AIM - To assess the effects of a rational prescribing intervention on specific prescribing areas.
DESIGN- Controlled 3-year retrospective review of the PACT catalogues (previously level three Prescribing Analysis and Cost data).
SUBJECTS AND SETTING- General practices within North and East Devon health authority area. Intervention practices had a rational prescribing day in 1995 (n=7). Control practices had either no intervention (n=7) or a rational prescribing day after the end of the evaluation period (n=7).
OUTCOME MEASURES- Changes in indicator drug prescribing before and after the intervention and for a similar time for the control practices. A composite prescribing improvement score (PIP) was calculated to assess overall changes.
RESULTS- There were improvements in effective prescribing over the 3-year period for both the intervention group and control groups. There was a greater improvement in PIP score in the intervention group, compared with the control group (P=0.0003, t=4.38). Improvements in specific prescribing areas were statistically significant for 2 out of 15 indicators and the intervention group demonstrated the greatest change in the desired direction for the other 13 indicators.
CONCLUSION- The rational prescribing intervention is successful at promoting change in the prescribing behaviour of general practitioners. Pharmacists can provide an economical and effective way of enhancing clinically effective prescribing activity.
The importance of rational prescribing cannot be over-emphasised. In financial terms alone, it is estimated that £600m could be saved annually through more cost-effective prescribing.1 Much research has been done to influence and change clinical behaviour. There is evidence that strategies to influence practising doctors needs to be specific, localised and targeted,2 and that the evidence base needs to be presented along with any guidelines or indicators.3 The environment in which the intervention takes place needs to be attractive and conducive to change4 and meetings within practices using standardised content and delivery have been shown to be useful.5 Guidelines and structured feedback have been shown to be somewhat successful on their own,6,7 but face-to-face visits to doctors by trained educators are also an effective and lasting method of disseminating guidelines for good clinical practice and influencing prescribing behaviour.8,9 It would seem logical that a combination approach may prove even more useful.
There is likely to be increased emphasis placed on cost-effective and clinically effective prescribing in general practice with the development of primary care groups (PCGs) and the introduction of unified budgets. Clinical governance will also require thought to be given to these areas. Given the economic benefits of centralisation of some services, PCGs may continue to seek the support of health authorities in an advisory role. It is therefore important to evaluate the benefits or otherwise of interventions the health authority can offer to PCGs. The aim of this study was to evaluate the effect of a rational prescribing intervention on the prescribing habits of general practitioners.
Indicators Since 1994, the North and East Devon health authority has attempted to influence the prescribing of local general practitioners in several key areas, based on the principals of good prescribing, clinical effectiveness and consideration of cost when two drugs are clinically equivalent. (Costs were taken from the British National Formulary and assessed in relation to the proven comparative clinical effectiveness of the drugs.)
Evidence-based indicators were developed by the prescribing department for specific drugs or combinations of drugs within the areas indicated below, where it was considered that a change in practice would result in either more clinically effective or more cost-effective practice. Evidence was collected by means of Medline literature searches for each area. Studies were assessed for suitability for inclusion according to published criteria.10 No opinions or consensus statements were used in the construction of the guidelines.
Gastrointestinal system Lansoprazole is as effective as omeprazole in most cases11 and is cheaper. The ratio of lansoprazole/ omeprazole should therefore increase on grounds of cost-effectiveness. Cimetidine has a comparable healing rate to the other available H2-receptor antagonists12 and is the cheapest. The ratio of cimetidine/H2-receptor antagonists should therefore increase on grounds of cost-effectiveness.
Cardiovascular system The ratio of bendrofluazide 2.5mg/bendrofluazide 5mg should increase as the 5mg dose does not exhibit further therapeutic effect and more side effects are reported.13 The percentage of combination diuretics to all diuretics should fall as these are not generally needed for uncomplicated hypertension11 and they are expensive. Expensive preparations containing a diuretic and potassium supplement are clinically ineffective in most cases.14 Beta-blockers have been shown to have a beneficial effect on cardiovascular mortality,15 hence their use should increase on grounds of clinical effectiveness.
Respiratory system Given the usefulness of inhaled corticosteroids in the management of asthma,16 the ratio of inhaled corticosteroids/inhaled bronchodilators should increase in accordance with the British Thoracic Society guidelines that inhaled corticosteroids be used earlier in the management of asthma.17 Fluticasone is the most expensive of the inhaled corticosteroids and its benefits over other inhaled corticosteroids for the broad majority of asthma patients are unclear,18 hence there should be a fall in the prescribing of this agent where there is an alternative.
