Variation in Intravenous Acetaminophen Use in Pediatric Hospitals: Priorities for Standardization

The Pediatric Health Information System (PHIS) database contains administrative and billing data from 49 tertiary, free standing, primarily academic (95%), pediatric hospitals affiliated with the Children’s Hospital Association (CHA) (Lenexa, KS). The median size of the pediatric hospital is 350 staffed beds with an average daily census of 230 patients. Encounter-level data include patient demographic characteristics; International Classification of Diseases, 10th Revision, Clinical Modification diagnostic and procedural codes; and date-stamped billing data for a wide range of clinical services including diagnostic tests and pharmaceuticals. Patients are assigned a unique identifier to permit tracking of multiple admissions to the same hospital. The quality of the data are managed with pre- and postsubmission audits that continue as a joint collaborative effort between CHA and the affiliated hospitals.¹³  We queried PHIS for inpatient and observation encounters for patients age 18 years or younger who were discharged between January 1, 2019, and December 31, 2019. Patients are assigned a unique identifier to permit tracking of multiple admissions to the same hospital.

Encounters from hospitals that did not submit complete data to PHIS were excluded. Encounters were categorized into clinically relevant groups by using All Patient Refined Diagnosis Related Groups (APR DRGs) v32.¹⁴ 

We used PHIS pharmacy billing data to identify calendar days on which patients were billed for IV acetaminophen. We defined IV acetaminophen use as the percentage of encounters associated with administration of at least 1 dose of IV acetaminophen. We defined days of therapy (DOT) as the number of calendar days for which an encounter was billed for IV acetaminophen. The 20 APR DRGs with the highest total DOT across all hospitals were designated “high-use APR DRGs.” We calculated the IV acetaminophen costs by multiplying billed charges for IV acetaminophen by the respective hospital’s pharmacy-specific ratio of costs to charges, as reported on their hospital cost report submitted annually to the Centers for Medicare and Medicaid Services.¹⁵ 

To adjust for differences in patient characteristics when comparing IV acetaminophen use across hospitals, we examined patient demographic and clinical variables that were potentially associated with higher use. These variables included age at admission (categorized into <1, 1–4, 5–12, 13–18 years), race and ethnicity (white, Black, Hispanic, other, and unknown), and primary insurance type (private versus public). International Classification of Diseases, 10th Revision, Clinical Modification diagnostic and procedural codes were used to identify patients with a complex chronic condition¹⁶ . The proprietary PHIS case mix index (CMI) served as another marker of disease severity.¹⁷  The CMI is an APR DRG–based resource intensity weight derived by CHA by using the HCUP (Healthcare Cost and Utilization Project) KID (Kid’s Inpatient Database) Database, a nationally representative sample of pediatric inpatient admissions.¹⁸  For a given APR DRG and severity of illness combination, the CMI is the ratio of the mean cost of hospitalization for that APR DRG severity of illness combination relative to the mean cost across all admissions to tertiary, pediatric hospitals across the United States. A CMI >1 indicates that the cost of that hospitalization is expected to be higher than average, and, conversely, a CMI <1 indicates an expected cost lower than average. We categorized patients in this study into deciles of CMI. Additional risk factors based on billed data included surgical procedure defined by the presence of an operating room charge, ICU admission, extracorporeal membrane oxygenation (ECMO), total parenteral nutrition (TPN), and mechanical ventilation.

We used logistic regression to estimate the univariate and multivariable associations between patient demographic and clinical characteristics and IV acetaminophen use. Generalized estimating equations were used to control for clustering within hospitals. The multivariable model was used to calculate rates of hospital-level IV acetaminophen use adjusted for variation in patient characteristics. We describe the hospital-level variation in IV acetaminophen use within each of the 20 high-use APR DRGs with medians and interquartile ranges (IQRs), defined as the difference between the 75th and 25th percentiles of hospital IV acetaminophen use. Within each APR DRG, hospitals with fewer than 20 encounters were excluded to ensure adequate volume of encounters. We defined high-priority APR DRGs for practice standardization as those with the high contributions to IV acetaminophen use (ie, top quartile in DOT across APR DRGs) accompanied by wide variation in IV acetaminophen use (ie, IQR across hospitals >50%). Statistical significance was achieved with a 2-sided P value <.05. Analyses were performed by using SAS System for Windows v9.4 (SAS Institute, Inc, Cary, NC).

In 2019, there were 879 869 inpatient and observation encounters from 49 hospitals in the PHIS database. A total of 13 523 encounters from 1 hospital that did not submit a complete year of data were excluded. The study cohort included 866 346 encounters for 677 497 unique patients (mean: 1.3 encounters per patient) from 48 hospitals with a total of 4 554 000 hospital days. The median number of encounters per hospital was 15 591 (range: 6500–40 428). Of those encounters, 125 456 (14.4%) received at least 1 dose of IV acetaminophen during the hospitalization, corresponding to a total of 287 935 DOT (6.3% of total hospital days) and a cost of $29.8 million.

