Family Practice Advance Access originally published online on February 18, 2005
Family Practice 2005 22(2):160-167; doi:10.1093/fampra/cmh734
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Antibiotic prescribing by ambulatory care physicians for adults with nasopharyngitis, URIs, and acute bronchitis in Taiwan: a multi-level modeling approach
a Department of Health Education, National Taiwan Normal University, Taipei, Taiwan, b Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, Room 406, 624 N. Broadway, Baltimore MD, 21205, USA, c Department of Social Medicine, National Yang Ming University, Taipei, d Bureau of National Health Insurance, Taipei, Taiwan and e Department of Infectious Diseases and Microbiology, Graduate School of Public Health, University of Pittsburgh, Pittsburgh PA, USA
Correspondence to Dr Laura Morlock, Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, Room 406, 624 N. Broadway, Baltimore MD, 21205, USA; Email: lmorlock{at}jhsph.edu
Received 18 May 2004; Accepted 26 November 2004.
Huang N, Chou Y-J, Chang H-J, Ho M and Morlock L. Antibiotic prescribing by ambulatory care physicians for adults with nasopharyngitis, URIs, and acute bronchitis in Taiwan: a multi-level modeling approach. Family Practice 2005; 22: 160–167.
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
Background. Imprudent prescribing of antibiotics in ambulatory care in Asia is of great concern. However, an adequate understanding of factors associated with antibiotic prescribing patterns in Asia has not been achieved.
Objective. Our aim was to identify patient and physician characteristics that influence antibiotic prescribing for adults with nasopharyngitis (common colds), upper respiratory tract infections (URIs) or bronchitis in Taiwan.
Methods. Generalized Estimating Equations (GEE) were used to analyze all 128 260 episodes of common colds, URIs and bronchitis generated by a random sample of 137 935 adult National Health Insurance (NHI) beneficiaries (
18 years old) in Taiwan in 2000.
Results. Multivariate analysis results revealed substantial variations across different physician groups. Physician age and accreditation level of the physician's practice setting were the characteristics most associated with prescribing of antibiotics at the initial encounters for these episodes of care. Urban practising physicians (adjusted OR 1.69, 95% CI 1.29–2.21) and those who were self-dispensing or with on-site pharmacists (adjusted OR 1.32, 95% CI 1.19–1.46) were also higher prescribers of antibiotics for adults. Other significant physician predictors included physician specialty, patient volume, and ownership of practice setting.
Conclusions. Results suggest that both accessibility to updated medical information and economic incentives of the attending physician may shape prescribing of antibiotics in ambulatory care in Taiwan. Interventions should be developed to influence these modifiable factors to reduce antibiotic prescriptions of questionable value.
Keywords. Adults, antibiotic prescribing, ambulatory care, Taiwan.
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
The rapid emergence of antimicrobial resistance in the community has become a major global health concern.1 Excessive use of antibiotics in ambulatory care settings is believed to be a major contributing factor to the emergence and spread of antibiotic-resistant bacteria in the community.2,3 In order to devise effective strategies for more rational prescribing of antibiotics in ambulatory care, a better understanding of factors influencing prescribing decisions is essential.
The decision to prescribe antibiotics is complex and multifactorial.1 In addition to clinical factors, many non-clinical influences can affect decisions to prescribe antibiotics; these include patient, provider and neighborhood characteristics, cultural influences and regulatory practices.4 A review of this literature suggests four main themes. First, only a limited number of patient and physician characteristics have been studied and results show relatively little agreement across the studies. In the case of adults, studies conducted in the US using 1992 and 1997–1999 NAMCS surveys, as well as studies of a specific HMO population, suggest that patient age is unrelated to antibiotic prescribing practices;5–7 however, other studies suggest positive effects;8 and still others show negative effects.9 A physician's practice location7,10 and physician specialty5,7,10 are two commonly studied physician characteristics, but results remain inconclusive.
