Using the Medical Expenditure Panel Survey (MEPS) to Assess Antibiotic Utilization in Acute Respiratory Tract Infections

Background: Respiratory tract infections (RTls) are considered to be the primary reason for prescribing antibiotics in community practice, accounting for 75% of prescribed antibiotics. Only some of these conditions, acute non pneumonic RTls (ARTls), are the focus of this dissertation. ARTls include acute cases of nasopharyngitis (common cold), sinusitis, pharyngitis (sore throat), laryngitis, and bronchitis. These conditions are, in many cases, caused by a viru~ and they usually resolve naturally without the use of any medication. Objectives: Many studies have demonstrated the inappropriate overutilization of antibiotics in certain ARTls. This practice may be associated with high unnecessary costs and potential for resistance from such use. Factors behind this practice are many. Some have been studied and many still need more exploration. The objectives of this study were to a) provide specific information on the influence of prescription drug coverage on antibiotic utilization; b) the influence of prescribing antibiotics on patient reported satisfaction; c) explore the effect of drug copayment on prescribing of antibiotics. The study also estimated the total cost in dollars that may be saved from the direct cost of antibiotics, to both the insurers and patients, when paying for these medications that may be used inappropriately. Methods: A cross-sectional retrospective study of prescription events associated with different ARTls was identified from the Household Component (HC) of the 1996 Medical Expenditure Panel Survey (MEPS). Descriptive and logistic regression analyses were conducted to explore the relationship between the different variables in the study with antibiotic utilization. The perspective taken in these studies was that of a healthcare insurer. Results: The study found that the likelihood of being prescribed an antibiotic, ' which may be of high cost and unnecessary, is greater when patients have prescription drug coverage. However, no effects were related to drug copayment on antibiotic utilization. The study also suggests that in this population the level of patient satisfaction with the quality of care provided is not influenced by whether antibiotics are prescribed or not. It was also found that millions of dollars may be wasted yearly on these medications. Conclusion: The findings from this study provide more insight to insurers and designers of drug coverage plans, who should closely monitor prescribing patterns for these conditions to avoid unnecessary cost as well as resistance from such antibiotics. They should not focus on antibiotic prescribing as a means of patient satisfaction, but on the appropriate use of such antibiotics and the avoidance of unwanted effects. More research is needed to examine the effect of the different copayment tier systems introduced in the last couple of years on the utilization of these drugs and whether they may help reduce the inappropriate use of these drugs.

Methods: A retrospective study of (N=2534) prescription events associated with ARTls have been identified from the Household Component (HC) of the 1996 Medical Expenditure Panel Survey (MEPS). Cases selected included acute nasopharyngitis (common cold), acute sinusitis, acute pharyngitis, acute tonsillitis, acute laryngitis and tracheitis, acute unspecified upper respiratory tract infections, and acute bronchitis and bronchiolitis. Antibiotic use and associated costs were determined by selecting oral antibiotics and the sum of payments associated with each prescription. Logistic regression was utilized to evaluate prescription drug coverage effect on prescribing an antibiotic and on the type of antibiotic prescribed.

Results:
In 1996, antibiotics accounted for 4 7% of the prescription events for ARTIS, of which 29% were for high cost antibiotics. When compared to patients with private prescription drug coverage, patients with no drug coverage were less likely to be prescribed an antibiotic (OR=0.49; 95% Cl = 0.37 _ 0.65). The chance of getting a high cost antibiotic was also lower in events associated with no drug coverage (OR=0.1 O; 95% Cl = 0.05 -0.19).
No significant difference was shown between public and private coverage.

Conclusion:
The results of the study indicate that the rates of prescribing an antibiotic for ARTls are still high, and that the likelihood of being prescribed an antibiotic, which may be of high cost and unnecessary, is greater when patients have prescription drug coverage. Providers of such coverage should closely monitor prescribing patterns for these conditions to avoid unnecessary cost as well as resistance from such antibiotics. 3

