OBJECTIVES:

There is a shortage of transplantable donor hearts to treat end-stage heart failure (ESHF). Further, high-cost left ventricular assist devices (LVADs) are restricted to last resort therapy. Two discrete event simulation (DES) models were developed; DES with queuing to represent restricted allocation of LVADs and the waiting list for heart transplants (HTx); and DES no queuing to cross-validate with a Markov model. The cost-effectiveness of various policies was assessed: ESHF policy without LVADs (Policy A); the current ESHF policy with restricted LVADs (Policy B), less restricted LVADs policy (Policy C) and less restricted HTx (Policy D).

METHODS:

The DES models were built using AnyLogic®. The DES model with queuing included a ‘match block’ which pairs a stream of waitlisted patients and the matching donor organ by blood type and weight. Time-to-event probabilities and costs were obtained from an observational dataset at St. Vincent’s Hospital Sydney, Australia. The incremental cost per quality-adjusted life-year was calculated over a 20-year time horizon.

RESULTS:

Comparing the two DES models, there was a lower proportion of patients receiving an LVAD under Policy B (39% DES with queuing vs. 85% DES no queuing). The predicted proportion of HTx differed between DES with queuing (Policy D> Policy A > Policy B > Policy C), compared to DES no queuing (Policy B=Policy C > Policy D > Policy A). Patients spent more time post-HTx in the DES with queuing model than the other models. The DES models with queuing produced more favourable ICERs compared to the Markov model.

CONCLUSIONS:

The DES with queuing model captured competition for LVADs and HTx and the match between the patient and the donor organ. The outputs (utilisation and waiting time) from the DES with queuing model better reflected real-world data than the DES no queuing or Markov model.