Introduction

Endoscopic retrograde cholangiopancreatography (ERCP) is a diagnostic and therapeutic procedure required for the management of pancreatic and biliary diseases in adults and children [1]. ERCP has been performed for adults since the 1960s [2], while the first successful ERCP performed on a child using the standard adult cannulating instrument was reported in 1976 [3]. Meanwhile, magnetic resonance cholangiopancreatography has largely replaced diagnostic ERCP to avoid post-ERCP pancreatitis (PEP); therefore, mainly therapeutic interventions are performed with ERCP in the pancreaticobiliary tract [4, 5]. PEP is the most common adverse event after ERCP, which can lead to a significant increase in mortality and healthcare costs, with the prevalence ranging from 1.6 to 10.9% among children [6,7,8,9,10]. Some single-center studies explored the factors associated with PEP in children; however, the national temporal trend of PEP in pediatric patients has never been reported [6, 11]. Barakat et al. conducted a national study using data from the National Inpatient Sample (NIS) and National Readmission Database to analyze the outcome and utilization trends over time in children, which was focused on ERCP indications, utilization, and readmissions; they used the updated data until 2014 [12]. Herein, we present our nationwide analysis of children undergoing ERCP in the USA using the NIS database from 2008 to 2017 to investigate the temporal trend and factors associated with PEP in those children.

Materials and methods

Study design and data source

A nationwide, retrospective cohort study was conducted using data from the NIS database from 2008 to 2017. The NIS has been developed by the Agency for Healthcare Research and Quality for the Healthcare Cost and Utilization Project (HCUP) in the USA [13]. It is the largest public database of all-payer inpatient care data that is used to estimate the regional and national costs of inpatient utilization, access, charges, quality, and outcome, and contains data on 7 million hospital stays, capturing approximately 20% of all inpatient hospitalizations in the USA [14]. The HCUP comprised the quality-control procedures to assess data quality and edit individual data sources [15], which enables the accurate and reliable estimation based on the NIS database. Several articles about ERCP using the NIS database have been published in authoritative magazines [16,17,18]. Because the NIS databases were deidentified for public availability, our study was exempted from the requirement of institutional review board approval.

Study population

The study population included all children aged ≤ 18 years who underwent ERCP. The procedures were identified using the International Classification of Diseases, 9th and 10th revisions-Clinical Modification (ICD-9-CM and ICD-10-CM) diagnostic and procedural codes (Supplementary Table 1).

Outcomes

Our primary outcomes were the temporal trend and factors of PEP in children, and secondary outcomes were in-hospital mortality, total charges (TC), and total length of stay (LOS). TC was adjusted for inflation to currency rates ($) of 2017. To distinguish PEP from acute pancreatitis at admission, we followed the methodology described in prior studies [19,20,21]. Patients with a primary or secondary diagnosis of acute pancreatitis were deemed as not associated with ERCP to avoid confusion. Conversely, patients diagnosed with acute pancreatitis that was not a primary or secondary diagnosis were considered PEP.

Study variables

Demographic variables included age (years), age groups (0–4, 5–9, 10–13, and 14–18 years), sex (male/female), and race (race/ethnicity: White, Black, Hispanic, Asian/Pacific Islander, Native American, and other). Patient-associated variables comprised median household income (first quartile-fourth quartile), primary expected payer (Medicare, Medicaid, private insurance, self-pay, or other), and admission day (Monday–Friday, Saturday–Sunday). Hospital-related variables included bed size (small/medium/large), location status (rural/urban), teaching hospital, and hospital region (Northeast, Midwest, South, and West). The diagnosis parameters included chronic pancreatitis, biliary obstruction, cholangitis, and choledocholithiasis. ERCP-related parameters included ERCP type (diagnostic or therapeutic), dilation of the ampulla and biliary duct, sphincterotomy and papillotomy, stent insertion into the bile duct, removal of the stone from the biliary tract, and all stents insertion into the pancreatic duct. Comorbidity variables included end-stage renal disease, obesity, hypertension, diabetes mellitus, nonalcoholic fatty liver disease, acute kidney injury, asthma, and Sickle cell disease. Even the discharge weight and hospital stratum data were retrieved from the NIS database, as well as the HCUP hospital identification number. The HCUP website [22] describes how an individual data element is coded in the NIS database and the uniform values, for example, bed size categories are based on hospital beds, and are specific to the hospital's location and teaching status.

Statistical analysis

Statistical analysis was conducted using SAS 9.4 (SAS Institute Inc., Cary, NC, USA). All reported probability values (P-values) were 2-tailed, and statistical significance was considered when P < 0.05. The sample size was estimated using 10 events per variable approach for the regression model [23], requiring about 290 PEPs or more. As the NIS contains sampling and weight variables that can be used to obtain national estimates, all results reported are weighted by NIS sampling weights. For quantitative variables, data were summarized as means with standard deviations or medians with interquartile ranges (IQR), and for categorical variables, data were summarized as frequency with percentage. The outcome variables were compared using t-tests or Wilcoxon signed-rank test, and the Rao-Scott χ2 test of patients with and without PEP in the univariate analysis. The temporal trends for PEP prevalence were tested with the Cochran-Armitage trend test.