Introduction:
Respiratory tract infections (RTls) are considered to be the leading cause of prescribing antibiotics in community practice, accounting for 75% of prescribed antibiotics. [1] In the year 1992, there were approximately 57 million prescriptions for antibiotics in the United States with colds, bronchitis and upper RTls accounted to 12 million of the total prescriptions. [2] Only acute nonpneumonic respiratory tract infections (ARTls) was the focus of this study.
ARTls include conditions where antibiotics are generally considered to be inappropriate such as in acute nasopharyngitis (common cold), acute unspecified upper respiratory tract infections, acute bronchitis, and acute laryngitis and tracheitis. [3,4] Most of these conditions are not life threatening and are usually caused by a virus. [3] Even when bacteria are the causative pathogen, the illness, in most cases, resolves naturally and rapidly. [4,5] It is also important to note that antibiotics have not been shown to improve the clinical outcome of patients with these conditions. [2,6] Other ARTls conditions like acute pharyngitis and acute sinusitis, antibiotics are in many cases not recommended, however it may be beneficial in some patients,. [3] Even when prescribed an antibiotic the choice should be for a narrow spectrum, low cost antibiotic (e.g., penicillin) to avoid the unnecessary cost and the potential of resistance. [4] The reasons for prescribing an antibiotic in acute pharyngitis is because it is hard to differentiate between streptococcal infection versus a viral infection, and also to prevent complications, such as acute rheumatic fever and acute glomerulonephritis. [4] However, the potential of being infected with the bacteria that causes rheumatic fever is rare in many population, [3,4] and for acute glomerulonephritis, one meta-analysis showed that the rates have not been reduced by the use of antibiotic. [?] It is also, important to note that the natural rates for resolution without any complications are >90% for streptococcal ' pharyngitis, [8] and approximately 69% for acute sinusitis. [9] Many studies have demonstrated the inappropriate overutilization of antibiotics in certain acute RTls. [10][11][12] In 1998 one study estimated that 76 million office visits for acute RTls resulted in 41 million antibiotic prescriptions. This was 55% (22.6 million) more prescriptions than were expected to be used, with an associated cost of approximately $726 million. [13] Another study estimated that between 1989 and 1999 there were 6.7 million annual office visits by sore throat (acute pharyngitis) patients, where approximately three quarters of adults (73%) were prescribed an antibiotic and 68% of the prescribed antibiotics were for non recommended use. The study also showed that broad spectrum and more expensive antibiotics are frequently used. [14].
Recent studies have shown that the rate of prescribing inappropriate antibiotics has decreased. [15,16] However, the rate is still high and is associated with high unnecessary costs, as illustrated in the previous paragraph. The practice is also associated with the potential and risk for resistance from such antibiotics, which is becoming an epidemic and a global problem. [17][18][19] Recognizing the problem has led to many studies that looked at factors leading to the inappropriate antibiotic prescribing. Studies that examined patients' preferences suggest that they are expecting to receive an antibiotic ' and consequentially pressure their physician to prescribe one. Patients may also be financially responsible for a low copayment or no copayment which may give them more incentive to fill the prescription. [20,21] Physicians, on the other hand, try to meet the expectation of their patients and attempt to avoid any liability in case of a treatment failure, leading them to prescribe broad spectrum antibiotics. [22] Other factors may include the health care system structure, including formularies and prescribing restrictions. [23] Directed advertising and marketing to physicians and consumers by pharmaceutical companies may also lead to more prescription of high cost broad spectrum antibiotics. [24] The influence of prescription drug coverage on the utilization is one of the potential factors that affect antibiotic prescribing. Many studies have demonstrated the influence of prescription drug coverage on the drug Utilization in different patient populations. [25][26][27][28][29][30][31] Many of these studies 6 showed an increase use of medications when the study population had prescription drug coverage. However, some suggest that by not having the coverage, underutilization of essential drugs may occur. [32] Studies that examine the effect of drug coverage on the utilization of drugs in general and in the utilization of antibiotics in ARTls in particular, are becoming more important, knowing that the strong possibility of having a drug benefit plan for the Medicare population and with different health care plans offering drug coverage.
The objective of this study is to explore the influence of prescription drug coverage on the utilization, both in terms of frequency and type, of antibiotics in ARTls, where studies are limited. The study will also identify the effect of other potential relevant factors such as socioeconomics and demographics.
The hypothesis is that the frequency and type of antibiotics prescribed for ARTls are greatly influenced by whether or not the person has a drug coverage policy. The results of this study will be of great value for policy and decision makers who are responsible for setting such benefits.

Study population:
The study population included cases of ARTls identified by the International Classification of Diseases, Ninth Revision Clinical modification codes (ICD-9 codes), that were associated with prescribed medicine events. Each prescribed medication was associated with up to three ICD-9 codes and was reported by patients, which then were conformed by pharmacy records. The study has excluded prescription events with more than one ICD-9 code to assure that the prescription medicine of interest is for a particular ICD-9 code.

'
The first dependent variable of interest was defined as being prescribed an antibiotic or not, and only oral antibiotics were included in the analysis. The second dependent variable was defined as the type of antibiotic being received (i.e., high or low cost antibiotic). Antibiotic type was created by summing payments for each antibiotic prescribed and calculating the median cost. An antibiotic is considered to be high cost if the median cost was more than $30.00, and was considered low cost if the antibiotic median cost was $30.00 or less.

Independent Variables (IV):
The main variable of interest was prescription drug coverage, which was categorized into public, private or no drug coverage. An event is considered to have public drug coverage if the event was paid for by one of the following '

Results:
The number of cases of ARTls included in the study is presented in Table1, were around 57% of the cases were identified as the common cold. The table also shows the rate of prescribing antibiotic in each ARTls condition and the overall rate of prescribing an antibiotic in the study population which was around 47% of the total events.  Table 4 illustrates the results of the cross-tabulation and the chi-square of all the variables in the study against antibiotic use and type. The unweighted chisquare test showed that when looking at the use of antibiotic (DV1 ), the variables that showed a significant difference in the study population were prescription drug coverage, race, age, poverty status, and self/family payment of prescription drugs. However, when the data were weighted using SUDAAN, the variables were shown to be significant were the prescription drug coverage, race, poverty status, and self/family payment of prescription drugs. Table 4 also shows that when the same unweighted chi-square test was carried out on the type of antibiotic used (DV2) the variables that showed a significant chi-square test were the prescription drug coverage, sex, age, census region, poverty status, and self/family payment of prescription drugs.
However, when the data were weighted using SUDAAN, the variables that showed to significant, were only the prescription drug coverage, poverty status, and self/family payment of prescription drugs.
The univariate logistic regression analysis showed that many of the variables had an independent effect on both antibiotic use and type. Interaction assessment conducted for both dependent variables have concluded that no interaction terms exist between the variables. The analysis also showed that ' no highly correlated variables exist. Testing for confounders concluded that when looking at the use of antibiotics, race, poverty status and self/family payments of antibiotic are potential confounders. However, when examining the type of antibiotic used, poverty status and self/family payments of antibiotic are potential confounders. The final logistic model shown in Table 5 indicated that, in 1996, before controlling for confounders, ARTls events with both public drug coverage and with no drug coverage were at a significant lower risk (i.e., chance) of receiving an antibiotic than in events with private coverage (Public: OR=0.47; 95% Cl= 0.34 -0.65) with a p-value of 0.000 and (No coverage: OR=0.66; 95% Cl=0.52 -0.83) with a p-value of 0.0004. However, when controlling (adjusting) for potential confounders (i.e., race, poverty status, and self or family payments of antibiotics), only events associated with no drug coverage 15 showed a significant lower chance of receiving an antibiotic (OR=0.49; 95% Cl = 0 . 37 _ 0.65) with a p-value of 0.000. The other logistic model shown in Table 6 also suggested that, in 1996, the chance of getting a high cost antibiotic before controlling for confounders, was also lower in events associated with public coverage events when compared to privet drug coverage (Public: OR=0.48; 95% Cl=0.27 -0.48) with a p-value of 0.0097 and no significant different was seen in events with no drug ' coverage (OR=0.75; 95% Cl=0.49-1.13) with a p-value of 0.1678. However, when controlling for potential confounders (i.e., poverty status and self or family payments of antibiotics) the opposite results were found in that the chance of getting a high cost antibiotic was only lower in events with no drug coverage (OR=0.1 O; 95% Cl = 0.05 -0.19) with a p-value of 0.000. No significant difference was observed between public and private coverage.

