Our literature search yielded 5092 articles. After removing duplicates and screening titles and abstracts, 4825 were excluded, leaving 267 articles for full-text review. Of these, 249 were excluded, and 18 articles were deemed eligible for analysis (Fig. 1). The agreement rate between the two reviewers on article selection was 90%, with a Cohen’s kappa value of k = 0.72.

Flow chart of study identification, screening, inclusion, and exclusion for meta-analysis

Table 1 describes the characteristics of the included studies. All eligible studies were published between 2001 and 2021, featuring a median sample size of 51 (interquartile range: 33–64), and collectively encompassing a total of 1,202 patients in the final analysis. Geographically, four of the studies (22.2%) were conducted in Europe [18,19,20,21], four (22.2%) in Asia [3, 22,23,24], eight (44.4%) in the North America [25,26,27,28,29,30,31,32], and two (11.1%) in the South America [4, 33]. Regarding the study design, 12 studies (66.7%) were prospective cohort studies [3, 4, 19,20,21,22,23,24, 27, 30,31,32], and six (33.3%) were retrospective cohort studies [18, 25, 26, 28, 29, 33]. Concerning the study setting, three (16.7%) studies were conducted in the pre-hospital settings [20, 21, 30], while 15 (83.3%) were in the in-hospital setting (14 in the EDs and one in the ICU) [3, 4, 18, 19, 22,23,24,25,26,27,28,29, 31,32,33]. For the types of population, 10 studies included medical patients [20,21,22, 24, 26, 27, 30,31,32,33], three included trauma patients [25, 28, 29], while five included a mixed population [3, 4, 18, 19, 23]. In total, ten studies assessed the outcome of ROSC, with percentages varying from 18.5% to 75% and a median of 41.7% [3, 4, 19, 22, 24, 26, 27, 29, 30, 33]. Twelve studies assessed SHA, with a range of 7.0% to 58.7% and a median of 18.9% [18, 20, 21, 23, 25,26,27,28,29,30,31,32]. Finally, nine studies assessed SHD, with percentages from 0% to 19.6% and a median of 2.7% [4, 22, 25,26,27,28,29,30, 33].

The overview of QUADAS-2 assessments is detailed in Appendix Table 3 and Fig. 2. Seven studies (39%) were rated as having a high risk of bias in patient selection, primarily due to the use of convenience or non-random sampling [3, 20, 24, 29,30,31,32]. In most studies (83%), unclear or high risks of bias were identified in the index test and reference standard, mainly due to the absence of pre-defined criteria for cardiac activity and a lack of a clear POCUS protocol [4, 18, 19, 21, 23,24,25,26,27,28, 30,31,32,33,34]. In the flow and timing domain, the absence of blinding in most studies (89%) raised concerns about the potential for decreased resuscitation efforts and self-fulfilling prophecies [3, 4, 18, 19, 21, 23, 25,26,27,28, 30,31,32,33,34,35]. Regarding applicability, twelve studies (67%) had an unclear risk of bias due to their focus on specific population [20, 24,25,26,27,28,