Essentially the whole of the pplus diffusion around the contact contributes to the resistance, but you're trying to approximate it by extending the narrow part of the dog-bone to touch the contact - which is a little tricky to do with layer derivation. Off the top of my head I can't think of an easy way of doing this in Dracula (it does tend to help people if you mention which verification tool you're using, by the way) - I'm a bit rusty with writing Dracula extraction rules though...
One alternative approach I've used in the past is to use the Flexible LPE capability in Dracula to compensate by adding on a factor for the end effects. For example:
Code:ELEMENT RES[Z] ZRDEV ZRTERM
PARAMETER RES[Z] 1200
LEXTRACT ZRD ZRDEV ZRTERM BY RES[H] ZRPAR &
EQUATION W=(W1+W2)/2 &
EQUATION L=AREA/W &
EQUATION RP=2 * ( 3.4 * 125.0 / W + 130 / (1 + ( W - 3.4 ) / 4.6 ) ) &
EQUATION R=(1200*L/W) + RP
Don't worry too much about the actual numbers or equation - but the point is that the RP equation is calculating end effects - based on an estimate of the diffusion region at the ends (in this case the diffusion at the ends of the resistor was differently doped than the body of the resistor), and some allowance for contact resistance (based on an estimate of the number of contacts).
Given that my resistors were always laid out using pcells, this approach worked quite well - and was easier than trying to extract absolutely everything from the layout (which is generally the goal) - especially as the contribution of the end parts of the resistor was generally quite small compared with the body of the resistor.
Regards,
Andrew.