Discussion:
The study demonstrates that the rate of prescribing an antibiotic for ARTls is still high, and in 1996 it was around 13.6 million prescription, in which around 29% were for high cost broad spectrum antibiotics which should not be used even if antibiotics may be recommended since in many ARTls cases they are not the drug of choice [4]. These prescriptions are, in many cases, considered unnecessary and may be associated with excessive costs and bacterial . t nt that can be easily avoided with a better management of such events. res1s a That can be accomplished by a better understanding of the different socioeconomic and behavioral factors behind such a problem.
The study also, demonstrated that the prescription drug coverage is an important and a key factor having a significant role in the utilization of antibiotics in ARTls. Results showed that prescription drug coverage is associated with an increased utilization of antibiotics, and that was also ' associated with more consumption of high cost antibiotics. Findings are consistent with many of the studies that indicate the effect of having a drug coverage on the increase utilization of drugs in different diseases and populations. [26][27][28][29]41] However, this study is different in that it looks at the utilization of antibiotics in a specific group of conditions (i.e., ARTls), using a national database allowing us to make national estimates of utilization.
The study illustrates the need for similar studies, which is becoming more important for many reasons. First, the potential of overutilization of unnecessary medications with the unwanted risk associated with it use (e.g., antibiotic resistance). Second, the need to lower the increasing cost of the health care system in regard to medication utilization which is about 13.1 % of the total health care cost.
[42] Finally, these kinds of studies will help many managed care organization and health insurers who are trying to lower their costs without affecting the quality of care they provide and will also help the United State government especially with the potential of having a Medicare prescription drug plan.
A number of potential limitations are associated with this study. First, the data used in the analysis does not provide information about the different benefits associated with each prescription drug plan, making it hard to examine at the influence of the degree of benefits (i.e., generosity level) on the utilization of antibiotics. Second, the study is cross-sectional in nature and may prevent ' analysts from capturing the influence of any change in the drug coverage policy over time. Third, the diagnosis information are imputed based on what was reported by the patient or the patient guardian, which may be biased (i.e., recall bias). Finally, the data does not provide information regarding the causative agent, whether it is a viral or bacterial, and it also does not provide a history of any antibiotic used.
Despite these limitations the study was able to show a significant result because we think that is reflecting what is really happening in regard of prescription practices in different clinical settings throughout the country. The results are very helpful to policy and decision makers, both in the government and the private sector, who are in charge in making and designing prescription drug benefits. It also illustrated that prescription drug coverage without the proper management may be reasons for misutilization and increasing cost that can be avoided.
Future studies using recent released MEPS or any other similar dataset should examine the effect of the different types of drug coverage on antibiotics utilization in ARTls and if possible to compare between the change in trend over time. It is also important to explore the effect of drug coverage in other disease states and conditions, which will be of great value for people in charge of designing drug benefit plans in both private and public sectors.

Conclusion:
The results of the study indicate that the rates of prescribing an antibiotic for ARTls are still high, and that the likelihood of being prescribed an antibiotic, which may be of high cost and unnecessary, is greater when patients have prescription drug coverage. Providers of such coverage should closely monitor prescribing patterns for these conditions to avoid unnecessary cost and consequently, resistance from such antibiotics.

