Generalisability of pharmacoepidemiological studies using restricted prescription data. (August 2016)

• Record Type: Journal Article
• Authors: Brown, C.; Barron, T.; Bennett, K.; MacDonald, D.; Dwane, F.; Sharp, L.
• Is Part Of: Irish journal of medical science. Volume 185:Number 3(2016)
• Page Start: 723
• Page End: 727
• Publication Date: 2016-08
• Subjects: Pharmacoepidemiology -- Prescriptions -- Ovarian cancer -- Routine data -- External validity -- Generalisability; Medicine -- Periodicals; Medicine -- Periodicals; 610
• DOI: 10.1007/s11845-015-1360-z
• Languages: English
• ISSNs: 0021-1265

Abstract

Abstract Background Linking medication databases to disease registries enables population-based pharmacoepidemiology research. In Ireland, country-wide dispensing data is available only from the means-tested government-medical cards scheme. This restriction may impact generalisability of analyses based on these data. Aim Gender was previously identified as predictor of card status so we aimed to compare women with and without medical cards at the time of ovarian cancer diagnosis. Methods Ovarian cancers diagnosed 2001–2010 were identified from the National Cancer Registry Ireland. Age, region, deprivation, smoking, employment and marital status were evaluated using logistic regression for associations with card status. Cumulative incidence of de novo card receipt post-diagnosis was assessed. Results 1778 (52 %) of 3396 women with incident ovarian cancer had a card at diagnosis (<70:33 %; 70+:87 %). Within those <70, all variables were significantly associated with card status at diagnosis. 52 % of those without a card at diagnosis received one post-diagnosis. Conclusions Although medical card coverage within ovarian cancer patients is similar to the general population, various factors predict card status. Particularly within those under 70, external validity needs to be considered when interpreting pharamcoepidemiological analyses using these data.