Using primary care prescribing databases for pharmacovigilance

Abstract

AIMS

In the UK, adverse drug reactions (ADRs) are responsible for over 6.5% of all hospital admissions, representing a significant morbidity and cost burden to the health service. We aimed to develop an ADR monitoring system capable of identifying the reasons for patient discontinuation of drug therapy within 6 months of the index prescription.

METHODS

Patients first prescribed amlodipine between 1 March 2004 and 28 February 2007 who discontinued their amlodipine medication within 6 months of the index prescription were identified from the practice team information (PTI) database. Once identified, reasons for amlodipine discontinuation were assessed by an electronic database search using relevant Readcodes and key words and by a direct approach to the primary care medical records.

RESULTS

The PTI database identified 995 patients [61.4% females, mean age 65.9 years (SD 12.4 years)] who discontinued amlodipine within 6 months of an index prescription. An electronic search of the database, using Readcodes, identified that 19.4% (193) of patients who discontinued their medication had an ADR recorded in the database. Six (20%) of 30 participating primary care practices, contributing to the PTI database, agreed to be approached directly and supply the reasons for discontinuation for the 51 patients identified as having discontinued amlodipine in their practices. Completed data were returned for all 51 patients, 98% of whom discontinued amlodipine due to an ADR or adverse drug event.

CONCLUSIONS

The results of this study confirm that primary care prescribing databases can be easily used to identify the frequency and nature of ADRs occurring in an ADR-enriched population identified through medication discontinuation.

WHAT IS ALREADY KNOWN ABOUT THIS SUBJECT

• The ‘Yellow Card Scheme’ is the main adverse drug reaction monitoring system in the UK.

• The Yellow Card Scheme, like other passive systems, suffers from under-reporting, long lag time for signal detection and generation of appropriate warnings, no precise denominator and lack of reports for special patient groups, such as children and the elderly.

WHAT THIS STUDY ADDS

• Routinely collected general practice prescribing data can be used to assess drug discontinuation rates/medication persistence in different patient populations.

• Routinely acquired prescribing data can be used to identify the specific reasons for patients discontinuing a chronic medication.

• Using the computerized practice team information prescribing database, the reasons for drug discontinuation were recorded for 20% of patients; however, reasons for discontinuation were recorded for 100% of patients in primary care medical records.

• Reasons for discontinuing prescriptions within 6 months from the index prescriptions can be used for pharmacovigilance and complement current adverse drug reaction monitoring systems.

Introduction

An adverse drug reaction (ADR) is a response to a medicine which is noxious and unintended, and which occurs at doses normally used in humans for the prophylaxis, diagnosis or therapy of disease, or for the modification of physiological function [1–3]. An adverse (drug) event (AE) or experience is defined as any untoward medical occurrence that may present during treatment with a medicine but which does not necessarily have a causal relationship with the treatment [1–3]. Adverse drug reactions are a common and often preventable cause of illness and disability, ranking among the top 10 causes of mortality in the USA, UK and Europe [4]. In the UK, ADRs are responsible for 6.5% for all hospital admissions, representing a significant cost burden to the health service [5]. In the UK, the principal system for pharmacovigilance, ‘the Yellow Card Scheme’, relies for success and completeness on the spontaneous reporting of possible ADRs. This process relies on the willingness of individuals, especially healthcare professionals, to report suspected ADRs for use by the drug regulatory authority in the pharmacovigilance process; however, healthcare professionals only report approximately 6–9% of those ADRs which should be reported [6–9].

