Risk Set Matched Difference-in-Differences for the Analysis of Effect Modification in an Observational Study on the Impact of Gun Violence on Health Outcomes.