Use of case-time-control design in pharmacovigilance applications: exploration with high-risk medications and unplanned hospital admissions in the Western Australian elderly

ABSTRACT Purpose To use a case–time–control design to derive preliminary estimates of unplanned hospitalisations attributable to suspected high‐risk medications in elderly Western Australians. Methods Using pharmaceutical claims linked to inpatient and other health records, the study applied a case–...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Pharmacoepidemiology and drug safety 2013-11, Vol.22 (11), p.1159-1170
Hauptverfasser: Price, Sylvie D., Holman, C. D'Arcy J., Sanfilippo, Frank M., Emery, Jon D.
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:ABSTRACT Purpose To use a case–time–control design to derive preliminary estimates of unplanned hospitalisations attributable to suspected high‐risk medications in elderly Western Australians. Methods Using pharmaceutical claims linked to inpatient and other health records, the study applied a case–time–control design and conditional logistic regression to estimate odds ratios (ORs) for unplanned hospital admissions associated with anticoagulants, antirheumatics, opioids, corticosteroids and four major groups of cardiovascular drugs. Attributable fractions (AFs) were derived from the ORs to estimate the number and proportion of admissions associated with drug exposure. Results were compared with those obtained from a more conventional method using International Classification of Diseases (ICD) external cause codes to identify admissions related to adverse drug events. Results The study involved 1 899 699 index hospital admissions. Six of the eight drug groups were associated with an increased risk of unplanned hospitalisation, opioids (adjusted OR = 1.81, 95%CI 1.75–1.88; AF = 44.9%) and corticosteroids (1.48, 1.42–1.54; 32.2%) linked with the highest risks. For all six, the estimated number of hospitalisations attributed to the medication in the exposed was higher (two to 31‐fold) when derived from the case–time–control design compared with identification from ICD codes. Conclusions This study provides an alternative approach for identifying potentially harmful medications and suggests that the use of ICD external causes may underestimate adverse drug events. It takes drug exposure into account, can be applied to individual medications and may overcome under‐reporting issues associated with conventional methods. The approach shows great potential as part of a post‐marketing pharmacovigilance monitoring system in Australia and elsewhere. Copyright © 2013 John Wiley & Sons, Ltd.
ISSN:1053-8569
1099-1557
DOI:10.1002/pds.3469