I tried to track down this factoid. First I went to the world bank: https://data.worldbank.org/indicator/SI.POV.DDAY?locations=U...

which does show 1% for the US. In the notes it says " Data for high-income economies are from the Luxembourg Income Study database." My next stop was there. From http://www.lisdatacenter.org/frontend#/home I found that the source of their data was Current Population Survey (CPS) - Annual Social and Economic Supplement (ASEC) (United States). The Current Population survey has a table creator here: https://www.census.gov/cps/data/cpstablecreator.html

Although there is nothing that quite matches the 3.2 million figure cited, the LIS did some harmonization so the most likely ultimate source is the 3.446 million people living in households that were categorized as having "no income" in 2013. This is separate from the 2.775 million people that were categorized as living in households that have "$1 to $2,499 or loss" in income.

I'm very skeptical that this is an appropriate use of this data. The world bank poverty number is supposed to be about consumption--it has to do with people consuming very little in the way of goods and services. But no one at all can live on no goods and services. They'd die of hunger and thirst in far less than a year. It could in theory be the case that millions of Americans were living on less than $1.90/day in consumption but it cannot possibly be the case that millions of Americans are living on $0/day.

To the extent that the figure cited for "no income" means anything (and isn't just noise in the data), it doesn't mean what the world bank means when it says living on $1.90/day.