Nephrotoxic Exposures and Acute Kidney Injury in Noncritically Ill Children Stratified by Service

This is a single-center, retrospective chart review of children admitted to the pediatric wards at a large quaternary care center in the Rocky Mountain region of the United States from January 2019 to June 2020. We included patients who met criteria for a NINJA trigger as described previously.⁸  Children with end-stage renal disease, history of kidney transplant, or who were admitted to the intensive care units, nephrology and hematology, oncology, or bone marrow transplant (BMT) services were excluded. The hematology or oncology or BMT services were excluded on the basis of existing single center studies in these populations that have described high NTMx epidemiology and demonstrated reductions in both NTMx exposures and NTMx-AKI rates.^12,14  This study was approved by the Organizational Research Risk and Quality Improvement Review Panel and the local institutional review board with a waiver of informed consent.

Primary outcomes of this study include NTMx exposure prevalence and type as well as prevalence of NTMx-AKI categorized by service normalized to 1000 patient days. Secondary outcomes of this study include serum creatinine monitoring compliance, NTMx-AKI duration as well as progression of NTMx-AKI to CKD.

Automated weekly and monthly reports were used to track NTMx exposures and NTMx-AKI episodes. Service was evaluated as the first service of the day. Manual chart review was performed to assess the indication for admission, duration and severity of NTMx-AKI, and hospital length of stay. Therapeutic drug monitoring levels for vancomycin were also evaluated. Data reported as “absence of supratherapeutic vancomycin level” indicate that either no vancomycin levels were collected or levels were collected but were not supratherapeutic.

High NTMx exposure was defined as administration of: (1) ≥3 nephrotoxic medications administered within 48 hours, (2) ≥ 3 days of intravenous vancomycin, or (3) ≥ 3 days of intravenous aminoglycoside as has been previously described.⁸  A NINJA trigger was defined by the initial occurrence of high NTMx exposure, which prompted recommendation of daily SCr monitoring, strict intake and output documentation, and daily weights during the exposure period and continued for 48 hours after discontinuation of the nephrotoxic medication(s). Hospital days is defined as the total number of days for all patients evaluated in a service area during the study period.

Baseline SCr was assessed using the lowest measured value in the preceding 6 months or back-calculated by using the bedside Schwartz equation¹⁵  using normative estimated glomerular filtration rate (eGFR) for age for children <18 years of age (Supplemental Table 4).¹⁶  For patients ≥18 years of age without a baseline SCr, one was calculated using the modification of diet in renal disease equation as has been previously described and validated.^17,18  NTMx-AKI was defined using the Kidney Disease Improving Global Outcomes SCr criteria¹⁹  with stage 1 AKI defined as an increase in SCr by ≥50% or absolute increase in SCr by 0.3 mg/dL exceeding a threshold of 0.5 mg/dL as defined by the Solutions for Patient Safety operating definition (unpublished guidance document), stage 2 defined as an increase in SCr by ≥100% from baseline, and stage 3 defined as an increase in SCr by ≥200% or eGFR ≤ 35 mL/min/1.73 m² (if age  < 18 years) or receipt of dialysis. Severe NTMx-AKI was defined as stage 2 or 3 AKI. AKI duration is defined as the number of days of NTMx-AKI. Duration of AKI was determined when available SCr measurements demonstrated the patient no longer met AKI criteria for 48 hours. If SCr values were not measured after the initial SCr value demonstrating AKI occurrence, the duration of AKI was determined by the patient’s discharge date. Serum creatinine compliance is defined in Supplemental Table 5. SCr was assessed at 1-year follow-up in AKI patients. AKI recovery was defined as SCr <150% of previous baseline. Based on the 1-year SCr, eGFR was calculated using the bedside Schwartz formula.¹⁶  Patients with an eGFR <90 mL/min/1.73 m² were classified as having CKD.

