Prevalence of potentially inappropriate long term prescribing in general practice in the United Kingdom, 1980–95: systematic literature reviewBMJ 1996; 313 doi: https://doi.org/10.1136/bmj.313.7069.1371 (Published 30 November 1996) Cite this as: BMJ 1996;313:1371
- Stephen A Buetow, research fellowa,
- Bonnie Sibbald, senior research fellowa,
- Judith A Cantrill, clinical senior lecturerb,
- Shirley Halliwell, researchera
- a National Primary Care Research and Development Centre, University of Manchester, Manchester M13 9PL
- b Department of Pharmacy and National Primary Care Research and Development Centre, University of Manchester
- Correspondence to: Dr Buetow.
- Accepted 8 October 1996
Objective: To determine the prevalence of potentially inappropriate long term prescribing in general practice in the United Kingdom.
Design: Review of 62 studies of the appropriateness of prescribing identified from seven electronic databases, from reference lists, and by hand searching of journals. A nominal group of 10 experts helped to define the appropriateness of prescribing.
Setting: General practice in the United Kingdom.
Main outcome measures: Prevalences of 19 indicators of inappropriate long term prescribing representing five dimensions: indication, choice of drug, drug administration, communication, and review.
Results: Prevalences of potentially inappropriate prescribing varied by indicator and chronic condition, but drug dosages outside the therapeutic range consistently recorded the highest rates. The lowest rates were generally associated with indicators of the choice of the drug, except cost minimisation. Communication is studied less frequently than other dimensions of prescribing appropriateness.
Conclusions: The evidence base to support allegations of widespread inappropriate prescribing in general practice is unsound. Although inappropriate prescribing has occurred, the scale of the problem is unknown because of limitations associated with selection of a standard, publication bias, and uncertainty about the context of prescribing decisions. Opportunities for cost savings and effectiveness gains are thus unclear. Indicators applicable to individual patients could yield evidence of prescribing appropriateness.
The United Kingdom government has emphasised its determination to reduce costs and volumes of prescription drugs and improve prescribing quality.1 2 National initiatives, such as the Prescription Pricing Authority's reports on prescribing by individual general practitioners, have sought to improve general practitioners' prescribing practices and contain drug expenditure. However, wide variations in prescribing rates and costs have persisted geographically and in practices1 3 and been only partly explained.4 5 Lack of consensus among doctors about the best way to practise medicine may help to account for the unexplained variations6 as general practice lacks accepted standards of appropriate prescribing.7
General practitioners have been rebuked, nevertheless, for practices such as overprescribing drugs of limited clinical value and underprescribing generic drugs and inhaled steroids for asthma.1 Such criticisms lack credibility.8 Because prescribing in general practice is often long term9 we systematically reviewed published work on the prevalence of potentially inappropriate long term prescribing in general practice in the United Kingdom.
The medication appropriateness index is the only generic measure of prescribing appropriateness with documented feasibility, reliability, and internal validity,10 but its generalisability to general practice in the United Kingdom is unknown. To define prescribing appropriateness in British general practice we set up a panel of 10 experts. Nominated by national opinion leaders in primary care, they represented professional, patient, research, and policy perspectives of prescribing.
Because of the incomplete evidence for assessing the quality of prescribing the panellists met as a nominal group. Their one day remit was to use professional opinion to suggest consensus based criteria for screening the appropriateness of prescribing recorded in individual patients' medical notes. The group functioned through a highly controlled process that facilitates idea generation but has unknown reliability and validity.11 Using transcripts of the meeting, we refined the 19 indicators of prescribing appropriateness in general practice that were identified as being relevant to individual patients and as not specific to a given disease or drug. They cover five dimensions: indication, choice of drug, drug administration, communication, and review (table 1).
ELIGIBILITY OF STUDIES
Unlike Cochrane style reviews of the effects of an intervention, our review focuses on the prevalence of inappropriate prescribing. We sought studies published in English, or with an English abstract, between 1980 and July 1995 owing to the lesser relevance of earlier studies. Studies were excluded if they did not explicitly refer to drug treatments,12 13 14 15 as in diabetic care16 17 18 19 20 21 22 23 24; if they referred to prescribers' knowledge of and attitudes towards prescribing rather than their prescribing practices25 26 27 28 29 30; if they did not make clear the doctor's prescribing role in shared care arrangements31 32 or after discharge of patients from hospital33; or if they had poor or uncertain validity because information on appropriateness was unclear34 35 36 or unsupported by the British National Formulary.37 38 39
A search of our personal reference files was followed by a computerised search of seven electronic databases including Medline. Because none of its medical subject headings directly encompasses the subject area of our review we used combinations of such headings, text words, and wild cards. We also identified studies from the online public access catalogue in the Manchester region. The British Journal of General Practice and Family Practice were searched by hand by scanning their contents pages for key words and reading the abstracts of relevant articles. Experts were consulted through, for example, the electronic list server GP-UK. We scrutinised the reference lists of retrieved studies.