In univariate analyses, patient characteristics associated with higher rates of IV acetaminophen use included older age (5–12 and 13–18 years versus <1 year: 18% vs 9%, P < .0001), non-Black race or ethnicity (white and Hispanic versus Black: 15% and 16% vs. 10%, P = .004), private versus public insurance (16% vs 13%, P = .01), complex chronic condition versus no complex chronic condition (21% vs 11%, P < .0001), surgical procedure during admission versus no surgical procedure (46% vs 5%, P < .0001), ICU admission versus no ICU admission (26% vs 12%, P < .0001), ECMO versus no ECMO (57% vs 14%, P < .0001), TPN versus no TPN (37% vs 13%, P < .0001), mechanical ventilation versus no mechanical ventilation (34% vs 13%, <.0001), and higher CMI (P = .0001) (Table 1). Age, payer, surgical procedure, ICU admission, TPN, and CMI remained significantly associated with IV acetaminophen use in the multivariable model.

The study cohort comprised 312 APR DRGs. The 20 APR DRGs (6% of APR DRGs) with the highest total DOT accounted for 47% of DOT (135 910 days) and 48% of IV acetaminophen costs ($14.2 million). Nineteen of these APR DRGs included a surgical procedure and 1 was a medical diagnosis. Four were neonate related (Table 2).

Appendectomy had the highest DOT for IV acetaminophen use (18 447 days), with 32% more DOT than that of the second-ranked APR DRG, major small and large bowel procedure (12 452 days). Craniotomy, tonsil and adenoidectomy, dorsal, and lumbar spinal fusion were also in the top quartile in terms of DOT. Appendectomy is an example of an APR DRG with a large volume of treated encounters (n = 9958) with short average duration of therapy (1.9 days per encounter). In contrast, major small and large bowel procedure had fewer treated encounters (n = 2714) but longer average duration of therapy (4.6 days per encounter).

The observed rate of IV acetaminophen use at the hospital-level ranged from <0.1% to 43% across hospitals (Fig 1). Even after accounting for patient demographic and clinical factors associated with IV acetaminophen use, wide variation in adjusted hospital-level use persisted (range <0.1%–31%), which suggests that peer pediatric hospitals are using IV acetaminophen at varying rates in similar populations. Furthermore, 19% of the hospitals (9 PHIS hospitals) use IV acetaminophen in less than 5% of all patient observed and adjusted encounters despite similar clinical populations.

In Figure 2 we display boxplots of the distribution of hospital IV acetaminophen use within each of the 20 highest using APR DRGs. In 3 APR DRGs (dorsal and lumbar fusion, major repair of heart anomaly, and cardiac valve procedure [noncatherization]), median hospital IV acetaminophen use was >90%, indicating that IV acetaminophen use is high in these populations at the majority of hospitals. In contrast, in septicemia, the median hospital IV acetaminophen use was 17%, indicating consensus away from use. We also observed wide hospital variation in IV acetaminophen use, with IQRs ranging from 17% for septicemia to 67% for tonsil and adenoidectomy. Fifteen of these APR DRGs had IQR >50%, another indication of lack of consensus.

To identify priority populations for reduction of IV acetaminophen use and future standardization, we propose an APR DRG prioritization framework that takes into account the potential magnitude for reduction in use (total DOT) and variation in practice (IQR of use across hospitals). We defined high-priority APR DRGs as those with the greatest contributions to IV acetaminophen use (ie, top quartile in DOT) accompanied by wide variation in IV acetaminophen use (ie, IQR across hospitals >50%). In Figure 3 we illustrate the prioritization framework. Appendectomy is a high contributor to use (18 447 DOT), with wide variation in IV acetaminophen across hospitals (IQR 55%, median 60%), and a large number of encounters (20 452). Another high-priority population is tonsil and adenoidectomy, with a high contribution of use (DOT = 9629), with wide variation across hospitals (IQR 67%, median hospital use 40%) and a large number of encounters (19 669). Appendectomy, tonsil and adenoidectomy, and craniotomy are strong candidates for future reduction and standardization efforts to mitigate excessive IV acetaminophen use and its associated cost.

The cost associated with IV acetaminophen use was $29.8 million among 866 346 children hospitalized in 2019. We found substantial variation in IV acetaminophen use across the 48 study hospitals, ranging from <0.1% to 31% of all hospital encounters after adjusting for patient factors. Of the 312 diagnostic groups examined, 20 diagnostic groups accounted for 47% of the total use and 48% of the total cost of IV acetaminophen in 2019. Within these 20 high-use diagnosis groups, we found high variability across pediatric hospitals in use of IV acetaminophen for each diagnosis, especially for appendectomy, tonsil and adenoidectomy, and craniotomy.

