Subgroup analyses for use of acid-suppressive agents and risk of pneumonia using random-effects model for observational studies