From the article "4.9 percent of children tested for lead turned out to have elevated levels." What is 'elevated'? Is that 30ug/dl? or 3ug/dl when the average is 2ug/dl. When specific information is absent, I generally doubt the accuracy.
But I guess that's typical news media anyways. If it's not bursting in to flames, then it isn't news.
Edit: added "From the article..."
https://www.washingtonpost.com/news/wonk/wp/2016/01/15/this-...
Flint has an actual problem.
'So the Virginia Tech researchers took 30 different readings at various flow levels. What they found shocked them: The lowest reading they obtained was around 200 ppb, already ridiculously high. But more than half of the readings came in at more than 1,000 ppb. Some came in above 5,000 -- the level at which EPA considers the water to be "toxic waste."'
> The city opted out of Detroit's water supply
As I understand, it the City of Detroit basically kicked them out (or raised the rates prohibitively high). Flint was already on track to transition to a regional water system (not the Flint river system), but it wasn't completed yet. This statement places all of the blame onto Flint for pulling out of using Detroit's water system.
1) Detroit was gouging Flint, well into 7 figures per year, (plus the supply pipes from Detroit to Flint have lead)
2) Flint signed a new deal for a new supply direct from Lake Huron with lead-free supply lines at an 8-figure cost savings per decade (or less) -- better water, much cheaper. Construction to take ~4 years.
3) Detroit hears this, cuts off Flint at earliest opportunity (1-year notice opt-out), apparently assuming that they could later gouge even more under the assumption that Flint had no alternative;
4) Flint resorts to using Flint River in the interim (until new Huron feeds from new water system are complete), prepared for the bacterial risks but unaware of lead risks.
Statistics are the best tool we have but it's also important to realize that when the incidence rate is low enough, you need a lot of data for statistical methods to be reliable. We see this in A/B testing of conversion rates. If your base conversion rate is only 0.1% and you are looking for a 15% lift (e.g. 15% higher chance of lead poisoning), then you need a sample size of over 700,000 for each branch (i.e. 700,000 kids "exposed" to lead and 700,000 not exposed). That kind of sample size is hard to come by.