Frank 2004

Methods	Purpose: comparison of opportunistic on‐screen reminders to physicians about preventive care compared to no reminders Design: RCT Duration of study: 9 March 1998 to 8 March 1999 Interval between intervention and when outcome was measured: between 9 March 1998 to 8 March 1999  Power computation: not performed Statistics: univariate binomial regression with GEE; ITT analysis (Very helpful e‐mail from Dr. Frank, 23 August 2008: "Our study looked at whether each opportunity to provide a preventive service in a consultation was taken. This is a different way of looking at the question from the more usual approach of asking what proportion of participants who had attended during the influenza immunization season had received the vaccine by the end of the season (in other words, efficacy), or from asking what proportion of participants of the practice had received the vaccine by the end of the season (effectiveness) We were interested in what happened in each consultation in which influenza vaccination was indicated and due for the patient. We were able to do this very data‐intensive exercise only because we set out to use a practice that kept all clinical and billing data electronically and because I custom wrote software to analyze the practice's electronic data automatically. To my knowledge, this study is unique in its intensive automated analysis of each consultation The GPs actually performed slightly worse when reminded to give influenza vaccine. We don't know why this occurred, but it may be because the rate of giving influenza vaccine to participants 65 years and over in Australia was already quite high, possibly making our reminders redundant In our approach, we were not interested in numbers of participants, but in the number of opportunities that arose in consultations for the participants who did attend. Our approach to examining the question of opportunistic performance of preventive services is almost unique, in that we looked closely at every opportunity that arose, and did not take a snapshot of the practice population at one point in time, which is what almost all other studies have done. In retrospect, it would have been useful to collect data about efficacy so that we could compare our results more easily with those other studies.")
Participants	Country: Australia Setting: urban practice with 10 GPs Eligible participants: (health status): 10,507 for all reminder activities, of whom 1847 were ≥ 65 and eligible for the influenza intervention Age: ≥ 65 Gender: 57% f
Interventions	Intervention: computer‐generated reminder Control: no intervention
Outcomes	Outcome measured: % influenza vaccination Time points from the study that are considered in the review or measured or reported in the study: 9 March 1998 to 30 June 1998 (these dates are from e‐mail from author)  % vaccinated by 30 June 1998
Notes	Funding: not stated (PhD thesis)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	All quotes are from e‐mail from author 18 August 2008: "Randomization of participants was automated. Patients were randomised by the last digit of their family's five digit number within the practice. Family numbers had been allocated sequentially by the practice's computer system without regard to any characteristics of the patient or the family. We were satisfied that this method was not likely to cause any bias in the randomization."
Allocation concealment (selection bias)	Unclear risk	"Allocation was not concealed. However, I believe that in the daily rush of seeing participants, most of the GPs were unlikely to have had time or energy to look at the patient's family number in order to work out to which group the patient had been randomised."
Blinding (performance bias and detection bias) All outcomes	Unclear risk	"Blinding, in the sense of blinding the investigators, was not necessary because the judgement of whether a preventive activity (including the administration of influenza vaccine) had been performed was made by searching the practice's electronic clinical record automatically". "Vaccinations were recorded by the doctors in their clinical record system's immunization module which used coded data entry to make the entries consistent and therefore machine‐searchable.  If our search found a record of influenza vaccine being given between 9th March (the start of our trial) and the end of June (the end of the useful immunization season), this was counted as influenza immunisation having been performed"
Incomplete outcome data (attrition bias) All outcomes	Low risk	"We analysed all data by intention to treat. All participants who were enrolled and randomised (both of which occurred automatically at their first visit during the trial) were included in the analyses."
Selective reporting (reporting bias)	Low risk	No selective reporting