You've got your correlation v causation mixed up there in interpreting those results. It's just as likely that negative well being means people take more shelter in digital interaction as opposed to the opposite (which is why the paper is careful to avoid actually suggesting causality).

Given that loneliness and isolation actually is toxic [0], for many it's not that online interactions are replacing offline interactions; they're the only interactions these people can get.

[0]http://journals.sagepub.com/doi/full/10.1177/174569161456835...