To describe the implementation and impact of a centralized statin switch program at a large academic medical center.

Patients on atorvastatin were identified from electronic medical records and pharmacy claims data. Relevant information was sent to physicians for approval of the proposed switches. Approved patients were then contacted via phone and offered the opportunity to switch to simvastatin; those who switched received a new prescription for simvastatin. To assess the independent impact of the active switch process, conversion rates within a single insurance plan were compared for patients who participated in this program versus those who were contacted only by mail.

Physicians approved 3207 of the 3677 patients identified for this program. A total of 1710 approved patients accepted the switch, 704 declined, and 170 became ineligible. Information packets were mailed to 623 patients who could not be contacted. Within the single insurance plan, the generic dispensing rate for statins among the 1867 patients included in our program was significantly higher than that for the 2472 patients in the mail-only group (59.2% vs 35.8%, 

 <.001). Over 8 months, the direct cost of the program was $131,000 with projected annual cost savings of up to $1.14 million to payers and up to $250 for each patient who switched.

A proactive and voluntary statin switch program to promote the use of a lower cost generic alternative can be successfully implemented in a fee-for-service health system setting with benefits to patients, providers, and payers.

A statin switch program was successfully implemented in a fee-for-service patient population by a large physician group in an academic healthcare system setting.

This collaborative and voluntary program allowed physicians and patients to have complete control of the switch process.

A centralized switch program saved time and money, while promoting the use of a lower cost generic alternative.

Over an 8-month period, 1710 of 3677 patients identified for switching from a brand to a generic statin did so at a direct cost of $131,000 with projected annual cost savings of up to $1.14 million to payers because of lower drug costs. In addition, each patient who switched saved up to $250 because of lower copayments.

Prescription drug spending has significantly increased in the past decade, totaling $217 billion in 2006 and representing 10.3% of total healthcare expenditures in the United States.

 Of this amount, 22% was paid by patients out-of-pocket, 34% by public health insurance plans, and 44% by private insurance plans. In addition, prescription drug spending is predicted to increase by 8% to 9% annually through 2016.

Targeted medication conversions to more cost-effective alternatives represent a safe and effective way to reduce drug expenditures for patients, providers, and payers. Studies have shown that lower prescription drug copayments are correlated with higher levels of medication compliance for 3-hydroxy-3-methylglutaryl coenzyme-A reductase inhibitors (statins), which are commonly used cholesterol-reducing drugs.

 Gibson et al studied the impact of statin copayments on patient adherence and healthcare expenditures.

 They found that every $10 increase in copayment resulted in a 1.8% reduction in medication adherence among new users and a 3% reduction among continuing users. The study also found that among continuing users adherent to statins, there was a significantly lower incidence of coronary heart disease—related hospitalization.

 Similarly, Schultz et al found that a $15 increase in copayment was associated with a 10% decrease in statin compliance in a retrospective analysis of claims data.

 Results from these and other pharmacoeconomic studies suggest that interventions to reduce patient copayments will improve statin adherence and, in turn, improve outcomes and reduce overall healthcare costs, especially in high-risk populations.

All currently available statins have been shown to reduce the incidence of cardiovascular events, which supports the belief that these benefits are class effects.

 Charts comparing doses of statins that provide similar levels of low-density lipoprotein cholesterol (LDL-C) reduction have facilitated the substitution of one statin for another.

 In June 2006, simvastatin (Zocor) became the third statin available as a generic, and many insurers instituted programs such as tiered copayments and step therapy requirements to promote its use and to encourage their members to switch from brand-name statins, if clinically appropriate. For example, the University of Michigan Benefits Office moved all brand-name statins, except atorvastatin (Lipitor) 80 mg, to the highest copayment level and added a step-therapy edit requiring patients to try a generic statin before any other brand-name statin could be covered.

 When such a change occurs, patients are typically notified via mail and asked to contact their physicians to reassess therapy and adjust medications if necessary. When a patient calls to request a medication change, the new prescription has to be approved and then called in or faxed to the patient’s pharmacy. This process involves multiple message relays that require staff and clinician time and effort, which can disrupt clinic workflow and may result in patients being temporarily without their medication. All of this causes additional, avoidable costs to patients, providers, and payers.

To assist our patients and physicians, the University of Michigan Medical School Faculty Group Practice, which includes approximately 1400 physicians who are directly involved in patient care, initiated a voluntary program to switch eligible ambulatory patients from A to Z (atorvastatin [Lipitor] to simvastatin [Zocor]). Although this type of switch program has proven effective in organizations with fixed formularies, such as the Department of Veterans Affairs and Kaiser Permanente, our program is the first we know of that targeted a large multipayer population, including fee-for-service patients, through a physician group.

