Corelation does not imply causation. Maybe its that people who are prone to depression have some factors that make them less likely to believe in religion.