Dear Sasan Zamani,

Following parameters may used in mesoscopic model:

1. LOS

2.Time

for other Mesoscopic Simulation Models pl visit this link.

http://ops.fhwa.dot.gov/trafficanalysistools/tat_vol1/sectapp_a.htm#a6

In mesoscopic models vehicles (or better packets of vehicles) are traced explicitly, while the performances of links  are expressed in an aggregate way. Therefore, as for supply, you need the length, capacity and maximum density of each link of the network; as for demand, you need the flows (vehicles per time unit) at each node of the network. If you're interested in, I can send you a paper of mine about a mesoscopic model of Dynamic Network Loading.  

Crash Severity on a urban highway, as on other roads, mainly depends on the kinetic energy. The kinetic energy by the speed. The speed by traffic flow, density, lane width, tortuosity of the route etc.

I would check the following parameters in function of the crash  severity:

number and width of lanes, traffic flows per lane (in passenger car equivalents per hour), composition of traffic flow, speed (85th percentile or speed diagrams for the entire path), speed differences among different types of vehicles (such as cars, vans, trucks etc).

The interdependence of these parameters and the crash severity should give you some good results.

Good work and good luck

In my opinion and according to previous answers - speed is the main factor. But the severity also may be influenced by headway and spacing. Lower distance or  shorter time for reaction of the driver = shorter time of breaking = higher level of the crash  severity. And hope that we are discussing about crash between two or more vehicles.

Iam totally agree with Mauro

Dear, i think this ref is suitable for you. regards.

I agree with the explanation of Dr. Mauro where you analyze platoons of vehicles .  The relationship between crash severity and speed is directly proportional for the same class of vehicles