Assessing the dose-response relationship between number of office-based visits and hospitalizations for patients with type II diabetes using generalized propensity score matching