This was from a thread regarding the OP considering Mini-Vans for purchase. In that thread I asserted that mini-vans are not very safe as compared to the average sedan, which I assign an arbitrary value of 100 for comparing to other vehicles out of class. Thus a Dodge 2500 at 210 and the Honda Odyssey at 68 from the 2017 cycle. The major factors I consider are mass, 60" contact crush, 6" contact crush on all 4 sides, supplemental restraint systems, sacrificial crush zones, trip over, high speed rollover. Then I take the IIHS, Manufacturer and 3rd party data to calculate occupant G-Loads due to 15mph, 30mph and 60mph front and rear impacts. Search the NHTSA database for safety issues related to MVAs. There is of course, assumptions based on my reading of the data, IIHS crash tests, review of accidents on the same model and several days every year in the salvage yards looking at cars for which I have obtained accident reports and injury data. If you are interested in what I get from 3rd party companies, here is a sample report of one factor set from one company I buy reports from. https://www.neptuneeng.com/SSF/SSFChevrCavalier01.pdf One has to understand a few things about the various entities that provide crash data, the first I alluded to in the prior thread, which is that the comparisons available to the public are for in class types. Meaning about the same weight and structural elements. So 2 door subcompacts vs 2 door sub compacts. Next, the financial drivers of the two groups that fund the tests (Manufacturers and Insurance companies) are at odds with each other...sales and risk mitigation. But one thing I have learned over 25 years, is that a lot more people are living with significant lower extremity injuries than did when I started. I tell folks that if you want to compare two cars, get them to equivalent dollars in trim level and ask for an insurance quote on both. The one that costs more to insure has a history of more occupant injuries...it is just that simple. When pick-ups went from solid front axles to IFS, the occupant injury rate went up, while at the same time, the vehicle repair (relatively) costs did not. The two biggest drivers are $ and Mass. More $ means more demand for safety and the manufacturers comply. If you think about it in terms of insurance, a deceased insured from a $100K care costs them 8 to 10 times what a deceased insured from a $20K costs, so the insurance carriers demand safety above the .gov requirements in more expensive vehicles. Dollars only comes into my analysis in terms of the safety features the manufacturer installs. In so many of the 2 car MVA fatals I work, the descendants are in vehicles with half the mass of the survivors. When I drive, and teach others to drive, I always try to have a smaller vehicle in front of, and behind me, avoid left hand turns at controlled intersections, and stay as far away from Semis as possible. Those practices will, in large part mitigate the base safety ratings of the vehicles you drive. Crash severity is a large part of what I do for insurance carriers who hire me to evaluate an accident. The fault part is really pretty easy. I use an accident severity scale to classify the accident for each vehicle based on the damage to the car, and what I know about the safety factors for the vehicle. I do that knowing nothing about the actual injuries to the occupants. Then I provide that in a report (usually) to my clients. They then couple that with a Medical professional who evaluated the injuries without seeing my work. If I go to depo or trial, the IME report is reviewed by me in detail to make any minor adjustments for occupant physiology, and vice versa. So, I don't look at all cars, but every 4 years, I take 3 sedans that I "assume" are the mid-pack and assign them the 100 value for that year. Then every MVA I work on, when I get crash data, I compare it. When I have friends or clients ask me to evaluate a vehicle, I have them buy the reports I need and run those. So over 4 years, I get a pretty clean look, and hitting my 5th cycle in 2021, my assumptions are getting better. But still some art to it, as well as a lot of monte carlo type calculations.