This program was implemented at the University of Michigan Health System (UMHS) in Ann Arbor. The UMHS is a nonprofit teaching institution that provides more than 1.6 million outpatient visits annually between 3 hospitals and 30 health centers, which include 120 primary care or specialty clinics. It serves a large multipayer population.

The switch program was scheduled to begin approximately 6 months after the introduction of generic simvastatin as the price was expected to drop at that time with multiple manufacturers entering the market. The program was approved by the UMHS Ambulatory Formulary Committee and the Faculty Group Practice. The analysis of the program, reported here, was approved by the University of Michigan Medical School Institutional Review Board.

Patients received letters notifying them about the forthcoming changes in prescription copayments and preferred medications from their respective insurers. Physicians received several e-mail messages informing them about the changes in their patients’ prescription copayments, reasons for the switch program, and how it would be conducted. In addition, presentations about the program were given to primary care physicians, cardiologists, and endocrinologists at their departmental meetings. A clinical pharmacist also visited several high-volume clinics to answer any additional physician or staff questions. Health center personnel were informed of the program by their medical and Administrative directors.

Eligible patients were identified from electronic medical records and pharmacy claims data provided by 2 managed care organizations, 1 fee-for-service insurer, Michigan Medicaid, and the University of Michigan Benefits Office. Patients were eligible if as of January 2007, they were (1) on atorvastatin 10, 20, or 40 mg/day, (2) enrolled in a collaborating prescription drug plan, and (3) receiving care from a UMHS primary care physician, cardiologist, or endocrinologist. 

All patients enrolled in the University of Michigan Benefits Office’s prescription drug plan received identical letters in November 2006 about the coverage changes for statins, but only patients with a UMHS provider were included in the active switch program. Those patients without a UMHS provider served as the control group to help evaluate the benefits of this active switch program over the passive approach of sending patients letters.

All patients on atorvastatin 80 mg/day were excluded from the switch program because we could not obtain similar levels of LDL-C reduction with simvastatin and because there were randomized controlled trials demonstrating benefits of atorvastatin 80 mg/day for certain patient populations. Patients on medications with possible differential response to simvastatin compared with atorvastatin (ie, carbamazepine, cyclosporine, danazol, delavirdine, efavirenz, imatinib, oxcarbazepine, risperidone, verapamil, warfarin) were eligible for the program, but their physicians were notified of these potential drug interactions.

An electronic database was created using Microsoft Access 2002 that included (1) pharmacy claims data from participating insurance plans; (2) patient demographics (eg, name, age, sex, address, phone), select medication history, and latest laboratory values (total cholesterol, high-density lipoprotein cholesterol, LDL-C, triglycerides, alanine aminotransferase test) from our health system’s electronic data warehouses; and (3) the physician identified for each patient using an algorithm based on the number of clinic visits the patient had in 2006. From the database, reports were produced for each physician with information on each of their eligible patients and, for patients using a mail-order pharmacy service, a preprinted prescription. Physicians reviewed the proposed medication switches, changed the statin dose as needed, and approved or declined their patients to be contacted regarding the switch program. To facilitate this process, clinical pharmacists per formed a preliminary review of the reports for 7 high-volume physicians (ie, more than 50 assigned patients, range 55-130), as well as several others who asked for help. A total of 972 (26%) patients were reviewed by 3 clinical pharmacists. Physicians returned their reports to the switch team and the information was entered into the program’s database.

It was recommended that patients switch to an equipotent or greater dose of simvastatin, with respect to LDL-C reduction: (1) from atorvastatin 10 or 20 mg to simvastatin 40 mg, based on the published results supporting moderate dosing of statins,20 and (2) from atorvastatin 40 mg to simvastatin 80 mg.

A computer-aided telephone interview program was developed to guide pharmacy students in their discussions with eligible patients. Using a standardized script, the students informed patients about the program, told them that their physicians had approved the switch, and asked whether patients wanted to switch from atorvastatin to simvastatin. A maximum of 6 attempts were made to contact each patient, including 2 during weekdays, 2 during weeknights, and 2 on Saturdays. An information packet about the program was mailed to patients who could not be contacted. 

Patients who elected to switch received a new prescription for simvastatin, which was either called in to their local pharmacy within 24 hours or sent to them if they used a mailorder pharmacy. Patients using mail-order pharmacies had the option of having a 30-day supply of simvastatin called in to a local pharmacy if they had fewer than 28 days of atorvastatin remaining. Each patient also received a letter confirming the medication change and a laboratory requisition form for a fasting lipid profile and serum alanine aminotransferase test, to be drawn approximately 8 weeks after initiating simvastatin. The patients’ electronic medical records were updated. At the program’s conclusion, physicians were given a report summarizing their patients’ switch status.

