Estimation of optimal dynamic anticoagulation regimes from observational data: a regret‐based approach

A complication of long-term anticoagulation is that the optimal dose level varies not only between patients but over time within patients, in response to short-term changes in lifestyle. Consequently, doseage needs to be adaptive but there are as yet no accepted decision rules. Since anticoagulant use is increasing worldwide there is a need for more objective and routine procedures. In this paper, we describe an analysis of observational longitudinal anticoagulant data, aimed at determining an optimal reactive dose-changing strategy. We use the regret parameterization approach advocated by Murphy (J. R. Stat. Soc. Ser. B 2003; 65:331-366). Practical problems encountered in the implementation of the approach are discussed and illustrated.