Drug discontinuation or drug therapy discontinuation refers to the end of a patient's current drug therapy prescription. Drug discontinuation may also be described as planned or unplanned/abrupt. Planned discontinuations, for instance the end of a course of antibiotic therapy, do not usually signify a problem; however, unplanned drug discontinuations, for instance the discontinuation of an antihypertensive drug intended for lifelong therapy, may indicate a problem, such as lack of efficacy, poor adherence or the occurrence of an ADR or AE. Drug therapy persistence implies continuous refill of a prescribed drug (usually a long-term drug, such as an antihypertensive drug) until permanent discontinuation or end of a prescription period [10]. Persistence with medication is a key factor in ensuring the successful outcome of therapy in the treatment of chronic disease states; however, the assessment of drug persistence during tightly regulated clinical trials bears little relevance to the real-life situation. The value of assessing drug persistence during randomized clinical trials is limited because of short study duration, small number of study patients, close patient scrutiny and the frequent absence of special patient groups, such as children, or those with co-morbidities or concurrent multiple medications. While long-term clinical trials suggest that the frequency of discontinuation for antihypertensive medications is in the range of 5–10% per annum, in routine clinical practice the figure has been reported to be significantly greater at 20–60% [11–17]. Burke et al., using discontinuation data obtained from the UK general practice research database (GPRD), reported a mean risk for discontinuation of any antihypertensive drug of 20.3% [95% confidence interval (CI), 20.0–20.5] at 6 months and 28.5% (95% CI, 28.2–28.7) at 1 year [11]. Similar discontinuation figures of 11.9% within 1 month of the index prescription, 23.8% within 3 months and 43.3% by 3 years were reported by Gregoire et al. for a prospective cohort study involving 173 Canadian pharmacies [12]. Similarly, Bloom et al. reported that the levels of drug persistence at 1 year varied between 38% for thiazide diuretics and 64% for angiotensin II receptor blockers, with other agents lying in between (angiotensin-converting enzyme inhibitors 58%, calcium channel blockers 50% and β-blockers 43%) [13]. Similar levels of persistence have been reported in a number of observational studies of antihypertensive medication [14–17]; however, none has determined the reasons for medication discontinuation, although lack of efficacy or adverse drug effects are the most commonly postulated reasons.

Hypertension treatment guidelines recommend monitoring a patient's response to any new antihypertensive agent and titrating the dose over time. It is therefore unlikely that a patient will discontinue an antihypertensive medication due to a lack of efficacy within the first 6 months of the index prescription, before the titration process has been completed. In a previous study, we demonstrated that discontinuation rates could be used as a surrogate marker for ADR/AE rates [18]; however, we were unable to identify the reasons for such discontinuations. The aims of the present study, therefore, were to develop an ADR monitoring system capable of identifying the reasons for patient discontinuation of drug therapy within 6 months of the index prescription. Using amlodipine as the exemplar (amlodipine was chosen as the exemplar because it has a well-established safety profile, a high frequency of use and has a mid-range 6 month discontinuation rate in between that of angiotensin II receptor blockers and thiazide diuretics of 15–60% [13–16]), we proposed two approaches centred on the use of an electronic prescribing database to identify a patient population who had discontinued a medicine of interest within 6 months of the index prescription, followed either by direct extraction of relevant data regarding the reason for discontinuation from a primary care electronic healthcare database or by direct extraction of relevant data from the individual patient's primary care medical records held by their respective primary care physician.

Methods

All patients who had discontinued amlodipine within 6 months (≤185 days) of the index prescription were identified from the computerized practice team information (PTI) database. The PTI database is an electronic database which contains routinely collected primary care healthcare data, including prescribing information for approximately 224 000 patients registered with a primary care physician, and is representative of the Scottish population with regards to age, sex, deprivation and urban/rural ratio mix [19].

Patient prescribing and contact data were extracted from the PTI database for the 5 years from 1 March 2003 to 29 February 2008. To determine accurate start and discontinuation dates, the periods 1 March 2003 to 28 February 2004 and 1 March 2007 to 29 February 2008 were used as washout periods, ensuring that each patient had a minimum 12 month study drug-free period prior to their index prescription and following their last prescription for amlodipine.

All patients first prescribed amlodipine between 1 March 2004 and 28 February 2007 were identified and then tracked as an exposed cohort. Any records for apparently inactive patients (e.g. dead or moved from practice) were excluded from the analysis. All patients who discontinued their amlodipine medication within 6 months of the index prescription were then identified. Following identification of these patients, the following two approaches were used to determine the reasons for discontinuation.

