Data source
A retrospective cohort study was performed using anonymized data extracted from the Maccabi Healthcare Services (MHS) database. MHS is one of four state-mandated healthcare insurer-providers in Israel, covering approximately a quarter of the population nationwide. The MHS database contains more than two decades of longitudinal data on > 2.7 million people (1.8 million adults) since 1998, with similar demographics to the national population and a high retention rate (> 98%/year). Data are automatically collected, including comprehensive laboratory data from a single central laboratory, full pharmacy prescription and purchase data (including all medications patients have been prescribed, except ones given during hospitalization), and extensive demographic data on each patient. MHS uses the International Classification of Diseases, Ninth Revision, and Clinical Modification (ICD-9-CM) coding system as well as self-developed coding systems to provide more granular diagnostic information beyond the ICD codes. Medications are coded according to the Israeli coding system with translations to the Anatomical Therapeutic Chemical (ATC) classification system, when available. Procedures are coded using Current Procedural Terminology (CPT) codes.
Study design & population
Identification of HBV, HDV, and HBV mono-infection
The study used ICD-9 diagnosis (Dx) and laboratory data collected from 1 January 1998 to 31 December 2021 to identify patients with HBV and HDV. Chronic HBV infection was defined based on any of the following criteria: (a) ≥ 1 HBV inpatient Dx; (b) ≥ 2 HBV outpatient Dx ≥ 30 days apart; (c) ≥ 1 HBV panel result indicative of chronic infection [HBV surface antigen (HBsAg) positive and total antibody to hepatitis B core antigen (anti-HBc) positive and immunoglobulin M (IgM) anti-HBc negative and antibody to HBV surface antigen (anti-HBs) negative]; or (d) ≥ 2 hepatitis B surface antigen-positive results ≥ 90 days apart. The ‘HBV date’ was defined as the date of the earliest HBV Dx or HBsAg positive result in the database. Patients with HDV infection had (a) an HDV diagnosis (≥ 1 inpatient Dx or ≥ 2 outpatient Dx ≥ 30 days apart), (b) ≥ 1 positive HDV polymerase chain reaction (PCR) result, or (c) ≥ 2 positive HDV antibody tests ≥ 90 days apart. The ‘HDV-positive date’ was defined as the earliest positive HDV antibody or PCR test result in the database. The ‘HDV date’ was defined as the date of the earliest HDV Dx or HDV-positive date in the database. Patients with HBV mono-infection had chronic HBV infection as defined above but no HDV Dx or positive HDV test results and had ≥ 1 HDV negative test result. The ‘HDV-negative date’ was defined as the earliest HDV negative antibody or PCR test result in the database. The study period and populations of interest were defined separately by study objective.
Study design & population for objective A
To estimate the annual prevalence and incidence of HDV infection (Objective A), the population of interest included adults aged ≥ 18 years and enrolled ≥ 1 day in the most recent decade available in the database, 2010–2021, which was also the time period for which PCR test results were available in the database. Patients with incident HDV or HBV mono-infection had an HDV or HBV date, respectively, in 2010–2021 at age ≥ 18 years. Additionally, a cross-sectional analysis of patients with prevalent HDV or HBV mono-infection (HBV-Mono) was performed in 2021, the end of the study period.
Study design & population for objectives B & C
The description of HDV management and treatment (Objective B) as well as the comparative analysis of disease progression (Objective C) utilized data from 2005 to 2021, when antibody test results were available and to maximize follow up. The HDV cohort included patients with HDV infection in 2005–2021 (index date = HDV-positive date). The HBV-Mono cohort included patients with HBV mono-infection with an HDV-negative date and no HDV-positive date in 2005–2021 (index date = HDV-negative date). The HDV-negative date (rather than the HBV date) was chosen as the index date for patients in the HBV-Mono cohort to (a) avoid immortal time bias related to delays between HBV diagnosis and HDV testing, (b)minimize potential indication bias related to reasons for HDV testing referral, and (c) aid comparability to the HDV cohort when patients had several years between HBV diagnosis and HDV-negative date. All patients were aged ≥ 18 years at index date and continuously enrolled in MHS for ≥ 12 months pre-index. The follow-up period was defined from the index date through the end of the study period on 31 December 2021, the date of leaving MHS, or death, whichever came first. Common among database studies, the high level of uncertainty in the timing of HDV infections (i.e., unknown delay between HDV infection and available laboratory/diagnosis evidence) meant that a time-dependent study design was not feasible.
Study variables and definitions
Study variables used the same definitions for all objectives unless otherwise specified.
Demographics
Sociodemographic characteristics included age, sex, and residential socioeconomic status (SES) based on place of residence documented in MHS. Age was recorded at the HBV or HDV date (incidence), assessment date (prevalence), or index date (disease management and progression analyses). Residential SES used a neighborhood-level score ranked from 1 (lowest) to 10 (highest) at an individual’s place of residence [21, 22]. This residential SES measure has been widely used in other Israeli database studies and was originally derived by the Israel Central Bureau of Statistics using national census data and augmented by POINTS Location Profiling Ltd. using aggregated data on housing prices, motorization level, education, employment, and financial resources. Patient-level SES data, such as income, were not available.
