Are joint replacement registries associated with burden of revision changes? A real-world panel data regression analysis