Depends on what you consider "meaningful" I suppose...but yes, as long as the sample's representative. You can do a lot with n=100 - unless the authors cherry-picked it would be incredibly unlikely for them to find 83/100 non-compliant hospitals by chance if there are only, say, a few hundred non-compliant hospitals.

    $ R -q -e "binom.test(17, 100, p=0.9, conf.level=0.999)"
    > binom.test(17, 100, p=0.9, conf.level=0.999)
    
     Exact binomial test
    
    data:  17 and 100
    number of successes = 17, number of trials = 100, p-value < 2.2e-16
    alternative hypothesis: true probability of success is not equal to 0.9
    99.9 percent confidence interval:
     0.06925006 0.32125658
    sample estimates:
    probability of success
                      0.17

Of course, there would be potential for "not fully compliant" to be insignificant for practical purposes (e.g. if it included hospitals that just missed a couple of billing codes), but based on TFA that doesn't seem to be the case.