Drug interactions avoided – a useful indicator of good prescribing practice

Abstract

Aims

To develop an index of quality prescribing in general practice by investigating the incidence of potential drug interactions when medicines were coprescribed within the State supported General Medical Services (GMS) in Ireland.

Methods

We determined an odds ratio (OR), as a measure of the relative risk of being exposed to a potential interaction, comparing the use of the H₂-receptor antagonist, cimetidine, with that of the noninteracting agents ranitidine, famotidine and nizatidine in users and nonusers of warfarin, phenytoin and theophylline.

Results and conclusions

In 86 510 prescriptions for the H₂–receptor antagonists potentially interacting drugs were dispensed to 8188 (9%) patients in the Eastern Health Board Region of the GMS. We found that prescribers were significantly less likely to use cimetidine (OR = 0.20,95% CI 0.17–0.21, P < 0.001) in those patients who were coprescribed warfarin, suggesting good prescribing practice within the GMS. Similarly there was preferential use of the noninteracting H₂–receptor antagonists in patients receiving phenytoin or theophylline and the extent of this selective prescribing was in keeping with the rank order of severity of interaction with these drugs. This novel pharmacological index may be a sensitive marker of good prescribing practice.

Introduction

Hitherto pharmacoepidemiology has concentrated more on the quantity than on the quality of prescribing. Prescribing quality has become an important issue in the provision of health-related services. Drug prescribing is an integral part of the health care service and represents a relatively safe, effective and inexpensive mode of treatment. Therefore some measure of the quality of drug prescribing is needed. The majority of indicators of prescribing quality assume that a consensus has been reached about evidence-based prescribing. An indicator has been defined as a measurable element of practice performance for which there is evidence or consensus that it can be used to assess the quality, and hence change in the quality of care provided [1]. Prescribing indicators can be used as objective measures of prescribing allowing comparison between different prescribers. They may be used to improve the quality and effectiveness of prescribing in general practice and also to reduce prescribing costs.

While a broad range of measures to assess quality prescribing have been advocated such as the ratio of inhaled corticosteroid to bronchodilator, the percentage of drugs which are prescribed generically or the use of drugs of questionable therapeutic efficacy [2, 3], recently it has been suggested [4] that potentially serious interactions might be included as an indication of prescribing appropriateness. Such interactions have both clinical and financial implications. The Adverse Drug Event prevention Study estimated an overall event rate of 6.5 per 100 admissions in US hospitals, of which 28% were judged preventable [5]. In a recent hospital based study, drug interactions, which were deemed preventable accounted for 4.6% of all adverse drug events [6]. In a related study [7] the cost of medical errors was estimated to be $4685 per preventable adverse drug event although the relative cost of interaction based events was not computed, nevertheless emphasising the financial impact of adverse drug events.

We therefore attempted to develop a quantifiable index of prescribing using a pharmacoepidemiological database based on the potential for interactions. A previous study had identified a significant level of potentially dangerous drug interactions with warfarin [8]. Evidence of good prescribing practice can be inferred where a prescriber chooses as cotherapy within a particular drug group (for example H₂-receptor antagonists), a noninteracting rather than an interacting drug. For a sensitive index it is important that the drug(s) are in common use and for the interaction to be clinically relevant. Warfarin was chosen because of its potential for serious drug interactions.

Cimetidine, but not ranitidine, famotidine or nizatidine inhibits the metabolism of not only warfarin, but also phenytoin and theophylline [9]. As a measure of good prescribing we examined the relative risk of being prescribed an interacting combination by determining an odds ratio comparing coprescribing with cimetidine to that of other noninteracting H₂-receptor antagonists.

Methods

The General Medical Services (GMS) scheme provides free health service to ≈ 35% of the Irish population and is described in detail elsewhere [10]. Eligibility for free health care is means tested, and is confined to those who are unable without undue hardship to arrange general practitioner services for themselves and their dependants. All medicines are dispensed to such patients without charge. While the GMS population cannot be regarded as representative of the general population since socially disadvantaged persons, children and the elderly are overrepresented, however, it comprises just over a third of the entire Irish population who receive some 70% of all medicines prescribed in general practice [11] and thus represents a valuable resource for epidemiological purposes. Ireland is divided into eight regions known as Health Boards for administration of health services. Prescriptions are dispensed through community pharmacies operating within the scheme and a computer system is used for processing pharmacists' claims which in addition to providing details on prescription claims also contains unlike PACT data, demographic data on patients such as age and sex.

The region with the largest population is the Eastern Health Board (EHB) which includes the counties of Dublin, Wicklow and Kildare. The number of eligible persons within this Health Board was 338 025 for the year ending 1998. The total number of eligible persons within all the Health Board Regions at the end of 1998 was 1.2 m and the total cost of prescriptions for the entire GMS scheme in 1998 amounted to £206.9 m [12].

