A simple method for analyzing data from a randomized trial with a missing binary outcome