Multivariable logistic regression accounting for the survey design was applied to investigate the independent factors of PEP. Variables with P ≤ 0.2 in univariate analysis were used for multivariable logistic regression analysis, in which unadjusted odds ratios (ORs), adjusted odds ratios (aORs), and corresponding confidence intervals (CIs) were calculated. For the adjusted model, we used Wang’s model selection macros for complex survey data to conduct the model selection (stepwise method, SL entry = 0.05, SL stay = 0.05) [24]. A forest plot was also provided with this model.

Missing variables and sensitivity analysis

If a variable had ≥ 5% missing rates, we would impute the missing value before conducting multivariable logistic regression in the sensitivity analysis. This was the case for one variable, i.e., race, which had an 8.9% missing rate (Table 1). We imputed missing values using mode imputation (White race), and then used the same covariates and methods in the adjusted model to evaluate the robustness of our results.

Table 1 Missing rate of variables in multivariable logistic regression
Full size table

Results

Patient demographics

This study analyzed 45,268 hospitalized pediatric patients who underwent ERCP during 2008–2017. Of them, 2043(4.5%) were diagnosed with PEP. A selection flow diagram of the target population is shown in Fig. 1. Most patients who underwent ERCP were aged 14–18 years, female, and of White or Hispanic race. More than 90% of the expected primary payers were Medicaid or private insurance. The baseline characteristics are summarized in Table 2.

Fig. 1
figure 1

Flow diagram of the target population

Full size image
Table 2 Baseline characteristics of children underwent ERCP
Full size table

Temporal trends of PEP

The overall PEP prevalence in children was 4.5%, and the temporal trend of PEP prevalence decreased from 5.0% in 2008 to 4.6% in 2017 (P = 0.0002) (Fig. 2).

Fig. 2
figure 2

Temporal trend of PEP prevalence

Full size image

Mortality, TC, and LOS

Of 2043 pediatric patients with PEP, 16 (0.8%) died during hospitalization, compared with 0.1% of patients without PEP (P = 0.0325). The mean TC of PEP patients was $132,356 ± 12.384.7, which was higher than that of patients without PEP ($65,768 ± 1833.3; P < 0.001). The median LOS of patients with PEP was 5.4 (IQR: 3.2–10.2) days, which was higher than that of patients without PEP (3.2 days, IQR: 1.9–5.4 days; P < 0.001, Table 2).

Multivariable analysis of factors associated with PEP

By multivariable logistic regression analysis, unadjusted risk factors associated with PEP were the 0–4-year-old age group (OR 2.61, 95% CI 1.51–4.50; P = 0.0090) compared with the 14–18 age group, hospitals located in the West (OR 1.94, 95% CI 1.26–2.98; P < 0.0001) compared with hospitals in the Northeast, insertion of stent into the bile duct (OR 1.47, 95% CI 1.07–2.02; P = 0.0178), and end-stage renal disease (OR 6.41, 95% CI 1.11–37.16; P = 0.0382). Unadjusted protective factors associated with PEP were increasing age (OR 3.49, 95% CI 1.87–6.52; P < 0.0001) and hospitals located in the South (OR 0.51, 95% CI 0.29–0.90; P < 0.0001). The results are shown in Table 3.

Table 3 Multivariable logistic regression results of factors associated with PEP
Full size table

Adjusted risk factors associated with PEP were hospitals located in the West (aOR 2.09, 95% CI 1.36–3.20; P < 0.0001), bile duct stent insertion (aOR 1.49, 95% CI, 1.08–2.05; P = 0.0040), and end-stage renal disease (aOR 8.05, 95% CI 1.66–39.16; P = 0.0098). Adjusted protective factors of PEP were increasing age (aOR 0.95, 95% CI 0.92–0.98; P = 0.0014) and hospitals located in the South (aOR 0.53, 95% CI 0.30–0.94; P < 0.0001). The results are summarized in Table 3 and Fig. 3. In the sensitivity analysis, the adjusted results were the same as above.

Fig. 3
figure 3

Adjusted factors of PEP

Full size image

Discussion

In this nationwide, retrospective cohort study, we observed a decreasing prevalence of PEP in children of the USA during 2008–2017. We identified multiple risks and protective factors associated with PEP as well and noted that this is the first study to explore the national temporal trend of PEP in children.

