Wide variation in the number of different drugs prescribed by general practitioners A prescription database study

Transcription

1 ORIGINAL PAPER Wide variation in the number of different drugs prescribed by general practitioners A prescription database study Lars Bjerrum 1,2 and Ulf Bergman 2,3 1 Research Unit of General Practice, University of Southern Denmark, Odense, Denmark, 2 Department of Clinical Pharmacology, University of Southern Denmark, Odense, Denmark, 3 Division of Clinical Pharmacology, WHO Collaborating Centre for Drug Utilization Research and Clinical Pharmacological Services, Karolinska Institutet, Huddinge University Hospital, Sweden. Scand J Prim Health Care 2000;18: ISSN Objecti es To investigate the number of different drugs prescribed per dispensing unit and to analyse the influence of practice characteristics on this number. Design Register analysis based on the Odense Pharmacoepidemiological Database (OPED). Setting General practices in the County of Funen, Denmark. Subjects 173 general practices (99 single-handed and 74 group practices). Main outcome measures Number of different drugs prescribed per dispensing unit. Results The number of different drugs prescribed per dispensing unit varied nearly fourfold (range ) and four practice characteristics were able to predict 74% of this variation. Practices with several doctors, a high number of patients listed per doctor, a high percentage of elderly individuals, and a heavy workload showed the highest number of different drugs prescribed. Conclusion As the quality of drug prescribing is associated with the use of a limited number of drugs, it is suggested that GPs should agree on a formulary containing the most essential drugs in primary health care. Key words: general practice, variation, drug use, formularies, data- base. Lars Bjerrum, Research Unit of General Practice, Uni ersity of Southern Denmark, Winsløwparken 19, DK-5000 Odense C, Den- mark. Studies have indicated that doctors can master the prescribing of only a limited number of drugs. On exploring the effects of limitations of knowledge on physicians decision-making behaviour, Chinburapa et al. (1) discovered that when presented with a large and complex number of choices of therapy, physicians shifted from using compensatory (i.e. efficient) processes to using more inefficient, non-compensatory processes. The result was a lower quality of prescribing when limits of knowledge were exceeded. High quality prescribing is therefore associated with the use of a relatively limited number of pharmaceutical products. When the number of prescribed drugs increases, it may be difficult for the prescriber to keep abreast of effects, potential adverse drug reactions, interactions and possible contraindications. Good quality prescribing in a general practice may be facilitated by the use of only those drugs well known to the prescriber, and sufficient to provide rational treatment for medical problems prevalent in the primary health care system (2). In particular, the number of analogue drugs (drugs consisting of various chemical substances, but used for the same condition) should be limited (3). Considerable variations in practice prescribing patterns have been demonstrated but only a few studies have focused on the variability in the number of drugs used per dispensing unit, and our knowledge about the factors behind this variation is scarce (3,4). It has been shown that there is a correlation between the number of prescribed drugs and the number of doctors in practice (5). However, there are few studies focusing on the influence of other practice-related factors, such as practice demography, practice workload and practice activity patterns. The aim of this study was to investigate the number of prescribed drugs per dispensing unit and to analyse the influence of practice-related factors on this number. MATERIAL AND METHODS The study comprised all general practices in the County of Funen, Denmark (n=173), of which 99 were single-handed practices and 74 were group practices. In Denmark, most GPs work in group practices consisting of between two and six doctors and they share the caregiving (including prescribing) to all listed patients. Thus, prescriptions from GPs who