STANDARDS OF PRESCRIBING APPROPRIATENESS
We assessed evidence for violations of each indicator (table 1). Whenever possible our standard reference was the March 1995 issue of the British National Formulary because it is an authoritative source of drug information, widely available, and amenable to rapid reference.40 When information in the British National Formulary was lacking on communication and effectiveness we had to judge whether our appropriateness criteria had been met.
Table 2 is not published here owing to its large size but is available from SAB and on the BMJ's worldwide web page (www.bmj.com/cgi/content/full/313/7069/1371/DC1); it shows the prevalences of potentially inappropriate long term prescribing in general practice in the United Kingdom during 1980–95. Rates were extracted from 62 studies, of which two thirds were published in the BMJ or British Journal of General Practice. Table 2 shows 90 citations because eligible studies could be cited more than once. Also, 12 citations provided more than one rate.
Constructs underpinning the rates arose from published work and help to validate the dimensions suggested by the expert panel. Because our aim was to report the best evidence representing different dimensions of appropriateness, we have cited rates from up to five studies of the same violation. We omitted two out of seven eligible studies of inadequate inhaler technique41 42 and four out of nine studies of excessive durations of benzodiazepine prescribing.43 44 45 46 Omitted studies were older or were of lower quality than those we retained.
Table 2 also shows the prevalences of violations of 16 of our 19 indicators. No eligible rates were identified for drugs of limited value or drug regimen, and measures of cost effectiveness are beyond the scope of the review because they cannot be easily referenced against the British National Formulary or alternative standards. At least 10 studies represent each of our five dimensions of inappropriate prescribing. Rates from 53 of the 90 citations were based entirely or in part on information in general practice medical records, and about half of the practice based citations concerned single practices. Most subjects were elderly in about a quarter of the citations.
METHODOLOGICAL QUALITY OF STUDY EVIDENCE
Each rate was scored 1 or 0 on each of eight quality criteria (table 3), according to whether the criterion was met. The individual scores were summed to describe the overall ability of the rate to meet our quality requirements. The possible range of total scores from 0 to 8 was categorised into groups of increasing merit: low (0–2), moderate (3–5), and high (6–8).
Of the citations in table 2, 44 contributed rates scoring in the highest third and none had rates in the lowest. Without exception, study populations were clearly defined, and for only 19 citations was the subject selection flawed by the sampling frame or sampling procedure. In general, constructs were defined clearly and when possible in accordance with the British National Formulary. Missing data were rarely reported but selection bias was suggested by participation rates below 80% in 35 citations. Our quality criterion for sample size was met by 22 citations. No rate was generalisable to the population of the United Kingdom.
Prevalence varied by indicator and chronic condition. Drug dosages outside the therapeutic range consistently recorded the highest rates. Except for cost minimisation, the lowest rates were generally associated with indicators of inappropriate choice of drug(s).
In nine papers the indications for drug treatment were invalid or unstated; two related to antihypertensive agents47 48 and one to diuretics.49 The prescription of long acting benzodiazepines was criticised,46 50 and studies identified no valid reason for prescriptions for acid suppressing drugs,51 53 vitamin B-12 injections,54 and concentrators for long term oxygen treatment.55
Of 10 studies reporting general practitioners' failure to prescribe indicated drugs, eight described underprescribing for asthma.56 57 58 59 60 61 62 63 Others recorded no treatment for hypertensive patients.64 65 In six practices 234 out of 281 patients (83%) with heart failure were prescribed loop diuretics by their general practitioner but not angiotensin converting enzyme inhibitors.66
CHOICE OF DRUG
Studies documented underprescribing of generic substitutes49 67 68 69 and prescribing of potentially hazardous drug combinations.70 71 Concurrent prescribing of benzodiazepines45 46 50 72 73 characterised from 6% (25/450)46 to 21% (15/72)73 of patients receiving long term benzodiazepines. Three studies referred to prescriptions for contraindicated drugs but not to prescribers' knowledge and minimisation of risks.50 71 74
Underprescribing contributed to deaths from asthma75 and subtherapeutic carbidopa dosages in patients with Parkinson's disease.76 All 136 users of nitrazepam in 55 residential homes received >5 mg daily.50 Inappropriate drug dosages were inferred from serum concentrations of theophylline,59 77 78 79 thyroid stimulating hormone,80 81 82 digoxin,83 and lithium84 and from international normalised ratios outside normal therapeutic ranges.85 Raised random blood glucose concentrations were reported in patients with insulin dependent diabetes, possibly owing to factors other than insulin dosage.86 87
