Authors: Atu Agawu (1Department of Pediatrics; Children’s Hospital of Philadelphia; Philadelphia, PA, USA.), Christina Kanagawa (2Rowan University School of Osteopathic Medicine; Stratford, NJ, USA.), Janeline Wong (3Department of Medicine; Perelman School of Medicine at the University of Pennsylvania; Philadelphia, PA, USA.), Justine Shults (1Department of Pediatrics; Children’s Hospital of Philadelphia; Philadelphia, PA, USA.; 4Department of Biostatistics, Epidemiology, and Informatics; Perelman School of Medicine at the University of Pennsylvania; Philadelphia, PA, USA.), Chris Feudtner (1Department of Pediatrics; Children’s Hospital of Philadelphia; Philadelphia, PA, USA.), Meenakshi Bewtra (3Department of Medicine; Perelman School of Medicine at the University of Pennsylvania; Philadelphia, PA, USA.; 4Department of Biostatistics, Epidemiology, and Informatics; Perelman School of Medicine at the University of Pennsylvania; Philadelphia, PA, USA.)
Categories: Article, gallstones, children, cholecystectomy, freestanding children’s hospital
Source: Journal of pediatric gastroenterology and nutrition
Authors: Atu Agawu, Christina Kanagawa, Janeline Wong, Justine Shults, Chris Feudtner, Meenakshi Bewtra
Previous studies have shown increasing hospitalizations for pediatric cholelithiasis, but recent trends are unknown. We conducted a national study of pediatric cholelithiasis to characterize recent hospitalization rate trends.
Retrospective repeated cross-sectional analysis of pediatric (age < 18) cholelithiasis-associated hospitalizations combining data from the 2006 through 2019 Kids Inpatient Database (KIDS) releases. The primary outcome of interest was the national hospitalization rate (per 100,000 children). We examined rates stratified by age group and sex and characterized hospitalization outcomes and characteristics for pediatric cholelithiasis.
29,102 hospital records representing 42,282 gallstone-associated hospitalizations were identified. The hospitalization rate declined from 12.9 (95% CI: 12.6 – 13.2) in 2006 to 9.1 (95% CI: 8.8–9.3) in 2019. Consistent with the literature, hospitalizations occurred most often amongst teenagers (71%) and individuals with female sex (72%). The proportion of hospitalizations at freestanding children’s hospitals increased significantly (from 18.2% to 35.1%). Finally, the proportion of hospitalizations involving a potentially medical predisposing condition increased significantly.
The estimated US hospitalization rate for pediatric cholelithiasis declined by 30% between 2006 and 2019. Female patients and teenagers had the largest decline, and hospitalizations increasingly occurred at freestanding children’s hospitals. Potential explanations include potential changes in delivery of care as well as changes in population disease burden.
Gallstones (cholelithiasis; abnormal calcifications of bile) are common with an estimated lifetime prevalence of 16%. While most people with gallstones are asymptomatic, every year 2–3% develop complications (common bile duct obstruction, gallstone pancreatitis, cholecystitis, cholangitis) requiring inpatient hospital care.^(1–4)^ Hospitalizations for gallstones and their complications have increased significantly over time in adults. For example, between 1997 and 2009 national hospitalizations for acute cholecystitis in adults increased 44%.^(5)^ The putative mechanism for this increase is a rise in obesity prevalence and resultant cholesterol gallstones.