Our study reported an overall prevalence of 4.5% for PEP in children, which is consistent with previous pediatric studies. The prevalence declined from 5.0% in 2008 to 4.6% in 2017; this downward trend may be beneficial because of several reasons. First, this may have stemmed from various mechanisms involving the interruption of biochemical reactions involved in proteolytic enzyme activation, acinar secretion impairment, autodigestion, and other key mechanisms that are central to the development of pancreatitis or pancreatic injury during ERCP, including damage from mechanical, thermal, chemical, hydrostatic, enzymatic, and microbiologic factors [25,26,27]. Second, the widespread use of non-steroidal anti-inflammatory drugs (NSAIDs), including diclofenac and indomethacin suppository to prevent PEP among adolescents as recommended by the European Society of Gastrointestinal Endoscopy, may have reduced the prevalence of PEP during the latter part of the study period. Third, the increase in the number of research studies on PEP prevention among children, which identified potential protective and risk factors of PEP, may have contributed to the declining prevalence [4, 6, 8, 11]. Therefore, these efforts would be useful for endoscopists to conduct pre-ERCP risk assessments.

We identified several protective and risk factors associated with PEP. Among the protective factors, older age of children was a protective factor for PEP, while a younger age was associated with higher PEP prevalence. Our result was in line with the report of Limketkai et al. [28], who reported a higher prevalence of PEP in the youngest (0–6) age group than in the 7–12- and 13–17-year-old age groups. The bile and pancreatic ducts are more vulnerable in younger children, and they have unique anatomical challenges in the small duodenal lumen, smaller papillary orifice, and narrow caliber pancreatic duct that present technical difficulties in the cannulation of the intended duct and other therapeutic interventions. Among the risk factors, we identified end-stage renal disease as a risk factor for PEP, which has not been recognized as a risk factor in children until now but is a risk factor for adults [21]. Its association with higher PEP may be due to papillary edema from fluid overload, which causes difficult biliary cannulation [29]. Even stent insertion into the bile duct was concluded as a procedure-related risk factor. No previous study has reported any association between biliary stent insertion and higher prevalence of PEP in children. However, in the adult population, several studies have found that biliary stent placement is associated with a high prevalence of PEP [30, 31]; this happens because biliary stent placement oppresses the orifice of the main pancreatic duct or common channel, which obstructs the pancreatic duct [32].

We also found associations between hospital locations and the PEP prevalence; hospitals located in the South were associated with a lower prevalence of PEP, while hospitals located in the West were associated with higher PEP prevalence in the USA. One study on adults with PEP that used the NIS database obtained similar results as ours [21], but they did not explain their results. To determine the possible reasons behind this result, we conducted an exploratory analysis. We found that the number of ERCP procedures and the proportion of teaching hospitals in the South region were both high, while in the West, the number of teaching hospitals was low even though a higher number of ERCP procedures had been conducted. Therefore, we ascribed that the number of ERCP procedures and that of teaching hospitals in different regions may be attributed for our obtained result.

In our study, the overall post-ERCP mortality was 0.14%, which was similar to that reported by Barakat et al. (0.1%) [12]. The mortality, TC, and LOS of PEP patients were significantly higher than those without PEP, which meant that PEP would result in higher mortality and more medical burden for children. Based on these results, we consider it critical for endoscopists to provide extensive care for preventing PEP and adopt appropriate clinical practice methods. Several studies have explored methods for preventing PEP in adults, including the assessment of patient-related factors, procedural techniques for prevention, and chemoprevention with NSAIDs. However, for children, especially non-adolescents, the effective chemoprevention of PEP has been seldom reported. Troendle et al. provided some encouragement that intravenous ibuprofen might help prevent PEP in children, but their evidence was very weak regarding the reduced rates of PEP with NSAIDs [33]. Therefore, assessing patient-related and procedure-related factors may most effectively prevent PEP in children.

Our study had several limitations. First, although ICD codes used for defining PEP were used by several previous studies, the methodology has not been validated. And potential coding errors in the database were inevitable. However, due to the large sample size of the NIS, these errors might be random. Second, because NIS was not specifically designed for our research, we were unable to analyze some uncollected covariates, such as use of NSAIDs, aggressive hydration, and prophylactic pancreatic duct stents. The frequencies of these means to decrease PEP may differ between younger and older children (especially adolescents), and may differ in Southern and Western hospital practices, which may contribute to differences in PEP rates between these groups. Third, we were unable to obtain detailed information on ERCP procedures and characteristics of endoscopists, such as which duct was intended to be treated, whether there were difficult cannulation situations, and qualifications of endoscopists, which might affect PEP rates. Further prospective and multicenter studies are warranted. Finally, the database does not capture patients who develop PEP after discharge; therefore, we may have underestimated the prevalence of PEP in children during the study period.

In conclusion, our study shows a decreasing temporal trend and identifies multiple protective and risk factors for PEP in children from 2008 to 2017. This study may help endoscopists evaluate the relevant protective and risk factors before performing ERCP in children to prevent PEP and reduce the associated healthcare economic burden.