Acute kidney injury associated with increased costs in the neonatal intensive care unit: analysis of Pediatric Health Information System database

While an increasing number of studies suggest AKI is frequent and imparts significant morbidity and increased risk for mortality in neonates cared for in the NICU, the economic burden associated with neonatal AKI remains largely unknown. In this singular examination of costs associated with AKI in a broad cohort of surviving neonates cared for in the NICU, our findings support our primary hypotheses. In this cohort, AKI development is associated with significant increases in hospitalization costs after controlling for differences that suggest those with AKI are simply a sicker population (including race, ethnicity, sex, gestational age, Feudtner CCC, and RRT). Neonates who developed AKI had, on average, $58,807 greater costs than those who did not develop AKI. Furthermore, those greater costs are most strongly driven by degree of illness, here quantified by Feudtner CCC status, and gestational age, which is strongly, inversely associated with LOH. However, contrary to our hypothesis, RRT receipt was not a key driver of AKI-associated costs, likely due to the low incidence of RRT receipt in our cohort (<0.15%).

Neonatal AKI is common, affecting an estimated 30% of neonates in the NICU, and is associated with both short- and long-term morbidity as well as mortality [1]. Despite this high prevalence, studies to improve our understanding of the economic impacts of AKI are limited, and few publications are available examining the costs associated with AKI development, particularly in neonates. In our recent exploratory pilot project, our group found AKI development was independently associated with significant increases in hospitalization costs in neonates with PDA in the NICU, a group at particularly high risk for AKI [20,21,22]. Though we found AKI to be associated with higher costs in this high risk NICU subpopulation, a deeper examination of the costs attributable to AKI in a broad NICU cohort was, until now, missing.

In pediatric and adult patients, AKI is independently associated with prolonged hospitalizations and excessive hospitalization costs [3,4,5]. In the NICU, however, hospitalizations costs are markedly variable based on gestational age [23,24,25]. With improving technology, increasing numbers of extremely premature and peri-viable neonates are surviving to hospital discharge, but the costs for care of these fragile neonates are high [26]. In all 18 studies included in a review published in 2014 assessing the cost of prematurity according to gestational age at birth, costs were inversely related to gestational age, and these findings are corroborated in subsequent publications [23, 25, 27, 28]. In the short-term (i.e., during the first year of life), mean costs for extremely premature neonates ranges from $12,910 to $297,627 [23, 27, 29]. However, when focusing solely on NICU costs, extremely preterm neonates account for only a small percentage of the total NICU costs, with total NICU expenditures skewed toward the care of moderate and late preterm infants [24, 25]. With these findings in mind, here we have examined cost in pre-specified gestational age-based subgroups; in both those with and without AKI, adjusted costs decreased with increasing gestational age. Similarly, median LOH decreased with increasing gestational age in both those with and without AKI. Significant differences in the populations studied, may, in part, explain the significantly lower costs attributable to AKI noted here compared to our previous findings in neonates with PDA and AKI which were noted in a much smaller cohort with significantly fewer older patients with shorter lengths of stay.(9). Additionally, modeling strategies employed in this analysis were selected, recognizing that LOH and costs are endogenous, i.e., LOH affects costs but costs also influence LOH. Using LOH as a predictor or covariate when modeling costs can result in spurious correlations and inaccurate estimates of the true relationships between the modeled predictors and the outcomes (here, total costs) [30].

Costs of other morbidities associated with prematurity have been examined [28, 31]. Russell et al. found mean hospital costs for preterm infants with common morbidities of prematurity including respiratory distress syndrome (RDS), bronchopulmonary dysplasia (BPD), intraventricular hemorrhage, and necrotizing enterocolitis (NEC) are four to seven times higher than costs in gestational-age equivalent healthy controls [28]. In this study using data from the 2001 Nationwide Inpatient Sample from Healthcare Cost and Utilization Project, the single costliest comorbidity of prematurity, expressed as average cost per discharge, was BPD with an average cost of $115,000. However, this comorbidity was only noted in 4.4% of analyzed cases. In contrast, the average cost for stays with RDS was less at $56,800, however 23.3% of infants studied were impacted. When examining direct costs for the initial NICU hospitalization and potentially preventable comorbidities of prematurity, Johnson et al. found costs were increased by $12,048 with the presence of brain injury, $15,400 with NEC, $31,565 with BPD, and $10,055 with late-onset sepsis after controlling for birthweight, gestational age, and sociodemographic characteristics [31]. Notably, costs associated with AKI were not examined in either of these important studies. Encouragingly, despite the high costs of care in some circumstances, studies confirm that neonatal intensive care is very cost-effective at $1000 per term infant per quality-adjusted life year (QALY) and $9100 per extremely premature neonatal survivor per QALY [25].

