Research that provides an evidence base for the pharmacotherapy of people with mental disorders is needed. The abundance of digital data has facilitated pharmacoepidemiology and, in particular, observational research on the effectiveness of real-world medication. Advantages of pharmacoepidemiological research are the availability of large patient samples, and coverage of under-researched subpopulations in their naturalistic conditions. Such research is also cheaper and quicker to do than randomised controlled trials, meaning that issues regarding generic medication, stopping medication (deprescribing), and long-term outcomes are more likely to be addressed. Pharmacoepidemiological methods can also be extended to pharmacovigilance and to aid the development of new purposes for existing drugs. Drawbacks of observational pharmacoepidemiological studies come from the non-randomised nature of treatment selection, leading to confounding by indication. Potential methods for managing this drawback include active comparison groups, within-individual designs, and propensity scoring. Many of the more rigorous pharmacoepidemiology studies have been strengthened through multiple analytical approaches triangulated to improve confidence in inferred causal relationships. With developments in data resources and analytical techniques, it is encouraging that guidelines are beginning to include evidence from robust observational pharmacoepidemiological studies alongside randomised controlled trials. Collaboration between guideline writers and researchers involved in pharmacoepidemiology could help researchers to answer the questions that are important to policy makers and ensure that results are integrated into the evidence base. Further development of statistical and data science techniques, alongside public engagement and capacity building (data resources and researcher base), will be necessary to take full advantage of future opportunities.

需要为精神障碍患者的药物治疗提供证据基础的研究。大量的数字数据促进了药物流行病学的发展，尤其是对现实世界药物有效性的观察性研究。药物流行病学研究的优势是可获得大量患者样品，并覆盖处于自然状态的未充分研究的亚群。这样的研究比随机对照试验更便宜，更快捷，这意味着更可能解决有关仿制药，停药（处方）和长期结果的问题。药物流行病学方法也可以扩展到药物警戒，并有助于开发现有药物的新用途。观察性药物流行病学研究的弊端来自于治疗选择的非随机性质，从而导致适应症混淆。解决此缺陷的潜在方法包括主动比较组，内部设计和倾向性评分。许多更严格的药物流行病学研究已经通过三角剖分的多种分析方法得到了加强，以提高对推断因果关系的信心。随着数据资源和分析技术的发展，令人鼓舞的是，指南开始包括来自稳健的观察性药物流行病学研究以及随机对照试验的证据。指南作者和参与药物流行病学的研究人员之间的合作可以帮助研究人员回答对政策制定者重要的问题，并确保将结果纳入证据基础。为了充分利用未来机会，有必要进一步发展统计和数据科学技术，以及公众参与和能力建设（数据资源和研究人员基础）。