It isn't the number of people that die that determines whether it is worthwhile, it is the cost/benefit ratio. Fortunately, TFA provides some of the needed information, but it doesn't seem very consistent.
"But regulators say that 95 to 112 deaths and as many as 8,374 injuries could be avoided each year by eliminating the wide blind spot behind a vehicle." (Compared to the 200/17000 numbers, it looks like they believe the cameras will about halve the number of accidents.)
"...regulators predicted that adding the cameras and viewing screens will cost the auto industry as much as $2.7 billion a year, or $160 to $200 a vehicle." Wikipedia says 5.5 million vehicles sold in USA in 2009. (I presume this is new sales only.) This would imply about $500 per vehicle to reach $2.7 billion.
"For the 2012 model year, 45 percent of vehicles offer a rearview camera as standard equipment." Is that 45% of vehicles sold, or 45% of models? If 45% of vehicles, then only 55% are going to have extra cost if the cameras are required.
Optimistic cost/benefit ratio: 112 deaths prevented per year, 55% of 5.5 million vehicles at $160 per vehicle = 484 million dollars per year = $4.3 million dollars to save one life and 75 injuries. (75=8374/112)
Pessimistic cost/benefit ratio: 95 deaths prevented per year at a cost of $2.7 billion per year = $28 million to save one life and a bunch of injuries.
(Note that the cost is up-front, but the benefit is spread out over the ~10 year lifetime of the vehicle, which makes the investment a little less attractive, but I'm not trying to account for this.)