Post-hoc analysis of the 22–26 weeks of gestational age cohort was performed given this subgroup represents 19% of those with AKI. However, these results should be interpreted with caution as this subgroup represents merely 3.8% of the overall cohort. Notably, the scale of the cost differences between the overall cohort ($58,807) and this subgroup ($143,212) is remarkable. In the overall cohort, AKI increased costs 96.9%, however in this subgroup, AKI increased costs only 35.7%. Additionally key drivers of AKI-associated costs differed in this subgroup from the overall cohort suggesting the costs associated with AKI development in this subgroup may be distinct.

Establishing the costs associated with neonatal AKI is a critical first step before true cost-effectiveness studies can be undertaken. Only with these costs quantified can we begin to understand which of several potential high-value practices for prevention of neonatal AKI, such as systemic surveillance programs to monitor nephrotoxin exposure and creatinine for example, are most cost-effective. In the future, this groundwork may also help to identify the most cost-effective therapeutic options for this costly pathology.

A number of strengths and limitations require acknowledgement. Few studies have examined the costs associated with development of specific comorbidities in the NICU, and to our knowledge, none have examined the costs of AKI specifically in a population of neonates cared for in the NICU. As such, this data is the first of its kind, and studies designed to identify high-value practices and interventions for prevention of neonatal AKI and cost mitigation strategies can be performed. We utilized novel strategies to ascertain the costs associated with AKI and employed both multiple imputation analysis for missing gestational age data and examined cost differences by gestational-age based groups given the association between gestational age and LOH. Additionally, we used the Feudtner CCC system to quantify degree of illness in these analyses. This system has been used in prior, noteworthy publications, but we believe this is an important point of novelty for these analyses [32, 33]. Additionally the Feudtner CCC system allowed us to include, into the assessment of degree illness, comorbidities which occurred throughout the index hospitalization rather than datapoints from solely the first 12 hours of life, as is often the case in other scoring systems frequently used for this purpose in neonates. As our group found in our pilot study, though the PHIS database is an exceptional administrative database for cost estimation, clinical information including AKI incidence is limited due to the retrospective nature of the database. Here AKI is determined by the presence or absence of an ICD-10-CM diagnosis code for AKI, and multiple recent studies confirm AKI recognition in the NICU often remains quite low despite increasing study [34, 35]. As such, our AKI incidence (3%) is likely an underestimation of the true AKI incidence, though it is noteworthy that AKI incidence was slightly higher in the 2019–2021 groups than in the 2015–2018 groups. Our study excluded non-survivors which impacts anticipated AKI incidence but also improves the generalizability of our findings. Other limitations include the retrospective nature of this project and the non-inclusion of infant with and without AKI who did not survive. Additionally, though we utilized multiple imputation techniques to account for missingness of the gestational age variable, missingness of other data points was not addressed.

In this analysis of the economic burden of AKI in the NICU, AKI is independently associated with increased hospital costs. The key drivers of costs of hospitalization in neonates with AKI in the NICU included four Feudtner CCCs (cardiovascular, congenital or genetic, gastrointestinal, medical technology) and gestational age. Importantly, associations between gestational age and LOH were noted. The estimated hospitalizations costs decreased with increasing gestational age in those with and those without AKI. These findings add to a slowly growing body of literature informing the costs of care of critically ill neonates in the NICU and, in particular, costs of comorbidities associated with prematurity. Knowledge of these cost and their key drivers can help in identifying high-value practices and interventions that may provide cost mitigation and inform future studies.