Using indirect standardization to adjust for case mix reduces apparent AU variation and may enhance stewardship efforts by providing adjusted comparisons to inform interventions.

Racial, ethnic, and socioeconomic disparities in paediatric critical care in the USA

In an era of tremendous medical advancements, it is important to characterise and address inequities in the provision of health care and in outcomes. There is a large body of evidence describing such disparities by race or ethnicity and socioeconomic position in critically ill adults; however, this important issue has received less attention in children and adolescents (aged ≤21 years). This Review presents a summary of the available evidence on disparities in outcomes in paediatric critical illness in the USA as a result of racism and socioeconomic privilege. The majority of evidence of racial and socioeconomic disparities in paediatric critical care originates from the USA and is retrospective, with only one prospective intervention-based study. Although there is mixed evidence of disparities by race or ethnicity and socioeconomic position in general paediatric intensive care unit admissions and outcomes in the USA, there are striking trends within some disease processes. Notably, there is evidence of disparities in management and outcomes for out-of-hospital cardiac arrest, asthma, severe trauma, sepsis, and oncology, and in families' perceptions of care. Furthermore, there is clear evidence that critical care research is limited by under-enrolment of participants from minority race or ethnicity groups. We advocate for rigorous research standards and increases in the recruitment and enrolment of a diverse range of participants in paediatric critical care research to better understand the disparities observed, including the effects of racism and poverty. A clearer understanding of when, where, and how such disparities affect patients will better enable the development of effective strategies to inform practice, interventions, and policy.

Variation in Arterial and Central Venous Catheter Use in Pediatric Intensive Care Units

OBJECTIVE

Describe patient and hospital characteristics associated with Arterial Catheter (AC) or Central Venous Catheter (CVC) use among pediatric intensive care units (ICUs).

DESIGN

Hierarchical mixed effects analyses were used to identify patient and hospital characteristics associated with AC or CVC placement. The ICU adjusted median odds ratios (ICU-AMOR) for the admission ICU, marginal R², and conditional intraclass correlation coefficient were reported.

SETTING

166 PICUs in the Virtual PICU Systems (VPS, LLC) Database.

PATIENTS

682,791 patients with unscheduled admissions to the PICU.

INTERVENTION

None.

MEASURES AND MAIN RESULTS

ACs were placed in (median, [interquartile range]) 8.2% [4.9%-11.3%] of admissions, and CVCs were placed in 14.9% [10.4%-19.3%] of admissions across cohort ICUs. Measured patient characteristics explained about 25% of the variability in AC and CVC placement. Higher Pediatric Index of Mortality 2 (PIM2) illness severity scores were associated with increased odds of placement (Odds Ratio (95th% Confidence Interval)) AC: 1.88 (1.87-1.89) and CVC: 1.82 (1.81-1.83) per 1 unit increase in PIM2 score. Primary diagnoses of cardiovascular, gastrointestinal, hematology/oncology, infectious, renal/genitourinary, rheumatology, and transplant were associated with increased odds of AC or CVC placement compared to a primary respiratory diagnosis. Presence of in-house attendings 24/7 was associated with increased odds of AC placement 1.32 (1.11-1.57). Admission ICU explained 4.9% and 3.5% of the variability in AC or CVC placement, respectively. The ICU-AMOR showed a patient would have a median increase in odds of 55% and 43% for AC or CVC placement, respectively, if the same patient moved from an ICU with lower odds of placement to an ICU with higher odds of placement.

CONCLUSIONS

Variation in AC or CVC use exists among PICUs. The admission ICU was more strongly associated with AC than with CVC placement. Further study is needed to understand unexplained variation in AC and CVC use.

Variations in Medicaid Payment Rates for Radiation Oncology

PURPOSE

Interstate variations in Medicaid reimbursements can be significant, and patients who live in states with low Medicaid reimbursements tend to have worse access to care. This analysis describes the extent of variations in Medicaid reimbursements for radiation oncology services across the United States.

METHODS AND MATERIALS

The Current Procedural Terminology codes billed for a course of whole breast radiation were identified for this study. Publicly available fee schedules were queried for all 50 states and Washington, DC, to determine the reimbursement for each service and the total reimbursement for the entire episode of care. The degree of interstate payment variation was quantified by computing the range, mean, standard deviation, and coefficient of variation. The cost of care for the entire episode of treatment was compared to the publicly available Kaiser Family Foundation (KFF) Medicaid-to-Medicare fee index to determine if the pattern of payment variation in medical services generally is predictive of the variation seen in radiation oncology specifically.

RESULTS

Data were available for 48 states and Washington, DC. The total episode reimbursement (excluding image guidance for respiratory tracking) varied from $2945 to $15,218 (mean, $7233; standard deviation, $2248 or 31%). The correlation coefficient of the KFF index to the calculated entire episode of care for each state was 0.55.

CONCLUSIONS

There is considerable variability in coverage and payments rates for radiation oncology services under Medicaid, and these variations track modestly with broader medical fees based on the KFF index. These variations may have implications for access to radiation oncology services that warrant further study.

Successful merging of data from the United Network for Organ Sharing and the Pediatric Health Information System databases

Data routinely collected through United Network for Organ Sharing (UNOS) lack the detailed information on medical resource utilization and treatment costs required to accomplish for center-level comparisons of quality of care and cost for pediatric heart transplantation. We aimed to overcome this limitation by merging UNOS with the Pediatric Health Information System (PHIS) database, an administrative database containing inpatient, emergency department, ambulatory surgery, and observation unit information from over 40 not-for-profit, tertiary care pediatric hospitals. Utilizing a probabilistic match based on center, date of birth, recipient gender, and transplant date within ±2 days, more than 90% of eligible UNOS patients (N = 2264) were successfully merged to their corresponding PHIS records. Thirty-day and 1-year mortality rates observed for the merged cohort (3.2% and 9.0%, respectively) were compared with those previously reported for pediatric heart transplants, as were the significant predictors of increased mortality. These results demonstrate that the established UNOS-PHIS cohort will provide a valid platform for subsequent research aimed at identifying center-level differences that could be exploited to optimize quality of care while minimizing cost across institutions.