Original ArticleAgreement between self-reported data on medicine use and prescription records vary according to method of analysis and therapeutic group
Introduction
Prescription records and self-reported data from surveys are often used to measure various aspects of the population's medicine use patterns at the individual level [1], [2]. However, little is known about the agreement of these two data sources: which is preferable, and in what situations?
Most prescription records have the advantage of being administered electronically and thus accuracy is high [3], [4]. Prescriptions can be collected at the general practitioner level or at the pharmacy level. If the register has full cover, the pharmacy level will be more accurate [4], as not all medicine prescribed by a general practitioner is purchased at the pharmacy, and further, not all medicines dispensed are actually used. Reviews have reported that adherence among patients suffering from chronic diseases such as hypertension, depression, and asthma averages only 50% [5], [6]. Prescription records do not include information on purchase of over-the-counter (OTC) medicine, and information on individuals is usually sparse.
Due to the high accuracy, prescription records are usually seen as the gold standard compared with self-reported medicine use [4].
Self-reported data provide information on which medicines are actually used, but self-reporting is prone to various forms of bias (e.g., recall bias) and respondents may be reluctant to report sensitive data such as medicine use [7], [8]. Surveys have the advantage of possible inclusion of extensive information on the respondent such as use of OTC medicines, herbal medicines, and other types of health care services as well as disease and sociodemography.
Few large, population-based studies dealing with the correlation of registers and self-reported data are reported in international literature. Most studies are rather small and often conducted in specific populations such as participants in case-control studies or specific patient and age groups [9], [10], [11]. We have identified only two large, population-based studies comparing self-reported data on medicine use and pharmacy records [12], [13].
In a population-based Dutch cohort (n = 7,568), sensitivity of pharmacy records was calculated with a questionnaire as the gold standard and using a 1-year fixed-time window [12]. Good sensitivity was found for antihypertensives, lipid lowering drugs, oral contraceptives, hormone replacement therapy (HRT), and oral antidiabetics, whereas nitrates and OTC painkillers had low sensitivity [12]. A population-based American study (n = 1,430) found that the congruence between self-reported medicine use and pharmacy records was high at the overall level using a legend time method analyzing the pharmacy records [13]. Congruence was better for medicines used for serious conditions or on a regular basis [13].
The Danish Register of Medicinal Product Statistics provides a unique opportunity to study individual medicine use at the national level, as the register covers all Danish pharmacies and thus captures all prescriptions purchased by every resident of the country in a Danish pharmacy.
The aim of this study was to analyze the agreement of data on medicine use at the individual level by comparing national self-reported data and national prescription records.
No study has previously applied national representative data when dealing with this topic.
Section snippets
Methods
Data on self-reported medicine use were retrieved from The Danish Health and Morbidity Survey 2000. The survey was conducted by the National Institute of Public Health on a large sample of the Danish population aged 16 years and above (n = 22,486). Respondents were estimated to be representative of the Danish population. The respondents were interviewed in their homes by trained interviewers, and the response rate was satisfactory (74.2%, n = 16,688) [14]. The respondents were sampled irrespective
Method (a)—legend time method
Table 1 shows the point prevalence detected in the two data sources for different ATC groups.
The lowest point prevalence for the register and the survey was detected for antimigraine preparations (0.3 and 0.7%, respectively). The highest prevalence in the register was found for HRT among women aged 45 years and above (15.8%), and in the survey for medicines for the cardiovascular system (12.8%) (Table 1).
Agreement ranged from fair to very good [19]. The lowest agreement between the two data
Discussion
As far as we know, this is the first study on this subject in a general population at the national level, adding knowledge to the smaller and more specific studies previously conducted. The main finding is that agreement between self-reported data and prescription records differed according to method of analysis and therapeutic group. Applying the legend time method to medicine groups mainly used on a chronic basis revealed good to very good agreement between the two data sources, whereas
References (34)
- et al.
Identifying parents in pharmacy data: a tool for the continuous monitoring of drug exposure to unborn children
J Clin Epidemiol
(2004) - et al.
A review of the use of health care utilization databases for epidemiological research on therapeutics
J Clin Epidemiol
(2005) - et al.
Validation of pharmacy records in drug exposure assessment
J Clin Epidemiol
(1997) - et al.
An interview strategy was critical for obtaining valid information on the use of hormone replacement therapy
J Clin Epidemiol
(2004) - et al.
Influence of question structure on the recall of self-reported drug use
J Clin Epidemiol
(2000) - et al.
Prescription and non-prescription medicine use in Denmark: association with socio-economic position
Eur J Clin Pharmacol
(2003) How should one perform pharmacoepidemiology studies? Choosing among the variable alternatives
- et al.
Interventions for enhancing medication adherence
The Cochrane Library
(2006) Adherence to long-term therapies—evidence for action
WHO
(2003)- et al.
Validity of pharmacoepidemiology drug and diagnosis data
Survey methodology
Agreement between questionnaire data and medical records
Am J Epidemiol
A validation study of patient interview data and pharmacy records for antihypertensive, statin, and antidepressant medication use among older women
Am J Epidemiol
Reliability of recall of drug usage and other health-related information
Am J Epidemiol
Pharmacy data in epidemiological studies: an easy to obtain and reliable tool
Pharmacoepidemiol Drug Saf
Congruence of medication information from a brown bag data collection and pharmacy records: findings from the Seattle Longitudinal Study
Exp Aging Res
Sundhed og sygelighed i Danmark 2000. & udviklingen siden 1987
Cited by (70)
How Many Steps Per Day During the Early Postoperative Period are Associated With Patient-Reported Outcomes of Disability, Pain, and Opioid Use After Lumbar Spine Surgery?
2021, Archives of Physical Medicine and RehabilitationCitation Excerpt :Participants were asked if they were currently taking opioid medication at 1 year after surgery. The response to the question was recorded as yes or no. Prior work suggests that self-report is comparable with data collected from a Prescription Drug Monitoring Program or national prescription records.24,25 Daily walking at 6 weeks after spine surgery (ie, baseline) was objectively measured using an accelerometer (ActiGraph GT3Xa).
Self-reported medication use among coronary heart disease patients showed high validity compared with dispensing data
2021, Journal of Clinical EpidemiologyCitation Excerpt :Results are less consistent for medications used as needed [9,10,12,13]. Cardiovascular medications, such as antihypertensive drugs, lipid-lowering drugs (LLDs) and antiplatelet drugs, are normally used on a daily basis, and agreement and validity between different data sources are usually found to be high [4–7,9,10,12–17]. However, few studies have investigated this in a population with established coronary heart disease (CHD) or compared data from complete prescription registries with self-reported data from a large population study.