Second, as most studies have used a self-administered survey or the National Ambulatory Medical Care Survey (NAMCS), conducted annually by the US National Center for Health Statistics, they are susceptible to three inherent limitations: small sample sizes, selection bias, and a possible Hawthorne effect.11 Moreover, while multilevel modelling has been applied to other research targeting provider behaviour, it has seldom been used in studies examining antibiotic prescribing patterns, although a hierarchical structure is clearly present in these data. Ignoring this hierarchical structure may lead to underestimated standard errors and misleading inferences.
Fourth, most existing literature is limited to the experiences of Western countries, while antibiotic resistance is of equal if not greater concern in Asia.1 In many Asian countries such as China, Hong Kong, Japan, Malaysia, South Korea, Taiwan and Thailand, where physicians' earnings are linked to drugs dispensed, strong incentives to increase drug prescriptions may be a major cause of the widespread prescription of antibiotics in Asia.12–15 For example, in South Korea, where 59% of patients receive antibiotic treatment, approximately 86% of S.pneumoniae is resistant to penicillin.15 In Hong Kong, several surveys have shown that antibiotics are being prescribed for approximately 60–80% of cold and flu outpatient visits.15
In Taiwan, drugs are dispensed in approximately 97% of all ambulatory care visits; almost half of the doctors in Taiwan prescribe 4–5 drugs per visit for URIs, and 10% prescribe more than 8 drugs.15 Compared to 40% in the US, 53% in France and less than 10% in Germany,16 about 50%–70% of S.pneumoniae in Taiwan are penicillin-resistant.17,18 Due to escalating antibiotic resistance rates in Asia, and rapid transmission of antibiotic resistant bacteria beyond political and geographic borders, it is critical that we expand the scope of research to Asian countries in order to successfully control global antibiotic resistance.
The National Health Insurance (NHI) program in Taiwan provides mandatory universal health insurance, offering comprehensive medical care coverage to all civilian Taiwanese residents. For ambulatory care services in either clinics or hospital outpatient settings, patients only pay minimal user fees and co-payments. Patients have complete freedom of provider choice. Traditionally, physicians in Taiwan both prescribe and dispense. Their earnings are directly linked to the sale of pharmaceuticals. This raises a serious concern that physicians might prescribe and dispense excessive amounts of drugs, which could result in harmful consequences such as antibiotic resistance. Therefore, a policy was instituted in 1997 to separate the prescribing and dispensing functions of clinic-based physicians. In practice, however, to ease resistance from physicians, a concession was made to allow clinics with on-site pharmacists to continue dispensing drugs, while those without on-site pharmacists were required to use outside pharmacies. On average, over 60% of clinics hired on-site pharmacists after the separation policy was instituted.
In this study, our objective was to identify important patient and physician factors influencing antibiotic prescribing for adults with nasopharyngitis, upper respiratory infections (URIs) or acute bronchitis in Taiwan. By conducting this study in Taiwan, which shares many common features in both health care delivery systems and culture with other Eastern Asian countries, we can help identify factors influencing the relatively high rates of antibiotic prescribing in this region.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
This is a cross-sectional study, using the NHI sample files and the Family Registration Files for the fiscal year 2000. The NHI sample files, constructed and managed by the National Health Research Institute, consist of comprehensive utilization and enrolment information for a random sample of 185 889 NHI beneficiaries out of a population of 21 400 826 enrollees throughout Taiwan in 2000. The NHI sample files include the NHI claims data, as well as data from the enrolment files and provider files. Relevant fields in the NHI claims data include diagnosis, date of service, drugs prescribed and filled, dispensing method and anonymous identifiers for the patient, the hospital/clinic and the physician providing the service. The NHI claims data also record the birth date and gender of each patient. In addition, three diagnosis fields in the NHI claims files were used to construct the Resource Utilization Band (RUB) of the Johns Hopkins University ACG Case-Mix System Version 6.0 to measure each patient's co-morbidity level.19 The RUB algorithm assigned each individual to one of the five morbidity levels: no morbidity, low morbidity, moderate morbidity, high morbidity and very high morbidity.
