Hollis 1993.
| Methods | Study design: Randomized controlled trial Country; USA Recruitment: People visiting outpatient internal medicine and family practice offices in a group practice Health Maintenance Organiation. |
|
| Participants | 2707 smokers who received provider (physician, physician assistant or nurse practitioner) advice to quit Therapists: Project nurse or health counsellor | |
| Interventions | If people refused to see a counsellor they were mailed information appropriate to their assignment 1. Advice ‐ In addition to provider advice, given brief pamphlet by health counsellor 2. Self‐quit ‐ Cessation advice, CO assessment, 10‐min video, stop‐smoking kit, and choice of S‐H manuals. Encouraged to set quit date. 1 follow‐up telephone call and series of mailings 3. Group referral. Cessation advice, CO assessment. Video encouraged use of intensive (9 meetings over 2 m) group programme, and waiver of fee. Effort made to schedule attendance 4. Combination. Participants shown video explaining both S‐H and group approaches, and encouraged to choose one | |
| Outcomes | 1 yr 2 x PPA (7 days at 3 m and 12 m) Validation: Saliva cotinine at 1 yr. Most conservative outcome is used in which self‐reported non‐smokers who did not provide saliva samples are recorded as smokers | |
| Notes | 3 vs 2 in comparison 1.1.2 vs S‐H programme. 3 vs 1 in 1.3.1 vs brief advice control | |
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Random sequence generation (selection bias) | High risk | "Two random digits contained in the patient's health record number were used to assign patients" |
| Allocation concealment (selection bias) | High risk | Use of health record number precludes concealment. All who received initial provider advice were considered participants, and providers who delivered initial message stated to be blind to assignment, so possibility of selection bias may be low |
| Blinding of outcome assessment (detection bias) All outcomes | Low risk | Biochemical validation of abstinence |
| Incomplete outcome data (attrition bias) All outcomes | Low risk | 14% lost to follow‐up at 12 m. Response rates not significantly different across conditions, all participants included in analysis |