The analysis considered patients who had multiple high NTMx encounters as independent. Descriptive statistics summarized demographic and clinical data with continuous data reported as median with interquartile range and categorical data reported as frequency with percent. The outcomes of interest included prevalence of NTMx exposure, prevalence of NTMx-AKI occurrence, severity of NTMx-AKI, and duration of NTMx-AKI measured in days. More detailed study measure definitions appear in Supplemental Table 5. A retrospective case-control analysis assessed potential associations between measures of interest and NTMx-AKI. The measures of interest included anthropometrics, service line, classification of NTMx exposure, specific NTMx exposure, therapeutic drug monitoring, admission indications, and duration of the NINJA encounter. Univariable associations between categorical hospital measures, and outcomes of interest were assessed via χ² test or Fisher’s exact test for small sample size as well as unpaired t tests or Wilcoxon rank tests for continuous measures. Measures with a univariable P value of .15 or less were considered for covariate adjustment in multivariable logistic regression models. Logistic regression assessed associations with NTMx-AKI occurrence as dichotomous outcome. Measures associated with severity were examined in the 44 records with indication of NTMx-AKI. The outcome of NTMx-AKI severity was dichotomized as either stage 1 or stage 2 or 3 for univariable and multivariable logistic regression model. As the validity of logistic regression models came into question with small cell counts and adjustment for multiple covariates,²⁰  multivariable models for the subcohort analysis adjusted for only age, weight, height, and days of exposure. Statistical significance was set at an α .05 level. Estimated odds ratios are reported with 95% confidence limits. All analyses were conducted in SAS version 9.4 (SAS Institute Inc, Cary, NC, USA).

There were 609 NINJA encounters for 565 unique patients. The median patient age was 11 (interquartile range: 3.8 to 16) years. A summary of demographics, admission indication, and type of NTMx exposures are included in Table 1.

NTMx prevalence rate was highest for the gastroenterology, liver, and neurosurgery services, all exceeding 18 NTMx exposures per 1000 patient days (Table 2). Among the included service lines, NTMx exposures resulted from administration of ≥3 nephrotoxins in 73.5%, ≥3 days of vancomycin in 19.1%, and ≥3 days of an aminoglycoside in 7.4%. In patients who received ≥3 nephrotoxins, 84% of exposures included nonsteroidal antiinflammatory drugs, 56% vancomycin, and 49% contrast. Figure 1 summarizes the relative incidence proportion of the type of NTMx exposures by service line. Supplemental Table 6 summarizes the classification of NTMx exposure between service lines.

The top 3 nephrotoxic medications categorized by service are summarized in Fig 2. Given the frequent use of vancomycin, the indications were assessed and are as follows: (1) empirical use before de-escalation to another antibiotic (57%), (2) sepsis “rule-out” (22%), (3) use supported by microbiology culture data (15%), (4) surgical prophylaxis (6%), and (5) empirical treatment per infectious disease service recommendations without definite culture data (1%).

Across all NINJA encounters, there were 44 (7.2%) episodes of NTMx-AKI in 41 patients (3 patients had 2 NTMx-AKI episodes). NTMx-AKI prevalence rates per 1000 patient days were high and varied by service (0 to 4.28 NTMx-AKI events per 1000 patient days) (Table 2). Among those exposed, the rate of NTMx-AKI was highest for patients admitted to pulmonology (14.29%), gastroenterology (11.48%), and neurology or psychiatry (11.11%) services. Despite an overall SCr compliance of 70%, compliance categorized by service was highly variable ranging from as low as 47.6% for neurology or psychiatry to 85.1% for pulmonology (Table 2).