Similar to adults, national data of hospitalization rates for children with gallstones between 1997 and 2009 revealed a 34% increase.^(6)^ An important difference between pediatric and adult gallstone disease is that non-obesity-related medical predispositions, such as congenital anomalies (e.g. choledochal cyst) and hemoglobinopathies (e.g. sickle cell disease), are more common in the pediatric age-range.^(1–3)^ However, given that studies demonstrate increasing pediatric obesity prevalence over time, the proportion of pediatric gallstone disease related to a non-obesity-related predisposition would be expected to decrease.^(7–10)^
Using a cohort derived from a national sample of inpatient hospitalizations for children we sought to evaluate the population hospitalization rate for pediatric cholelithiasis. We had two specific
This was a retrospective cross-sectional study using the Kids’ Inpatient Database (KID), a large all-payer pediatric inpatient care database produced by the Healthcare Cost and Utilization Project (which was previously used to demonstrate increases in hospitalizations for gallstones between 1997 and 2009).^(6, 11, 12)^ We created the sample by combining data from 2006, 2009, 2012, 2016 and 2019 KID releases. Data available in KID includes diagnostic and procedure codes, hospital length of stay, discharge status, patient demographics, and hospital characteristics. The study was reviewed and considered exempt by the Institutional Review Board at the University of Pennsylvania and follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (Table, Supplemental Digital Content 1).^(13)^ To protect confidentiality and comply with restrictions specified in the data use agreement, individual cells with a frequency count < 10 are not displayed (raw numbers were used for statistical analysis).
Inclusion criteria included age < 18 years at the time of admission and an International Classification of Disease-Clinical Modification (ICD-CM) diagnosis code for cholelithiasis (ICD-9-CM: 574.XX, ICD-10-CM: K80.XX) at any position on the hospitalization record. These diagnosis codes were selected based on review of the literature and clinical expertise. Hospitalizations involving a transfer to another acute care hospital were included in the analysis.
The primary outcome of interest was the national rate of pediatric gallstone-associated hospitalizations (per 100,000 children). Based on previous literature on cholelithiasis epidemiology, we calculated stratified hospitalization rates based on sex (male vs. female) and age-group (newborn, infant-toddler, school-aged child, teenager).^(1–3, 14, 15)^
A secondary outcome of interest was the proportion of gallstone-associated hospital admissions that involved individuals with a potential medical predisposition for cholelithiasis. The potential medical predispositions of interest sickle cell disease, hereditary spherocytosis, cystic fibrosis, and structural biliary anomalies (e.g., choledochal cysts). Hospital admissions were characterized as involving a medical predisposition based on the presence of a diagnostic code (See Table, Supplemental Digital Content 2) at any position on the hospitalization discharge list.
KID contains several characteristics regarding each hospitalization, including demographic data (age and sex), expected primary insurance payer (Medicare, Medicaid, self-pay, no charge, other), hospital length of stay, in-hospital death, elective admission status, hospital location/teaching status (Rural, urban non-teaching, urban teaching, or missing), hospital size, and hospitalization charges. These characteristics were stratified by year to examine visual trends.
Data cleaning, analysis and figure/table creation were performed in Stata 16.1 (College Station, TX). We used two-sided hypothesis tests with a p-value of 0.05 as the criterion for statistical significance.
Due to the sampling frame and design of KID we used the svy commands in Stata to generate national estimates of pediatric gallstone hospitalizations with 95% confidence intervals. ^(16, 17)^ The hospitalization rate (per 100,000 children) was calculated by dividing the national estimates by the mid-year pediatric-aged population of interest (all children, or adjusted population of interest (e.g. age subgroup).^(18)^
We summarized patient and hospital characteristics with frequencies and percentages for categorical variables, and medians and interquartile ranges (IQR) for continuous variables.
To evaluate whether there was a trend in the annual hospitalization rate we plotted the annual hospitalization rate versus the year of hospitalization. Although not perfectly linear, the data was well approximated by a superimposed linear model with 95% confidence intervals (Figure, Supplemental Digital Content 3). To assess the relationship between hospitalization year and hospitalization rate we used the sign of the coefficient for year in the linear model to judge trend direction (positive of negative) and, tested the coefficient for significance with a Wald test. We evaluated the model fit using the coefficient of determination (R^2^, adjusted and unadjusted), and residual analysis (plotting studentized residuals vs. predicted values and vs. year). We tested for independence between year and proportion of gallstone hospitalizations with a medical predisposition using a chi-squared test.
