Alternating recurrent event data arise frequently in clinical and epidemiologic studies, where 2 types of events such as hospital admission and discharge occur alternately over time. The 2 alternating states defined by these recurrent events could each carry important and distinct information about a patient's underlying health condition and/or the quality of care. In this paper, we propose a semiparametric method for evaluating covariate effects on the 2 alternating states jointly. The proposed methodology accounts for the dependence among the alternating states as well as the heterogeneity across patients via a frailty with unspecified distribution. Moreover, the estimation procedure, which is based on smooth estimating equations, not only properly addresses challenges such as induced dependent censoring and intercept sampling bias commonly confronted in serial event gap time data but also is more computationally tractable than the existing rank-based methods. The proposed methods are evaluated by simulation studies and illustrated by analyzing psychiatric contacts from the South Verona Psychiatric Case Register.
Bibliographical noteFunding Information:
The authors gratefully acknowledge the use of the anonymous South Verona, Italy, psychiatric case register data for illustrating the proposed method, provided by Dr Michele Tansella. The authors also thank the University of Minnesota Supercomputing Institute and the Texas Advanced Computing Center (TACC) at The University of Texas at Austin for providing computing resources that have contributed to the research results reported within this paper. This research was supported by NCI R01CA193888 to Huang, NSF SES-1659328, DMS-1712717, and NSA H98230-17-1-0308 to Xu, and NCI R03CA187991 and NIMH R03MH112895 to Luo.
- accelerated failure time model
- alternating renewal process
- gap times
- recurrent events