Most kinematic structures in robot architectures for medical tasks are not optimal. Further, the workspace and payloads are often oversized which results in high product prices that are not suitable for a clinical technology transfer. To investigate optimal kinematic structures and configurations, we have developed an adaptive simulation framework with an associated workflow for requirement analyses, modelling and simulation of specific robot kinematics. The framework is used to build simple and cost effective medical robot designs and was evaluated in a tool manipulation task where medical instruments had to be positioned precisely and oriented on the patient's body. The model quality is measured based on the maximum workspace coverage according to a configurable scoring metric. The metric generalizes among different human body shapes that are based on anthropometric data from UMTRI Human Shape. This dexterity measure is used to analyze different kinematic structures in simulations using the open source simulation tool V-REP. Therefor we developed simulation and visualization procedures for medical tasks based on a patchwork of size-variant anatomical target regions that can be configured and selectively activated in a motion planning controller. In our evaluations we compared the dexterity scores of a commercial lightweight robot arm with 7 joints to optimized kinematic structures with 6, 7 and 8 joints. Compared to the commercial hardware, we achieved improvements of 59% when using an optimized 6- dimensional robot arm, 64% with the 7-dimensional arm and 96% with an 8-dimensional robot arm. Our results show that simpler robot designs can outperform the typically used commercial robot arms in medical applications where the maximum workspace coverage is essential. Our framework provides the basis for a fully automatic optimization tool of the robot parameters that can be applied to a large variety of problems.