Excessive durations of benzodiazepine prescribing were reported for 0.4% to 2.6% of all patients in practices.45 72 73 88 89 Median durations of use ranged from 2.546 50 to five years.45 88
In five studies58 90 91 92 93 from 8% (n = 126)93 to 38% (27/71)90 of treated patients had an inadequate inhaler technique. Metered dose inhalers were most commonly used inadequately by patients with asthma (30%, n = 262).91
Patients' understanding of their asthma treatment was often poor,58 94 95 and for 70% (32/46) of patients in one clinic, the general practitioner's version of the drug regimen differed from that reported during interviews with patients.96 Inadequate writing of prescriptions97 was exemplified by a lack of instructions98 and errors in the prescription that were not potentially serious to patients.99 In London 12 007 out of 181 100 (0.07%) anomalies were potentially serious and detectable only by a pharmacist.100
Prescribed drugs failed to control conditions such as hypertension48 64 65 101 and asthma as defined by the results of lung function tests,57 59 nocturnal waking at least once weekly,58 60 and high morbidity.92 102 Our experts recommended holding a review of drug treatment at least annually or in accordance with the British National Formulary. Failure to fulfil this criterion was reported in studies of elderly patients receiving repeat prescriptions103 and patients prescribed antihypertensive drugs,47 64 104 105 digoxin,106 drugs for chronic asthma,107 108 lithium,84 antiepileptic drugs,109 and thyroxine.107
Potentially inappropriate prescribing, in which appropriateness is relative to what prescribers should normally do,110 occurred in general practice during 1980–95. However, the scale of the problem is unknown and subsequent improvements or deterioration would have gone undetected. The external validity of the studies reviewed is unknown, and there is no standard against which we can assess the relative importance of each of our quality criteria. Our assessments are qualified by our choice of standard, publication bias, and uncertainty about the context of prescribing decisions.
There is no saliently better standard for our purposes than the British National Formulary. However, it refers to more drugs than general practitioners could expect to become familiar with; is revised less frequently than, say, the Monthly Index of Medical Specialities; may require supplementation with specialised publications such as manufacturers' data sheets; lacks information on drug cost effectiveness; and is constrained by the relative paucity of well designed comparative studies on drug efficacy and safety in general practice.111 Retrospective application of the March 1995 issue of the British National Formulary to studies during earlier periods is problematic, but we can only sensibly assess prescribing appropriateness by current standards.
Our indicators are relevant to the full range of conditions amenable to drug treatments, and they emphasise patients rather than conditions or drugs. However, published work has focused on specific long term, and inappropriate rather than appropriate, drug treatments for a few conditions—namely, asthma, hypertension, insomnia, and anxiety. Some dimensions of prescribing appropriateness, most notably drug administration, accounted for many more publications than other dimensions such as communication.
Investigators might not know or report the context in which apparently inappropriate prescribing decisions were made. They seldom refer, for example, to coexisting medical conditions or circumstances that might invalidate the uniform application of standard prescribing criteria in individual patients. We required that investigators report an explicit and valid justification for prescribing decisions that contradict our prescribing criteria. Because published work does not reliably show when the violations of our criteria are appropriate in a given case, we have reported the prevalence of potentially inappropriate prescribing or even of inappropriate recording of prescribing.
Though the amount of inappropriate prescribing in general practice in the United Kingdom is unknown, we have reported ample evidence to support assertions that inappropriate prescribing occurred during 1980–95. Our findings, however, overestimate the prevalence of inappropriate prescribing because published studies (a) provide no opportunity, beyond medical records, for prescribers to explain their decisions and (b) favour reports of bad prescribing, particularly for vogue conditions and without reference to comorbidity, which cannot be generalised to the whole of general practice in the United Kingdom. There is no evidence for widespread inappropriate prescribing in general practice, and, contrary to the claims of the Audit Commission,1 the scope for cost savings and effectiveness gains is unknown.
Prescribing indicators applicable to individual patients can yield evidence of prescribing appropriateness. Such indicators should address the overall appropriateness of each patient's drug regimen, complementing guidelines for managing specific conditions. Consequently, we have developed indicators that are applicable to patients' medical records and not specific to a given condition or drug and have assessed their feasibility, reliability, and validity.112
We thank Nick Freemantle for advice on the review.
Funding Department of Health (National Primary Care Research and Development Centre).
Conflict of interest None.