> The survey was designed to reach adults who currently have health care debt or have had health care debt in the last five years. In order to do this, the probability panel sample included an oversample of panel members with annual household incomes less than $30,000, individuals with current or past health care debt (n=641) were included in the sample. In addition, 52 adults with past or current health care debt were reached by calling back respondents that had previously completed an interview on the SSRS Omnibus poll (and other RDD polls) and reported an annual household income of less than $25,000.

> The combined online and telephone samples, excluding oversamples, were weighted to match the sample’s demographics to the U.S. population using data from the Census Bureau’s 2021 Current Population Survey (CPS). On the basis of this weighted sample, parameters for the debt and no-debt populations were extracted. Weighting parameters for each group included sex, age, education, race/ethnicity, region, education, and income, as well as patterns of civic engagement from the September 2017 Volunteering and Civic Life Supplement data from the CPS. The sample was also weighted to match population density parameters from the Census Planning Database 2020. The weights take into account differences in the probability of selection for each sample type. This includes adjustment for the sample design and geographic stratification of the cell phone sample, within household probability of selection, and the design of the panel-recruitment procedure.

So they intentionally targeted people with debt, so the "63% percent of respondents" is not able to be read as "63% percent of Americans".

The 100M in the title is extrapolated, and is likely correct, but people are not reading the percentages as they should be read.

The key takeaway from the study is that not all medical debt is reported as such, and poor people have difficulty paying bills, one type of which is medical bills.