The first sensitivity analysis evaluated the risk of distorted trends due to diagnostic coding system changes, specifically ICD-9-CM to ICD-10-CM which officially occurred in October of 2015. To perform the ITS, we identified gallstone-associated hospitalizations in a separate data set (the Nationwide Inpatient Sample) in a time range that spanned the coding switch (2012 – 2019) using the diagnosis codes as listed above.^(19)^ Again, we used the HCUP weights to generate monthly hospitalization counts (the outcome of interest).^(20)^ The primary exposure of interest for this analysis was the classification system switch which was represented as a dummy variable (ICD-9-CM coded pre-intervention (January 2012 – September 2015); or ICD-10-CM, coded 1, post-intervention (October 2015 – December 2019). We used negative binomial regression since our outcome was monthly hospitalization count with evidence of overdispersion (variance > mean), with regression logYt=β0+β1T+β2Xt+β3TXt and the following Yt – the monthly count of GBD hospitalizations at time T, Xt – indicator variable for the pre/post intervention period (0/1), T: month of observation (count starting January 2012 as month 1), β0 – coefficient for baseline count rate, β1 – change in outcome associated with a 1-month time increase (pre-intervention trend), β2 – level change following ICD coding switch, β3 – slope change following intervention with 1-month time increase (post-intervention trend). This model was proposed because we hypothesized that the impact of the coding switch could affect either or both the slope and level of monthly hospitalizations.^(21, 22)^ Finally, since local trends may be obscured by long-term trends, we fit four additional models increasingly with restrictive time windows of observation (24, 18, 12, and six months pre/post intervention) and compared the beta coefficients for the analysis including all available data and the restricted month evaluations for substantive differences.
The second sensitivity analysis was performed to estimate the risk of misclassification bias due to coding errors. For this study data was obtained from the clinical data warehouse (CDW) of an urban pediatric healthcare system that includes over 2 million clinical encounters. Data was obtained by review of clinical documentation, laboratory results, and radiology reports. Using ICD-9-CM and ICD-10-CM diagnostic codes, we randomly selected candidate inpatient hospitalizations and performed chart review to confirm the presence of cholelithiasis during the hospitalization (date range January 2010 – December 2019). Each hospitalization was categorized as 1) definite case 2) possible case, or 3) non-case. Definite cases were those that demonstrated radiologic or histologic (e.g., pathology specimen) evidence of gallstones. Possible cases were those with radiologic evidence of gall bladder sludge without gallstones. Non-cases were those with no radiologic/histologic evidence of gall bladder sludge or stones. We calculated the positive predictive value (PPV) of a diagnosis code for pediatric cholelithiasis as the sum of definite and possible cases divided by the total number of candidate cases. The validation study protocol was reviewed by the Institutional Review Board at the Children’s Hospital of Philadelphia and determined to be exempt from review.
There were 15,940,948 records initially available. After elimination of 2,799,888 records of adults (age ≥18), and 13,111,958 non-gallstone hospitalizations, there were 29,102 records remaining. After application of the HCUP discharge weights, these records represented 42,293 hospitalizations (Table 1 - 95% CI: 41,159 — 43,428). Broken down by year, the total number of hospitalizations increased from 2006 (n = 9,486, 95% CI: 9,289 — 9,683) to 2009 (n = 10,107, 95% CI: 9,921 — 10,292) but subsequently decreased with the lowest count occurring in 2019 (n = 6,644, 95% CI: 6,419 — 6,870).
The population hospitalization rate in 2006 was 12.9 hospitalizations per 100,000 children (95% CI: 12.6 – 13.2; Figure 1 Panel A) and decreased to a rate of 9.1 (95% CI: 8.8 — 9.3) in 2019, a 29.5% decrease. In regression analysis there was a significant negative association between hospitalization rate and hospitalization year (β = −0.35, 95% CI: −0.59 — −0.11, p = 0.02). The R^2^ explained more than 80% of the variance of hospitalization rate (R^2^ = 0.8742, adjusted R^2^ = 0.8322). Residual analysis ((Supplemental digital content, figure 2), suggested that assumptions of equal variance (non- heteroscedasticity) and linearity were adequate.