9.
Del Mar, C., P. Glasziou, and A. Spinks, Antibiotics for sore throat (Cochrane review    -Studies that look at satisfaction are part of a larger area of research known as outcomes research, which is a multi-dimensional approach of evaluating the outcomes of the health care system, of which not only clinical outcomes are of interest, but also information about the patient functional outcomes, patient satisfaction, and cost-effectiveness of the health care provided. [1,2] One of these dimensions that has attracted a great deal of interest in the last couple of years is ,on studies that uses satisfaction as an indicator for the quality of care. [3] Studies that examine the patient perspective are considered an important method in the evaluation and assessment of the quality of the health care system. [4] Satisfaction shows the realities about care, as well as the expectations and preferences of the patient. Both expectations and preferences can be viewed as the determinants of satisfaction. [5] The expectation of patients have been linked to different level of satisfaction with the health care system, patients' tendency to sue the health care provider, compliance to the suggested therapy, the physician's decision for ordering different tests, and the ability to achieve the desired clinical effect. [6,7] Patient satisfaction can be viewed as a performance evaluation of any health service.
[8] Patients' satisfaction has been defined as "the reaction of the patient to the salient aspects of the context, process, and result of the service experience" [2,9] Satisfaction and other health care outcomes information are relevant to patients, the payers and to the providers. [10,11] The consumers of the health care system are demanding a better service from providers and payers and decision makers are trying to follow their expectations, [12] by aiming to help patients achieve the best satisfying health outcome in the most efficient way. [7] So both the provider and the receiver of the health care service may be making decisions based on the level of satisfaction as measurable by outcome studies. [13] ' The information gathered from such studies can be used in some of the following areas: to compare between different providers, examine at the provider performance over time, set priorities for projects that may improve quality, and determine the influence of changes in quality of care with changes in the healthcare delivery system. [14] Satisfaction studies have been conducted in many disease states and conditions, for example, studies have examined patient satisfaction in HIV/AIDS, diabetes, and cancer. [15][16][17] Information from such studies are usually gathered from a single item or multiple item questionnaires, where the multiple item questionnaires tend to be more reliable in capturing different levels of satisfaction. [2] However, the main problem with these kinds of studies is that high satisfaction has been reported utilizing many of them and there is a potential bias involved in the collection of such data. [18] 36 The influence of prescribing drugs on patient satisfaction, in general, has been shown in a study that explored whether patient expectations drive drug prescribing and if that has any effect on the level of satisfaction. The study showed that meeting the expectations of the patient is not associated with higher satisfaction. [19) Several satisfaction studies have been conducted in patients with acute nonpneumonic respiratory tract infections (ARTls). The objective of these studies was to explore the influence of both the prescribing of an antibiotic and the patients expectations on their satisfaction. One study showed that 65% of patients with respiratory infection expected an antibiotic, 63% received them, and 97% of the patients were satisfied with their visit. [20) No association was found between an antibiotic prescription and patient satisfaction. However, the strongest association with patient satisfaction was based on whether the physician spent sufficient amount of time explaining to the patient about his illness.
A second study explored the influence of reducing antibiotic prescribing for uncomplicated acute bronchitis on patient satisfaction. [21) The study concluded that reducing antibiotic prescribing through educating both patients and clinicians was not associated with a reduction in the satisfaction of care. A third study has also concluded that in patients with upper respiratory tract infections, patients want to be diagnosed but their satisfaction with their medical care depends on the personal interest and reassurance they receive from their provider. [22] The over prescribing of antibiotics in ARTls, that in many cases may be unnecessary, has been demonstrated in many studies. [23][24][25] This overutilization is not without the risk of getting antibiotic resistance and the high potential cost associated with such practice. [26][27][28] For that and for other reasons, it is important to study the influence of any factor that may lead to such a practice.
The objective of this study was to explore the influence of prescribing antibiotics to patients with ARTls on their reported satisfaction with the provided quality of care. The hypothesis behind the study is that patients receiving antibiotics for ARTls are more likely to be satisfied with their reported quality of care than in patients not receiving them.
The results of this study will be of a great value for both the providers and the payers, who are trying to please and retain patients in their plans and practices and, at the same time, lower their cost and avoid unnecessary use of medications. This study is unique compared to studies conducted in the past, because it attempted to study the patient satisfaction from a national dataset that allowed calculation of national estimates about satisfaction. For the purpose of this analysis both files were merged. This allowed us to append personal characteristics such as demographic and health insurance coverage to each prescribed medicine record. Information about merging was derived from the data codebooks. [33,34] The data were then transposed to make the unit of interest to be patients, instead of prescription events.
Study population: The study population was (N=514) patients with ARTls, who answered the question about their satisfaction with the quality of care provided to them.
ARTls were identified by the International Classification of Diseases, Ninth 40 .. n Clinical modification codes (ICD-9 codes), that were associated with Rev1s10 r ibed medicines. Each prescribed medication was associated with up to presc three ICD-9 codes. The prescribed medicines were reported by patients and were then confirmed from pharmacy records. The study only included prescriptions with one ICD-9 code to assure that the prescribed medicine of interest is for a particular ICD-9 code.

Dependent Variable (DV):
The dependent variable of interest in this study was the reported patient satisfaction with their quality of care. Information about this variable was gathered from the satisfaction question only asked in the second round of the 1996 MEPS and therefore the study population was limited to only this round.
This variable was reported in four different satisfaction levels and, for the purpose of this study it was dichotomized to very satisfied and to some degree of dissatisfied.

Independent Variable (IV):
The main variable of interest was whether or not an oral antibiotic was utilized tor any of the ARTls conditions explored in the study, and if that had any effect on patients' satisfaction level.

Other variables:
Demographic and socioeconomic variables were also examined for any ' difference in the satisfaction level and to determine whether they may be potential confounders. Variables included: sex, race, age, marital status, census region, poverty line, type of health insurance, the number of prescriptions and refills in 1996, and total health care expenditure in 1996.
The last two variables (number of prescriptions and total expenditure) were used in the analysis to try to capture both the burden of other co-morbidities and extent of overall health care utilization on the satisfaction level. The information about other conditions were not available in the data used for this study. The information about other co-morbidities can only be gathered for the medical condition file. However, the problem with that is that not all records in the medication file have a match in the prescription event file used in the study.

Statistical analysis:
This study utilizes both descriptive and analytical techniques in its analysis. A chi-square test was carried out to examine the difference between the various independent variables on the patient satisfaction level, with p-values reported.
The level of significant used for a p-value was alpha=0.05.
Logistic regression was used to determine the relative risk of the utilization of antibiotics on the level of patient reported satisfaction with their quality of care.
The final logistic regression model was run while controlling for confounders.
Odds ratios and the 95% confidence interval were used.

