$89 billion is surely false precision, but it's not unreasonable to put a value on lives when you have an economic decision to make. None of them work all that well, but it's better than just flailing in the dark which is the alternative.
For example, you can look at how much it would cost to save those lives another way (eg, through spending on road or rail safety or other, known, healthcare spending). This might give you a figure of a few million. But you tend to find huge discrepancies between spending in different areas - eg, much more in air and rail safety or terrorism prevention than in road safety - depending on how the public responds to those things. Refusing to put a cost on lives this way kills - governments spend huge sums on rail (eg, after the UK Hatfield rail crash) and terrorism prevention when spending a lot of it on healthcare and road safety would save more lives. eg, according to this http://www.theguardian.com/uk/... the UK government was prepared to spend three times as much (~£3m) per life saved on rail compared to road, and more like £15m in an expensive system after the crash - in effect letting 15 people die to save each one.
You can also recognize that we're not talking about certain death, we're talking about risks to life - and people implicitly put a value on risks to life all the time. Car vs train, one car vs another, a dangerous job vs a safe one, driving further to buy something more cheaply or commute from somewhere different. You can come to estimates based on how much they're prepared to spend to avoid risk. But, of course, people are quite irrational about risk and you get widely varying numbers.
And, as another commenter has said, you can estimate from economic output lost, but that's not very satisfactory. In theory it produces a minimum value (assuming that the economy isn't overproducing, spending people's time on producing things less valuable than the time). But it confuses the purpose of an economy - to give people the best quality of life it can, not to produce as much stuff as it can.