Vollmer 2006.

Methods	Randomised clinical trial
Participants	Patients: Central to this study was the aim of identifying individuals at greatest risk for acute exacerbations. Patients from Kaiser Permanente Northwest KPNW, a large group health maintenance organisation in Portland Oregon, were used for this study. Patients were eligible if they were aged 18 or older on January 1st 2003 and were on the KPNW asthma registry or had had at least 180 days of anti‐asthma medication dispensed. 850 individuals who had COPD listed on their problem list were excluded. Resulting sample size was 6948 patients.
Interventions	Intervention: This consisted of 3 rounds of telephone calling about 5 months apart. The calls consisted of a series of questions to assess any recent emergency care for which they had not had follow up; their current stage of asthma control and medication use and whether they could identify a primary care provider whom they usually saw for asthma care. Patients were asked questions to screen for COPD. Next they were given optional tailored feedback regarding their overall level of asthma control and use of medications. Suggestions might include advice on the control of night waking or the need to continue inhaled corticosteroids daily. An offer of further KPNW services and a repeat call in 5 months was given. 3389 patients received calls using automated speech recognition technology. 192 patients received the call from a live caller. The call generated alerts for the provider as to which participants were at high risk of a future exacerbation, thereby hopefully triggering follow up. Controls: Received usual care, i.e. no additional telephone calls
Outcomes	Primary outcome measures: data were collected by surveys of patients and a selection of providers. Baseline data came from administrative data, KPNW data and a survey mailed to a random sample of 549 health plan members in November 2002. This had an 83% response rate. Follow‐up data came from a survey of 1583 randomly selected participants 1 month after final call, response rate 65% 1. Healthcare utilisation (KPNW data) 2. Medication use (Asthma Therapy Assessment Questionnaire (ATAQ), dispensing of antiasthma medications) 3. Quality of life (ATAQ, 5‐point scale reflecting the number of control problems in the last month, mini‐Juniper Asthma Quality of Life Questionnaire and Asthma Impact survey) 4. Demographic data, measures of health status, self‐management practices, attitudes about asthma and satisfaction with care
Notes	At first the strength of this study seems to be its large number of participants. However, on closer inspection it transpires that the data sources are often fragmented and chosen as representative data or randomly selected representative group rather than actual data from the full number of participants. Little effort seems to have been made to confirm that the samples used were indeed representative of the groups they were selected to represent, i.e. no confirmation of baseline data etc. In addition, the very low participation rate in the intervention calls is a high risk factor for the introduction of bias Also of note was the ethically dubious decision not to inform participants that they were taking part in research. The authors justified this by stating this was "in order to mimic the results of real‐world implementation.”
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Insufficient information
Allocation concealment (selection bias)	Unclear risk	Insufficient information
Blinding (performance bias and detection bias) All outcomes	Unclear risk	No detail given as to who amassed results from questionnaires and whether blinding procedures were in place
Incomplete outcome data (attrition bias) All outcomes	High risk	Outcome data was only collected from a random sample of both the intervention and usual care groups, i.e. incomplete. For healthcare utilisation data, it was only reported for patients who had had at least 6 months of cover by the plan.
Selective reporting (reporting bias)	High risk	"although the overall intent‐to‐treat analyses gave non significant results, post hoc analyses that compared the control participants to participants who actually used the intervention found numerous significant, albeit small differences."
Other bias	High risk	Very low participation rates in intervention. Follow‐up data were collected from only 38% of participants.