Project site selection
We evaluated HBV case-reporting in Fujian, Hainan, and Gansu Provinces (Figure 1). Fujian is located on the eastern coast and has a population of 37 million; Gansu is located in the west and has a population of 26 million; and Hainan, is the smallest province and an island, with a population of approximately 9 million in 2010 . The estimated population prevalence of HBsAg+ is 4.4% in Gansu, 11.9% in Hainan, and 15.5% in Fujian .
In China, NHC rates hospitals from grade 1 to 3based on the level of service availability. Grade 3 hospitalsoffer the highest comprehensive level of service. In each province, we selected one grade 2 and one grade 3 hospitals for a total of six project hospitals using the following criteria: 1) the hospital had an advanced laboratory information system (LIS) with access to HBsAg test results; 2) the hospital had an electronic hospital information system (HIS) that can link the LIS to the inpatient medical record number; and 3) in 2015, the hospital reported a greater number of hepatitis B cases compared to the hospital-based provincial mean. Additionally, administrators at each hospital agreed to participate in the project. Because more than 90% of hepatitis B cases are reported from grade 2 and 3 hospitals, we elected not to include other types of health care facilities in this project.
Case inclusion criteria and data collection
We included hepatitis B cases from the six participating hospitals that were reported to NNDRS from 1 January to 31 December 2015 and had a LIS record indicating HBsAg+ test result for the same time-period. We downloaded all HBV cases that were reported to NNDRS from the six hospitals and linked these case reports to LIS data by name, gender, and birthdate. NNDRS records that could not be linked to LIS data and outpatients without record numbers were excluded from the analysis. We developed a standard abstraction form to collect medical record data from the six project hospitals for the reported HBV cases that were successfully linked to LIS. Staff from the Chinese Center for Disease Control and Prevention (China CDC) and the six hospitals used the form to collect the following information: reason for admission, admission ward (HBV-related wards included internal medicine, infectious disease, gastroenterology, and liver disease; non-HBV-related wards included surgery, pediatrics, obstetrics and gynecology, pulmonology, endocrinology, nephrology, orthopedics, neurosurgery, and traditional Chinese medicine), discharge diagnoses, clinical information including signs and symptoms of viral hepatitis, liver function tests, HBV DNA viral load, and hepatitis B sero-markers including HBV core antibody IgM (anti-HBc IgM), IgM antibody for hepatitis A virus (anti-HAV IgM) and hepatitis E virus (anti-HEV IgM), and hepatitis C virus antibody (anti-HCV). We obtained information on a patient’s classification as an acute or chronic hepatitis B case from the hospital’s information system.
Data management and analysis
We double-entered and verified data abstracted from the medical record in EpiData (version 3.1, Denmark). The resulting EpiData database was merged with the NNDRS-LIS case data using the patient medical record number and imported into in SPSS (version 23, IBM, New York, USA) for analysis. We generated descriptive statistics on HBV cases by province, hospital grade, ward type, and patient characteristics. We compared the classification of hepatitis cases reported to NNDRS with the national HBV case definitions.
According to the national case definitions, hepatitis B patients should be classified as follows:
(1) HBV carriers: HBsAg-positive patient with no signs or symptoms of liver disease and normal alanine aminotransferase (ALT) levels (≤40 IU/mL).
(2) Acute HBV: HBsAg-positive patient, who has signs or symptoms of liver disease or abnormal ALT levels (>40 IU/mL) without chronic inflammatory changes reported on abdominal ultrasound and negative hepatitis A and hepatitis E serologies, and who has clear evidence of HBV infection for less than 6 months or anti-HBc IgM positive.
(3) Chronic HBV: HBsAg-positive patient, who has signs or symptoms of chronic liver disease or abnormal ALT levels (>40 IU/mL), and with at least one of the following: previous history of HBV infection ≥6 months before, chronic inflammatory changes reported on abdominal ultrasound, or anti-HBc IgM negative.
(4) Cirrhosis: HBsAg-positive patient with liver cirrhosis as reported via abdominal ultrasound, computerized tomography (CT), or magnetic resonance imaging (MRI).
(5) Hepatocellular carcinoma (HCC): patient with liver lesion(s) that are suggestive of HCC reported via abdominal ultrasound, CT, or MRI together with alpha fetoprotein (AFP) >400 μg/mL.
(6) Non-classifiable HBV: HBsAg-positive patient with only signs and symptoms of liver disease and no additional information, or an HBsAg-positive patient with insufficient information for classification.
We generated descriptive statistics on the clinical characteristics and laboratory test results from HBV patients who were included in the final database. Patients who were reported to NNDRS as having acute HBV infection were considered to be correctly classified if they met the case definition for acute HBV based on the medical record review. Similarly, patients reported to NNDRS as having chronic HBV were considered to be correctly classified if they met the case definition of chronic HBV, cirrhosis, or HCC, based on the medical record review. The remaining patients were considered misclassified, or unable to be classified because of missing laboratory data. We assessed the percent of patients who were correctly and incorrectly classified by province, hospital grade, gender, age, ward type, and primary discharge diagnosis (HBV or other). We stratified this analysis by NNDRS reporting status (i.e., reported as an acute or chronic HBV patient). In this bivariate analysis, all hospital and patient characteristics, except for province (three categories) and age (four categories), were dichotomized. An alpha level of 0.05 was applied to assess statistical significance.
Using classification status—NDRSS case-reports that were correctly classified and case-reports that were misclassified (including unable to classify)—as our binary dependent variable, we developed a logistic regression model to estimate the crude association of each of the above characteristics (e.g., province, hospital grade, and ward type) with an incorrect HBV case classification. Characteristics that were statistically significant in the crude analysis were included in our multivariable logistic regression model. Unadjusted and adjusted odds ratios (OR) and 95% confidence intervals (CI) were calculated. Finally, using available medical record information, we described the characteristics of HBV case-reports that were misclassified as acute and chronic in NNDRS.
The China CDC Ethics Review Committee approved the project as a program evaluation. The United States Centers for Disease Control and Prevention approved the project as a routine surveillance activity. We maintained all project-related data on a secure and password-protected computer.