The notion of communicability is a broad one, and there are numerous concepts of "epidemic" which extend beyond the notion of purely biologically infectious (bacterial, viral, parasitic) diseases, including "epidemic of obesity", "epidemic of violence", etc.
It was an article regarding the latter which made the general case (and I'll see if I can't turn it up). Effectively: if you have a phenomenon which includes transmission from carrier to susceptible populations, then you're dealing with a case of epidemic.
This generalises in some interesting (and quite probably controversial) ways. Richard Dawkins's notion of the meme -- a unit of transmissible cultural information -- falls squarely within the model, and suggests a further generalisation:
Epidemics exist where there is an infectious agent (biological, material, energetic, informational) originating from some source or reservoir which causes a negative functional or behavioural change in a susceptible population spreading along vectors of transmission. They further have characteristics of exposure-response, latency, transmissivity (e.g., source => individual, sometimes individual => individual), susceptibility or resistance, symptomatic / asymptomatic periods, morbidity and mortality rates, etc.
This means we can look at agents including the familiar bacteria, aemoebae, parasites, and viruses, but also more exotic notions: prions, chemical, physical, and radiological triggers (say, heavy metals, asbestos / PM2.5 matter, radon or plutonium), and cultural or informational ones: cults, religions, mythologies, ideologies, psychological disorders, stress, and the like.
What the model offers are notions of potential containment. I've done some research to see if there's an underlying theory or model of public health or epidemiology (and would expect it to closely correspond to the above). What I've found looking through introductory texts is that there largely is not, but instead the focus is on teaching statistical monitoring methods used within the field. I'm not sure if I'm looking in the wrong places, or if there simply is no such generalised model.
But again: if you can identify specific agents (biological, material, informational, energetic), sources, vectors, susceptibility factors, and the like, then it's possible to apply fairly standard epidemiological notions to prevention.
Note that the direct medical intervention of cure is not on that list -- it's an expensive and uncertain intervention, and one that comes with very high costs. Far better to practice avoidance, increase resistance, and reduce transmission, as well as monitor aggressively for outbreaks. All of which are what we're seeing in the case of 2019-nCoV.
In the case of mental health, the study of suicide itself is foundational to sociology (Emil Durkheim's book of the same title, Suicide: https://archive.org/details/DurkheimEmileSuicideAStudyInSoci...), for numerous reasons. It turns out that rates are very consistent across time, with a few notable exceptions. There's considerable variance across place, whether triggered by culture or natural environment isn't clear. As David Simon, author of The Wire notes, dead bodies are hard to hide (see his excellent "The Audacity of Despair", ignore the recording and speaking quality issues: https://www.invidio.us/watch?v=nRt46W3k-qw). And suicide itself is an extreme act. So by looking at trends with time, you're finding a exceptionally strong signal of underlying social dysfunctions.
Looking at the epidemiological model, isolating yourself might reduce some dynamics of infection (vectors of transmission), but would quite likely increase others (susceptibility through social isolation). The model's underlying premise is that you simply cannot assess these factors in isolation, which would point out the flaw in your reasoning.