2 Wide ariation in drugs prescribed by GPs 95 work in the same practice derive from the same dispensing unit. For each practice, prescription data for the first quarter of 1995 were retrieved from the Odense Pharmacoepidemiological Database (OPED) (6). OPED contains identifiable records of all prescription refunds for inhabitants in the County of Funen (n= , 1 January 1995). Drugs were classified according to the anatomical, therapeutical, chemical (ATC) classification system developed by the World Health Organisation (WHO) (7). Each drug was identified by brand name, generic name (5th level of the ATC code), pharmacological subgroup (4th level of the ATC code) and therapeutic group (2nd and 3rd levels of the ATC code). Synonyms were defined as drugs consisting of the same chemical substance (same ATC code at the 5th level) but having different brand names. Analogues were defined as drugs consisting of different chemical substances, but belonging to the same therapeutical group. The prescribed number of different generic drugs was calculated for each dispensing unit, and the number of analogues and synonyms was analysed for selected groups. The number of drugs accounting for 90% of the volume in numbers of prescribed defined daily doses (DDD), (the drug utilisation 90% segment, DU90%) has been found to be an inexpensive, flexible and simple method for assessing the quality of drug prescribing in routine health care (8). The number of products in the DU90% segment may serve as a general quality indicator and has been used to compare prescription data between primary health care centres in Sweden. We calculated the DU90% for selected therapeutic groups. We applied a 3-month time-window as catchment period, as most prevalent drug users redeem their prescriptions within intervals of less than 3 months (9). Practice characteristics were retrieved from the Health Insurance Service (HIS) and each dispensing unit was described by a number of characteristics which were analysed by bivariate analysis as possible explanatory variables to the range drugs prescribed. To adjust for potential confounders, a multiple regression analysis was performed (SPSS for Windows Vers. 7.5) (10). A 95% significance level was used as a criterion for including a practice characteristic as a predictor. RESULTS The number of generic drugs (ATC 5th level) prescribed per dispensing unit ranged from 102 to 381 (median 236). The number of different pharmacological subgroups (ATC 4th level) ranged from 81 to 219 (median 149) and the number of brand names from 135 to 614 (median 314). Table I gives the median number of prescribed analogues, synonyms and the DU90% segment for selected therapeutic groups. A median of five or more analogues was prescribed for NSAIDs, tricyclic antidepressants, neuroleptics, beta- Table I. Median number (range) of prescribed analogues and synonyms in 3 months for selected therapeutic groups. For each group, the number of generic drugs accounting for 90% of the volume in numbers of defined daily doses (the DU90% segment) is calculated. (Number of practices, n=173). Therapeutic group (ATC code) Median no. (range) of analogues Median no. (range) of synonyms Median (range) of DU90% segment Anti-ulcer drugs (A02B) 5 (2 8) 7 (2 14) 3 (2 5) Oral antidiabetics (A10B) 4 (1 6) 5 (1 11) 3 (1 4) Nitrate vasodilators (C01D) 3 (1 3) 5 (1 10) 2 (1 3) Thiazide diuretics (C03A) 1 (1 3) Loop diuretics (C03C) 2 (1 3) 6 (2 10) 1 (1 2) Beta-blockers (C07) 7 (1 16) 13 (1 30) 5 (1 9) Calcium channel blockers (C08) 6 (2 8) 15 (4 28) 4 (1 6) ACE inhibitors (C09) 6 (2 12) 8 (2 16) 4 (1 9) Tetracyclines (J01AA) 3 (1 4) 4 (1 8) Penicillins with extended spectrum (J01CA) Macrolide antibiotics (J01FA) 4 (1 5) 3 (1 4) 5 (1 8) 5 (1 9) NSAIDs (M01A) 10 (4 14) 23 (7 47) 5 (1 9) Neuroleptics (N05A) 10 (2 18) 11 (2 24) 6 (1 10) Anxiolytics (N05BA) 3 (1 7) 3 (1 19) 2 (1 5) Hypnotics (N05CD) 2 (1 6) 2 (1 12) 1 (1 4) Tricyclic antidepressants (N06AA) 6 (1 10) 8 (1 15) 4 (1 7) SSRI antidepressants (N06AB) 3 (1 5) 3 (1 5) Bronchodilators (R03A) 4 (2 6) 6 (3 11) Antihistamines (R06A) 4 (1 9) 5 (1 10) 3 (1 5)

