Trends and Disparities in Pediatric Nonalcoholic Fatty Liver Disease-Associated Hospitalizations in the United States

Abstract

Objective: 

Non-alcoholic fatty liver disease (NAFLD) represents a spectrum of disease characterized by accumulation of fat in the liver and is associated with co-morbidities linked to metabolic syndrome. The prevalence of NAFLD in children has increased in the United States over time and with marked racial differences observed in geographically limited studies. This study aims to provide a current, nation-wide analysis of temporal trends of pediatric NAFLD-related hospitalizations and associated co-morbidities as well as assess for racial/ethnic disparities.

Methods: 

A cross-sectional study was conducted using the National Inpatient Sample (NIS) from 2004 to 2018 and included NAFLD-associated hospitalizations of children ages 0–17 years of age based on ICD-9/10 diagnosis codes. Rates and patient characteristics analyzed via descriptive statistics and associations via survey logistic regression. Temporal trends assessed via joinpoint regression.

Results: 

There was an overall increase in pediatric NAFLD-associated hospitalizations with an average annual percent change (AAPC) of 6.6 with highest rates among Hispanic patients (AAPC = 11.1) compared to NH-White (AAPC = 4.1) and NH-Black (AAPC = 2.1). Analysis of race/ethnicity and NAFLD hospitalization showed an increased association in Hispanic patients (odds ratio [OR] = 1.64, 95% confidence interval [CI] = 1.51–1.77) and a decreased association in non-Hispanic (NH)-Black patients (OR = 0.49, 95% CI = 0.45–0.54) when compared to NH-White patients.

Conclusion: 

Utilizing a nation-wide database we demonstrated significant increases in NAFLD-associated hospitalizations with highest prevalence and rates seen in Hispanic patients. In addition, sex and comorbidities showed notable correlation to these hospitalization rates displaying the need for further studies on these relationships and highlights the potential for interventions aimed at high-risk groups.

An infographic is available for this article at:https://links.lww.com/MPG/C656.

What Is Known/What Is New

What Is Known

• Non-alcoholic fatty liver disease (NAFLD) is characterized by accumulation of fat in the liver and is associated with co-morbidities linked to metabolic syndrome.
• The prevalence of NAFLD in children has increased in the United States over time and with marked racial differences observed in geographically limited studies.

What Is New

• There was an overall increase in pediatric NAFLD-associated hospitalizations with an average annual percent increase of 6.6% with highest rates among Hispanic patients.
• Analysis of race/ethnicity and NAFLD hospitalization showed an increased association in Hispanic patients (odds ratio [OR] = 1.64) and a decreased association in non-Hispanic (NH)-Black patients (OR = 0.49) when compared to NH-White patients.

Non-alcoholic fatty liver disease (NAFLD) represents a spectrum of diseases characterized by the accumulation of fat within the liver that can range from steatosis and nonalcoholic steatohepatitis (NASH), to advanced fibrosis and cirrhosis ⁽¹⁾. NAFLD has been shown to be strongly associated with obesity and other markers of metabolic syndrome including insulin resistance, hypertension, and hyperlipidemia ⁽²⁾. In parallel to the global obesity pandemic, the estimated worldwide pooled prevalence of NAFLD has increased over the last decades with current US-specific prevalence studies estimated at 21–24.7% of the adult population ⁽³⁾.

As in the adult population, NAFLD in children is increasingly being recognized as the most common form of pediatric chronic liver disease ⁽⁴⁾. While difficult to accurately assess the true burden of disease, post-humous pathological examination revealed prevalence of fatty liver to be 9.6% in normal/underweight children versus 38% in obese children, with prevalence increasing with age and differing significantly by race and ethnicity with higher rates seen in children of Hispanic ethnicity ⁽⁵⁾. Analysis of temporal trends of pediatric NAFLD-related hospitalizations in the United States demonstrated that from 1986 to 2006 the prevalence of disease increased from 0.9 to 4.3/100,000 ⁽⁶⁾. Similarly, a California-specific database found an increase over time in pediatric NAFLD-related hospital admissions versus other pediatric liver diseases; and noted striking ethnic and racial disparities with a greater burden of disease again noted in Hispanic patients ⁽⁷⁾.

