Ab Malik 2017.

Study characteristics
Methods	Cluster RCT randomised at hospital level, Malaysia Study recruitment and setting details: see Table 4
Participants	Inclusion criteria: all registered nurses caring for people with stroke were invited to take part; informed written consent Exclusion criteria: not reported OHC training group: 277 registered nurses General stroke care training group: 270 registered nurses Details of participants are shown in Table 5
Interventions	OHC training Intervention: Internet‐based continuing professional development programme Materials: secure internet portal, computer Agent: none Procedures: online training programme was specific to provision of oral hygiene care in stroke patients. Programme covered oral health knowledge, attitudes, subjective norms, means of behavioural control and intention. Contents included information on good oral condition and importance of having good oral health, consequences of poor oral hygiene and importance of nurses' roles and care of people with stroke. Provided by: stroke physicians (rehabilitation medicine) and dentists, and followed good practices of computer‐aided learning for oral health. Development of contents was guided by the definition of the theory of planned behaviour domains and scope of the study. Training: as described above Delivery: online; 1‐to‐1 Location: unclear Regimen: participants were reminded and encouraged to complete the Internet‐based continuing professional development programme every 6 weeks; no details about length of programme Tailoring: not reported Modification: not reported Adherence: not reported General stroke care training Intervention: Internet‐based generic continuing professional development programme not specific to OHC Materials: secure internet portal, computer Procedures: programme related to 'bundles of care' for people with stroke that included some details on oral hygiene care but not specific to theory of planned behaviour. Provided by: stroke physicians and physicians Training: as described above Delivery: online; 1‐to‐1 Location: unclear Regimen: participants were reminded and encouraged to complete the Internet‐based continuing professional development programme every 6 weeks; no details about length of programme Tailoring: not reported Modification: not reported Adherence: not reported
Outcomes	Outcomes: questionnaire on practice of providing oral hygiene care to people with stroke (contained 12 items specific to attitudes, subjective norm, perceived behaviour control and general intention to providing oral hygiene care related to Theory of Planned Behaviour), knowledge of OHC (assessed using 5 items related to dental plaque, gum bleeding, consequences of dental plaque, how to prevent gingivitis and how oral health affects general health) Data collection: baseline, 1 and 6 months postintervention
Funding	Authors declared no conflicts of interest. Study funded by The University of Hong Kong
Notes	Dropouts are detailed in Table 9 Statistical data included within the review meta‐analyses
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "Hospitals were first stratified by size into either large, medium, or small in terms of number of health care providers. From each stratified group, hospitals were block‐randomized in groups of 4 ('ABBA') by a computer‐generated randomization method."
Allocation concealment (selection bias)	Low risk	Quote: "5 hospitals were assigned to the test group (277 registered nurses) and 5 hospitals were assigned to the control group (270 registered nurses). The allocation sequence was concealed from the investigator coordinating the trial (who had contact with the centers)."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "Participants were also blind as to what groups they were assigned to, as both received a form of Web‐based CPD."
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "the assessor was 'blind' as to what group participants had been assigned."
Incomplete outcome data (attrition bias) All outcomes	High risk	Comment; high attrition rate (82/277 in OHC training group and 92/270 in general stroke care training group). Total dropouts accounted for (quote: "loss to follow‐up was because nurses were transferred to other wards or hospitals"). However, dropouts were only reported in CONSORT diagram as loss to follow‐up, so unclear when participants dropped out (month 1 or 6). Study authors reported that ITT analysis employed but not all participants appeared to be included in the final analyses.
Selective reporting (reporting bias)	Low risk	Comment: all prespecified outcomes reported.
Baseline data comparable?	Unclear risk	Comment: insufficient information. Unable to compare baseline demographics for healthcare provider and environmental characteristics as data were combined. Quote: "no significant difference between the response rate among those in the test and control groups."
A priori power calculation	Low risk	Comment: yes Quote: "With the assumption that this practice is at 50% and that it will not change without education intervention, whereas there will be a 25% improvement in practices following CAL intervention (i.e. 63% of nurses will practice oral care in rehabilitation). Then a proposed sample size of 247 in each group is required with sample power at 80%. Allowing for nonparticipation and a dropout rate of ~20%, thus it was prudent to attempt to recruit over 600 nurses (300 per group) in total to test the hypothesis."
Other bias	Unclear risk	Comment: limited information supplied about how long each group were exposed to the intervention (i.e. how frequently they used the programme and total duration).