This paper combines functional principal component analysis with autoregressive frailty models to jointly analyze sparse longitudinal measurements and recurrent events. The approach accommodates changing relationships over time and is evaluated through simulations and applications to blood-pressure and kidney-stone studies.