When stratified by sex, the hospitalization rate for females declined by 36% from 19.7 in 2006 (Figure 1 Panel B — 95% CI: 18.9 — 20.4) to 12.7 in 2019 (95% CI: 12.2 — 13.2). In contrast, the hospitalization rate for males was similar in 2006 (5.9, 95% CI: 5.6 — 6.2) and 2019 (5.6, 95% CI: 5.3 — 5.9). Teenagers had a different pattern of change to other age groups. The population hospitalization rate for teenagers declined by 36% (2006: 32.2 per 100,000 children; 95% CI: 31.0 — 33.4 | 20.7 per 100,000 children; 95% CI: 19.8 — 21.5) whereas the hospitalization rate did not significantly change for the other age groups (Figure 1 panel C).
Overall, the number of hospitalizations involving individuals with a medical predisposition decreased between 2006 and 2019 (1,225 to 1,032); however, the proportion of hospitalizations involving such individuals increased from 13.2% in 2006 to 15.5% in 2019 and there was a significant (p < 0.0003) association between hospitalization year and the proportion of cholelithiasis hospitalizations with a medical predisposition.
The median age of admitted patients was 15 years (IQR 12 — 16 years, Table 1) and the group was predominantly female (72%). When stratified by age group, the largest age group was teenagers (71%). Most individuals resided in urban areas (86%) and the expected payor for most hospitalizations was Medicaid (55%) with many commercial insurance payers (35%).
Most hospitalizations occurred in non-children’s hospitals (75%, Table 2), and most of these hospitalizations occurred in in the South (39%) and the West (29%) census regions. Additionally, most hospitalizations occurred in large hospitals (65%), and Urban teaching hospitals (73%). The median length of stay was 3 days (IQR 2 — 5 days), and the rate of in hospital death was low (0.2%). Cholecystectomy was performed during most hospitalizations (61%) with ERCP performed at a much lower rate (6.3%). When stratified by year of hospitalization, there was no significant difference in the rate of in-hospital death or length of stay. The proportion of admissions that were elective declined significantly from 23% in 2006 to 14% in 2019 (p <0.001, Table 2). The proportion of hospitalizations that included a cholecystectomy declined from 64% in 2006 to 54% in 2019 (p <0.001), and the proportion of hospitalizations that occurred in freestanding children’s hospitals increased from 18% in 2006 to 35% in 2019 (p <0.001).
A total of 300 randomly selected records were selected for review. For cholelithiasis the diagnosis codes with the highest positive predictive value were ICD-9-CM 574 (PPV: 84%) and ICD-10-CM K80 (PPV: 77%) (Table, Supplemental Digital Content 4).
There were 4,973,754 cholelithiasis-associated hospitalizations identified between 2012 and 2019. The mean monthly hospitalization count was 51,116 (standard deviation (SD) – 1,782) and was slightly higher in the ICD-9-CM era (52,495; SD: 2,148) than the ICD-10-CM era (51,116; SD: 1,782). Plotted over time the monthly hospitalization count was non-linear but continuous with no abrupt changes (Figure 2). In ITS regression analysis the baseline coefficient was nonzero (β0= 10.90 95% CI: 10.30 — 11.48, p < 0.001), however the coefficient for the coding switch was not significantly different from zero (β1= −0.104, 95% CI: −1.5782 — 1.3706, p = 0.89) and none of the other model coefficients was significantly different from zero (Table, Supplemental Digital Content 5). Results were similar when the time window around the coding switch was restricted from 24 to 6 months in 6-month increments (See Table, Supplemental Digital Content 5, which shows all the regression coefficients from the analysis).
In this national study of hospitalizations for gallstones in children, we found that the overall hospitalization rate declined by nearly 30% between 2006 and 2019 due to a decrease in hospitalization frequency. Additionally, the proportion of gallstone-associated hospitalizations that involved individuals with a possible medical predisposition increased significantly contrary to our initial hypothesis. Notably, there were significant changes in the demographic characteristics of children with gallstone-associated hospitalizations with the population of patients becoming younger and with less female predominance over time, demonstrated by the declining hospitalization rates amongst females and teenagers. There was also a significant shift in location in care with a third of gallstone-associated hospitalizations occurring in freestanding children’s hospitals. Finally, the rate of cholecystectomy during gallstone hospitalizations decreased during the study period.
