To enable better human-machine and human-robot interaction in healthcare applications, there is a need to improve haptic devices, assistive and companion robots to touch humans in a way that can be perceived as comfortable, safe and trusting. Towards this vision, we developed a flexible, wearable matrix force-sensing garment capable of recording dynamic force behaviour of caring touch from four healthcare experts in their natural therapy environment, and enable the touch recipient to provide feedback on the perceived quality of the same touches. Instead of assuming particular touch gestures are caring, we explored the application of the sensing and analysing system to characterise a caring quality of touch across different gestures. We were able to characterise distinctive features for touch aimed for sensory aspects and for moving the body. Our findings demonstrate that caring touch is not limited to specific gestures or a single parameter but is rather characterised by how the touch is performed across multiple features, with deliberateness - the careful and smoothness in the increasing and decreasing of force, emerging as a shared characteristic of caring touch regardless of the specific gesture employed. The study also demonstrates the feasibility of using raw tactile sensor data to assess subjective touch quality.