Table 1
Adapted Risk of bias summary
|
Selection bias (a)
|
Performance bias
(Assignment to intervention) (b)
|
Performance bias
(Adhering to intervention) (c)
|
Attrition bias (d)
|
Detection bias (e)
|
Reporting bias (f)
|
Other bias (g)
|
Overall
|
Brug et al,[32]
|
|
|
|
|
|
|
|
|
Casebeer et al,[33]
|
|
|
|
|
|
|
|
|
Clark et al,[34]
|
|
|
|
|
|
|
|
|
Cooper et al,[35]
|
|
|
|
|
|
|
|
|
De Ruijter et al,[36]
|
|
|
|
|
|
|
|
|
El-Sayed et al,[37]
|
|
|
|
|
|
|
|
|
Flocke et al,[38]
|
|
|
|
|
|
|
|
|
Karvinen et al,[39]
|
|
|
|
|
|
|
|
|
Kinmonth et al,[40]
|
|
|
|
|
|
|
|
|
Moore et al,[41]
|
|
|
|
|
|
|
|
|
Pelto et al,[42]
|
|
|
|
|
|
|
|
|
Presseau et al,[43]
|
|
|
|
|
|
|
|
|
a) Extent participants were randomly allocated to groups how this was concealed throughout enrolment (Random sequence generation, Allocation concealment)
b and c) Extent to which the researchers/participants were blinded to the group they are allocated to
d) Assessment of the completeness of outcome data
e) Extent the outcome assessors were blinded to the intervention/control
f) Measurement of how selective researchers have been when reporting outcomes (Selective reporting)
g) Any forms of bias not covered in the five domains (Research quality)
In summary, the high overall risk of bias was heavily influenced by the high risk of reporting bias found in all twelve studies whereby multiple outcome measures were used to assess delivery quality (a mean 11 outcome measures were used per study), with little consideration for statistical correction (31). As well as this major source of bias, other sources were also identified. Specifically, six studies (50%) failed to report the allocation sequence or its concealment in detail, causing some concern towards their selection bias. ‘Some concern’ regarding attrition bias was recorded for one study (8%), due to a moderate drop-out rate (23.8%). When analysing detection bias, four-studies (33%) presented a high-bias risk due to use of self-report and patient recall, while one-study (8%) provided some concern due to limited blinding of assessors.
Table 2
Adapted risk of bias table
a) Extent participants were randomly allocated to groups how this was concealed throughout enrolment (Random sequence generation, Allocation concealment)
b and c) Extent to which the researchers/participants were blinded to the group they are allocated to
d) Assessment of the completeness of outcome data
e) Extent the outcome assessors were blinded to the intervention/control
f) Measurement of how selective researchers have been when reporting outcomes (Selective reporting)
g) Any forms of bias not covered in the five domains (Research quality)
Narrative synthesis
Within the 12 included studies a total of 132-outcomes were identified. Sixty-eight outcomes from eight-studies related to communication style, 64-outcomes from seven-studies related to delivery of intended intervention content. Communication style and delivery of intended intervention content improved significantly for 54% (37 of 68) and 55% (35 of 64) of outcomes respectively.
The two types of training intervention, communication training and retraining, produced similar results, significantly improving 50% and 49% of outcomes, respectively. Training using both educational and practical elements seemed to be more effective, significantly improving 52% of outcomes, whereas education only training significantly improved only 23% of outcomes. Effectiveness was also highest in the medium-term (between six to 12-months) with significant improvements in 81% of communication and 66% of content-delivery outcomes), compared with the long-term (over 12-months), where significant improvements were found in 50% of communication outcomes and 42% of content-delivery outcomes.
Effect on communication skills short-term
Across the three-studies (19, 25, 28) that assessed the effect of healthcare professional training on communication skills short-term, 38% of outcomes significantly improved. All three training interventions included educational and practical elements, however the two training interventions (19, 25) that specifically focused on communication style provided the greatest benefit. Of these two-studies, “teachable moment” communication training (25), significantly improved 90% (nine of ten) outcomes measured. However, these data only relate to the delivery of smoking cessation interventions. Moreover, the limited effect of 0% (zero of six measures improving versus controls) within the other study (28) could be explained by high detection bias (use of patient recall).
Effect on communication skills medium-term
Across the three studies (19, 24, 29) that assessed communication skills in the medium-term 81% of outcomes significantly improved. All of the interventions included educational and practical elements. Assessment of effectiveness short-to-medium term was performed by one study (19) and showed communication outcomes improved (25% vs 58%).
Effect on communication skills long-term
Three studies (21, 22, 27) assessed effect on communication style in the long-term, 50% of outcomes significantly improved, a decrease compared to medium-term. The least effective training intervention by this measure (22) in the medium term (which showed improvement in 25% (two of eight) of outcomes) only provided educational elements. The most effective training intervention (21), recording an improvement in 75% (six of nine) of outcomes, provided both educational and practical elements. Additionally this study used a large sample size, 74 Paediatricians, and a validated communications skills questionnaire (21).
Effect on delivery of intended intervention content short-term
Three-studies (20, 26, 28), were identified as assessing the effect on delivery of intended intervention content in the short-term. Across these studies, 44% of the assessed outcomes significantly improved. The two (26, 28), that encompassed educational and practical activities (e.g. evaluating case studies, quizzes) were the most effective, significantly improving 40% (two of five) and 67% (two of three) of outcomes, respectively.
Effect on delivery of intended intervention content medium-term
Across the three-studies (23, 24, 29), that assessed the medium-term effect of healthcare professional training on delivery of intended content, 66% of outcomes significantly improved. The two most effective training interventions (24, 29), which improved 79% (22 of 28) and 100% (two of two) of outcomes, respectively, included both educational and practical elements.
Effect on delivery of intended intervention content long-term
Across the two-studies (21, 30) that assessed the effect of healthcare professional training on delivery in the long-term, 42% of outcomes significantly improved. Both training interventions included educational and practical elements.
Meta-anaysis of health behaviour change
A total of six studies, with 2802 participants, were included in the meta-analysis. The results are shown in Figure 2.
The analysis indicated that healthcare professional training resulted in a small significant improvement in patient health behaviour compared to control (SMD: 0.20; 95% CI 0.11 to 0.28; P<0.0001, I2= 0%). Sub-group analysis highlighted a greater effect in the short-term (median six-months) (SMD: 0.25; 95% CI 0.10 to 0.41; P=0.001; I2= 24%) compared to the long-term (median 12 months) (SMD: 0.15; 95% CI 0.03 to 0.28; P=0.02, I2=0%), however this difference was not significant (P=0.31).