Central nervous system Sertraline has no more benefit than other selective serotonin reuptake inhibitors (SSRIs) for most patients,19 hence prescribing of this drug should be reduced and it should not generally be initiated before other SSRIs are tried. Most migraine patients find relief on a 50mg dose of sumatriptan, and a 100mg dose is often inappropriate,20 hence the ratio of sumatriptan 50mg/sumatriptan 100mg should increase. Prescribing of brand name compound analgesics containing paracetamol and codeine phosphate should fall as prescribing of separate analgesics allows for dose titration, and it is far cheaper to use generic paracetamol and codeine phosphate.
Musculoskeletal system The danger of gastrointestinal side effects with non-steroidal anti-inflammatory drugs is well known and reported.21 The safest NSAIDs in terms of side effects are, in order of safety, ibuprofen, naproxen, diclofenac and indomethacin,21 hence the prescribing of these agents as a proportion of all NSAID prescribing should increase on grounds of safer clinical practice. The more expensive modified release NSAIDs may have a greater chance of adverse reactions,22 hence a standard NSAID should be considered as an alternative. Topical NSAIDs do reduce gastrointestinal side effects,23 however, skin is a barrier to effective absorption and there is likely to be a high placebo effect.24 The cost of a course of treatment with topical compared with oral NSAID is many times higher, so alternative oral treatment should always be considered and the number of prescriptions for these drugs should fall.
Antibiotics Antibiotics are often prescribed for patients who have no evidence of bacterial illness.25,26 The use of antibiotics for minor, self-limiting viral problems is to be discouraged. Total antibiotic use should therefore fall. There should be a practice antibiotic formulary, endorsed by expert microbiologists to facilitate effective antibiotic use. Where there is no clinical indication to use a cephalosporin over a penicillin, the latter should be considered on grounds of cost effectiveness.
Intervention The rational prescribing intervention consisted of the production of a standardised practice report based on the prescribing data for the indicators mentioned above. This was followed by an "away day" to discuss prescribing patterns with the whole practice (general practitioners, practice manager, practice nurses) as well as a local community pharmacist if appropriate. The content of the away day followed a standard format and was delivered by trained educators. The educators were health authority pharmacists experienced in presentation skills who had been exposed to a training day outlining the objectives and structure of the intervention. Presentation of practice-specific information was compared with the health authority averages, and areas for possible improvement were highlighted. Data on the individual partners were presented if requested. A considerable amount of time was allocated for detailed discussion and locum staff were funded by the health authority to enable many members of the practice to be present. The intervention was thus local, practice-specific and targeted.
Practice selection All practices within the health authority area were made aware of the availability of the intervention. Practices selected themselves for intervention, and this was arranged via the pharmaceutical advisor to the health authority. No practices were identified by the health authority for intervention on grounds of poor performance. For the purposes of this study, all practices which had a rational prescribing intervention during 1995 were included (n=7), except one for which there was no PACT record due to an administration error.
Control practices were randomly selected from health authority lists. In order to control for the self-selecting nature of the intervention group, half the control practices were selected at random from those who had elected to have a rational prescribing intervention, but after the end of the evaluation period (n=7). During the review period there was no exposure to the intervention or report for these practices. The remaining controls (n=7) were selected at random from the practices which have not had an intervention to date.
Data collection For all practices, paper copy level 3 PACT data were retrieved for the financial year March, 1994, to April, 1995 (pre-intervention) and the financial year March, 1996, to April, 1997 (post intervention). The data were reviewed for all of the indicators. In all cases except that of antibiotics, the unit of measurement was PACT "items" multiplied by PACT "quantity". For antibiotics, the units of measurement were PACT "items".
The following practice variables, thought to be of possible importance in prescribing behaviour, were recorded from health authority records to assess the comparability of the groups in respect of size, fundholding status, number of partners (wholetime equivalent), Jarman deprivation score, training status, rate of growth in practice size, dispensing status, changes in partners, closeness to target budget factors, and urban or rural location. An ordinal assessment of practice policy towards pharmaceutical representatives was made, on a scale of 0 (pharmaceutical representatives never seen) to 7 (pharmaceutical representatives see most partners most days, on both a drop-in and appointment basis).
Estimated costs The estimated cost of a rational prescribing intervention is as follows:
Statistical methods For each indicator, rates per ASTRO-PU (age, sex, temporary resident oriented prescribing unit) or ratios as appropriate, were calculated for each practice. Analysis of covariance was used to assess the difference between the intervention and control group post-intervention, using the baseline prescribing rate (or ratio) as a covariate. The contribution of other possible explanatory variables detailed in the "Data collection" section was examined in a multivariable analysis.
