Hepatitis D-Associated Hospitalizations in the United States: 2010–2018

Paul Wasuwanich; Catherine W. Striley; Saleem Kamili; Eyasu H. Teshale; Eric C. Seaberg; Wikrom Karnsakul


J Viral Hepat. 2022;29(3):218-226. 

In This Article

Materials and Methods

Study Population

After receiving approval from the institutional review board of the University of Florida, we utilized data from the National Inpatient Sample, a database on inpatient stays and hospital discharges in the United States from the Healthcare Cost and Utilization Project (HCUP). The National Inpatient Sample is a stratified sample of approximately 20% of hospitals participating in HCUP (2011 and prior) and a stratified sample of approximately 20% of more than 35 million discharges annually from all hospitals participating in HCUP (2012 and later).[16,17] An approximately six-year interval of 1 January 2010 to 1 September 2015 was selected for full analysis due to the consistent use of the International Classification of Diseases, Ninth Revision (ICD-9) diagnosis codes which was retired on 1 September 2015 and replaced with the International Classification of Diseases, Tenth Revision (ICD-10) diagnosis codes. The interval between September 1st of 2015 to the end of 2018 (the most recent National Inpatient Sample database available) was also analysed, but the analysis was separated from the 2010–2015 period and was limited to hospitalization rates and trends due to the large difference in ICD diagnosis code versions.

Data Extraction

Data were extracted for analysis of two main groups of interest: hepatitis D and hepatitis B without HDV infection (HBV only). The hepatitis D group was defined by ICD-9 diagnosis codes 070.21, 070.23, 070.31, 070.33, 070.42 and 070.52, and by ICD-10 diagnosis codes B16.0, B16.1, B17.0 and B18.0. The HBV only group was defined by ICD-9 diagnosis codes 070.20, 070.22, 070.30 and 070.32, and by ICD-10 diagnosis code B16.2, B16.9, B18.1, B19.1, B19.10 and B19.11. For each of the three groups, data on in-hospital deaths were obtained.

For the 2010–2015 period, we extracted demographic and geographic data including age, sex, race/ethnicity, region of hospital and type of hospital (rural, urban non-teaching or urban teaching). The following clinical data were obtained: length of hospital stay, deaths, liver failure, acute kidney failure, other organ failure (i.e. heart, lung and brain), hepatic neoplasm, non-alcoholic cirrhosis, alcoholic cirrhosis, biliary cirrhosis, portal hypertension, hepatic encephalopathy, ascites, jaundice, chronic kidney disease, anorexia, haematemesis, thrombocytopenia, coagulopathy, HIV infection, HCV infection, history of injection drug use, diabetes, solid organ transplantation, pregnancy, maternal adverse events and foetal/neonatal adverse events. History of injection drug use was determined indirectly. Hospitalizations with the diagnosis of dependence of drugs commonly injected intravenously including opioids, sedatives, amphetamines, hallucinogens or combinations of these were classified as having a history of injection drug use.

Statistical Analysis

Hospitalization rates were analysed for trends in the 2010 to 2015 period and the 2015 to 2018 period using Poisson regression and reported as incidence rate ratios (IRR) per year. Hospitalization rate calculations incorporated discharge weights provided by HCUP to provide the true number of hospitalizations, compensating for the 20% sampling of the database. Changes in discharge weights prior to 2012 due to database redesign were accounted for in the hospitalization rates. The denominator used in the calculation of hospitalization rates was the total United States population residing within the country on July 1st of the relevant year. These data were extracted from the database of the United States Census. Case-fatality rates were calculated as the percentage of deaths that resulted from all hospitalizations for a given disease. Quantitative data were tested for normality using the Kolmogorov–Smirnov test. The non-normal data were summarized using the median and interquartile range (IQR) and compared using the Mann–Whitney U test. Normally distributed data were analysed with the Student's t-test. Frequencies were compared using the chi-squared test. Risk factors for mortality among hepatitis D hospitalizations were analysed by logistic regression in both univariate (crude) and multivariate (adjusted) models. The dichotomous outcomes were discharged alive versus death. Results of the logistic regression were reported by odds ratio (OR) with 95% confidence intervals (CIs).

All results reported were weighted, using discharge weights provided by HCUP in order to present the true number of hospitalizations in the United States. Results for a category that contain 10 or fewer hospitalizations but greater than zero hospitalizations were displayed as ≤10 due to the data use privacy policy of HCUP. p-values were calculated from weighted data using actual numbers regardless of whether ≤10 hospitalizations were reported.

All reported p-values were two-tailed p-values. Statistical significance was defined as p < .05. Statistical calculations were performed using STATA® software (StataCorp., 2013. Stata Statistical Software: Release 13; StataCorp LP.) and R program (R Core Team, 2020. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/).[18,19]