Clark 1998.

Methods	Study design: parallel, open‐label, cluster RCT Length of observation: 22 months Setting: 74 general practices in Michigan and New York, USA
Participants	Population: 74 physicians were randomised and 69 completed the trial. It is not clear how many were randomised to each group, but the study states that 637 children were recruited in total, and outcome data were available for 472. Baseline characteristics Baseline data were reported for the whole population rather than for each group. 60% of physicians and 70% of children were male. Physician and child ages were reported in brackets rather than as a mean per group. 30% of families were Latino/Hispanic (15%) or African American (15%). Inclusion criteria:Physician criteria: primary specialty of general paediatrics; licensed no earlier than 1960; providing direct patient care; if board‐specialised, certified only in paediatrics; willing to take part in the interactive seminar if randomised to the treatment group. Child criteria: 1 to 12 years of age; diagnosis of asthma made by a physician; no other chronic disorders with pulmonary complications; at least 1 emergency medical visit for asthma in the previous year. An emergency visit was a hospitalisation, emergency department (ED) visit, or physician office visit on an emergency basis defined as administration of epinephrine subcutaneously or bronchodilators by aerosol. Exclusion criteria: none in addition to inclusion criteria
Interventions	Intervention: shared decision‐making seminars for clinicians Interactive seminar based on self‐regulation theory to guide physicians in NAEPP care and to engage in interactive conversations with patients to derive information for making therapeutic decisions, create a supportive atmosphere, reinforce self‐management, give a view of the long‐term therapeutic plan, and build patients’ confidence in controlling symptoms and using medicines. Materials included brief lectures from respected asthma specialists; a video depicting effective clinician teaching and communications behaviour; case studies presenting troublesome clinical problems; a protocol by which physicians could assess their own behaviour regarding patient communications; and review of messages to communicate and materials to use when teaching patients. Resources: The seminar comprised 2 face‐to‐face group meetings, each lasting 2 ½ hours, held over a 2‐ to 3‐week period. Control: usual care Physicians in the control group were randomly assigned a date corresponding to 1 of the 3 seminar time points, to determine when follow‐up interviews of their patients should begin.
Outcomes	Physician survey (items related to using clinical practice methods/medicines, encouraging self‐management, and providing patient teaching and communications). Analysis of data illustrated close correlation between physician and parent descriptions of behaviour. Questions on the parent interview form related to symptom status of the child, medicines prescribed, use of healthcare services for asthma (ED visits, hospitalisations, physician office visits), parents’ observations and opinions of physicians’ teaching and communications behaviours, other aspects of the clinician–patient interaction
Notes	Trial registration: not reported Funding: supported by MD/Family Partnership ‐ Education in Asthma Management grant number HL‐44976 from the Lung Division of the National Heart, Lung, and Blood Institute
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomized, controlled study design" but no description of how this was done
Allocation concealment (selection bias)	High risk	"Names of patients meeting criteria were selected by the investigators at random from the roster provided by physicians", which may have introduced recruitment bias within practices, even if practices were themselves randomised adequately to groups
Blinding of participants and personnel (performance bias) All outcomes	High risk	"Patients and their parents were blind to physicians’ involvement in the intervention." "A potential source of bias in the study was that physicians would give positive reports of their behavior to be consistent with good clinical and communications practices. To guard against such bias, data were collected from parents of patients regarding physician behavior as a means of corroborating physician reports."
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	"Patients and parents were blind so outcomes measured by them can be considered low risk of bias. Outcomes measured or self‐assessed by the physicians taking part in the study are at high risk of detection bias." "A potential source of bias in the study was that physicians would give positive reports of their behavior to be consistent with good clinical and communications practices. To guard against such bias, data were collected from parents of patients regarding physician behavior as a means of corroborating physician reports." Patients and parents were blinded to their physician's participation in the intervention. Depends who is reporting the outcome, and to whom. Will be assessed separately when GRADE is applied
Incomplete outcome data (attrition bias) All outcomes	High risk	"Data were collected from physicians at baseline, and 69 (93%) provided follow‐up data 5 months after the program. Data were also collected from 637 of their patients at baseline, and in a 22‐month window after the intervention, 472 (74%) of this number provided follow‐up data." Unclear how many were randomised to each group and whether dropout was balanced, but nonetheless quite high attrition overall
Selective reporting (reporting bias)	High risk	Study does not report methods fully, for example, number of people assigned to each group and participant flow. In terms of data, uncertainty regarding the number of participants per group means that data are difficult to analyse reliably in meta‐analyses. Some data relevant to this review are presented narratively. We did not identify a study protocol or trial registration
Other bias	Low risk	None noted