Results

A total of 86 510 prescriptions were dispensed for the H₂-receptor antagonists in 1998 in the EHB region of the GMS. This number represented 1.6% of all prescriptions dispensed or 8.6% of those patients who received prescriptions in that year. An odds ratio was derived comparing the ratio of the use of the interacting drug cimetidine with that of the noninteracting H₂-receptor antagonists in the population at risk and those not at risk of the potential drug interaction (users and nonusers of warfarin, phenytoin and theophylline, respectively) (Table 1).

Table 1.

Number of coprescriptions with H₂-receptor antagonists and odds ratios comparing the use of the H₂-receptor antagonists in users and nonusers of warfarin, phenytoin and theophylline for 1998.

	H2-receptor antagonist
	Cimetidine	Ranitidine/Famotidine/Nizatidine	Odds ratio	95% confidence interval	P-value
Warfarin	443	2475	0.20	(0.17–0.21)	< 0.001
Not on warfarin	39865	43727
Phenytoin	252	473	0.60	(0.52–0.70)	< 0.001
Not on phenytoin	40056	45729
Theophylline	1595	2950	0.60	(0.57–0.64)	< 0.001
Not on theophylline	38713	43252

An odds ratio of 0.5 indicates that the interacting drug cimetidine is half as likely to be coprescribed compared with the noninteracting drugs.

Prescribers were six times less likely to prescribe cimetidine in users of warfarin and half as likely in users of phenytoin and theophylline.

In addition we determined odds ratios for each month in 1998 to determine whether there was any seasonal variation in the results (Figure 1). While the values for users of warfarin and theophylline remained consistent, the odds ratios for users of phenytoin were variable largely due to the smaller number of patients on this combination.

Figure 1.

Odds ratio for users of warfarin (□), theophylline (▵) and phenytoin (○) for 1998

Discussion

Prescriptions for medicines should be necessary, safe, efficacious and economical. However, in practice there is considerable variation in prescribing standards. There is a need for agreed standards of prescribing performance, both to inform individual prescribers and to provide assurance that public expenditure is used efficiently.

In 1994 the Audit Commission recognized that general practitioner prescribing could not be considered in isolation from other general practitioner activities and services [13]. It proposed that more rational prescribing by general practitioners would lead to better quality care for patients and to major economies in drug expenditure. More recently performance indicators which include prescribing indicators for primary care groups, using an evidence-based approach have been described [14–16]. Such indicators cover not only measures of efficiency, but also measures of access to health care and effective delivery of appropriate health care to patients. Indeed the National Primary Care Research and Development Centre advocates a wide range of measures for assessing quality in general practice. It emphasizes the need for Primary Care Groups (PCGs) to develop their own priorities for quality improvement. In addition Campbell et al. have assessed the validity of quality indicators being proposed for use in general practice by health authorities [17]. Indicators of prescribing quality are therefore of importance and may be divided broadly into three categories [18].

1. Indicators which assess the appropriateness of prescribing specific drugs or combinations in selected conditions where there is a sound evidence base, e.g. ACE inhibitors for cardiac failure, anticoagulation with aspirin/warfarin for atrial fibrillation.

2. Descriptive prescribing indicators, which do not attempt to define optimal values, e.g. number of items prescribed per patient.

3. Indicators which are based on unnecessary or potentially harmful prescribing, e.g. duplication of medications, drug interactions.

The former, however, requires knowledge of individual patient diagnosis and only the latter two categories are amenable to study with currently available prescribing databases, which do not include diagnosis or indication for treatment. Since drug interactions may prove costly in financial terms, their avoidance indicates cost effective prescribing.

The prescribing index described is both qualitative and quantitative and suggests selective prescribing to avoid therapy-related toxicity. Furthermore it provided consistent results for warfarin and theophylline on a monthly basis albeit to a lesser extent with phenytoin due to the relative small number involved in such a short period. Recently Suissa has described a relative excess risk as an alternative measure of comparative risk which may be applied to pharmacoepidemiological studies [19].

The odds ratios, which we obtained, indicate that for patients receiving warfarin, phenytoin or theophylline, there was a significant shift to the use of the noninteracting ranitidine, famotidine or nizatidine compared with the use of cimetidine when the choice of an H₂-receptor antagonist was made. Furthermore the extent of this selective prescribing is in keeping with the rank order of severity of interaction with these drugs. While these drugs have a narrow therapeutic ratio and for warfarin, interactions have led to life-threatening haemorrhage, for phenytoin (gait disturbance) and theophylline (nausea and tachycardia), the interaction has less severe consequences [20–22].

The odds ratio developed in this study, although explored in one national database, we believe may be applied to other health care systems and drug groups where a physician has a choice between an interacting and noninteracting drug as a sensitive marker of good prescribing practice. This indicator should be applied to other prescription databases and incorporated into the range of quality indices used to audit prescribing in primary care.