Expected net benefit of clinical pharmacy in intensive care medicine: a randomized interventional comparative trial with matched before-and-after groups

Rationale, aims and objectives: This study evaluated clinical pharmacy costs against drug costs.

Method: We conducted a randomized interventional comparative trial at the surgical intensive care unit (ICU) of Ghent University Hospital, Belgium (period B: group B1 with pharmacist consultation; control group B0). We obtained before (period A) and after (period C) control groups using 1:1 propensity score matching with B1 and B0. Mean daily ICU drug costs with standard error of the mean (SEM) were compared between B1 and B0 (primary analysis) and between matched pairs (AB1, AB0, CB1 and CB0; secondary analysis). For B, we performed a 1000 bootstrapping (by resampling B1 and B0), calculated the benefit-cost ratio using pharmacy time (gross salary) as cost (euros) and drug cost savings as benefit. We performed sensitivity analysis with and without outlier drug costs (i.e. twice the standard deviation).

Perspective: Belgian health care payer.

Results: In period B, 135 patients were randomized: B0, n = 60; B1, n = 75. Pharmacists provided recommendations in 148/706 (21.0%) therapies with 83.1% acceptance. Mean drug cost difference between B0 (430.6 euros, SEM 406.0) and B1 (221.2 euros, SEM 58.7) (P = 0.870) became significant after excluding outlier drug costs (B0, 184.4 euros, SEM 42.5; B1, 90.5 euros, SEM 17.7; P < 0.001). Recommendations were cost-beneficial (break-even drug costs or savings) in 53.8% of patients with a benefit-cost ratio of 25:1 (confidence interval -5:1 to 94:1). In sensitivity analysis excluding outlier drug costs, B0 costs were significantly higher than both A and C, indicating high baseline expenses in B0.

Conclusions: The randomized interventional comparative trial in a small ICU patient group suggested the potential cost-benefit of clinical pharmacy on daily ICU drug costs. However, after matching, this benefit was attenuated. A final conclusion demands a larger randomized trial adopting a similar design with matched controls. Future research should include clinical impact of recommendations.