A risk analysis based on a two-stage delayed diagnosis regression model with application to chronic disease progression

This paper presents a two-stage regression model for quantifying different stages of a disease progression with delayed diagnosis time and for identifying the risk factors associated with each stage. Conventional chronic disease progression studies reported replied on the assumption that the time of the confirmation of a disease state by diagnosis is the start time of this disease state. Clearly this will lead to biased estimates of progression since the disease state should have already occurred before the diagnosis, but the true occurrence time is unknown. This later confirmation is called the delayed diagnosis in this paper and a delay-time modelling procedure is developed for the identification of the unknown stages of progression. A hazard-based regression model is also proposed for a further risk analysis. We apply the developed methods to hepatitis C data and the analysis shows that considering the delayed diagnosis significantly improved the model fit in comparison with the conventional model. We also find that the risk factors associated with each stage are more significant, particularly in the second stage of progression, than those based on the conventional model. We conclude that such delayed phenomena in diagnosis should be taken into account when modelling the chronic disease progression process and conducting related risk analysis.

► Developed a two-stage delay-time-based regression model for chronic disease progression.
► Developed a novel technique for the two-stage delayed model for model diagnostic.
► Illustrated the correlated two-stage model to hepatitis C data from Edinburgh cohort study.
► Identified the risk factors associated with each stages of progression of hepatitis C.
► Developed “Hazard Ratio Matrix Reference” for the risk analysis of two-stage progression of hepatitis C disease.