As previous nation-wide study was conducted almost a decade and half ago ⁽⁵⁾, this study aims to provide a current analysis of the temporal trends of nation-wide pediatric nonalcoholic fatty liver disease-related hospitalizations and associated co-morbidities. In addition, we aim to assess racial/ethnic and socio-economic disparities with respect to hospitalization patterns on a nation-wide scale.

METHODS

Study Design and Data Collection

We conducted a retrospective study using the National Inpatient Sample (NIS), the largest all-payer database of hospital admissions in the United States of America (USA) from January 1, 2004 through December 31, 2018. The NIS contains discharge data from more than seven million hospitalizations annually (approximately 35 million when weighted) and constitutes a 20% stratified sample of all US non-federal, non-rehabilitation, short-term community hospitals ⁽⁸⁾. Specifically, for each hospitalization, up to 30 diagnoses and procedures are able to be captured using International Classification of Diseases, Ninth Edition, Clinical Modification (ICD-9-CM) through the third quarter of 2015, and International Classification of Diseases, Tenth Edition, Clinical Modification (ICD-10-CM) codes from October 1, 2015 to the present. The NIS lacks patient identifiers that allow the linkage of hospitalizations for the same patient over time; as such, the unit of analysis is the hospitalization, or discharge, as opposed to the individual person. Since NIS data are publicly available and de-identified, this study was exempt by the Baylor College of Medicine Institutional Review Board. The study sample consisted of hospitalizations of patients’ ages 0−17 years of age.

NAFLD associated hospitalizations were identified based on the diagnoses codes—ICD-9-CM: 571.8x; ICD-10-CM: K76.0, K75.81. Comorbidities included in this study were obesity (ICD-9-CM: 278.00x, 278.01x; ICD-10-CM: E66.01, E66.09, E66.9), insulin resistance (ICD-9-CM: 277.7; ICD-10-CM: E88.81), hypertension (ICD-9-CM: 401.9; ICD-10-CM: I10.x) and dyslipidemia (ICD-9-CM: 272.4; ICD-10-CM: E78.5).

For each hospitalization, the NIS also captures individual-level sociodemographic and hospital characteristics. We categorized patients’ ages in groups of 0–4 years, 5–9 years, 10–14 years and 15–17 years. Ethnicity was initially stratified by reported ethnicity (Hispanic, non-Hispanic); and the non-Hispanic group was subdivided into White, Black, or other. The primary payer for the hospitalization was classified into Medicare, Medicaid, private, self-pay, and other (including under/uninsured). As a proxy for socioeconomic status, the Healthcare Cost and Utilization Project (HCUP) provides zip-code-level estimates of median household income, grouped into quartiles based on the patient's residence. Hospital factors included census region (Northeast, Midwest, South, West), bed size (small, medium, large), and hospital type (rural, urban-nonteaching, urban-teaching).

Statistical Analysis

We used descriptive statistics to capture the frequency distribution of patient and hospital characteristics among NAFLD and non-NAFLD related hospitalizations. In each patient and hospital characteristic subgroup, we also calculated the prevalence of NAFLD per 10,000 hospitalizations. We also explored temporal trends of NAFLD among the study cohort to asses any changes in rates of hospitalizations by race/ethnicity. To assess temporal trends in the national incidence of inpatient deaths among all NAFLD hospitalizations, we used joinpoint regression, a statistical modeling approach specifically designed to detect changes in the rates of events over time ⁽⁹⁾. Our model assumed 0 as the minimum and 2 as the maximum number of joinpoints based on the total data points. The model started by assuming zero joinpoints (straight line) to fit the model and tested whether more joinpoints were statistically significant and must be added to the model (up to the maximum number). The tests of significance used a Monte Carlo Permutation method. The final model containing an optimal number of joinpoints generates the Average Annual Percentage Change (AAPC) and 95% confidence interval (CI) of the rates of outcome (in this case, NAFLD hospitalizations) over the study period. The outcome in our case, was log transformed to smoothen the trend. Next, we calculated the prevalence of each of the comorbidities among NAFLD hospitalizations by race/ethnicity. To examine the top reasons for hospitalizations among pediatric NAFLD patients, we enlisted the top 10 principal diagnoses with their frequencies among NAFLD hospitalizations during the study period. Furthermore, we used survey logistic regression to generate adjusted odds ratios (ORs) and 95% CIs that measured the associations between various patient sociodemographic characteristics, hospital factors and comorbidities versus the likelihood of NAFLD hospitalization. Sensitivity analysis was conducted with and without missing data in the model variables. This was done to test the generalizability of the association model. We adopted a 5% type I error rate for calculation of CIs, and used appropriate survey weighting to generate national prevalence estimates considering the complex sampling design of the NIS. Statistical analyses were performed using R (version 3.6.1) and RStudio (Version 1.2.5001) while the trends analyses were run using Joinpoint Regression Program, version 4.7.0.0 (National Cancer Institute).