IV acetaminophen has been widely incorporated into clinical practice across pediatric hospitals despite its high cost and the availability of alternative formulations. One of the clinical applications of IV acetaminophen is as a nonopioid analgesic. Although the initial marketing campaign focused on the opioid-sparing effects, a majority of pediatric studies suggest only a modest clinical benefit of the IV formulation. Authors in a few single-center studies reported small decreases in opioid equivalents when using IV acetaminophen compared with placebo in the neonatal and pediatric populations.^8,19,20  However, researchers in other pediatric studies evaluating the opioid-sparing effects of IV acetaminophen compared with enteral or rectal formulations indicated a lack of superiority of IV acetaminophen.^10,21,22  The second clinical application of IV acetaminophen is as an antipyretic. In an observational single-center study, IV acetaminophen was compared with oral acetaminophen in the management of fevers. The authors demonstrated patients receiving IV acetaminophen had faster fever descent compared with the oral formulation; however, beyond 4 hours, there was no difference between the 2 groups.²³ 

In this study, we do not evaluate the opioid-sparing effect of IV acetaminophen, including, but not limited to, the presence of an ileus, length of stay, and total opioids received. The PHIS database documents the presence of opioid charges per day rather than the granular doses received by the patient. In previous studies, researchers using the PHIS database have evaluated opioid exposure as a dichotomous variable.²⁴  Future investigation is needed to understand whether institutions with high use of IV acetaminophen correspond to lower opioid DOT and shorter length of stays.

Despite the limited evidence for enhanced efficacy, with our study, we demonstrated there remains a portion of hospitals that use IV acetaminophen upwards of 2 times the median hospital rate of 15.5% of adjusted encounters. Because the data do not clearly demonstrate a benefit, the high degree of variation is unexplained. We speculate it is due to a lack of consensus on which patient populations warrant use of IV acetaminophen. Given the high price point of the IV formulation and variation in practice, further research is imperative to identify patient populations that benefit from IV acetaminophen as well as from practice standardization to control use. Future research focusing on practice patterns and standardization in appendectomy, tonsil and adenoidectomy, and craniotomy will create a platform for implementation for the remaining high-use DRGs.

With our study, we demonstrated a skewed pattern of use toward 20 diagnosis groups accounting for 47% of the total use and $14.2 million. We further demonstrated that pediatric hospitals use IV acetaminophen at varying rates for each diagnosis. Appendectomy, tonsil and adenoidectomy, and craniotomy are candidate diagnosis to further evaluate hospital factors that contribute to use. Other potential contributors to the hospital variability include presence of IV acetaminophen on hospital formulary, incorporation into order sets, and pediatric hospitals with formulary management systems. In the adult literature, widespread use of IV acetaminophen has been associated with hospital formulary adoption, extending pharmacy approval from 24 to 48 hours of use, and use of postoperative order sets incorporating IV acetaminophen.²⁵  As hospitals aim for provider efficacy, the implementation of order sets in the electronic medical record has increased the prescribing of certain medications and laboratories.²⁶  Future research is needed to define the factors contributing to varying use rates across the PHIS hospitals to inform cost reduction and practice standardization initiatives.

One of the strengths of this study is the large number of pediatric hospitals in the United States that contribute data to the PHIS network. By querying the PHIS database, we were able to evaluate clinical and resource use in pediatric hospitals. Another strength of this study is evaluating hospital IV acetaminophen use across multiple diagnosis groups to understand the depth of variation in practice patterns. A major implication of this study is the need for standardization of IV acetaminophen use. Within a given DRG, hospitals are using IV acetaminophen in vastly different patterns without demonstrated patient benefit.

The results of this study must be considered in the context of certain limitations. We used each hospital’s pharmacy-specific ratio of costs to charges to calculate cost of IV acetaminophen on the basis of billed charges. The $29.7 million incurred by IV acetaminophen use at these 48 pediatric hospitals in 2019 is an estimate of resources expended by the institutions and does not reflect actual costs incurred by the institutions or payments received by the hospitals. The PHIS network’s quality assurance is a collaborative effort between the hospitals and the CHA. Errors in coding and billing are still possible despite the rigorous audits. In addition, we were unable to assess the total doses of IV acetaminophen received by patients, precluding further analysis of doses of IV acetaminophen per encounter. Another limitation is using billing data to account for a patient’s clinical complexity. CMI is used as an indicator for disease severity. CMI is subject to the accuracy of the health provider documentation of the patient and the coder extracting the data.²⁷  Mendez et al demonstrated that CMI’s can be higher for patients who cost more money to care for rather than those with more severe disease.²⁸  A cofounding factor of IV acetaminophen use is the patient’s enteral status. PHIS does not document the enteral status of patients or the enteral practice patterns of the health care institutions. Depending on a patient’s clinical status, preoperative and postoperative surgical procedure, TPN, ECMO, mechanical ventilation, and a complex chronic condition are not contraindications to enteral medications. Finally, PHIS data are limited to the hospitals within the network; we are unable to calculate the entire cost to the health care system.

We observed large variation in IV acetaminophen use across pediatric hospitals and within specific diagnosis groups. The high cost of IV acetaminophen coupled with limited evidence demonstrating superiority to available alternatives support the implementation of practice reduction and standardization efforts. Candidate conditions for such initiatives include appendectomy, tonsil and adenoidectomy, and craniotomy. Identifying specific policies and practices that effectively decrease unnecessary hospital use of IV acetaminophen could help guide initiatives targeting these conditions.