The enrollment files provide information on patient's socio-economic status (SES) and disability status. SES was inferred by linking the patient identifier and birth date to the NHI enrolment files and was defined by the policy holder's insurable wage and occupation. NHI enrolment is mainly through payroll deduction for people with well-defined monthly wages, and per individual tax payments for farmers, fishermen and others without a well-defined monthly wage. People with well-defined monthly wages were classified into 3 categories: <US$640; US$640–US$1279; and US$1280. Those without a well-defined monthly wage were categorized based on their occupation: (1) union worker; (2) farmers and fishermen; (3) others, which include veterans, low-income people and individuals enrolled in the NHI through local government offices. Dependants were classified in the same categories as the primary insured persons. The Family Registration File was used to identify the ethnicity (aboriginal status) of each patient.
NHI also maintains data on each physician and medical facility in Taiwan. The NHI medical personnel registry records the physician's birth date, gender and specialty. The NHI hospital registry provides information on accreditation level, ownership and location of each physician's practice setting. Unique, but anonymous, physician and hospital identifiers were assigned to link the claims data to these provider files.
In the random sample of 185 889 NHI beneficiaries, 137 935 were adults aged 18 and above who made 1 939 249 ambulatory care visits in 2000. From these visits, the study sample was drawn from 283 904 visits made by adult patients who received a primary diagnosis of either nasopharyngitis (i.e. the common cold, ICD-9-CM code 460.xx), acute upper respiratory tract infection (URI) of unspecified and multiple sites (ICD-9-CM code 465.xx), bronchitis and bronchiolitis (ICD-9-CM code 466.xx), or unspecified bronchitis (ICD-9-CM code 490.xx). We excluded 6455 visits (2%) involving a concomitant diagnosis with other common outpatient infections for which antibiotic treatment may have been indicated.3,7 The exclusion criteria are listed in Table 1. Of the remaining 277 449 visits included in the study, 115 571 (42%) included a prescription for antibiotics.
|
Next, since it is debatable whether the prescribing of antibiotics during follow-up visits is appropriate, we studied only the initial encounter for each episode. We defined and constructed an episode of care and the ‘window period’ by following the Ambulatory Episodes of Care methodology.20 A window period of 30 days was assigned to each of the conditions as the maximum amount of time for which follow-up of the condition would be reasonable. Beginning with January 1, 2000, each initial visit had an empty run-in period and additional visits occurring within the 30 day window were treated as part of the same episode and not as a new condition. The 277 449 visits were grouped into 128 260 episodes with, on average, 2.16 ambulatory care visits per episode. Only the initial visit of each episode was analysed.
Analyses
The outcome (filled antibiotic prescription or none) was modelled as a binary variable. Because of the hierarchical nature of the data, cluster effects are likely. Population average models were constructed to address the analytical issues raised by data of this type. They applied the Generalized Estimating Equations (GEE) to adjust for the correlation of data as a nuisance parameter. The models included two levels: patient visit and physician. Seasonal variables were included as possible confounders, in order to account for seasonal variations and different disease patterns within each season. A significance level of 
= 0.05 was selected. Robust estimators were used for standard errors in all analyses. All analyses were conducted using SAS 8.2 and STATA 8.0 statistical software packages.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
Of 128 260 initial visits made by adult patients with common colds, URIs or acute bronchitis in 2000, 40% included a filled antibiotic prescription. By condition, the proportions were 37% for common colds, 39% for URIs, 52% for acute bronchitis and bronchiolitis, and 35% for unspecified bronchitis. Table 2 presents distributions of the total sample and number of visits with antibiotic prescriptions by patient and physician characteristics.
|
Table 3 shows the results of univariate and multivariate analyses. Univariate analysis results indicate that common cold, URI and bronchitis patients, for whom physicians prescribed antibiotics on their first encounters, varied significantly according to patient characteristics (age, gender, ethnicity, SES, disability status, co-morbidity level and diagnosis) and the physicians' personal and practice setting characteristics (physician age, gender, specialty, accreditation level of practice setting, ownership of the practice, volume of episodes, geographic area and urban area practice).