RESULTS

Temporal and Racial/Ethnic Trends in Pediatric Non-Alcoholic Fatty Liver Disease-Associated Hospitalizations

Analysis of rates per 10,000 hospitalizations during our study period showed an overall increase in pediatric NAFLD-associated hospitalizations with an average annual percent change (AAPC) of 6.6 (Fig. 1). Further delineation of trends by race/ethnicity revealed disproportionate increase in rates among Hispanic patients (AAPC = 11.1) compared to NH-White (AAPC = 4.1) and NH-Black (AAPC = 2.1) (Fig. 1).

FIGURE 1: Trends in rates of NAFLD hospitalizations- overall and by race/ethnicity: 2004–2018. NAFLD = non-alcoholic fatty liver disease.

Patient Characteristics Associated With Non-Alcoholic Fatty Liver Disease-Related Hospitalizations

A total of 42,218 NAFLD-associated hospital admissions in the pediatric population was captured during the study period, accounting for 0.05% of all admissions. Patient socio-demographic and hospitalization characteristics are displayed in Table 1. The rate of NAFLD-associated hospitalizations was noted to be highest in 10–14 years age group and 15–17 years age group with rates of 25.19 and 22.92/10,000 hospitalizations, respectively, in contrast to the 5–9 and 0–4 year age groups with rates of 11.57 and 1.43/10,000 hospitalizations, respectively. Rates based on sex showed a slight male predominance (4.97/10,000 hospitalizations vs 4.56/10000 hospitalizations for female patients). Analysis by race/ethnicity demonstrated the highest rates among Hispanic patients with a rate of 8.54/10,000 hospitalizations compared to 4.34/10,000 hospitalizations for NH-White patients, and 3.13/10,000 hospitalizations for NH-Black patients.

TABLE 1 - Patient socio-demographic and hospitalization characteristics among those with NAFLD

Patient characteristics	Total (n = 88,567,161)	NAFLD (n = 42,218, % = 0.05)	Prevalence of NAFLD per 10,000 hospitalizations
Age
0–4 y	7,3020,492	10,452	1.43
5–9 y	4,455,629	5159	11.57
10–14 y	4,926,627	12,443	25.19
15–17 y	6,164,413	14,164	22.92
Sex
Male	45,170,399	22,445	4.97
Female	43,158,141	19,686	4.56
Missing	238,620	86	3.60
Race
NH-White	37,950,760	16,476	4.34
NH-Black	11,060,347	3466	3.13
Hispanic	16,435,436	14,042	8.54
Other	8,383,544	3763	4.49
Missing	14,737,073	4470	3.03
Discharge status
Routine	83,657,569	370,87	4.43
Transfer	2,094,197	2171	10.36
Died	328,040	485	14.76
DAMA	66,125	34	5.14
Other/missing	2,421,229	2440	10.07
Zip income quartile
Lowest quartile	21,576,551	11,847	5.49
Second quartile	18,671,415	9832	5.26
Third quartile	17,733,697	8485	4.78
Highest quartile	15,949,680	6661	4.17
Missing	14,635,817	5392	3.68
Primary payer
Medicare	3,803,978	2138	5.62
Medicaid	32,578,565	16,002	4.91
Private insurance	32,896,844	12,761	3.88
Self-pay	5,273,924	2599	4.93
Other/missing	14,013,849	8717	6.22
Hospital characteristics
Hospital region
Northeast	15,007,189	6054	4.03
Midwest	18,888,386	6907	3.66
South	33,984,654	16,857	4.96
West	20,686,931	12,399	5.99
Hospital bed size
Small	12,010,231	5393	4.49
Medium	23,851,959	10,104	4.23
Large	52,348,995	26,447	5.05
Missing	355,975	274	7.69
Hospital location and teaching status
Rural	8,543,977	1215	1.42
Urban non-teaching	27,402,121	5189	1.89
Urban teaching	52,265,088	35,540	6.80
Missing	355,975	274	7.69

All bivariate associations were statistically significant with P < 0.01.DAMA = discharged against medical advice; NAFLD = non-alcoholic fatty liver disease; NH = non-Hispanic.

