USE OF CATALYTIC MODELING TO ESTIMATE HEPATITIS A INCIDENCE IN A LOW ENDEMICITY COUNTRY: IMPLICATIONS FOR MODELING IMMUNIZATION POLICIES

To Register SMDM Homepage

Tuesday, 19 October 2004 - 2:15 PM

This presentation is part of: Oral Concurrent Session B - Public Health 2

USE OF CATALYTIC MODELING TO ESTIMATE HEPATITIS A INCIDENCE IN A LOW ENDEMICITY COUNTRY: IMPLICATIONS FOR MODELING IMMUNIZATION POLICIES

Hong Chen¹, Chris Bauch, PhD², Richard Foty, MSc¹, Ba Pham, MSc¹, and Murray Krahn, MD, MSc³. (1) Glaxo SmithKline Canada, Biomedical Data Science, Oakville, ON, Canada, (2) University of Guelph, Mathematics and Statistics, Guelph, ON, Canada, (3) University Health Network, Department of Medicine, Toronto, ON, Canada

Purpose In order to evaluate the benefits of Hepatitis A (HA) universal immunization programs, it is important to estimate the true incidence of infection. Because HA infection is significantly underreported, it is necessary to use serologic data and epidemiological methods to arrive at an estimate of true incidence.

Methods The catalytic model used (e.g., Armstrong & Bell, 2002) involved several steps: 1) obtaining seroprevalence estimates; 2) obtaining reported incidence estimates; 3) fitting a model that relates prevalence to reported incidence (adjusted), factors for underreporting, and change in incidence over time; and 4) deriving estimates of the latter factors from the fitted model.

MEDLINE and EMBASE were searched [“hepatitis” and “Canada”(n=365)] to identify studies that reported HA epidemiological data. Prevalence rates from 13 studies (a total of 28 seroprevalence studies from 55 epidemiological ones) involving persons with no known risk factors were used (Step-1). HA reported incidence data were obtained from the National Notifiable Disease Registry System (Step-2). Average reported incidence (1980-2000) was adjusted for the probability of remaining at risk (i.e., 1 - prevalence of past infection) and being asymptomatic when infected (i.e., derived from a published model estimating the age-specific likelihood of jaundice, given HA infection) (Step-3).

Results Approximately 1/3 of the Canadian-born population had serologic evidence of past infection, with a prevalence rate ranging from < 1% in ages < 10, 3% in 10-19, 1%-10% in 20-29, 7%-15% in 30-39, and above 30% in ages 40+.

The average annual reported incidence rate of HA was 6.2/100,000 (range 4.3, 9.5) and 6.8 (range 2.9, 10.8) from 1980-1989 and 1990-2000, respectively.

The model estimated an annual average of 11,244 (95% CI [8,148 - 13,591]) HA cases in Canada, 6.61 [4.79 - 7.99] times the average annual reported incidence of 1,701 cases from 1980-2000.

For a typical birth cohort (e.g., n=403,434 in 1990), the model predicts 3,114 HA cases by age 1; 16,707 by age 20; and 29,930 by age 30 (i.e., a prevalence of 7.4% in 2020).

Conclusion Reliable estimates of true incidence are important for estimating the health benefits and costs of immunization programs. In cases when there is discordance between true and reported incidence of disease, catalytic modeling provides a useful framework for synthesizing fragmented epidemiological data to derive true incidence.

See more of Oral Concurrent Session B - Public Health 2
See more of The 26th Annual Meeting of the Society for Medical Decision Making (October 17-20, 2004)