|
Multivariate model results indicate that after adjusting for patient's gender, ethnicity, SES, disability and diagnosis, and physician demographic and practice characteristics, the prescribing of antibiotics decreased with the patient's age and co-morbidity level. Patients aged 70 years or older had the lowest rate of antibiotic prescribing (OR 0.70, 95% CI 0.67–0.74). Patients with very high co-morbidity levels had the lowest antibiotic prescribing (OR 0.86, 95% CI 0.81–0.91) when compared to those with high co-morbidity (OR 0.90, 95% CI 0.86–0.93), moderate co-morbidity (OR 0.91, 95% CI 0.87–0.94) and low co-morbidity levels (OR 0.94, 95% CI 0.91–0.98). Unlike patient's age and co-morbidity, SES and disability status in the multivariate analysis were no longer significant predictors of antibiotics being prescribed. In addition to patient's age and co-morbidity level, female gender (OR 0.94, 95% CI 0.92–0.97) remained associated with lower antibiotic prescribing where aborigines (OR 1.09, 95% CI 1.00–1.20) remained associated with higher prescribing, but the magnitudes of the influences were both marginal.
More importantly, multivariate results indicate that physician characteristics played an even more significant role than many of the patient characteristics. Older physicians were associated with significantly greater antibiotic prescribing and these associations increased after adjusting for detailed patient, provider and practice setting characteristics. Physicians aged 55–64 and 65 and older had more than twice the prescription rates of physicians younger than 35. Physician's specialty is another significant predictor. Internists (OR 0.90, 95% CI 0.82–0.99) were significantly less likely to prescribe antibiotics than other physicians after adjusting for patient, physician and practice setting factors. ENT physicians (OR 1.58, 95% CI 1.42–1.76) still remained high prescribers when compared to physicians in other specialties.
In terms of the influence from practice characteristics, doctors practising in clinics, district hospitals and regional hospitals were respectively 2.16, 1.68, and 1.52 times more likely to prescribe antibiotics to adults than doctors practising in medical centres. Doctors practising in private settings had a 1.22 times higher risk of prescribing antibiotics than those in public settings. Furthermore, the antibiotic prescription rate of doctors dispensing drugs or having hired on-site pharmacists (OR 1.32, 95% CI 1.19–1.46) became significantly higher than that of non-dispensing doctors after adjusting for all other patient and physician characteristics. Physicians with a high patient volume (OR 1.20, 95% CI 1.09–1.32) and those in urban practices (OR 1.69, 95% CI 1.29–2.21) were more likely to prescribe antibiotics than their rural counterparts. Antibiotic prescription variations due to the geographic location of the physician practice remained significant after controlling for various characteristics and clustering effects.
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
This study has revealed several significant patterns regarding antibiotic prescribing in a typical Asian country with high prescription rates and expenditures. First, the results suggest that prescribing antibiotics to adults with respiratory infections in ambulatory care in Taiwan was widespread in the year 2000. Second, other than differences by patient age and disease pattern, the results of marginal odds ratios suggest that the prescribing of antibiotics for nasopharyngitis, URIs and bronchitis did not vary dramatically across patient characteristics and was widespread among all patients. Third, after adjusting for patient characteristics and other factors, substantial variations were observed across different physician groups.
Older physicians and those practising in clinics rather than medical centres were significantly more likely to prescribe antibiotics to patients during initial visits for nasopharyngitis, URIs and bronchitis. This pattern could be explained by the continuing education and more up-to-date medical information accessible to the physicians practising in medical centres. Older physicians are further removed from training and clinic-based physicians generally lack academic affiliations after graduating from medical school, so their knowledge, with respect to appropriate, up to date practices, may decrease with time. The results suggest the possible importance of educational efforts regarding the judicious use of antibiotics in continuing medical education forums to assist in combating antibiotic resistance.
Significantly higher prescribing among dispensing doctors in contrast to those without dispensing privileges or on-site pharmacists could result from the direct link between physician income and the sale of drugs, as the new policy separating the prescribing and dispensing functions still allows doctors to dispense drugs under a few conditions and also to hire on-site dispensing pharmacists. These dispensing doctors or those who dispense through on-site pharmacists have an incentive to prescribe more antibiotics. The large discrepancy in antibiotic prescribing between dispensing and non-dispensing doctors highlights the role of financial incentives, regardless of the standardized application of accurate clinical knowledge. Public health professionals and policy makers must be made aware of and take steps to reform policies that indirectly encourage antibiotic overuse and undermine other efforts in reducing possibly inappropriate prescribing of antibiotics.
