Statistical analysis
The distributions of patient, tumor, and treatment characteristics were compared. Categorical variables were analyzed with the Pearson chi-square or Fisher exact test, and continuous variables were compared by one-way analysis of variance. Difference-in-differences analysis for evaluating a policy change that apply to some but not all members was used in this series.11,12 In period 1, there was no specific intervention or policy difference for oral cancer treatment in target group and control group. In period 2, LNPA was applied for target group alone. The difference of lymph node yield between different periods (period 1 and 2), and different departments (target group and control group) was analyzed. Among difference-in-differences method, the effects of the target group (target group in period 1), the time period (period 2), and target group in the post-intervention period (target group in period 2) was analyzed through liner model for LN yield and regional recurrence.13 In order to reach the requirement of strict exogeneity, propensity score matching with clinical tumor category and node category was performed.15 A two-side test with a P value of < 0.05 was set as representing statistical significance.