Thirty-three RCTs (n>7,400) were included.
Twelve studies reported adequate methods of randomisation: the outcome assessors and care providers were blinded in 8 studies, while the patients were blinded in 3 studies. All but one study achieved a follow-up rate greater than 75%, and 18 studies performed an intention-to-treat analysis. Ten studies were considered to be ‘high quality’.
Compared with usual care, DMPs were associated with a statistically significant reduction in the likelihood of mortality (OR 0.80, 95% CI: 0.69, 0.93, p=0.003; based on 5,308 patients in 28 studies) and all-cause hospitalisation (OR 0.76, 95% CI: 0.69, 0.94, p<0.00001; based on 7,387 patients in 32 studies).
Compared with usual care, DMPs were associated with a statistically significant reduction in the likelihood of HF-specific mortality (OR 0.37, 95% CI: 0.21, 0.73, p<0.0002; based on 700 patients in 4 studies), HF-related hospitalisation (OR 0.58, 95% CI: 0.50, 0.67, p<0.00001; based on 3,817 patients in 20 studies) and fewer number of all-cause days in hospital (WMD –1.49 days, 95% CI: -2.03, -0.95, p<0.00001; based on 2,356 patients in 12 studies). DMPs were also favourable with regards to HF-related number of days in hospital, length of hospital stay and number of hospital readmissions, however, the authors stated that there were insufficient data to reliably pool the results.
Evidence of statistical heterogeneity was shown in the main analyses. It was largely accounted for in the sensitivity analyses or following the exclusion of a single outlying study. The results of the sensitivity analyses suggested that results were similar for mortality, all-cause and HF-related hospitalisation rates, across different DMP approaches. High-quality studies and those of 3- to 6-month duration were consistently associated with significant reductions in all main outcomes. Nurse-led interventions did not impact on all-cause mortality.
Sixteen studies evaluated quality of life. Of these, eight reported a statistically significant improvement in quality-of-life score, four reported discordant results and four did not report any statistical significant difference between DMP and usual care.
The funnel plot was fairly symmetrical, suggesting little evidence for publication bias.