Mobile healthcare (m-health) systems have attracted a great deal of attention due to their potential to improve the quality of diagnosis, reduce medical costs and help address the challenges posed by the aging society. A generic m- health service platform has been developed and specialized to deal with emergency settings such as epileptic seizure detection, cardiac care and trauma care. The m-health infrastructure and services can be in this case described as both mission-critical and safety critical. Dynamic context-aware adaptation mechanisms are required in order to meet the stringent requirements on such mission critical applications. The use of mobile devices means that attention must be paid to resource optimization and in emergency situations timeliness of response is also a critical factor. These factors lead to critical performance requirements on the system across several dimensions. This paper presents an m-health application scenario requiring rapid response and identifies the system performance measures that are key to the success of such m-health solutions. As an extension to the m-health service platform we propose an adaptive middleware framework based on dynamic task redistribution. In particular, we present a computational model to estimate the QoS of m- health system given a particular task assignment and further to select the optimal assignment.