Results:
The study population of unweighted and weighted data of 514 patients with ARTls is shown in Table 1. It demonstrates the high rate of satisfaction of the study population regardless of any medication use, which was around 75% versus 25% for all the other levels of satisfaction. Satisfaction, which was dichotomized into complete satisfaction (i.e., very satisfied) or to some degree 44 of dissatisfaction, was also reported. The table also shows the rate of antibiotic use in patients in the study population, which was around 63%.
A simple cross-tabulation, without controlling for any other variables, was conducted between the satisfaction level and antibiotic use as shown in Table   2 , in which patient's satisfaction is very high whether or not antibiotics were used. It was around 75% for both groups.

'
The demographic characteristics of the study population are presented in Table 3, for both the weighted and unweighted data, which shows that 85% of the patients are white Americans and 57% are female. Children 17 years old and younger accounted for 45% of our study population and that was probably why the majority of patients in the study were not married (71 %). A chi-square analysis was also conducted between the main independent variable (antibiotic utilization) with all the other independent variables, in order to help verify any potential confounder. The variables that showed a significant ' difference were race, the type of health insurance coverage and poverty status. However, when the data was weighted using SUDAAN, the variables that showed a significant difference were only race and the type health insurance coverage.
From the results of the conducted chi-square tests, we concluded that the only potential confounder for the unweighted data was race, since it showed a significant difference when looking at both antibiotic utilization and the satisfaction variables. However, no potential confounders were found when the data was weighted with SUDAAN.
The univariate logistic models showed that race and the total number of prescriptions in 1996 was the only variables which had an independent effect on the level of satisfaction. The interaction assessment conducted concluded that no interaction terms exist between the variables. The analysis also showed that no highly correlated variables exist.
The final logistic regression model presented in Table 6 indicated that, before controlling for other variables, antibiotics effect on the level of patient satisfaction with their quality of care is not significant with an odds ratio (OR) = and a 95% confidence interval (Cl) = 0.63 -1.63 and a p-value of 0.9707. Similar results were found, when trying to control for race, total healthcare ' expenditure in 1996 and the total number of prescription in 1996.

Discussion:
The results of the study demonstrate that overall satisfaction with quality of care in patients with ARTls in 1996 were very high, which was around 75% of patients. The results also showed that patients with different demographics and socioeconomic background were very satisfied, regardless of any antibiotic use.
Our study suggested that the utilization of antibiotics in patients with ARTls is not considered to be an important factor that may influence patient satisfaction with their quality of care. Results from our logistic regression analysis indicated that no changes in the level of satisfaction were found whether antibiotic were used or not. Findings were consistent with many of the studies that examined the influence of medication utilization on patient satisfaction in general and in ARTls in particular. [19][20][21][22] A number of potential limitations are associated with this study. First, the study is a cross-sectional, which prevented examination of the change in the reported patient satisfaction over time. Second, the satisfaction level was very high in all the variables used in the study, which may suggest the presence of a bias introduced by the survey administrators (i.e., interviewer bias) or by the instrument used in the study. Third, the satisfaction question in the study was asked based on the satisfaction with the quality of care, reported by patients or by their parents, and was not specifically based on antibiotics utilization.
Despite these limitations, the methodology used in the study was able to utilize a national dataset to capture the influence of antibiotic utilization on the reported satisfaction on a national level. The results of the study may be very helpful for physicians and providers of the health care system, who are trying to use antibiotics appropriately and at the same time to satisfy their patients.
The results illustrated that they should not worry about patient satisfaction when treating these illnesses and should instead concentrate on the appropriate use of antibiotics in their patients.
This research points out the need tor similar studies that will take into account the problems and limitations associated with studies measuring patient · of their health care. More standardized and disease specific perceptions instruments should be used, only after being tested for both validity and reliability, which in many cases has not been done. [18,40] That will help avoid, or at least minimize the biases introduced, and will also help interpret the study findings accurately. [18] Researchers and people involved in the design of a future national database, whether it is MEPS or any other, should pay more attention and try to introduce new questions that will capture the influence of the medication use on patient satisfaction. That will help provide national estimates, instead of using small studies that in many cases can not be generalizable and are of limited use to health care decision makers.

CONCLUSION:
The results of the study suggest that the level of patients '

' 62
The Influence of Drug Copayment on Antibiotic Utilization in Acute

Respiratory Tract Infections
Abstract Background: Identifying the influence of prescription drug copayments on the decision to obtain or prescribe an antibiotic is essential and may play an important role in the health care decision making process due to the associated ~ost and the potential for misutilization.
Objectives: The main objective of this study was to explore the effect of the drug copayment of the prescribing of antibiotics in acute respiratory tract infections (ARTls) and if that may increase or decrease the inappropriate use of such antibiotics. The study also tried to estimate the total savings in dollars to both the insurers and patients when paying for these unnecessary medications prescribed for different ARTls.
Methods: A retrospective study of (N=1635) prescription events associated with different ARTls was identified from the 1996 Medical Expenditure Panel Survey (MEPS). Antibiotic use was determined by selecting oral antibiotics and the associated source of payment (e.g., copayments) with each prescription was also identified. A logistic regression model was developed to evaluate prescription drug copayment effect on prescribing an antibiotic. The total cost that may have been spent on these antibiotics was also estimated .

Results:
In 1996, the total spending on potentially inappropriate antibiotics for different ARTls was around $394 million, which could be saved for both the patient when paying copayments and to the health care insurer. The results of the logistic regression analysis concluded that medication copayments do not have a significant effect on the utilization of antibiotics in these conditions, with ' an odds ratio (OR) = 1.03 and a 95% confidence interval (Cl) = 0.99 to 1.08 and a p-value of 0.1269.

