Statistical analysis

We used the Review Manager software package (RevMan) version 5.4.1 (The Nordic Cochrane Centre, Copenhagen, Denmark, 2020) for outcome analyses. Forest plots of the outcomes were created using the Mantel–Haenszel statistical method and random effect analysis model due to the diverse methodologies used in the included studies. Funnel plots were constructed to examine any exaggeration of effect estimates from low-quality studies. Risk of bias was assessed by RoB 2.0 (a revised tool to assess risk of bias in randomized trials) according to the Cochrane Handbook for Systematic Reviews of Interventions Version 6.3, 2022 [12] and Newcastle–Ottawa Scale (NOS) for Assessing the Quality of Non-Randomized Studies [13]. Heterogeneity was quantified by I² statistic. The extent of heterogeneity was categorized into mild (I² < 30%), moderate (30 ≤ I² < 50%), and substantial (I² > 50%). Network meta-analysis (NMA) was employed for pairwise comparison between patients with and without care bundles and stratification based on biomarkers. We used MetaInsight V4.0.0 [National Institute for Health and Care Research (NIHR)—Complex Reviews Support Unit (CRSU), United Kingdom, 2023] [14], a tool adapted from the R software to conduct NMA. Surface under the cumulative ranking curve (SUCRA) was used to show the hierarchy of the treatment effects in a rank-heat plot, with the preferential treatment having the highest SUCRA value. Trial sequential analysis (TSA) was employed to reduce the likelihood of type 1 and 2 errors after repetitive significance analysis of the study data (Copenhagen Trial Unit, Centre for Clinical Intervention Research, Denmark, 0.9.5.10 Beta software). This statistical methodology also assesses the need for further trials to clarify the effect of an intervention [15, 16]. TSA was used to confirm the impact of biomarker incorporation in AKI care bundles.