52 Chapter 3 Kingdom [30, 40, 42], three in Australia [33, 35, 44], and one in Singapore [38]. The studies were published between 1996 and 2016 (median publication year=2010; IQR=2003 to 2013). Despite the shared focus on the elderly patient population, included studies varied in population by age, health condition, and place and time of exposure to the intervention. Of the 16 studies, 10 included patients aged ≥65 years [30-32, 35, 36, 38, 40, 41, 43, 44], three included patients aged ≥ 75 years [33, 37, 42], two included patients aged ≥70 years [29, 34], and one included patients aged ≥60 years [39]. Seven studies further specified their inclusion criteria to: patients admitted to the ED with a medical diagnosis [35], a chronic disease [44], a traumatic injury [39], fall injuries [30, 34], increased risk of ED readmission [29, 35], and requiring outpatient follow-up – like ADL assistance at home – after ED discharge [29, 36]. The sample size ranged from 39 to 3,850 participants for the intervention groups and from 43 to 3,850 participants for the control groups. Of the 16 studies, four (25%) focused on measuring ED throughput time: i.e., ED LOS [39, 43, 44], and time until patients are reviewed by a geriatrician [42]. Effect on ED throughput time was the primary outcome in three studies [39, 42, 44]. Thirteen studies compared the ED revisit rates for intervention groups with the controls [29-38, 40-42]. Follow-up measurement periods varied within and between studies from 7days to 18months after the patient’s initial ED visit. Risk of bias in included studies Overall judgement scores on each risk of bias item are presented in Fig. 2. Of the 16 studies, twelve studies (75%) had a high risk of bias, most commonly due to inadequate randomization, allocation concealment, and blinding [30, 31, 33, 36,37, 38,39,40,41,42,43,44]. Two studies (13%), both RCTs, had a low risk of bias [29, 34]. Two other RCTs (13%) had a moderate risk of bias [32, 35]. The reviewers could not ascertain whether one study was protected against contamination [35], and whether two studies were free from selective outcome reporting [32, 35]. Inter-rater agreement for the individual domains of risk of bias varied – before resolution of disagreements – between slight agreement for ‘Study protected against contamination’ (kappa of 0.30) and very good agreement for ‘Random sequence generation’ (kappa of 0.88). Intervention characteristics Based on our qualitative assessment, seven core elements central to the studied interventions were identified and adherence of each intervention to these elements are summarized (supplement 3), with all studies being organized according to the primary setting of their investigated interventions, namely: the hospital setting and the community setting. Eleven studies examined interventions that
RkJQdWJsaXNoZXIy MjY0ODMw