180 6 Chapter 6 3.9. Hospital days per patient - short-term Two studies with a total of 273 participants reported on the di erence in mean hospitalisation days per patient per group within the rst 6 months (Boxall 2005; Trappenburg 2011). Pooling showed a signi cant reduction in days spent in the hospital per patient in favour of IDM (MD -4.36, 95% CI -6.41 to -2.31) (Analysis 1.23). 3.10. Hospital days per patient - medium-term Ten studies including 2994 participants assessed the di erence in mean hospitalisation days per patient per group from 6 to 15 months’ follow-up (Bourbeau 2003; Engstrom 1999; Farrero 2001; Kessler 2018; Ko 2016; Kruis 2014; Lenferink 2019; Rea 2004; Silver 2017; Vianello 2016). Pooling showed a non-signi cant reduction in hospitalisation days in favour of IDM (MD -1.73, 95% CI -3.71 to 0.25), with moderate heterogeneity (I² = 71%). Heterogeneity could not be explained by di erences in the quality of studies. Three studies showed a signi cant e ect in favour of IDM (Farrero 2001; Ko 2016; Rea 2004), and one study showed a signi cant e ect in favour of control (Smith 1999). Smith 1999 reported increased attention to disease and symptoms by the COPD nurse as a possible explanation. Mean hospitalisation days also varied substantially between studies and within the IDM study groups, with an average hospital stay ranging from 2 days in Silver 2017 to 25.5 days in Vianello 2016 (Analysis 1.23). 3.11 Hospital days per patient - long-term Two studies with 346 participants reported the di erence in mean hospitalisation days after 15 months’ follow-up (Titova 2017; van Wetering 2010). There was no signi cant di erence between groups (MD -1.60, 95% CI -6.12 to 2.92) (Analysis 1.23). 3.12 Emergency department Twelve studies assessed the number of participants with at least one ED visit (Bourbeau 2003; Fan 2012; Farrero 2001; Lou 2015; Rea 2004; Rice 2010; Rose 2017; Sanchez-Nieto 2016; Silver 2017; Smith 1999; Trappenburg 2011; Wakabayashi 2011). To account for clustering, we reduced the study size in Lou 2015 to its ‘e ective sample size’. We were able to pool the data from nine studies with 8791 participants (Bourbeau 2003; Fan 2012; Lou 2015; Rea 2004; Rice 2010; Rose 2017; Sanchez-Nieto 2016; Silver 2017; Smith 1999), which revealed a signi cant reduction in the number of participants with at least one ED visit in favour of IDM, with considerable heterogeneity (OR 0.69, 95% CI 0.50 to 0.93; I² = 68%) (Analysis 1.24). A sensitivity analysis including only high-quality studies showed that the risk of an ED visit was still signi cantly reduced with IDM (OR 0.69, 95% CI 0.50 to 0.94; I² = 64%) but could not explain the heterogeneity. Further exploration to assess reasons for heterogeneity revealed that seven trials had decreased risk of ED visits in favour of IDM (Bourbeau 2003; Fan 2012; Lou 2015; Rea 2004; Rice 2010; Sanchez-Nieto 2016; Silver 2017), of which three were statistically signi cant (Bourbeau 2003; Lou 2015; Rice 2010). Two studies showed a non-signi cant increase in risk of ED visits for the IDM group (Rose 2017; Smith 1999). Silver 2017 reported in the discussion that lack of e ect on ED visits “may be due to the emergency department functioning
RkJQdWJsaXNoZXIy MTk4NDMw