I approach the findings reported by Matthew Roché and colleagues  in the same way as I would any other peer-reviewed science results covering the topic of violence in relation to a diagnosis / label / condition: cautiously and minus the need for sweeping generalisations (see here), but without shying away from potentially important findings.
Roché et al discuss results based on their analysis of "intake records of 63,572 patients diagnosed with SMIs [serious mental illness] (i.e., schizoaffective disorder, schizophrenia, bipolar disorder, and unipolar depression), substance use disorders, and non-SMI psychiatric disorders" in relation to the risk of violent ideation and behavior (VIB). As well as looking at the frequency of VIB among their cohort, they also looked for other variables outside of a diagnosis of SMI that may impact on VIB.
Results: "patients with SMI conditions had higher rates of VIB than both patients with non-SMI psychopathology and those with substance use disorders only." No, this does not make for great PR for SMIs but is a reality of their observations. Further: "patients with SMI and comorbid substance use pathology were responsible for the majority of VIB within each SMI condition." This equation - SMI plus substance abuse equals greater risk of violence - is something that is becoming rather important based on the peer-reviewed science literature. It follows other independent findings  too and might even link into other areas.
Appreciating that those diagnosed with a SMI are also at greater risk of being a victim of crime (see here) including crime with a violent element attached to it, there are some important lessons to be learned from the Roché data. Not least is the potential focus on reducing comorbid substance abuse in the context of a diagnosis of serious mental illness so as to potentially modify the heightened risk of VIB and also, other less than desirable outcomes . This is not something that can be done easily (see here) but does not mean it cannot be attempted at all.
I might also add that violence, as and when it does occur in the context of SMI, is likely to be related to other social and situational factors as well as being influenced by something like substance abuse (disorder). Indeed, in the context that various aspects of life can very much be *altered* by the experience of an SMI (diet, physical activity, etc) I'm minded to direct your attention to other variables potentially important to VIB such as nutritional factors for example (see here and see here). Science might also perhaps look to other diagnoses that potentially complicate the clinical picture in SMI as perhaps also exerting any effect on the risk of VIB (see here) and the [developmental] importance of transitioning risk from one label to another (see here) again, minus any sweeping generalisations. Finally, and minus passing the buck, the findings reported by Patel and colleagues  further complicate the clinical picture...
 Roché MW. et al. Prevalence and Risk of Violent Ideation and Behavior in Serious Mental Illnesses: An Analysis of 63,572 Patient Records. J Interpers Violence. 2018 Mar 1:886260518759976.
 Fazel S. et al. Schizophrenia, substance abuse, and violent crime. JAMA. 2009 May 20;301(19):2016-23.
 Skalisky J. et al. Prevalence and Correlates of Cannabis Use in Outpatients with Serious Mental Illness Receiving Treatment for Alcohol Use Disorders. Cannabis Cannabinoid Res. 2017 Jun 1;2(1):133-138.
 Patel RS. et al. Is Cannabis Use Associated With the Worst Inpatient Outcomes in Attention Deficit Hyperactivity Disorder Adolescents? Cureus. 2018 Jan 7;10(1):e2033.