Algorithm for resolving discrepancies between claims for smoking cessation pharmacotherapies during pregnancy and smoking status in delivery records: The impact on estimates of utilisation

Background: The linkage of routine data collections are valuable for population-based evaluation of smoking cessation pharmacotherapy in pregnancy where little is known about the utilisation or safety of these pharmacotherapies antenatally. The use of routine data collections to study smoking cessation pharmacotherapy is limited by disparities among data sources. This study developed an algorithm to resolve disparity between the evidence of pharmacotherapy utilisation for smoking cessation and the recording of smoking in pregnancy, examined its face validity and assessed the implications on estimates of smoking cessation pharmacotherapy utilisation. Methods: Perinatal records (n = 1,098,203) of women who gave birth in the Australian States of Western Australia and New South Wales (2004–2012) were linked to hospital admissions and pharmaceutical dispensing data. An algorithm, based on dispensing information about the type of smoking therapy, timing and quantity of supply reclassified certain groups of women as smoking during pregnancy. Face validity of the algorithm was tested by examining the distribution of factors associated with inaccurate recording of smoking status among women that the algorithm classified as misreporting smoking in pregnancy. Rate of utilisation among smokers, according to original and reclassified smoking status, was measured, to demonstrate the utility of the algorithm. Results: Smoking cessation pharmacotherapy were dispensed to 2184 women during pregnancy, of those 1013 women were originally recorded as non-smoking as per perinatal and hospital data. Application of the algorithm reclassified 730 women as smoking during pregnancy. The algorithm satisfied the test of face validity—the expected demographic factors of marriage, private hospital delivery and higher socioeconomic status, were more common in women whom the algorithm identified as misreporting their smoking status. Application of the algorithm resulted in smoking cessation pharmacotherapy utilisation estimates ranging from 2.3–3.6% of all pregnancies. Conclusion: Researchers can use the algorithm presented herein to improve the identification of smoking among women who use cessation pharmacotherapies during pregnancy. Improved identification can improve the validity of safety analyses of smoking cessation pharmacotherapy—providing clinicians with valuable evidence to use when counselling women on the role of pharmacotherapy for smoking cessation during pregnancy.