Moreover, significant prescribing variations between physicians practising in urban and rural areas and across different geographic regions reflect the asymmetrical distributions of a wide range of factors. Competition may be one influence. Physicians practising in urban areas may face stronger competitive pressure due to a greater supply of physicians, so they may be more likely to prescribe antibiotics to retain patients and satisfy their expectations. In addition, the distribution of environmental factors such as pharmaceutical marketing efforts, access to educational materials and up to date information, and local attitudes towards antibiotics, may also vary substantially among different geographical regions. These significant small area variations in antibiotic prescribing patterns may warrant a variety of interventions targeted to different geographic areas. Future research may advance our knowledge in this regard.
This study has several strengths. First, using the National Health Insurance data of a large national representative sample in Taiwan can help us avoid some of the known shortcomings of survey data. Second, in contrast to other studies on this topic, the large sample size and relatively large number of patient and physician characteristics available in the dataset allow us to study a wide array of factors in the general population. Third, in order to obtain more accurate estimates of standard errors and avoid misleading inferences, our study uses multilevel modelling to test a large number of factors simultaneously, and also takes into account the clustering of data at different levels.
Several limitations should also be noted. First, this study could suffer from limitations inherent in the use of claims data such as limited diagnostic and clinical information on symptoms. Chart review may be more informative in this regard. Also, since we measured prescriptions filled, the true prevalence of antibiotic prescribing may be underestimated. Second, a possible misclassification bias may be of concern. Because a claim may have up to 3 diagnoses, assuming that the antibiotics prescribed were related to the primary diagnosis for that particular encounter may be questionable. Therefore, sensitivity analyses were conducted on visits containing only one diagnosis of cold, URI or bronchitis. All major statistical results remained significant. Third, diagnostic consistency across physicians could be another concern. However, there is no evidence of differential diagnosis assignment across different physician groups. On the other hand, non-differential rates would create a bias in our estimates towards zero. The odds ratios could be underestimated. Fourth, using a single measure (i.e. the NHI program's payroll and occupation-based categories) to analyse the association between a patient's SES and the antibiotic prescribing decision did not allow us to fully explore this association.
In conclusion, although it is difficult to fully unravel the complex interplay among the various influences affecting antibiotic prescribing practices, this study demonstrates substantial variations among physician groups in the practice of prescribing antibiotics for viral respiratory infections after taking patient characteristics into consideration. The results clearly identify several types of physicians at high risk of prescribing antibiotics with questionable value in a typical Asian country with high drug expenditures.15 Efforts to reduce the potentially inappropriate prescribing of antibiotics should target these physician groups through educational, financial and regulatory interventions. Antibiotic resistance is a global issue and reaches beyond political and geographic boundaries; without a better understanding of the factors influencing antibiotic prescribing in Asia, efforts in other parts of the world could be seriously undermined. Taiwan's experience can help motivate public health authorities to devise appropriate educational and policy interventions which will resonate with medical professionals, both in Asia and around the world.
|
Declaration |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
Funding source: none declared.
Ethical approval: this research is approved by the Committee on Human Research of the Johns Hopkins Bloomberg School of Public Health in the US and by the Bureau of National Health Insurance in Taiwan.
Conflicts of interest: none.
|
Acknowledgments |
|---|
We wish to thank David Bishai MD, PhD, Christopher Forrest, MD, PhD and Peter Pronovost, MD, PhD for their helpful comments on previous drafts.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
1 Kunin CM. Resistance to antimicrobial drugs—a worldwide calamity. Ann Intern Med 1993; 118: 557–561.
2 McCaig LF, Besser RE, Hughes JM. Antimicrobial drug prescription in ambulatory care settings, United States, 1992–2000. Emerg Infect Dis 2003; 9: 432–437.[ISI][Medline]
3 Steinman MA, Gonzales R, Linder JA, Landefeld CS. Changing use of antibiotics in community-based outpatient practice, 1991–1999. Ann Intern Med 2003; 138: 525–533.