1. A PTI electronic database search for the reasons for discontinuation using relevant Readcodes (standard of coding for disease, investigation and clinical sign and symptoms) and free text search for adverse drug reactions and the reasons for discontinuation [20, 21]. All clinical event entries after discontinuation of drug of interest were scrutinized.

2. A direct approach to a patient's primary care physician to determine the reasons, if any, for discontinuation recorded in the primary care practice medical file/records. Invitations to participate were sent to 30 general practices contributing to the PTI database. Practices that agreed to participate were sent a list of their patients who had discontinued amlodipine within 6 months of the initial prescription. Patients were identified by a unique identifier known only to their primary care practice. The reasons for drug discontinuation were then reported to the research team by participating practices. No personally identifiable patient information was collected.

Study approval

The project was reviewed by the primary care clinical informatics unit (PCCIU) advisory group [22] and was approved by the North of Scotland Ethics Committees and relevant regional National Health Service (NHS) management.

Results

Interrogation of the PTI electronic prescribing database identified 3646 patients who were initiated on amlodipine during the study period. Of these, 995 (27.29%) patients [61.4% females, mean age 65.9 years (SD 12.4 years)] discontinued amlodipine within 6 months of their index prescription. The discontinuation rates observed were 16.2% within 1 month, 6.3% within 2–3 months and 4.8% within 4–6 months.

Identification of reasons for discontinuation

Direct search of the PTI database using Readcodes and keywords

Approximately one-fifth (193, 19.4%) of patients who discontinued amlodipine had an ADR/AE recorded in the PTI database as the reason for discontinuation. As expected, the most frequent ADR recorded electronically, using Readcodes and keywords, was ankle swelling, which accounted for almost half (48.2%) of all reports. The reasons for amlodipine discontinuation are listed in Table 1.

Table 1.

Reasons for amlodipine discontinuation obtained from the practice team information database and primary care medical records

Recorded reasons for discontinuation of therapy	Number of patients (%) Electronic prescribing database (30 practices)	General practice patient records (6 practices)
Ankle swelling	93 (46.3)	23 (38.3)
Dizziness, headache, head rush, nausea	23 (11.4)	4 (6.7)
Knee swelling, peripheral oedema	–	4 (6.7)
Medication review	–	4 (6.7)
Nonspecific side-effect or unwell	41 (20.4)	4 (6.7)
Back ache, leg pain, leg cramps, loin pain, myalgia	3 (1.5)	4 (6.7)
Dry mouth	–	3 (5.0)
Sweats	–	2 (3.3)
Diarrhoea, indigestion	–	2 (3.3)
Rash, dermatitis, pruritus	23 (11.4)	2 (3.3)
Poor compliance	–	1 (1.7)
Others (flushing, fatigue, vivid dreams etc.)	18 (9.0)	7 (11.7)
Total	201*	60†

^*Two reasons for discontinuation given for eight patients.

^†Two reasons for seven patients and three reasons for one patient.

Primary care medical records

Of the 30 general practices approached to take part in the study, six (20%) agreed to participate. Patients identified from the PTI database as having discontinued amlodipine within 6 months of an index prescription were then matched to these practices. A total of 64 patients [64.7% females, mean age 65.3 years (SD 13.3 years)] registered with these primary care practices fulfilling the inclusion criteria were identified. Participating general practitioners returned data for 51 patients (80%). Thirteen patients were reported as having deregistered, and notes were unavailable. The most frequent reason for discontinuation was ankle swelling/oedema (45%). Reasons for amlodipine discontinuation are listed in Table 1. Reported reasons for discontinuation were then assessed and recorded as an ADR/AE or other. Ninety-eight per cent of patients were recorded in the primary care records as having discontinued amlodipine due to the occurrence of either an ADR or an AE, and one patient due to poor compliance (Table 2). No patients were recorded as having had amlodipine discontinued due to a lack of therapeutic efficacy.

Table 2.