In regards to hospitalization outcomes, pediatric NAFLD-associated diagnosis rates were highest in hospitalizations that resulted in patient death with a rate of 14.76/10,000 hospitalizations, and lowest in routine hospitalizations (rate = 4.43/10,000 hospitalizations, Table 1). The highest rate of NAFLD-associated hospitalizations was found in patients with the lowest patient zip code income quartile, with prevalence decreasing as income quartiles increased (5.49, lowest quartile; 5.26, second quartile; 4.78, third quartile; 4.17, fourth quartile). Rates by primary payor showed highest rates in other/missing (including uninsured patients), followed by Medicare-covered patients, those on self-pay, Medicaid, and finally private insurance with rates of 6.22, 5.62, 4.93, 4.91, and 3.88/10,000 hospitalizations, respectively.

NAFLD-associated hospitalization rates by hospital characteristics displayed the greatest burden in the US geographical West region, followed by the South, Northeast, and Mid-west regions with rates of 5.99, 4.96, 4.03, and 3.66/10,000 hospitalizations, respectively (Table 1). In addition, rates were higher in large hospitals (by bed size), followed by medium and then small hospitals with rates of 5.05, 4.49, and 4.23, respectively. Finally, we observed higher rates in urban teaching hospitals compared to urban non-teaching and rural hospitals (6.80, 1.89, and 1.42/10,000 hospitalizations, respectively (Table 1).

Associated Comorbidities

Identification of patient co-morbidities revealed the highest rates of association with obesity co-diagnosis (Fig. 2). The overall rate of co-morbidity diagnosis of obesity was 24.3% among patients with NAFLD-related hospitalizations. When differentiated by race/ethnicity, Hispanic patients were noted to have a higher rate of co-morbid obesity compared to NH-White and NH-Black patients with rate of 32.3% versus 20.6% and 17.2%, respectively. Rates of co-morbid insulin resistance (overall 3.2%) and dyslipidemia (overall 3.3%) were highest in Hispanic patients (3.9% and 3.8% respectively). The co-morbid diagnosis of hypertension was found in 7.8% of patients overall, with the highest rate observed in NH-Black patients, 10.8%, followed by Hispanic patients, 8.2%, and NH-White patients, 7.0% (Fig. 2).

FIGURE 2: Rates of comorbidities among NAFLD patients, overall and by race/ethnicity. NAFLD = non-alcoholic fatty liver disease.

Top Reasons for Hospitalizations in Non-Alcoholic Fatty Liver Disease Patients

Examination of the top 10 principal diagnoses in patients with a NAFLD-associated hospitalization demonstrated a mixture of both acute and chronic conditions including hepatobiliary, oncologic, infectious processes (Table 1, Supplemental Digital Content, https://links.lww.com/MPG/C654).

Factors Associated With Non-Alcoholic Fatty Liver Disease-Associated Hospitalizations

The results from the association model containing missing data in the variables are shown in Table 2, Supplemental Digital Content, https://links.lww.com/MPG/C655 (n = 88,567,161). Since the results were very similar to the ones obtained from the model where missing data was removed (Table 2, n = 5,082,777), we have only reported and discussed results from the model where missing data were excluded. Utilizing a logistic regression model to identify patient characteristics associated with NAFLD-related hospitalization, age was noted to have significant effect with higher adjusted odds ratios in older pediatric patients compared to the 0–4 years age group (Table 2). A positive dose-response pattern was observed between patient age and the likelihood of NAFLD-related hospitalization (ORs of 5.88, 10.41, and 10.52 for age groups 5–9 years, 10–14 years, and 15–17 years, respectively). In addition, there was a slight male predominance noted in NAFLD-related hospitalizations, with female sex associated with OR = 0.8, 95% CI: 0.76–0.84 compared to male sex. Analysis of the effect of race/ethnicity showed an increased association noted in Hispanic patients (OR = 1.64, 95% CI = 1.51–1.77) and a decreased association in NH-Black patients (OR = 0.49, 95% CI = 0.45–0.54) when compared to NH-White patients.