4 Avorn J, Solomon DH. Cultural and economic factors that (mis)shape antibiotic use: the nonpharmacologic basis of therapeutics. Ann Intern Med 2000; 133: 128–135.
5 Gonzales R, Steiner JF, Sande MA. Antibiotic prescribing for adults with colds, upper respiratory tract infections, and bronchitis by ambulatory care physicians. J Am Med Assoc 1997; 278: 901–904.[Abstract]
6 De Sutter AI, De Meyere MJ, De Maeseneer JM, Peersman WP. Antibiotic prescribing in acute infections of the nose or sinuses: a matter of personal habit? Fam Pract 2001; 18: 209–213.
7 Steinman MA, Landefeld CS, Gonzales R. Predictors of broad-spectrum antibiotic prescribing for acute respiratory tract infections in adult primary care. J Am Med Assoc 2003; 289: 719–725.
8 Dosh SA, Hickner JM, Mainous AG III, Ebell MH. Predictors of antibiotic prescribing for nonspecific upper respiratory infections, acute bronchitis, and acute sinusitis. An UPRNet study. Upper Peninsula Research Network. J Fam Pract 2000; 49: 407–414.[ISI][Medline]
9 Cantrell R, Young AF, Martin BC. Antibiotic prescribing in ambulatory care settings for adults with colds, upper respiratory tract infections, and bronchitis. Clin Ther 2002; 24: 170–182.[CrossRef][ISI][Medline]
10 Oeffinger KC, Snell LM, Foster BM, Panico KG, Archer RK. Treatment of acute bronchitis in adults. A national survey of family physicians. J Fam Pract 1998; 46: 469–475.[ISI][Medline]
11 Mangione-Smith R, Elliott MN, McDonald L, McGlynn EA. An observational study of antibiotic prescribing behavior and the Hawthorne effect. Health Serv Res 2002; 37: 1603–1623.[CrossRef][ISI][Medline]
12 Abe MA. Japan's clinic physicians and their behavior. Soc Sci Med 1985; 20: 335–340.[CrossRef][ISI][Medline]
13 Chang SC, Chang HJ, Lai MS. Antibiotic usage in primary care units in Taiwan. Int J Antimicrob Agents 1999; 11: 23–30.[CrossRef][ISI][Medline]
14 Liu YC, Huang WK, Huang TS, Kunin CM. Extent of antibiotic use in Taiwan shown by antimicrobial activity in urine. Lancet 1999; 354: 1360.[CrossRef][ISI][Medline]
15 Chou YJ, Yip WC, Lee CH, Huang N, Sun YP, Chang HJ. Impact of separating drug prescribing and dispensing on provider behaviour: Taiwan's experience. Health Policy Plan 2003; 18: 316–329.
16 Harbarth S, Albrich W, Brun-Buisson C. Outpatient antibiotic use and prevalence of antibiotic-resistant pneumococci in France and Germany: a sociocultural perspective. Emerg Infect Dis 2002; 8: 1460–1467.[ISI][Medline]
17 Hsueh PR, Luh KT. Antimicrobial resistance in Streptococcus pneumoniae, Taiwan. Emerg Infect Dis 2002; 8: 1487–1491.[ISI][Medline]
18 McDonald LC, Lauderdale TL, Shiau YR et al. The status of antimicrobial resistance in Taiwan among Gram-positive pathogens: the Taiwan Surveillance of Antimicrobial Resistance (TSAR) programme, 2000. Int J Antimicrob Agents 2004; 23: 362–370.[CrossRef][ISI][Medline]
19 Johns Hopkins University School of Hygiene and Public Health. ACG Software Documentation and Users Manual (version 6.0). Baltimore: Johns Hopkins University; 2000.
20 Forrest CB. Opening the Gate to Primary Care: How First Contact Care Influences Costs for Ambulatory Episodes of Care. Ph.D. dissertation, 1995.
Read all E-letters
CiteULike 
Connotea 
Del.icio.us What's this?
E-letters:
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||