Univariable analyses suggest age, weight, height, length of admission, type of NTMx exposure, and duration of encounter are all associated with NTMx-AKI (all P < .001) (Table 1). In multivariable logistic regression analysis, after adjusting for age, weight, height, length of admission, and duration of encounter, there was no association between type of NTMx exposure and NTMx-AKI (P = .263). However, patients with supratherapeutic vancomycin trough levels had an increased odds of NTMx-AKI of 4.15 (95% [confidence interval] CI: 1.95–8.80; P < .001). The estimated increase in odds remained after adjusting for age, weight, height, length of admission, and duration of encounter (adjusted odds ratio [aOR] 3.8; 95% CI: 1.6–9.3; P = .003). All 9 of the NTMx-AKI episodes attributed to ≥3 days of aminoglycosides occurred within the pulmonology service. Twenty-six patients with NTMx-AKI (60%) were exposed to vancomycin; however, culture data only supported the use of vancomycin in 46.2% of occurrences. Vancomycin levels were documented for 258 of 370 (70%) records with vancomycin exposure. Of the patients with documented supratherapeutic vancomycin levels, 11 of 53 (21%) experienced NTMx-AKI compared to 12 of 317 (3.8%) AKI occurrences in those who did not have supratherapeutic vancomycin levels. Service line was not associated with NTMx-AKI (P = .143). In patients with NTMx-AKI, service line was not associated with duration of NTMx exposure (P = .351).

Table 3 compares stages of NTMx-AKI by service line, admission indication, classification of NTMx exposure, and other patient characteristics in the 44 patients who experienced AKI. Sixty-four percent (n = 28) of NTMx-AKI were categorized as stage 1 (mild) and 36% (n = 16) as severe (Table 3). In the 44 patients with NTMx-AKI, presence of a supratherapeutic vancomycin trough was not significantly associated with NTMx-AKI severity (OR [odds ratio] 2.8; 95% CI: 0.68 to 11.2; P = .155). After adjusting for age, weight, height, and days of exposure, the adjusted odds ratio (aOR) indicating an increase in severity of NTMx-AKI remained nonsignificant (aOR 3.0; 95% CI: 0.62–14.8; P = .172). There was no indication of an association between the type of NTMx exposure and AKI severity (P = .586). There was no association between admission indication and AKI severity (Fisher’s exact, P = .3766. Duration of exposure was not associated with AKI severity (P = .596).

SCr at 1-year follow-up was available for 25 (58%) patients who experienced NTMx-AKI. Among these patients, SCr was >150% baseline in 16 (64%). eGFR was >90 mL/min/1.73 m² in 14 (56%), 60 to 90 mL/min/1.73 m² in 10 (40%), and <60 mL/min/1.73 m² in 1 (4%) with an overall CKD rate of 44%.

In this single-center study, nephrotoxic medication exposures occurred most frequently in patients admitted to the liver, neurosurgery, and gastroenterology services. Although most episodes of NTMx-AKI were mild (stage 1), and did not differ across services, SCr compliance rates were highly variable (47.6% to 85.1%). The service with the lowest rate of SCr compliance was the neurology or psychiatry service, and the pulmonology service had the highest rate of SCr compliance. Across service lines, the high NTMx exposures prevalence rate per 1000 patient days ranged from 0.73 to 65.87, with the lowest number of exposures occurring on the neurology or psychiatry service and the highest occurring on the liver service. These data suggest that targeting services for NTMx exposure reduction is needed. Furthermore, NTMx-AKI rates may actually be higher than reported because of imperfect SCr compliance and presents an opportunity for improvement.

The most notable differences in this study compared to the existing literature⁹  are that we excluded the hematology or oncology and BMT services and that exposures in our study were much higher among neurosurgery, gastroenterology, and liver services (18 to 65 exposures per 1000 patient days). Our decision to exclude the hematology or oncology and BMT services were several-fold. Recently, two separate studies reported on their single-center quality improvement initiatives to reduce high NTMx exposures and NTMx-AKI in pediatric oncology patients by implementing changes to antimicrobial prescribing and duration of therapy.^12,14  Specifically, piperacillin-tazobactam was replaced with cefepime and antimicrobial stewardship restricted duration of vancomycin. Both measures resulted in a substantial reduction in high NTMx exposures and NTMx-AKI.^12,14  Although we have only recently begun to systematically track exposures by service, exposures on the hematology or oncology and BMT services are all <20 exposures per 1000 patient days in 2021, with SCr compliance rates exceeding 90% (Personal communication, Target Zero Initiative, Children’s Hospital Colorado). Thus, it was our intent to provide a description of other high NTMx services and begin to understand the types of medications prescribed. Understanding this will lay the foundation for single and multicenter studies that implement targeted interventions.