TABLE 2 - Factors associated with NAFLD hospitalizations using adjusted survey logistic regression model

	OR (95% CI)	P value
Patient characteristics
Age
0–4 y	Reference
5–9 y	5.88 (5.25–6.59)	<0.01
10–14 y	10.41 (9.24–11.73)	<0.01
15–17 y	10.52 (9.34–11.86)	<0.01
Sex
Male	Reference
Female	0.80 (0.76–0.84)	<0.01
Race
NH-White	Reference
NH-Black	0.49 (0.45–0.54)	<0.01
Hispanic	1.64 (1.51–1.77)	<0.01
Other	1.16 (1.06–1.27)	<0.01
Discharge status
Routine	Reference
Transfer	1.55 (1.39–1.73)	<0.01
Died	3.39 (2.72–4.22)	<0.01
DAMA	0.49 (0.23–1.04)	0.06
Other/missing	2.24 (1.96–2.56)	<0.01
Zip Income quartile
Lowest quartile	Reference
Second quartile	1.11 (1.02–1.22)	0.02
Third quartile	1.16 (1.07–1.27)	<0.01
Highest quartile	1.07 (0.99–1.16)	0.08
Primary payer
Medicare	Reference
Medicaid	0.94 (0.82–1.06)	0.31
Private Insurance	0.87 (0.77–0.99)	0.03
Self-Pay	0.90 (0.76–1.06)	0.21
Hospital characteristics
Hospital region
Northeast	Reference
Midwest	1.13 (0.98–1.31)	0.09
South	1.52 (1.33–1.73)	<0.01
West	1.66 (1.46–1.89)	<0.01
Hospital bed size
Small	Reference
Medium	1.03 (0.9–1.19)	0.68
Large	1.14 (1.01–1.28)	0.03
Hospital location and teaching status
Rural	Reference
Urban non-teaching	1.25 (0.97–1.61)	0.08
Urban teaching	3.13 (2.52–3.89)	<0.01
Comorbidities
Obesity
No	Reference
Yes	9.52 (8.85–10.24)	<0.01
Insulin resistance
No	Reference
Yes	6.20 (5.21–7.38)	<0.01
Hypertension
No	Reference
Yes	2.78 (2.5–3.10)	<0.01
Dyslipidemia
No	Reference
Yes	4.76 (4.02–5.64)	<0.01

CI = confidence interval; DAMA = discharged against medical advice; NH = non-Hispanic; OR = odds ratio.

Analysis by discharge status demonstrated a significant increase in NAFLD-related hospitalizations resulting in death compared to routine discharge (OR = 3.39, 95% CI = 2.72–4.22, Table 2). Zip code income quartile had a modest impact, with a slight increase in NAFLD-related hospitalizations in second and third income quartiles compared to the lowest quartile (OR of 1.11, 1.16, respectively). Interestingly, the highest income quartile did not have significantly increased NAFLD-related hospitalizations compared to the lowest quartile (OR 1.07, P = 0.08). Analysis of primary payer status did not reveal significant findings for Medicare, Medicaid, or Self-pay status. However, private insurance payors were noted to have lower likelihood of NAFLD-related hospital admissions (OR = 0.87, 95% CI = 0.77–0.99).

Analysis of hospital characteristics demonstrated geographical differences with increased NAFLD-related hospitalizations noted in the South (OR = 1.52) and West (OR = 1.66) geographical regions. In addition, NAFLD-related hospitalizations were more pronounced in hospitals classified as large compared to small ones (OR = 1.14, 95% CI = 1.01–1.28), as well as in hospitals classified as urban teaching versus rural hospitals (OR = 3.13, 95% CI = 2.52–3.89).

