Finally, our study probably underestimated the level of insulin nonadherence
in this population because it did not capture instances in which patients took an
injection but gave less than the amount of
insulin they knew was needed for optimal
glucose control.