Roecklein 2010.

Methods	Randomised parallel‐group study
Participants	N = 30 patients diagnosed with OSA by PSG, naive to CPAP and reporting intent to use CPAP. Inclusion criteria: age 18 to 65, CPAP naive, reported intent to use CPAP (other sleep, psychiatric or health problems were not exclusion criteria) Exclusion criteria: none reported. Baseline characteristics: 70% female. Mean age 46.3 (±11.2). Mean AHI 44.4. Mean ESS 11.6. Mean BMI 42.1. Country: USA
Interventions	Participants were randomised to standard education (SE, n = 16) or personalised feedback (PF, n = 14) group. PF: written personalised feedback report, including detailed information on severity of the disease, self‐reported daytime sleepiness, individually estimated risk of adverse health outcome and risk of motor vehicle accident, all compared with normative data. Feedback addressed barriers to using CPAP, ambivalence about treatment and difficulties of behaviour change and promoted self‐efficacy and personal responsibility for choosing to use CPAP SE (control): written information from the American Academy of Sleep Medicine on OSA, Snoring and PAP therapy for OSA Study duration: 3 months
Outcomes	Objective CPAP usage (total hours, average hours/night, number of sessions) Self‐reported CPAP usage
Notes	Participants were not provided machines but obtained them 'naturalistically', most commonly through insurance. Most participants were low‐income African Americans
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Information not available
Allocation concealment (selection bias)	Unclear risk	Information not available
Blinding (performance bias and detection bias) All outcomes	Unclear risk	"Physicians were blind to study participation and participants were blind to their study condition." Patients were aware that CPAP usage was monitored. Despite intended blinding, it is likely that participants would have been able to distinguish the two interventions
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Only two incidents of missing data in each group. However, in addition, participants who took longer to obtain machines (n = 5 in control group and n = 2 in intervention group did not obtain devices by two weeks) were included from the start and had CPAP usage recorded as 0 hours per session. It is possible that financial burden prevented some participants from acquiring CPAP machines in a timely fashion
Selective reporting (reporting bias)	Unclear risk	Information not available
Other bias	Unclear risk	Information not available
Bias arising from the randomisation process (ROB2, primary outcome)	Unclear risk	Only information provided as to the randomisation procedures used was within Methods: "Consecutive clinic patients were enrolled after receiving their OSA diagnosis and CPAP prescription. Participants were randomly assigned to feedback or standard information groups." Key baseline characteristics (age, gender, BMI, AHI, ESS) were reported and differences were insignificant (P values reported), consistent with chance.
Bias due to deviations from intended interventions (ROB2, primary outcome)	Low risk	Authors report, "Physicians were blind to study participation, and participants were blind to their study condition." No information is provided as to who distributed the written intervention materials to the participants and how blinding was maintained from physicians or other study personnel. No deviations documented; none suspected based upon review. In Methods (Measures) section, authors report, "Individuals who had not obtained machines (2 weeks: control, n =5 and feedback, n=2; 3 months: control, n =2 and feedback n =1) were recorded as having used CPAP for 0 hours per sessions. Individuals whose machines did not record use or who forgot to bring machines to the research session were treated as missing data (2 weeks: control, n = 1 and feedback, n= 0; 3 months: control, n =2 and feedback n = 2). Groups did not differ in the rates of missing versus present data at 2 weeks:X2(2,n=27)=3:86, ns; or 3 months: X2(2,n= 28) = 0:86, ns." In results, authors report, "Participants were 14 in the feedback group and 16 in the control group, and each group lost 1 to follow up." As the difference in Ns implies that the 'lost‐to‐follow‐up' participants were neither those who did not obtain machines nor those for whom machines did not record use or who forgot to bring machines, it is not clear whether the lost‐to‐follow‐up were assigned 0 hours/session or were treated as missing and excluded from analysis. Results tables do not indicate the Ns upon which the recorded results are based. Most likely, the results are based on mITT analysis.
Bias due to missing outcome data (ROB2, primary outcome) All outcomes	High risk	Authors do not report the number of participants randomised and no Ns were reported in the results table. In Methods (Measures) section, authors report number in each group who did not obtain machines and in each group whose machines did not record use or who forgot to bring machines to research session. It is not clear if those who didn't get machines, whose machines did not record or who forgot to bring machines were amongst the participants referenced in the results section, "Participants were 14 in the feedback group and 16 in the control group, and each group lost 1 to follow‐up." Thus, it is not clear whether the 15 total participants (N = 5 in intervention arm , N = 10 in control arm) who had no machine or no outcome data were amongst the 30 participants (N = 14 intervention, N = 16 control) referenced in the results or were in addition to them. Therefore, the total randomised number could have been as low as 15 (6,9) or as high as 45 (19, 26), depending upon whether the participant numbers in results section were numbers before or after accounting for those who did not obtain machines or had missing data. It is also not clear whether the one per group who were lost to follow‐up were amongst or excluded from the reported N values. Neither analyses to correct for bias nor sensitivity analyses were conducted. Missing outcome data occurred for documented reasons, some of which ('forgot' CPAP device/data) could be related to the outcome. There were non‐significant differences between groups in proportion of participants who met authors' definition of missing outcome data (n = 2 for each group at month 3). Reasons cited for missing data were either 'machines did not record use or who forgot to bring machines to the research session.' The latter reason (i.e. 'forgetting') can be related to outcome while machine failure would not be. The authors did not distinguish amongst those reasons in reporting the numbers of participants with missing outcome data for each intervention group. Thus, it is possible that there were differences between groups in the proportions missing because of 'forgetting'. Comparison of possible proportions is further complicated by the fact that we do not have the actual randomised denominators for each group. It is possible that both intervention participants with missing data had 'forgotten' to bring it for analysis while those missing data in control group were purely based on chance equipment failure. Depending on respective denominators, under the above‐noted conditions, there would be a difference of 14% to 0% (intervention vs. control) or 10.5% to 0% (intervention vs. control) in proportion missing due to 'forgetting,' depending on actual randomised Ns.
Bias in measurement of the outcome (ROB2, primary outcome) All outcomes	Low risk	Outcome measured using objective CPAP usage data. Each intervention group outcome data ascertained via automated CPAP device monitoring. Unable to confirm with study author that distribution of CPAP device makes did not differ between intervention arms. However, likely sufficiently similar (i.e. probably Respironics devices) based on authors note "95% of machines detected breathing; Respironics Inc., Murrysville, PA", suggesting most/all participants for whom data were available were using a Respironics device) across groups. Outcome "assessor" is CPAP device: no knowledge of allocation possible.
Bias in selection of the reported result (ROB2, primary outcome) All outcomes	Unclear risk	No protocol, abstract, clinical trials entry available for comparison. Results presented were in accordance with the plan specified in the Methods section of the publication. No information as to whether analysis plan was finalised before unblinded outcome data were available for analysis. Methods section indicates that multiple time points planned; each planned outcome reported. No evidence that multiple analyses (e.g. variable adherence 'thresholds') were conducted. No threshold‐based adherence outcomes reported.
Overall risk of bias (ROB2, primary outcome) Machine usage	High risk	‐