3 96 L. Bjerrum, U. Bergman Table II. Relationship between practice characteristics and the number of different drugs prescribed per practice. For each characteristic, the number of drugs prescribed is given for practices below (or at) and above the median practice (cut point). Practice characteristics Median value for practice Mean no. of different drugs prescribed (95% CI) characteristics (range) Practice characteristic Practice characteristic median value median value Practice structure No. of doctors in practice 1 (1 6) 207 ( ) 276 ( ) No. of patients per GP 1270 ( ) 217 ( ) 259 ( ) Percentage of female patients in 49.7 (38,9 77.7) 224 ( ) 252 ( ) practice Percentage of listed patients ( ) 230 ( ) 247 ( ) years Workload in practice Rate of surgery consultations per 23 (6 45) 220 ( ) 256 ( ) doctor per day Rate of telephone consultations 13 (2 51) 224 ( ) 251 ( ) per doctor per day Rate of home visits per doctor 0.85 (0 3) 227 ( ) 249 ( ) per day Clinical work profile Rate of surgical procedures per 8 (1 35) 239 ( ) 236 ( ) Rate of diagnostic procedures per 32 (7 100) 233 ( ) 243 ( ) Rate of referrals to specialists per 6 (1 16) 237 ( ) 239 ( ) Rate of admissions to hospital 0.5 (0 1) 229 ( ) 251 ( ) per blockers, calcium channel blockers, ACE inhibitors and anti-ulcer drugs. In particular, a high number of analogues and synonyms were found for NSAIDs (a median of 10 different analogues and 23 different synonyms). One practice prescribed 47 different synonyms of NSAIDs during the 3-month study period. Table II gives the relation between practice characteristics and the number of different generic drugs prescribed per dispensing unit. The median values of the practice characteristics were used as cut points, and the average number of generic drugs prescribed is given for the group of practices with practice characteristics below (or at) and above the median, respectively. To evaluate to what extent doctors who worked as partners in the same dispensing unit co-ordinated their pattern of prescribing, we compared the number of different drugs used in group practices with the number of different drugs if prescriptions from single-handed practices were aggregated in artificial groups of two, three and four or more doctors (Table III). The median number of generic drugs was 5 10% lower when doctors worked as partners, and the difference was significant (p 0.05) when four or more doctors worked together. The multivariate regression analysis (Table IV) showed that four practice characteristics were significant predictors of the number of different drugs prescribed. The multiple correlation coefficient was 0.86, corresponding to an explanation of 74% of the variation. The number of GPs in the practice was responsible for 52% of this variation. The beta coefficients of the predictors and their partial contributions to the squared multiple regression coefficient (Rsq) are given in Table IV. To illustrate the relationship between practice characteristics and the number of different drugs prescribed, the predicted number of generic drugs used in a typical two-doctor group Table III. Median number (range) of different drugs prescribed per dispensing unit in group practices (doctors working as partners) and randomly aggregated single-handed practices (doctors not working as partners). The number of drugs prescribed in solo practices is shown as a baseline figure. No. of doctors Working as partners Not working as partners One doctor 207 ( ) Two doctors 259 ( ) 273 ( ) Three doctors 306 ( ) 318 ( ) Four or more doctors 334 ( ) 365 ( )

4 Wide ariation in drugs prescribed by GPs 97 Table IV. Predictors of the number of different drugs prescribed in general practice. Adjusted beta-coefficients calculated from the multiple regression analysis. (Number of practices n=173, multiple regression coefficient R=0.86). Predictors of the no. of different drugs prescribed in general practice Beta coefficient (95% CI) Partial squared multiple regression coefficients (Rsq) p value No. of doctors in practice 41 (37 46) 52% 0.01 No. of patients per doctor (per 1000 listed 51 (32 69) 17% 0.01 patients) Percentage of listed patients 60 years 1.5 3% 0.01 Rate of consultations per day 1 2% 0.05 Constant ( corner value in the regression equation): 50 (26 74). Example: In a group practice with 2 GPs, 2400 listed patients (1200 patients per GP) of which 25% are 60 years of age, and a workload rate of 25 consultations per day, the predicted number of different drugs prescribed is: 50+(2 * 41)+(1.2 *51)+(25 *1.5) =256. practice is calculated based on figures from the model. DISCUSSION We found a considerable inter-practice variation in the number of different drugs prescribed. The number of generic drugs varied nearly fourfold from practice to practice. Rational pharmacotherapy is associated with the use of a limited number of drugs with which the prescriber is familiar. From a clinical