Of the 440 eligible physicians, 127 (28.9%) agreed to participate, and the first 107 were included in the study (). Participating physicians were slightly younger than those who did not participate (mean age 46.5 v. 49.4 years). However, participating and nonparticipating physicians were similar in the average number of prescriptions per elderly patient (35.6 v. 33.8) and the prevalence of inappropriate prescribing (18.9% v. 18.8%) in the 18 months before the study start date. There were no differences in characteristics between the CDS and control groups ().
Fig. 1: Selection and assignment of study population. List of Montreal general practitioners provided by the Collège des médecins du Québec. Random assignment was within strata defined by physician age (34–44, 45–48, (more ...)
Of the 20 109 eligible patients, 12 560 (62.4%) agreed to participate. Those in the CDS group were more likely than those in the control group to be men, to have made fewer visits to their primary care physician and to have received fewer prescriptions from their primary care physician ().
At the beginning of the study, there was at least 1 prescribing problem for 33.3% of the patients in the control group and 31.8% of those in the CDS group (). For 20.4% and 18.8%, respectively, the problems were attributable to a study physician, for 3.3% and 3.2% they were attributable to a study physician plus another physician, and for 8.3% and 9.1% they were attributable to another physician. In both groups, drug–disease contraindications were the most common prescribing problems, followed by drug–age contraindications and excessive duration of therapy ().
Two unforeseen factors influenced the effectiveness of the CDS. First, copayments for prescription drugs were increased when the study began, which resulted in a 9% reduction in prescription drug use by the elderly.40
Second, 22% of the physicians experienced frequent hardware or software failure in the early months of the study; the proportion declined to 4% by month 6. Physicians in the CDS group downloaded prescription information in 81% of the study weeks; however, those who had more computer problems downloaded information less often (r
During the study, the rate of initiation of an inappropriate prescription was significantly lower (18%) in the CDS group than in the control group (). This trend was evident for drug–disease contraindications, drug–age contraindications, excessive duration of therapy and therapeutic duplication and was significant for drug–age contraindications and excessive duration of therapy.
CDS had no significant impact on the discontinuation of pre-existing inappropriate prescriptions (). Although more patients in the CDS group than in the control group had all inappropriate prescriptions discontinued (47.5% v. 44.5%; or 35.5 v. 32.1 per 1000 visits; relative rate [RR] 1.14; 95% confidence interval [CI] 0.98–1.33), the 14% difference was not statistically significant. The only substantially higher discontinuation rate for a specific prescribing problem was for drug interactions: 68.6 v. 51.5 per 1000 visits in the CDS and control groups respectively.
Physicians in the CDS group were able to identify excessive duration of therapy, therapeutic duplication and drug interaction resulting from more than one source of prescribing for the same patient. Most of the therapeutic duplications and drug interactions occurred because prescriptions were written by both the study physician and another physician or another physician alone (). Discontinuation rates in the CDS group were systematically higher for problems created by the combination of prescriptions from study physicians and other physicians than for the other types of prescription problems. An exception was with drug interactions: the relative difference in discontinuation rates between CDS and control physicians was highest for problematic prescriptions written by the study physician, followed by problematic prescriptions written by both the study physician and another physician.
Adjusting for patient characteristics () did not modify differences in initiation and discontinuation rates between the CDS and control groups. However, a physician's previous computer experience influenced the effectiveness of CDS. Among experienced computer users the rate of initiation of inappropriate prescriptions was 30% lower in the CDS group than in the control group (RR 0.70, 95% CI 0.55–0.89). Among the computer beginners the rate of initiation of inappropriate prescriptions was virtually identical in the 2 groups (RR 1.03, 95% CI 0.82–1.29). The same trend was evident for discontinuation rates (RR for experienced users 1.17 and for beginners 0.93), but this apparent modification of the effectiveness of CDS by computer experience was not significant (interaction term: study group*computer experience, p = 0.32).