All patient co-morbidities investigated revealed significant positive associations with NAFLD-related hospitalization, with the greatest likelihood noted for obesity (OR = 9.52, 95% CI = 8.85–10.24), followed by insulin resistance, dyslipidemia, and hypertension (OR of 6.20, 4.76, and 2.78, respectively) (Table 2).

DISCUSSION

Utilizing a nation-wide database covering the period 2004–2018, we demonstrated significant increases in pediatric hospitalizations associated with a NAFLD diagnosis over time. In addition, we report racial/ethnic trends for these rates at a national level. Hispanic patients had both the highest prevalence and the highest increase in rates of pediatric NAFLD-associated hospitalizations over time. This is in line with previous studies that have focused on more geographically limited areas ⁽⁷⁾ and demonstrated the persistence of the racial/ethnic disparities at the national level.

Rates of NAFLD-associated hospitalizations showed slight predominance of male sex, as well as higher rates in older pediatric age groups, with the highest rates in the adolescent period as has been suggested with previous cross-sectional studies ⁽⁵⁾. This is contrast to a previous study of earlier time period in United States, which demonstrated higher incidence associated with female sex ⁽⁶⁾. These differences highlight the difficulty in studying such a heterogenous disease and is perhaps explained by differing epidemiological instruments. As expected, co-morbidities related to metabolic syndrome showed a positive association with NAFLD-associated hospitalizations, with highest observed rates of co-morbid obesity. We observed a predominance of co-morbid obesity in the Hispanic population versus NH-White and NH-Black patients.

Income zip quartile had modest influence on odds of NAFLD-associated hospitalizations with the lowest and highest quartile showing decreased rates. Interestingly, payor status was found to only be significant by comparison of Medicare versus Private insurance with decreased rates noted in private insurance group. Future studies would be necessary to further interrogate the interplay between NAFLD-associated hospitalizations and socioeconomic status. In addition, analysis of hospital characteristics showed increased rates present in large urban medical centers as well as increased rates based on geographical regions with the West and South regions of the United States. While the increased rates noted in the large urban centers likely follow the observed nation-wide trend of concentration of pediatric care in tertiary centers ⁽¹⁰⁾, it is unclear for the other examined characteristics if these represent true increases in prevalence, or rather increased NAFLD awareness and diagnosis.

Our study bears some limitations given the epidemiological survey used as well as difficulties inherent in the diagnosis of NAFLD itself. By utilizing the National Inpatient Sample database, we were only able to capture hospitalizations that carried a diagnosis of NAFLD and, given the under-recognized nature of the disease, we likely are failing to capture the true prevalence among all hospital admissions. In addition, by using ICD9/10 coding we relied on the clinical diagnosis of NAFLD and, given the complexities in the diagnosis of pediatric NAFLD, we were unable to ascertain a clinical standard of diagnosis. Some of the increase in rates noted might have been complicated by increased over-all awareness of pediatric NAFLD (or conversely, with underdiagnosis) and thus, different odds of the diagnostic code being added to an admission. In addition, the database does not reflect the clinical significance of the NAFLD diagnosis in the patient admission, which is whether it was an incidental finding versus complicating factor in the admission. Given the complexities of the diagnosis of NAFLD and likely center-to-center variation in regards to coding, further studies might be warranted to expand upon the outcomes associated with a NAFLD diagnosis.

Strengths of the study include the large sample size of the dataset, representing a nationally-representative sample of pediatric inpatient hospitalizations thereby making our study findings generalizable. Furthermore, to our knowledge, this is the first ever study to evaluate trends, comorbidities and factors associated with pediatric NAFLD hospitalizations in the US. While focusing on NAFLD-related hospitalizations does not capture the true population prevalence, it allows for evaluation of the burden of disease in the healthcare system. In light of the increasing recognition of the health care burden of pediatric patients with chronic diseases and increasing medical complexity of those patients ⁽¹¹⁾, this increase may have important implications in regard to patient outcomes and health care costs.

In conclusion, our data presents an important and troubling trend, namely, the rapidly increasing rate of pediatric hospitalizations associated with an NAFLD diagnosis nationally. This has important short and long-term implications, including contributing to increasing hospital admission complexity and costs, as well as the long-term health concerns associated with long standing NAFLD. Importantly, the noted racial and ethnic disparities highlight the potential for targeted interventions aimed at high risk groups.