Assessing Heterogeneity of Treatment Effect in Real-World Data