ACSPRI Conferences, RC33 Eighth International Conference on Social Science Methodology

Font Size:  Small  Medium  Large

Projecting probable missing hepatitis C reinfection data given a set of observations and intervals between study visits

Rachel Sacks-Davis

Building: Law Building
Room: Breakout 7 - Law Building, Room 028
Date: 2012-07-12 11:00 AM – 12:30 PM
Last modified: 2011-12-20

Abstract


Hepatitis C infection followed by spontaneous clearance and subsequent reinfection has been observed in people who inject drugs (PWID). The phenomenon is of interest for hepatitis C vaccine development, but the population at-risk (PWID) are difficult to follow longitudinally. A person can become reinfected with hepatitis C and clear spontaneously without being aware that they were reinfected. Indeed, unless blood is taken and tested for hepatitis C during the reinfection, there is no way of knowing that an individual who clears a hepatitis reinfection episode was reinfected. A number of longitudinal reinfection studies have been undertaken in PWID globally with considerable variation between study findings. We previously showed that a large proportion of this variation is likely to be due to variation in test intervals between studies, whereby studies with large test intervals miss reinfection episodes and therefore underestimate reinfection rates and spontaneous clearance proportions. We fit a markov chain model of hepatitis C reinfection and spontaneous clearance to data from a longitudinal study of PWID undertaken in Melbourne, Australia. Bayesian post estimation techniques are used to project likely reinfection rates, proportion of reinfections with spontaneous clearance, duration of reinfection, proportion of time infected, and proportion of participants with long-term chronic infection after one to ten years. With the study data used, it was not possible to derive precise estimates of reinfection rate and proportion of reinfections that spontaneously clear unless duration of reinfection was specified a priori. However, proportion of reinfections that spontaneously clear was high. When the duration of reinfection was not specified a priori, the credible intervals for reinfection rate, proportion of reinfections that spontaneously clear and duration of reinfection were 0.05-0.34 new infections per person-month, 0.91-0.99, and 0.25-1.16 months, respectively. Credible intervals for proportion of time infected and proportion of participants with long-term chronic infection after ten years were 0.05-0.13 and 0.29-0.66, respectively. Using simulated data we demonstrate that precise estimates could also be derived for reinfection rate, proportion of reinfections that spontaneously clear and duration of reinfection using these methods if the number of observations were increased. The required test interval is discussed. These methods can be used to estimate data on other types of sociological and biological events that occur between study visits.