Category	Details of previous reviews/meta‐analyses
Number of reviews and meta‐analyses	13 systematic reviews: Bauer 2018; Macedo 2014; Massey 2019; Milne 2016; Pallavicini 2016; Petriwskyj 2016; Reyes 2018; Robertson 2015; Skeffington 2013; Tams 2016a; Townshend 2016; Van Kessel 2014; Wainwright 2019 5 meta‐analyses, with only three being relevant due to meta‐analyses for psychological outcomes (Joyce 2018; Leppin 2014; Vanhove 2016). Deady 2017a conducted a meta‐analysis on psychological symptoms but included primary studies that did not explicitly mention resilience, while Pesantes 2015 conducted no pooled analysis for psychological outcomes.
Methodological characteristics	Eligibility critiera: heterogeneous eligibility criteria (e.g. concerning study design) and definitions of resilience training (e.g. the aim of fostering resilience was not always stated in the included primary studies
	Search strategy: Some reviews used rather simple, limited search strategies to identify relevant studies (e.g. only resilience/hardiness combined with training terms in, for example, Joyce 2018; Robertson 2015; restriction to English language), which may bias the search results.
	Review protocol/registration: A review protocol or PROSPERO registration was available for four publications only (Bauer 2018; Leppin 2014; Townshend 2016; Wainwright 2019).
	Review according to guidelines: Most reviews report having been conducted according to the PRISMA or alternative guidelines such as the guidance for undertaking reviews in health care (CRD 2009; e.g. Milne 2016; Van Kessel 2014).
	Quality assessment of included studies: Most reviews performed a quality assessment of the primary studies (the exceptions being Milne 2016; Pallavicini 2016; Reyes 2018; Skeffington 2013; Vanhove 2016, who only judged publication bias; we were also unable to verify if Tams 2016 conducted a quality assessment because we could not retrieve the full text). For studies included in several reviews, the reported risk of bias also differed between publications (e.g. detection bias for Abbott 2009 differed between Leppin 2014 and Robertson 2015).