A composite prescribing indicator was planned before the analysis - the prescribing improvement (PIP) score. A clinically significant improvement in prescribing rates or ratios was considered to be 10 per cent. For each indicator, improvements of less than 10 per cent and worsening performance scored 0 and improvements of 10 per cent or over scored 1. For the antibiotic formulary indicator, the presence of a formulary scores 1, otherwise 0. A t-test was used to assess the differences between means in the PIP score.
| Table 1: Practice characteristics | ||
| Characteristic | Control | Intervention |
| Mean 94/95 ASTROs | 25,122 | 23,726 |
| Mean 96/97 ASTROs | 28,289 | 26,972 |
| Fundholding | 43% | 43% |
| Dispensing | 14% | 14% |
| Mean Jarrman deprivation score | 1.05 | 4.8 |
| Mean wholetime equivalent | 3.75 | 3.9 |
| Urban location | 29% | 14% |
| Mean pharmaceutical representative score | 5 | 5.1 |
| Training status | 50% | 57% |
Characteristics of the intervention and control practices are shown in Table 1. There were no statistically significant differences in any of the variables between the groups. Table 2 shows the mean prescribing indicators for each group at baseline and post intervention. Figure 1 shows the difference in changes between the intervention and control groups adjusted for baseline rate, and with 95 per cent confidence intervals. No other factors were significant on univariable analysis. Within the control group, there was no statistically significant difference in practice characteristics or in performance between those who were to have an intervention after the end of the review period and those who have had no intervention to date. For all indicators except antibiotic prescribing, the intervention group had a greater change in the desired direction. This was statistically significant for bendrofluazide (P=0.003) and modified release NSAIDs (P=0.009).
| Table 3: mean prescribing improvement scores | |
| Group | Score (SD |
| Control | 6.6 (1.4) |
| Intervention | 9.9 (1.7) |
This retrospective evaluation of prescribing has found that those practices having a rational prescribing intervention and practice report do improve their overall prescribing more than the control group, as measured by the PIP score and by examination of the individual prescribing indicators that were discussed during the intervention. The small size of the sample, baseline variability and changes over time are possible reasons why not more individual indicators were significantly different between control and intervention groups. However, the finding of greater change in the intervention group in 14 out of the 15 indicators is quite striking.
This study did not evaluate the proportional value of the report as compared with the intervention day; rather, the two components were taken together as a single entity. Further comparative research could be carried out to find out the relative importance of the two aspects. However, it is likely that neither component on its own would be as effective as a combined approach.
The general improvements in prescribing seen in the control group could be explained by the effects of continuing medical education. It is likely that all of the general practitioners would make efforts to keep up to date with medical practice, hence prescribing in general would be expected to improve over time.
The penicillin/cephalosporin ratio, the sole indicator that did not improve more in the intervention group, may be explained by the use of private prescriptions for antibiotics. GPs are able to write a private prescription for their National Health Service patients for non blacklisted drugs provided that the patient is not exempt from prescription levies and is offered the choice of an NHS or private prescription. This allows GPs to make inexpensive items such as amoxycillin less expensive to the patient than the £6 prescription charge. Such prescriptions are not detected by PACT reports, and may amount to a substantial number. It is possible that the intervention group practices were more likely to use this approach as it was mentioned during the intervention that this might save money for the patient. If so, this could conceal any improvement.
The results of retrospective studies on heterogeneous groups of general practices always need to be interpreted with a degree of caution. Although the groups were comparable in terms of measured variables, it is always possible that non-measurable differences between the practices could have an effect on the results. The reasons for the intervention practices accepting the health authority offer of a rational prescribing intervention could indicate that these practices were better prescribers or more receptive to ideas of change than other general practitioners. The use of other practices which accepted the intervention, albeit later, within the control group attempts to control for this type of bias, but it is possible that this is not completely satisfactory. Nevertheless, such widespread findings in favour of the intervention group are encouraging.
The sample size was small, and this makes it difficult to evaluate the importance of the changes in individual drug prescribing. With the development of electronic PACT data, larger studies could be made easier. Prospective studies on the effects of follow-up interventions and the concept that prescribing in areas which had no direct indicator may improve as a result of a change in attitude to prescribing after the intervention could usefully be carried out.
The rational prescribing intervention does seem to be an effective way to influence and improve cost-effective and clinically effective prescribing. The one-off cost of the intervention would quickly be offset by prospective savings made by more cost effective prescribing. PCGs concerned with cost-effectiveness and clinical effectiveness may benefit from this kind of approach.
North and East Devon health authority has comparable prescribing figures and numbers of practices to other health authorities and, as the study practices come from a variety of locations (some rural and some urban), the generalisability of the intervention seems good. Utilising the evidence base to design interventions is a worthwhile approach for PCGs and fits well with the development of clinical governance. This study provides a workable example of how an evidence-based intervention may be run effectively. It also demonstrates simple and cost-effective monitoring and evaluation methods which could easily translate to other settings, and could be used routinely for the assessment of efforts to disseminate evidence-based practice.
itus Bradley is principal investigator, Alison Round is consultant in public health medicine and Marilyn Ramsden is head of the prescribing department at the North and East Devon health authority. Correspondence to Dr Bradley at 1 Campion Gardens, Campion Meadow,Exeter EX2 5RS
REFERENCES |
|