Clinical characteristics and comorbidities
Data were collected for the following liver complications (Dx and/or CPT codes): cirrhosis, cirrhosis decompensation (hepatic encephalopathy, esophageal variceal bleeding or ascites), HCC, and liver transplant. To adapt to local coding practices for HCC, liver cancer was described more broadly (Dx = 155.xx). Time since HBV or HDV diagnosis was calculated as the years from HBV or HDV date, respectively, to the assessment date (prevalence) or index date (disease management and progression analyses).
Chronic diseases were identified using previously validated MHS registries, which utilize all available data in the electronic health record since 1998, for diabetes [23], chronic kidney disease (CKD) [24], hypertension [25], chronic obstructive pulmonary disease (COPD), immunosuppressive conditions and diseases (including transplants and patients taking immunosuppressive therapy), and cardiovascular disease (CVD) [26]. Cancer history was obtained from the National Cancer Registry [27] and MHS cancer registry data, which draws from pathology reports and diagnoses linked to prior approvals for cancer treatments, since the start of available data. The Charlson Comorbidity Index (CCI) was calculated based on diagnosis codes for the following conditions: CVD (including myocardial infarction, chronic heart failure, peripheral vascular disease, cerebrovascular disease), dementia, rheumatic disease, peptic ulcer disease, liver disease (mild or moderate/severe); diabetes; diabetes complications; hemiplegia; renal disease; malignancy; metastatic cancer; and HIV.
Data were also obtained for HCV (based on HCV riboneucleic acid (RNA) positive tests) and nonalcoholic steatohepatitis (NASH) or other liver complications (Dx = 571.8). Drug dependence was defined according to ICD-9 codes for dependence on opioid, cocaine, Amphetamine and other psychostimulant, hallucinogen, other specified drug dependence, combinations of drug dependence, or unspecified drug dependence. History of alcohol abuse and depression/anxiety were also described according to ICD-9 diagnosis codes (≥ 1 Dx, ever).
Other clinical characteristics included body mass index (BMI) and history of smoking status. BMI was described based on the most recent measure, up to 5 years prior, and categorized using standard cut-points [28]. History of smoking status (ever vs. never, since the start of available data) was obtained from physician reports.
Disease management and treatment
For the description of HDV management and treatment, the physician specialty linked to the first diagnosis of HDV or HBV was collected. Follow-up visits to gastroenterologists and primary care physicians (PCP) were recorded. Treatment use was recorded based on ≥ 1 dispensed treatment on/after the index date, including (pegylated)-interferon, adefovir, entecavir, lamivudine, telbivudine, and tenofovir disproxil.
Disease progression
For the comparative analysis of disease progression, disease state was described at baseline and during the follow-up period (through 2021) as a mutually exclusive hierarchical variable representing the maximum severity at the time of assessment: (1) non-cirrhotic (NC), (2) compensated cirrhosis (CC), (3) decompensated cirrhosis (DC), (4) liver cancer (LC), (5) liver transplant (LT), and, for progression states, (6) death. Progression was described as any transition from a given disease state to a higher-order (advanced) disease state (including death).
Statistical analysis
Period and annual HDV incidence and prevalence rates were calculated among (1) all patients with chronic HBV infection and (2) all MHS members (general population; enrolled ≥ 1 day in the period). For the incidence calculation, the denominator excluded individuals with HDV diagnosis before the start of the period. Average annual rates were calculated from 2010 to 2021.
Descriptive statistics were reported by cohort (HDV and HBV-Mono) and, for the disease progression analysis, by baseline disease state (NC vs. CC or more advanced stage/death (CC+)). Numbers and percentages were provided for dichotomous and categorical variables. Missing data were captured as separate levels in categorical variables (e.g., BMI category = Missing). Continuous variables were inspected for normality (Kolmogorov-Smirnov test) and summarized accordingly as mean and standard deviation (SD) or median and inter-quartile range (IQR). Comparisons between groups were tested using the Wilcoxon rank sum test, Pearson’s Chi-squared test, or Fisher’s exact test. Standardized mean differences (SMDs) were presented.
Kaplan Meier plots and Cox regression models were used for the time-to-event disease progression analysis and were stratified by baseline disease state. Censoring occurred on 31 December 2021 (end of study period) or the date of leaving MHS or death (except where death was the outcome of the analysis). Multivariable Cox models were investigated, adjusting for baseline characteristics (age, sex, SES, BMI, cancer, CVD, diabetes, CKD, hypertension, HIV, HCV PCR positive, smoking, drug dependency, and alcohol abuse) at the index date. A sensitivity analysis limited the sample to patients with HDV PCR results. For some analyses, stratification by baseline disease state was not presented in aggregate for non-NC states due to small sample sizes. All analyses were performed using R statistical software v.4.0.2. Statistical significance was defined as P < 0.05.
Ethical considerations
The study was approved by the Institutional Review Board of Maccabi Healthcare Services (protocol number 0037-22-MHS). The study was conducted according to the Declarations of Helsinki and Istanbul. Written consent was waived due to the anonymized nature of the dataset. The STROBE statement checklist is included in Supplementary Table S1.