Conclusion:
The results of the study indicate that prescribing antibiotics for ARTls, which may be unnecessary, are associated with high medication costs and could be avoided . This is especially true with the limited resources available and with the rapid increase in the health care spending. The study also showed that the likelihood of being prescribed an antibiotic is not affected by the level of drug copayment. However, more research is needed to examine the effect of the different copayment tier systems introduced in the last couple of years on the utilization of these drugs.
Recently, the increase in the cost of health care has reached alarming levels.
In 2 001, health care cost was around $1.4 trillion, which accounted for 14% of the total United States gross domestic product (GDP), and prescription drugs accounted for about 10% of the total health care cost. The numbers are expected to rise to about $2.8 trillion in 2011, which will account for 17% of the estimated total GDP. The figures for prescription drugs will reach 14.7% and the total spending on pharmaceuticals will be 2.5% of the total GDP. [1] The above figures have led many people interested in this area to focus and conduct research to identify methods that may lead to the lowering of total spending on health care in general and on prescription drugs in particular.
Their research is aimed at minimizing spending without affecting the level of patient satisfaction and other health care outcomes. [2][3][4] Techniques that address prescription drug utilization were mainly developed by Pharmacy Benefits Management organizations (PBMs). These organizations use many tools and techniques to accomplish their goal of lowering the cost they incur for medications. The techniques include disease state management, formulary management, drug utilization reviews , delivery systems (e.g., mail order, retail), and by consumer sharing (e.g., copayments). [2,[5][6][7] This study focused on one of these cost sharing techniques, drug copayment.
Many copayment strategies have been initiated by public and private organizations to lower their share of prescription drug costs. For example, a survey showed that many employers intend to increase the cost sharing in their health care benefit structure, rather than eliminating drug coverage. [8] Many states are also limiting or restricting eligibility in their drug coverage programs. [9] ' Copayment is defined as a fixed amount of dollars that a patient must pay per prescription. [1 O] Health plans may use different tier system when applying copayments. In 2000, 49% of covered workers were in plans of two-tiered structure. The rate was down to 37% in 2001. [11] The reason behind using two tiers was to differentiate between generic and brand drugs and to encourage patients to use generic drugs at a lower copayment level. [1 O] However, generic drug dispensing rate of 42% has not change from 1996 to 2000, with a decline in the annual sales of generic prescriptions from 20.5% in 1996 to around 17.8% in the year 2000. [12] In order to increase the use of generics, different health care plans are now moving to simple multiple tier systems (e.g., three tier) and to more complex systems (e.g., four and five tiers) and even to coinsurance, where patient pay a percentage of the cost of the drug. [3, 1 O, 13, 14] Multitiered copayments were created to make a balance between what health plans need to control pharmaceutical costs and their need to avoid public criticisms for limiting access to drugs. [15] Research conducted in the past about cost sharing has suggested that overutilization may be decreased when patients are required to pay part of the expenses they incur. [16][17][18][19][20] One report that surveyed about 150 health plans with different cost sharing plans concluded that a direct relationship exists between the decrease in drug utilization with the increase in cost sharing. The ' respondents recommended a 5%-10% cost sharing to have an effect without having a negative outcome on the use of essential drugs. [21] Another study indicated that an increase in drug copayments from $5 to $7.50 was associated with 12.3% reduction in drug spending in health maintenance organizations. [22] Also, a study that examined the effect of three tier copay systems concluded that these techniques can control drug costs. [23] Information regarding the effect of drug cost sharing and copayments on the utilization of essential drugs are conflicting. One study showed that increased cost-sharing for prescription drugs in elderly and welfare recipients was associated with a decrease in the use of essential drugs and with higher rate of emergency department visits. [24] One survey suggested that higher out of pocket costs may causes a high non-compliance among users of prescription drugs. [25] On the other hand, a study demonstrated that compliance rates with chronic disease medications in the period of 6 months after a copay increase of $10 to $ 15 for brand drugs was not reduced when comparing the study group to a control group. [26] Other studies have also shown no effect on the use of essential drugs when copayments are increased in certain disease states (e.g., cardiac medications). [27,28] To date, no study has examined the influence of drug copayments on the The importance of exploring these issues is crucial, especially in conditions like ARTls where there are, in many cases, not only an overutilization and inappropriate use of antibiotic, but unnecessary costs resulting from both the cost of the drug and from the side effects associated with its use (e.g., antibiotic resistance The objective of this study was to explore the effect of the drug copayment on the prescribing of antibiotics in ARTls and if there is an increase or decrease the inappropriate use of such antibiotics. One hypothesis was that patients responsible for no copayments were more likely to receive an antibiotic for one ' of the ARTls conditions compared to patient with copayments. The study also tried to estimate the total savings in dollars to both the insurers and patients when paying for these unnecessary medications in the different ARTls. For the purpose of this analysis both files were merged. This allowed us to append personal characteristics, such as demographic and health insurance coverage, to each prescribed medicine record. Information about merging was derived from the data codebooks. [38,39] study population: The study population was for 1635 prescription events associated with one of the following. ARTls: acute nasopharyngitis (common cold), acute sinusitis, acute pharyngitis, acute tonsillitis, acute laryngitis and tracheitis, acute unspecified upper RTls, and acute bronchitis. These conditions were identified from 1996 MEPS by using the International Classification of Diseases, Ninth Revision Clinical modification codes (ICD-9 codes).

Exclusion criteria:
Since each prescribed medication was associated with up to three ICD-9 codes, any event with more than one ICD-9 code was excluded to assure that the prescribed medicine of interest was for a particular ICD-9. The study also excluded any prescription event where the patient or any family member paid for the prescription in full, since the study was only examining the influence of copayments.

