Accuracy of a Prescription Claims Database: DISCUSSION part 2

16 Nov
2010

Several previous studies have evaluated agreement between prescription records and interview-based medication history. However, those studies involved databases that were not centralized between pharmacies and were not available for general clinical applications beyond the individual pharmacy where the data were stored. In addition, dosing discrepancies were not reported, nor were reasons for discrepancies evaluated, in any of the studies reviewed. Sjahid and others17 evaluated the accuracy of information obtained from 1682 patients over 55 years of age in a Dutch population- based study database. These investigators found 80.4% concordance with interview-based medication history lists. Lau and others reported that 70% of medications present in a home inventory of 115 elderly patients were “active” according to pharmacy records and 96% were listed somewhere on the pharmacy record; however concordance was not calculated. In a study of antihyper- tensive drugs, there was full agreement between pharmacy records and a patient questionnaire for 321 (86%) of 372 patients interviewed. However, the defini­tion of agreement included instances in which specific antihypertensive drugs were listed neither by the patient nor in the pharmacy records; this category applied for most patients with “full agreement” (207/321 or 64%). More recently, Kaboli and others interviewed 493 patients receiving primary care through US Veterans Affairs and found complete agreement between the computerized medication profile and a medication interview list for only 5.3% of patients. Finally, Caskie and Willis evaluated agreement between an interview-based medication list and pharmacy records for 294 patients participating in a clinical trial evaluating cognitive triaging in elderly people who did not have dementia. Agreement was 49% to 81% depending on the drug class, but this was examined simply at the level of drug class, without taking into account the specific medications consumed. Notably, in all of these studies, including the present investigation, the computerized record under-reported the number of prescription medications used.
erectalis tablets

This study had several limitations. The medication his­tory information used for assessment of accuracy of the PharmaNet profile relied largely on patient responses to a survey that has not been validated scientifically. In particular, medications not filled at a British Columbia community pharmacy (e.g., drug samples provided by physicians) and not presented or mentioned at the interview could not be identified. On the other hand, focusing on medications used up to and including the day of the interview minimized recall bias. Furthermore, the use of multiple sources of medication history information (e.g., patient, spouse, PharmaNet profiles, vial labels, chart records) limited the possibility of patients overlooking medications that they were currently using. Medications were excluded from the analysis if it was not possible to determine whether they were currently active based on refill data (e.g., creams, liquids, and medication taken “as directed”). It is possible that additional discrepancies would have been identified if these medications had been evaluated. The accuracy of pharmacists’ transcription of prescription information into the PharmaNet database was not assessed; however, previous research evaluating other Canadian prescription claims databases indicates that this process is reliable. Previous research involving this clinic cohort has revealed a high rate of medication adherence, which might have influenced agreement between the 2 sources of medication history information. The attempt to recruit nonclinic patients through local community pharmacies, so that a wider range of adherence patterns would be represented in the study, was only marginally successful. Neither outcomes nor the potential for adverse consequences were assessed, which rendered it difficult to interpret the clinical relevance of the discrepancies identified. Lastly, this study evaluated a single prescription database for a specific cohort of patients. These results may not be applicable to other databases, where prescription information is entered under different circumstances, or to patients treated for different disease states in different settings. cialis super active

CONCLUSIONS

Most PharmaNet profiles reviewed for this cohort of patients with heart failure contained some inaccurate or misleading information, thus revealing PharmaNet’s short­comings as a principal source of information about patients’ prescription medication history in this setting. However, these results also suggest that PharmaNet can be an invaluable tool for improving the accuracy of information obtained during a medication history interview. Streamlining the thorough interview process employed in this study by briefly reviewing the PharmaNet profile with the patient, asking about overdue refills and side effects, and querying the use of samples and nonprescription medications appears to be an effective approach for medication reconciliation in this population.

top