Hacker News Reader

Comment by biophysboy | Hacker News Reader

Coming from a bio background, I’ve always been confused why auto fatality stats are normalized per miles driven. Epidemiological metrics like incidence or prevalence seem like they would work fine? Town A would be “safer” than town B if people’s commutes are 20% shorter, even if accidents occur w same frequency

View on HN

Because it yields a simple corollary that to make travelling safer you can reduce the number of miles driven. Mostly by giving people viable alternatives to driving, be it long-distance rail or bike lanes to move around quicker and safer in the city.

Pretty sure I've seen exposure-adjusted incidence rates used in clinical trials.

Miles is simply a proxy for exposure.

Given risk here does vary by exposure time and trip length varies so much, it seems reasonable to use - at least in combination with crude rates.

Fair point - a combo might be the best approach.. I understand the idea of accidents correlating w/ miles driven, but it seems to be optimizing for driving safety rather than human life? Does that make sense?

What are some other better ways to normalize?

Other common approaches:

1. Per capita

2. Per registered vehicle

3. Per trip

All of these have upsides and downsides (as does “per vehicle km”), and all will paint different pictures with different distortions.

Honestly, just per 1M person per year. If this normal incidence went up while the exposure incidence rate went down over 20 years, I'd wanna know.

Per trip?

Reply on news.ycombinator.com