Classification of reasons for discontinuation obtained from primary care medical records (adverse drug reaction, adverse drug event or other)

Classification for discontinuation	Number of patients (%)
Patients discontinued due to adverse drug reaction	44 (86.3%)
Patients discontinued due to adverse drug event	6 (11.7%)
Patient discontinued due to other reason	1 (2%)

Discussion

The discontinuation rates observed for patients prescribed amlodipine of 16.2% within 1 month, 6.3% within 2–3 months and a further 4.8% within 4–6 months are similar to those previously reported [23–27].

Two approaches, the use of Readcodes and keywords to search an electronic prescribing database and direct search of a patient's medical records at source, were used to identify the reasons for drug discontinuation and the nature of possible associated ADRs/AEs. The results of this study confirm that the use of an electronic prescribing database can be used to identify an ADR-enriched patient population (patients who have discontinued a medication of interest) and the reasons for discontinuation, including possible ADRs. As expected, using either approach, the most frequent ADR recorded either electronically or in the patient's medical records was ankle swelling/oedema [26, 27]. However, direct search of a patient's medical records provided more detailed and comprehensive data relating to drug discontinuation. We have previously reported that electronic prescribing databases can be used to identify patients who have discontinued their medication, and that such drug discontinuation rates can be used as a surrogate for ADR/AE rates [18]. However, this approach does not identify the reasons for discontinuation or the nature of possible ADRs/AEs. We therefore developed the current approach to permit identification of possible ADRs/AEs directly from the prescribing database or from patient medical records.

While interrogation of the electronic prescribing base identified a large number of patients who had discontinued their amlodipine within 6 months, this database was only able to provide reasons for discontinuation for approximately one-fifth of patients. As the prescribing database was designed as a general practice administrative tool rather than for pharmacovigilance or clinical/therapeutic purposes, recording the reasons leading to discontinuation of therapy is not compulsory. Furthermore, when a reason was recorded, free text rather than a Readcode [20, 21] was used, and the direct link between medication and possible ADR/AE was difficult to establish.

As a result of these limitations, we proceeded to the extraction of possible ADR/AE data directly from patient medical records stored in primary care practices. This approach permitted the accurate assessment of the reasons for drug discontinuation in almost all identified patients, with 98% of discontinuations being linked to an ADR/AE. Lack of efficacy was not given as a reason for discontinuation, confirming our original hypothesis that early drug discontinuation is most likely to be due to an ADR/AE.

Strengths and weaknesses

Both approaches to the identification of ADRs/AEs we describe have strengths and weaknesses. While the use of an electronic prescribing database readily permits the identification of a large number of patients who have discontinued their medication, generic reasons for discontinuation were only recorded for approximately 20% and specific reasons, such as ankle swelling, for 15% of patients. Furthermore, reasons for discontinuation were recorded as free text, making automated electronic word searches difficult. Finally, the temporal association between reasons for discontinuation and the drug of interest proved difficult to ascertain. Although these limitations are apparent, this approach was able to detect a large number of possible ADRs/AEs. The use of larger primary care patient databases, such as the general practice research database (GPRD) [28], with over 3.6 million active patients, or general practice administration system for Scotland (GPASS), with over 2.5 million active patients [29, 30], should prove more effective and yield a large number of reports.

The most reliable and readily assessed source of ADRs/AEs was the primary care patient medical records. Practices agreeing to participate were able to supply the exact reasons for patient discontinuation. The major weakness of this approach, however, was poor primary care practice participation, reportedly due to a lack of reimbursement.

The challenge is to develop a pharmacovigilance system with improved levels of reporting, reduced cost and resource requirement, improved cover of special populations, such as children, and reduced lag time between signal detection and regulatory action [7–9, 31, 32]. The approaches to ADR monitoring we have described readily permit selective targeting of specific medicines and patient groups. Use of healthcare databases in the way we describe to identify a population which is ADR/AE-enriched should increase the efficiency and the speed of ADR/AE detection and decrease the lag time between ADR detection, generation of appropriate warning and regulatory action.

Competing Interests

There are no competing interests to declare.