Our study is partially consistent with the literature surrounding pediatric healthcare utilization at a population level. A prior study that used a combination of national readmissions data and the NIS demonstrated an overall decrease in the annual hospitalization count for children and an increased readmission rate overall for children between 2010 and 2016.^(23)^ Similarly, in our study the population hospitalization rate for gallstone disease decreased. There are two potential mechanisms that would explain our observed phenomena. It is possible that the overall pediatric gallstone disease burden is unchanged or increasing but that care is shifting from inpatient to ambulatory settings. Since the study data focused exclusively on inpatient hospitalizations this is a putative limitation of this study. A second potential mechanism for the study findings would be if gallstones are less easily diagnosed with time. This is unlikely since gallstones are typically easily identified with non-invasive abdominal ultrasound, which has increased in availability over time.
Our study diverges from previous literature regarding the overall trend of pediatric cholelithiasis and the demographics of our study population. The most recent study of pediatric cholelithiasis demonstrated an increasing hospitalization rate through 2009.^(6)^ Notably, in our study there was a small increase in the overall hospitalization rate between 2006 and 2009 followed by a subsequent decrease with a nadir in 2019. Additionally, the demographics of patients admitted changed over time in contrast to previous studies. Specifically, the hospitalization rate decreases were larger amongst teenagers and individuals with female sex (Figure 1). Since our study is disease-focused, the difference in hospitalization trends in our study is likely disease specific. However, even though hospitalizations amongst teenagers and females declined the most over the study period, these demographic groups remained the largest groups among children admitted for gallstones.
A second aspect of our study that warrants discussion is the changes in site of care observed, specifically, the increase in hospitalizations at freestanding children’s hospitals. Previous studies of pediatric healthcare utilization have demonstrated a population with increasing medical complexity and increased regionalization of care.^(24–27)^ The trend towards increased hospitalization at freestanding children’s hospitals could be related to gallstones being considered a sign of medical complexity in children (and thus referral to a children’s hospital) or could be related to the increased medical complexity of patients cared for at children’s hospitals.^(28, 27)^ The fact that we found an increased proportion of hospitalizations with children with a medical predisposition supports the concept of increased medical complexity.
There were three primary limitations to our the risk for misclassification bias, the risk for surveillance bias, and the risk of trend distortion due to the ICD-9-CM to ICD-10-CM coding switch. The validity of ICD diagnostic codes for pediatric cholelithiasis is unknown and studies in adults have found poor PPV (< 80%) for diagnostic codes for biliary colic in adults.^(29)^ However, our sensitivity analysis (albeit based on a single institution) provides some confidence that misclassification bias is unlikely to significantly inflate the estimated population hospitalization rate. With respect to the risk for surveillance bias, although hospitalized children frequently receive medical imaging as part of a diagnostic work-up, if that was driving increased detection of gallstones, we would have expected either a trend towards increased hospitalization rate or no trend.
In other studies using administrative datasets, there has been evidence of trend distortion both in pediatric (congenital anomalies) and adult (trauma) populations related switch to the ICD-10-CM diagnostic codes.^(30, 31)^ The trend distortion is hypothesized to be related to the significant increase in complexity and number of available diagnostic codes between the two systems.^(32–36)^ In our study the risk of coding-related discontinuity was expected to be small given the one-to-one mapping for cholelithiasis between ICD-9-CM (574.XX) and ICD-10-CM (K80.XX)^(35, 36)^ and our sensitivity analysis supported the assumption that there are no statistically significant distortions to hospitalization counts related to the ICD-9-CM to ICD-10-CM coding switch (Figure 2).
In conclusion, hospitalizations for pediatric gallstone disease has been declining overall, the most so for female and for teenage patients, which may be related to evolving disease epidemiology or shift towards ambulatory care. Notably, the proportion of pediatric gallstone hospitalizations occurring at freestanding children’s hospitals has increased over time, likely related to the overall increasing medical complexity of the inpatient pediatric population and warrants further investigation.