pharmacological point of view, a few numbers of analogue drugs is sufficient within most therapeutic groups (11). The choice of which analogue drugs to recommend depends on the medical and pharmaceutical documentation, including potency and efficacy of the drug, the risk of adverse effects and the price. For NSAIDs, for instance, there is no clinical evidence of any difference in effect between different generic substances. The drug of choice should mostly be based on the risk of adverse effects, particularly gastrointestinal complications, and only a small number of different NSAIDs should be sufficient to provide a rational treatment (12). In the period 1980 to 1995 the number of registered brand names in Denmark rose by 20%, and in 1995 about 2200 brand names, distributed as 1200 different generic drugs, were registered. The use of a wide range of different chemical entities and a great number of different brand names may imply an increased risk of confusion to both the patient and the doctor, which in turn may reduce the compliance and thus the quality of drug use. Four practice characteristics were predictive of 74% of the variability in the drug assortment size. The number of doctors in practice had the most pronounced impact (partial Rsq 52%). Practices with a high number of listed patients per GP prescribed a high number of different drugs, and 17% of the variation in the size of the drug assortment could be explained by the number of patients per doctor. This relation may be explained by the fact that a high number of listed patients may imply a greater diversity of diseases and consequently a greater diversity of therapeutic needs. The demography of the practice population also influenced the number of prescribed drugs per dispensing unit. Practices with a high percentage of elderly people prescribed the highest number of generic drugs. Furthermore, practices with a heavy workload prescribed the highest number of various generic drugs. We found no other studies exploring the influence of a range of practice characteristics on the number of different drugs prescribed in practice. McCarthy et al. also found a significant correlation between the number of different drugs prescribed in a practice and the number of GPs working in that practice (5). In a Swedish study, Bergman et al. (8) focused the quality of drug prescribing on the number of drugs, accounting for 90% of the volume in numbers of DDD (the DU90% segment) in 24 primary health centres. A significant correlation was found between the number of GPs at the centre and the number of different drugs prescribed. However, in the DU90% segment, the number of GPs in practice only explained 34% of the variation, suggesting that much of the variation in this study is explained by the few or single prescriptions in the 10% segment in the tail of the prescribing. Tamblyn et al. (13) found that about one-quarter of inappropriate drug combinations resulted from contemporaneous prescribing by different physicians, and the presence of a single primary physician seemed to lower the risk of inappropriate combinations. In most group practices, doctors take turns at renewing prescriptions and often a GP is requested to renew a prescription, which was initiated by one of his colleagues. GPs working in the same practice should thus be familiar with all drugs prescribed in that practice as they share the caregiving to all pa-

5 98 L. Bjerrum, U. Bergman tients listed. If GPs who work within the same dispensing unit do not co-ordinate their choice of drugs, the size of the drug assortment may increase extensively due to an increasing number of analogues. In this study, comparison of prescription patterns from randomly grouped single-handed practices with prescription patterns from group practices indicated that doctors who worked as partners in the same practice to some extent co-ordinated their prescriptions. In large group practices (four or more doctors), the number of prescribed drugs was about 10% lower than the corresponding number of drugs in aggregated solo practices. It has been proposed that GPs working in the same practice should agree about a formulary covering most of the health care problems dealt with in general practice. An obvious advantage of a formulary is that all doctors in the practice use the same drugs for the same indications, each doctor becoming familiar with drug dosage, package