Objective: To implement a computer-based adverse drug reaction monitoring s
ystem and compare its results with those of stimulated spontaneous reportin
g, and to assess the excess lengths of stay and costs of patients with veri
fied adverse drug reactions.
Design: A prospective cohort study was used to assess the efficacy of compu
ter-based monitoring, and case-matching was used to assess excess length of
stay and costs.
Setting: This was a study of all patients admitted to a medical ward of a u
niversity hospital in Germany between June and December 1997.
Patients and participants: 379 patients were included, most of whom had inf
ectious, gastrointestinal or liver diseases, or sleep apnoea syndrome. Pati
ents admitted because of adverse drug reactions were excluded.
Methods: All automatically generated laboratory signals and reports were ev
aluated by a team consisting of a clinical pharmacologist, a clinician and
a pharmacist for their likelihood of being an adverse drug reaction. They w
ere classified by severity and causality. For verified adverse drug reactio
ns, control patients with similar primary diagnosis, age, gender and time o
f admission but without adverse drug reactions were matched to the cases in
order to assess the excess length of hospitalisation caused by an adverse
Results: Adverse drug reactions were detected in 12% of patients by the com
puter-based monitoring system and stimulated spontaneous reporting together
(46 adverse reactions in 45 patients) during 1718 treatment days. Computer
-based monitoring identified adverse drug reactions in 34 cases, and stimul
ated spontaneous reporting in 17 cases. Only 5 adverse drug reactions were
detected by both methods. The relative sensitivity of computer-based monito
ring was 74% (relative specificity 75%), and that of stimulated spontaneous
reporting was 37% (relative specificity 98%). All 3 serious adverse drug r
eactions were detected by computer-based monitoring, but only 2 out of the
3 were detected by stimulated spontaneous reporting. Thr percentage of auto
matically generated laboratory signals associated with an adverse drug reac
tion (positive predictive value) was 13%. The mean excess length of stay wa
s 3.5 days per adverse drug reaction. 48% of adverse reactions were predict
able and detected solely by computer-based monitoring. Therefore, the poten
tial for savings on this ward from the introduction of computer-based monit
oring can be calculated as EUR56 200/year ($US59 600/year) [1999 values].
Conclusion: Computer monitoring is an effective method for improving the de
tection of adverse drug reactions in inpatients. The excess length of stay
and costs caused by adverse drug reactions are substantial and might be con
siderably reduced by earlier detection.