A method is developed for the detection of parametric change for conditionally-linear stochastic systems. Such systems are quite prevalent in biology, economics and engineering, and may be synthesized as certain bilinear stochastic systems with output feedback (possibly nonlinear) through the controls. In other words, these systems are linear in the "unmeasured states" and nonlinear in the "measured states". The previously-derived conditionally linear filter forms a convenient part of the algorithm which estimates the time of change and the parameter values. The particular motivation here is for the application to immunology and clinical practice. In this regard, a simple example is presented.