Dependent Variable (DV):
The dependent variable of interest was defined as having been prescribed an antibiotic or not, and only oral antibiotics were included in the analysis.

Independent Variable (IV):
' Copayment was the independent variable of interest in this study. Information about this variable was constructed from prescription drugs payment information available in the data. Copayment was examined both continuously and categorically. In the latter case, the variable was designated as no copayment or copayment. Two categories were created because many of the copayments in the study population were between $5 and $10, and it was difficult to determine more break points.

Other variables:
Demographic and socioeconomic variables were also examined for any association with antibiotic utilization and to determine whether may confound the effect of copayment on the utilization of antibiotics in ARTls. Variables included: sex (male or female), race (white or non-white), age (less than 17 years old or older than 18), census region (northeast, west, south, or midwest), poverty status (poor to low income or moderate to high income), type of health insurance (any private, any public, or no insurance).

Statistical analysis:
A simple descriptive analysis of the study population was conducted, and included examining the different demographic and socioeconomics characteristics in this population.
The analysis also estimated the total savings to patients and to public or private drug coverage insurers, if antibiotics have been avoided in the different conditions of ARTls. The unweighted total savings were calculated by summing up all antibiotics payments, both by patients and by health insurers, and then the total number of antibiotic prescriptions and the total dollar value were weighted using SUDAAN to obtain national estimates.

Other analysis:
A bivariate analysis was conducted between the different study variables with antibiotic utilization, and a chi-square test was carried with p-values reported.
The level of significant used for a p-value was alpha=0.05.
logistic regression analysis was utilized to evaluate prescription drug copayment effect on prescribing an antibiotic. The final logistic regression model was run while controlling for confounders. Odds ratios and the 95% confidence interval were reported .
Several analytical techniques were used to build the final logistic regression model. First, variables were dichotomized and dummy variables created for the following variables: age, census region, health insurance, income level.
Second, univariate logistic regression was used to evaluate the relative contribution of each independent variable on predicting antibiotic utilization. -Information about the study population including demographic and socioeconomic characteristics are presented in Table 1, which shows both the weighted and unweighted data for (N=1635) medication events with different types of ARTls. The table also highlights that in 1996 around 73% of medication events for ARTls were associated with a copayment. The rate of antibiotic prescribing for these conditions was 49%.

'
The analysis also estimated the total direct cost in dollars that can theoretically be saved by eliminating unnecessary antibiotic use in ARTls (Table 2) and it shows that in 1996, the total savings to patients from copayments in conditions like common cold, acute tonsillitis, acute laryngitis and tracheitis, acute unspecified upper RTls, and acute bronchitis, were antibiotics are always considered unnecessary, was around $71 million. Potential savings for different drug coverage programs were estimated at $243 million. Table 2 also shows that in 1996, the total saving to patients from copayments in conditions like acute sinusitis and acute pharyngitis, where antibiotics may be used but in many cases are considered unnecessary was around $17 million and $63 million would accrue to the benefit of providers and payers of different drug coverage programs. The total cost that could have been saved in 1996 from inappropriate antibiotic prescribing in ARTls was approximately $394 million.
The results of the weighted cross tabulation between copayment and antibiotic utilization against the different study variables are presented in Table 3, which also shows the results of the chi-square test. The variables that demonstrated a significant effect (p-value less than 0.05) with drug copayment were health insurance, income level, age and race. Variables with a significant effect on antibiotic utilization were health insurance, income level and race.
The results from the different techniques for logistic regression showed no ' correlation between variables in the study population. Univariate logistic regression showed that the variables that have an independent effect on antibiotic utilization were only health insurance, income and race. Table 4, present information regarding the results of confounders assessment, in which income level, health insurance and race were potential confounders that need to be controlled for in the final logistic model.
The results of the final model showed that before controlling for the potential confounder, copayments had a significant effect on antibiotic utilization in different events of ARTls with an odds ratio (OR)=1.06 and a 95% confidence interval of 1.01 to 1.11 and a p-value of 0.0171. However, after controlling for potential confounders (i.e., health insurance, income and race) the effect of copayment on the utilization of antibiotics disappeared, with an odds ratio (OR)=1.03 and a 95% confidence interval (Cl) = 0.99 to 1.08 and a p-value of 0.1269.

--
The findings from this study suggest that in 1996, patients or a family member paid a copayment for antibiotics prescribed for different ARTls in around 66% of events. The study also showed that the effect of copayment level on the utilization of antibiotics in these conditions disappeared after controlling for potential confounders.

