•Although excess mortality in persons with mental disorders has been well documented, limited reports on changes in mortality trends have been conflicting.•We found based on a national-level data that decline trends in mortality in persons with mental disorders differ by primary diagnosis, while more than 50 % of the trend was counted by risk factors at individual level. Further evidence was found that multilevel risk factors as identified by the WHO MIF model were associated with mortality of individuals with psychiatric diagnoses and could have large impact on time trend of mortality.•Understanding changes in mortality trends and associated factors which are preventable and modifiable for individuals with specific psychiatric diagnosis help in developing targeted service and intervention strategies in bringing down mortality further for the population globally. This study aims to examine 20-year temporal trends in all-cause mortality among psychiatric patients and investigating impacts of risk factors on the time trends based on 218,703 Finnish adults with mental disorders who were discharged from 87 psychiatric hospitals between 1 Jan 1995 and 31 Dec 2014. The age-period-cohort analysis of Poisson model with random hospital effects estimated temporal trends in death rate and associated factors at individual, healthcare system, and society levels, following the WHO multilevel intervention framework model for six major psychiatric diagnosis. The adjusted annual mortality declined by 2.2 % annually (RR: 0.978 95 % CI 0.976–0.980) for all individuals, and by 2.8 % after adjusting for all risk factors, with varied decreasing rate between 2.0 % and 3.6 % by diagnosis. Individual level factors accounted for the declining rate by 54.5 % for all patients, with the highest impact on patients with personality disorders, followed by patients with affective disorders and patients with schizophrenia. Identified declining trends and associated factors which are preventable and modifiable for individuals with specific psychiatric diagnosis may lead to develop targeted service and intervention strategies in bringing down mortality further for the population.