Exposure measurement error when assessing current glucocorticoid use using UK primary care electronic prescription data

Abstract

Purpose To quantify misclassification in glucocorticoid (GC) exposure defined using UK primary care prescription data. Methods A cross‐sectional study including patients with rheumatoid arthritis prescribed oral GCs in the past 2 years. Glucocorticoid exposure based on electronic prescription records was compared with participant‐reported GC use captured using a paper diary. Prescription data (containing information about prescriptions issued but no dispensing information) was provided by the Clinical Practice Research Datalink. The following variables were defined: current use and dose of oral GCs and if (and when) participants had received a GC injection. For oral GCs, self‐reported use was taken to represent “true” exposure. A dataset representing a hypothetical population was generated to assess the impact of the misclassification found for current use. Results A total of 67 of 78 study participants (86%) were correctly classified as currently on/off oral GCs; 32/38 (84.2%) participants reporting current GC use and 35/40 (87.5%) participants not reporting current use were correctly classified. Estimated values of current dose were imprecise (correlation coefficient 0.46). Concordance between reported and prescribed GC injections was poor (kappa statistic 0.14). Misclassification bias was demonstrated in the hypothetical population: For “true” relative risks of 1.5, 4, and 9, the “observed” relative risks were 1.33, 2.48, and 3.58, respectively. Conclusions Misclassification of current use of oral GCs was low but sufficient to lead to significant bias. Researchers should take care to assess the likely impact of exposure misclassification on their analyses.

Prenatal medication exposure and epigenetic outcomes: a systematic literature review and recommendations for prenatal pharmacoepigenetic studies

“…When defining medication-exposed comparison groups, more than half of the reviewed studies relied partly or entirely on self-reported medication use during pregnancy [29,35,37,38,43,[45][46][47][48]51]. This measure does not necessarily reflect the actual medication use [66][67][68] and is vulnerable to recall bias if reported retrospectively [69]. In five studies, medication exposure was assessed at birth using maternal and/or neonatal blood concentrations of the medication [31,34,37,47,51].…”

Modeling of cumulative effects of time-varying drug exposures on within-subject changes in a continuous outcome

“…3 Large administrative health databases typically record only drug dispensations and not their actual use, which can lead to biased estimates and even spurious conclusions. 64 In general, some un-observed discrepancies between daily doses actually taken by a patient and the expected dose, reconstructed from the CPRD prescriptions database, are likely to occur due to a variety of patient-specific patterns of total or partial treatment non-adherence, and probably attenuated the strength of the estimated association. 64,65 In contrast, the measurement errors in the outcome are unlikely to have a material impact on our results as, in clinical practice, the body weight is typically measured with absolute errors of only 0.4–0.6 kg.…”

Long-term oral prednisolone exposure in primary care for bullous pemphigoid: population-based study

“…These data were often missing when prescriptions were issued with information restricted to the free-text field, such as ‘Take as indicated by your dermatologist’. Implausible and missing values were handled using the DrugPrep algorithm, 28 with the decisions described and validated by Joseph et al 29 (see Supplementary Table S2). Cleaning oral glucocorticoid prescriptions in this way has been found to have a sensitivity of 84.2% (95% CI = 68.7% to 94.0%) and a specificity of 87.5% (95% CI = 73.2% to 95.8%) for predicting patient-reported current glucocorticoid use.…”

“…Cleaning oral glucocorticoid prescriptions in this way has been found to have a sensitivity of 84.2% (95% CI = 68.7% to 94.0%) and a specificity of 87.5% (95% CI = 73.2% to 95.8%) for predicting patient-reported current glucocorticoid use. 29 …”

“…These data were often missing when prescriptions were issued with information restricted to the free text field, such as "Take as indicated by your dermatologist". Implausible and missing values were handled using the DrugPrep algorithm(17) with the decisions described and validated by Joseph et al (18) (Supplementary Table S2).…”