Fig. 2.

Adjusted differences in frequency of visits (via video and across all modalities) for SUD-diagnosed tablet-recipients (vs. recipients' baseline and SUD-diagnosed non-recipients) — event study estimates.

Note: We excluded months −1 and month 0 because treatment assignment i.e. tablet assignment likely occurred in these months and we did not want to attribute tablet assignment-related visits to the tablet-associated effects. All models adjusted for veterans' age, sex, race, number of physical and mental health chronic conditions, diagnoses of post-traumatic stress disorder and depression, Nosos risk adjustment score, VA priority-based enrollment, marital status, rurality, distance to VA primary care, homelessness indicator, high suicide risk indicator, cumulative monthly COVID-19 cases in the patient's county, and broadband availability in the patient's zip-code. All models included indicators for calendar month to adjust for events occurring in each month, indicators for patients' closest VA medical center to control for any time-invariant facility characteristics, and an indicator for being a tablet-recipient, which adjusted for any remaining fixed difference between tablet-recipients and non-recipients. In all models, standard errors accounted for clustering at the patient-level.