I'd like to add a few things here that might help you out. One thing that I have come to realize with software engineering is that there is never a "catch all" answer to a lot of these types of questions. Since each application is different, it is a good idea to look at the challenges and figure out what is the best tool for the job.
If you look at some of the performance tests done on the frameworks you notice that vanilla node is ultra performant. But most of the time your application isn't going to need to handle 8000 requests per second, and even if it does you can always scale your server horizontally nowadays. So you trade off a little bit of speed to gain a few benefits, such as:
- Much easier to write.
- The code base, in my opinion, is much easier to maintain.
- A lot of the little gotchas are taken care of by the framework so you don't have to be a Node.js god.
- Middleware.
Now, that doesn't mean that every single application should use express. A lot do because it is about as barebones as it gets. For instance, Walmart created Hapi because they claim that it is easier to maintain their code base with a 'configuration first' approach. So maybe if you are going to have this monster backend and a monster team, Hapi might be a good choice. If implementing something with real-time you might want to use something like socketcluster.io to help you scale websockets. It is all about picking the right tool for the right job.
That being said, there are advantages to building a server with vanilla Node, especially when learning to develop node apps. Since frameworks abstract a lot of the lower level stuff it takes away the opportunity to pick up some cool node tricks here or there. Or a lot of times if a framework is insufficient or an npm package is doing something weird it is good to have that knowledge of vanilla node under your belt to really understand what is going on. Having the skill of knowing how Node.js works is such a benefit when working with any framework.