Goldstein and colleagues found that NTMx exposures in their single-center study were most often because of antiviral agents, calcineurin inhibitors, and amphotericin,⁹  which differs from our findings likely because of our exclusion of patients admitted to the hematology, oncology, and BMT services, for the aforementioned reasons. The top 3 NTMx exposures among all services in our study were vancomycin, contrast, and nonsteroidal antiinflammatory drugs, excluding liver, pulmonology, and neurology or psychiatry services. The liver service saw the highest NTMx use of tacrolimus, valganciclovir, and aspirin. Although these medications may not be modifiable in patients undergoing liver transplant, future initiatives aimed at increasing compliance in serum creatinine, or alternatively, less invasive urine biomarker surveillance using neutrophil gelatinase associated lipocalin²¹  are warranted to improve detection of AKI in this high NTMx exposure service.

We did not find any difference in NTMx-AKI incidence or duration by service. Although our exploratory retrospective case-control analysis identified type and duration of NTMx exposure being associated with NTMx-AKI, there was no association with severity of NTMx-AKI. Both findings may be attributed to the overall low rate of NTMx-AKI when compared to previous literature.⁷  This can be explained by the excluded services as well as the highly variable SCr compliance across services and thus missed diagnoses of NTMx-AKI.

One of the biggest areas for improvement may be related to the use of vancomycin. Not surprisingly, our retrospective case-control analysis found that supratherapeutic vancomycin trough levels were associated with an increase in the relative risk of NTMx-AKI and suggests that therapeutic drug monitoring should be used in conjunction with SCr monitoring whenever possible. Although Young and colleagues and Benoit and colleagues demonstrated reductions in NTMx exposures even after implementation of a vancomycin trigger,^12,14  partnerships with antimicrobial stewardship and restricting prolonged use likely contributed. Indeed, vancomycin is used ubiquitously in pediatric hospitals,²²  but unfortunately, this is often without a clear indication. In our study, more than one-half of patients who developed NTMx-AKI were exposed to vancomycin (60%), and the indication was often empirical broad-spectrum coverage that was either narrowed or discontinued altogether after the sepsis rule-out period of 48 hours. Advances in rapid microbiology identification techniques and evaluation of culture data at 36 hours or shorter durations^23–25  have the potential to decrease the vancomycin exposure period and thus reduce the potential for NTMx-AKI. This would also potentially limit development of local or even wide-spread antimicrobial resistance.²⁶ 

Aligning with the report from Menon and colleagues,⁵  just under one-half of the patients who had follow-up SCr data had evidence of CKD. Ongoing surveillance of these children is necessary and recommended by the recent pediatric response to the 22^nd Acute Disease Quality Initiative consensus statement.²⁷  In fact, all patients who have experienced an episode of severe AKI or do not recover from an episode of AKI warrant nephrology follow-up for evolution of CKD within 3 months of the AKI episode.

The strength of this study is that we focus our attention on nonhematology or oncology and BMT services, thus exposing opportunities for improvement among other high-risk service lines. There are, however, several limitations of this study, most notably its single center retrospective approach. As the study is retrospective and observational, caution is advised when interpreting and generalizing results. Our findings of high NTMx exposures on the liver, gastroenterology, and neurosurgery services may not translate across institutions. We also excluded services known to have high NTMx exposures and high NTMx-AKI (hematology or oncology and BMT). We did not evaluate differences in prescribing patterns by month. This is relevant given the global coronavirus disease 2019 pandemic, which may have shifted prescribing patterns.

This study highlights other services for which NTMx surveillance should be improved, encompassing partnerships with antimicrobial stewardship and restricting use and duration of medications such as vancomycin to those with a firm indication. Although NTMx-AKI rates were low and most often stage 1 (mild), just under one-half of the patients in whom 1-year follow-up data were available had early CKD reinforcing the need for long-term systematic follow-up. Finally, our findings support the rationale for specific interventions on the gastroenterology, liver, and neurosurgery services.