'
The reason that in 1996 drug copayments has no effect on the utilization of antibiotics in ARTls may be due to 1) patients may have a perception that they need antibiotics regardless of copayment level, and 2) copayment levels were low, in many cases less than $10 per prescription.
This study is the first that has used a national dataset that allow the calculation of national estimates of the effect of drug copayment on antibiotic use in ARTls. The study also was able to calculate a national estimate of the direct cost in dollars that can be saved by both the patient and the prescription drug provider when these drugs are avoided, which was around $394 million in 1996. This figure was lower than the expenditure reported for 1998, which was around $726 million, suggesting an increase in the spending on antibiotics associated with treating these conditions from 1996 to 1998. [30) 77 A number of potential limitations are associated with this study. First, the data used in the analysis does not provide information about the different benefits and copayment tier systems associated with each prescription drug plan.
second, the study is cross-sectional in nature and may prevent analysts from capturing the influence of any change in the copayment tier system on antibiotic utilization over time. Finally, the data does not provide information regarding the causative agent, whether it is viral or bacterial, and also does not provide a history of other antibiotic use.
Despite these limitations the study was able to show results that will assist policy and decision makers, both in the government and the private sector, who are responsible for designing the different copayment tier systems. It also illustrated that without the proper management of these conditions, significant levels of waste in the health care system can occur. This is of particular concern especially with the rapid increase in the cost of health care system with limited and scarce resources. The results of the study indicate that prescribing antibiotics for ARTls which may be unnecessary are associated with high medication cost that should be avoided especially with the limited resources available and with the rapid increases in the health care spending. The study also demonstrated that the likelihood of being prescribed an antibiotic is not affected by drug copayment.
However, more research is needed to examine the effect on the different copayment tier systems introduced in the last couple of years on the utilization of these drugs.    antibiotics. [1] These infections can be divided, based on the location of the infection, into lower respiratory tract infections (LRTls) such as pneumonia and bronchitis, and to upper respiratory tract infections (URTls) which includes rhinitis (common cold), sinusitis, pharyngitis (sore throat), and laryngitis. [1,2] LRTls are considered to be one of the main reasons for morbidity and mortality, and for the increased health care utilization due to infections. It is estimated that 3.8 million deaths occurred in 1998 worldwide due to LRTls. [3] It is also estimated that in the United States, at least 500,000 hospitalizations occur each year with an estimated cost of $8.4 billion and with an additional $8.4 billion spent on community acquired pneumonia. [4] These infections are mainly caused by Streptococcus pneumoniae. Other pathogens may include Haemophilus influenzae, Moraxella catarhalis, or in some cases, viruses. [5] Antibiotics are usually recommended in these kinds of bacterial infections and the choice of antibiotic depends on the causing pathogen, and ranges from inexpensive first generation penicillins to more costly broad spectrum drugs such as fluoroquinolones. [5,6] 92 I The case is different for bronchitis and URTls since most of these conditions are not life threatening and they are usually caused by a virus, especially in acute illness. However, bacteria (e.g., Streptococcus pneumoniae) may be behind the infection in chronic illness. [7] For that reason, antibiotics are usually not indicated in acute infections, regardless on whether or not it is caused by a virus or a bacteria, since in most cases the illness resolves naturally and fast.
[8] It is also important to note that antibiotics have not been shown to improve the clinical outcome of these patients. [1,2] However, ' patients with chronic bronchitis and some patients with chronic sinusitis and/or pharyngitis may still benefit from an antibiotic. [7] Many studies have demonstrated the inappropriate overutilization of antibiotics in certain acute RTls. [9,10] In 1998 it was estimated that 76 million office visits for acute RTls resulted in 41 million antibiotic prescriptions. This was 55% (22.6 million) more prescriptions than were expected to be used, with an associated cost of approximately $726 million. [11] Another study showed that between 1989 and 1999 there were 6.7 million annual office visits by sore throat (pharyngitis) patients, where approximately three quarters of adults (73%) were prescribed an antibiotic and 68% of the prescribed antibiotics were for non recommended use. The study also showed that broad spectrum and more expensive antibiotics are frequently used. [12] The overutilization and inappropriate use of antibiotics in the community has not only led to higher treatment costs but has consequently led to the emergence of antibiotic resistant bacteria, which is becoming an epidemic and global problem. [7,[13][14][15] That is especially noted for one of the most common respiratory tract pathogen Streptococcus pneumoniae. However, resistance is not limited to one pathogen, it also includes resistance to Haemophilus influenzae, Moraxella catarhalis, and other bacteria. [13] The level and the mechanism of bacteria resistance to antibiotics vary between different ' antibiotic classes. [13] Many factors have played a role in the over prescribing of antibiotics. Patients, for instance, have an expectation of receiving an antibiotic and are consequentially pressure their physician to prescribe one. They may also be financially responsible for a low or no copayment for their medications which may give them more incentive to fill the prescription. [16,17] Physicians, on the other hand, try to meet the expectation of their patients and try to avoid any liability in the case of treatment failure, leading them to prescribe broad spectrum antibiotics. [18] Other factors may include the health care system structure, formularies and prescribing restrictions. [19] Also, directed advertising and marketing to physicians and consumers by pharmaceutical companies may lead to more prescription of high cost broad spectrum antibiotics. [20] With the high cost associated with the overuse of antibiotics and for the fear of more antibiotic resistant bacteria it becomes important, from the perspective of the decision and policy makers, to explore more of the factors and components that may affect the prescribing pattern of antibiotics. It was the intent of this study to explore the effect of drug coverage type, copayment, and patient reported satisfaction with their health care when prescribed an antibiotic. So far, a limited number of studies have been conducted on these factors.
Studies that explored the health insurance effect on prescribing antibiotics were either limited to a small group of patients, or did not look specifically at patients with acute respiratory tract infections. [21][22][23] Studies that examined the copayment effect demonstrated that an increase in the patient co-payment led to a decrease in the drug prescription and/or purchase. However, most of these studies were conducted on various classes of medications in a managed care organization setting and did not examine prescribing differences between different health care types. [24][25][26] The association between patient satisfaction and receiving an antibiotic was only demonstrated in two studies that have shown no association between getting an antibiotic and patient satisfaction. [27,28] However, these studies had a limited number of patients enrolled and only examined satisfaction based on the patient expectation, and not on overall health care service.

95
The uniqueness of this study derives from its ability to use a large national survey that is actually designed to help study the effect of drug coverage, copayments and patients satisfaction on the use of health care services. The result of this study provided decision and policy makers with more insight and information on what may be done in practice to avoid excess cost and antibiotic resistance caused by inappropriate use of antibiotics.