size, effects, interactions, etc. Furthermore, introduction of a formulary may decrease drug expenditures in practice (14). A few formularies for general practice are available and the number of proposed drugs varies between 100 and 300. An important limitation of this study is that we examined only subsidised prescription drugs purchased at pharmacies. The use of prescription drugs in Denmark accounts for about 80% of drug expenditures and about 60% of the number of packages sold (15). Studies have shown a high association between use of subsidised prescription drugs and non-subsidised drugs. Therefore, we do not believe that the inclusion of non-subsidised drugs would change the final regression relation very much. We used a population-based prescription database covering all inhabitants of Funen, corresponding to about 10% of the Danish population. The age and sex distribution of this population is similar to the total Danish population (16), and the total sales of various drugs correspond to the national average. Therefore, we believe that this study provides a fairly accurate picture of the country as a whole. CONCLUSION A considerable inter-practice variation exists in the number of different drugs used and 74% of this variation can be explained by practice-related predictors. Busy working practices, with several doctors, many patients listed per doctor and a high percentage of elderly individuals listed, prescribe the highest number of different drugs. As the quality of drug prescribing is associated with the use of a limited number of drugs, GPs should agree about a basic formulary consisting of the most important drugs for patients in the primary health care system. ACKNOWLEDGEMENTS We thank Malcolm Macclure, Lars F. Gram and Jakob Kragstrup for inspiring discussions. This study was funded by Sygekassernes Helsefond (Grant no. 22/076-95), The Danish Research Councils (Grant no ) and Hørsløvfonden. REFERENCES 1. Chinburapa V, Larson LN, Brucks M, Draugalis J, Bootman JL, Puto CP. Physician prescribing decisions: The effects of situational involvement and task complexity on information acquisition and decision making. Soc Sci Med 1993;36: Dukes M, Drug utilization studies. Methods and uses. 45th ed. Copenhagen: WHO Regional Publications European Series No. 45, Frolund F. Laegemiddelordinationer i almenpraksis (Drug prescriptions in general practice). Ugeskr Laeger 1991;153: Hartzema AG. Christensen DB. Nonmedical factors associated with the prescribing volume among family practitioners in an HMO. Med Care 1983;21: McCarthy M, Wilson Davis K, McGavock H. Relationship between the number of partners in a general practice and the number of different drugs prescribed by that practice. Br J Gen Pract 1992;42: Gaist D, Sorensen HT, Hallas J. The Danish prescription registries. Dan Med Bull 1997;44: Rønning M, Strøm H, Blix H, Harbo B, Litleskare I, Ullerud T. Guidelines for ATC classification and DDD assignment (5th ed.). Oslo: WHO Collaborating Centre for Drug Statistics Methodology, Bergman U, Popa C, Tomson Y, Wettermark B, Einarson TR, Aberg H. et al. Drug utilization 90% a simple method for assessing the quality of drug prescribing. Eur J Clin Pharmacol 1998;54: Hallas J, Gaist D, Bjerrum L. The waiting time distribution as a graphical approach to epidemiologic measures of drug utilization. Epidemiology 1997;8: Norusis MJ, SPSS/PC+ Professional Statistics. Chicago: SPSS Inc., SpeightTM, Holford HG. Avery s drug treatment (4th ed.) Adis International Limited, Henry D, Lim LL, Garcia RL, Perez GS, Carson JL, Griffin M, et al. Variability in risk of gastrointestinal complications with individual non-steroidal anti-inflammatory drugs: Results of a collaborative meta-analysis. Br Med J 1996;312: Tamblyn RM, McLeod PJ, Abrahamowicz M, Laprise R. Do too many cooks spoil the broth? Multiple physician involvement in medical management of elderly patients and potentially inappropriate drug combinations. Can Med Assoc J 1996;154: Beardon PH, Brown SV, Mowat DA, Grant JA, McDevitt DG. Introducing a drug formulary to general practice effects on practice prescribing costs. JR Coll Gen Pract 1987;37: Bjerrum L. Pharmacoepidemiological studies of polypharmacy. Methodological issues, population estimates and influence of practice patterns (Thesis). University of South Denmark: Research Unit of General Practice, Danmarks Statistik. Population in municipalities 1 January Copenhagen: Danmarks Statistik, 1995.