We took advantage of a natural experiment and compared the proportion of patients in 1 prescription drug plan who converted from atorvastatin to simvastatin in the active switch program with a similar group of patients in a passive switch program (ie, they received a letter notifying them of the copayment increase for brand-name statins). As in The active switch program, patients in the passive switch program who either were on a brand-name statin other than atorvastatin or were on atorvastatin 80 mg/day were excluded.

Total program expenditures include materials/supplies and the salary and benefits of personnel directly involved with the switch program. Personnel included an internist, a Clinical pharmacist, a data analyst, a programmer, an administrator, an administrative assistant, and 9 pharmacy and nursing students. The amount of time each individual spent on the program was determined to calculate the number of full-time equivalent staff needed for this program.

Total program savings were calculated from an insurer’s perspective and were based on the differences in estimated drug costs for atorvastatin and simvastatin for the patients who switched as part of this program. For a 30-day supply of atorvastatin 20 mg, we used the average wholesale price minus 10%, which was $110 in September 2007.

 For a 30-day supply of simvastatin 40 mg, we used the maximum allowable cost in Michigan, which was $10 in September 2007.

Descriptive statistics were utilized for the demographic data. The chi-square statistical test was utilized for comparative analyses and a 

 value of <.05 was considered to be significant. All analyses were performed with STATA version 8.1 (StataCorp, College Station, TX).

In January 2007, 3677 patients met all the inclusion criteria. Baseline characteristics were similar for patients who did and did not switch from atorvastatin to simvastatin (

). These patients were assigned to 182 physicians, 154 (84.6%) of whom responded with their decisions on 3336 (90.7%) patients (

). Of the nonresponding physicians, 24 (13.2%) were specialists (cardiologists or endocrinologists), accounting for 114 (3.1%) patients. Physicians approved the proposed switches for 3207 (87.2%) patients; of these patients, 1710 (53.3%) agreed to switch from atorvastatin to simvastatin. A total of 52 (1.4%) patients who initially declined the switch and 37 (1.0%) patients unreachable by phone were subsequently switched by their physicians during clinic visits.

Patients on lower doses of atorvastatin were more likely to switch to simvastatin (

 <.01). The average atorvastatin dose of patients who agreed to the switch was slightly lower than that of patients who did not switch (19.0 mg vs 20.4 mg, 

<.01). Almost all the switched patients (99.4%) were converted from atorvastatin to an equipotent or greater dose of simvastatin. Only 59 (3.5%) patients who switched were on 1 or more drugs considered to interact specifically with simvastatin; 39 (2.3%) of those patients were on warfarin. These patients were flagged and additional counseling regarding the potential effects of simvastatin on their anticoagulation tests were provided by a clinical pharmacist.

The University of Michigan Benefits Office had the most patients in this study, with a total of 4339 patients on atorvastatin 10, 20, or 40 mg/day. These patients all received a letter notifying them of the copayment increase for brand-name statins. Those University of Michigan employees or their dependents who had an UMHS provider were included in the active switch program (n = 1867) and the passive switch group consisted of the remaining 2472 patients who had a non-UMHS provider.

, patients in the active group were younger (58.1 vs 61.1 years old, 

 <.001) and were less likely to have drug interaction(s) with simvastatin (5.4% vs 8.3%, 

 <.001). By the third quarter of 2007, generic simvastatin was being used by 59.2% of patients in the active switch program compared with 35.8% of patients in the passive switch group (

). The 23.4% greater generic statin dispensing rate among switch program participants resulted in a $150,000 cost savings by the end of the third quarter of 2007 and an estimated $370,000 in cost savings annually, based on the University of Michigan Benefits Office’s prescription drug plan cost structure. Patients saved up to $168 annually due to lower copayment amounts ($7 for simvastatin vs $24 for atorvastatin), excluding any unmeasured indirect costs (eg, time spent to contact a physician regarding this switch program).

The total program cost was approximately $131,000, including $7500 for materials/supplies and $123,500 for the 1.9 full-time equivalents who worked on the program from November 2006 through June 2007. This total does not include the time that physicians and 2 ambulatory care clinical pharmacists spent reviewing patients for this program, although others might wish to include the opportunity costs of this time in such a calculation. This program was funded by the University of Michigan Medical School Faculty Group Practice, which received a nominal monetary amount for each patient converted from atorvastatin to simvastatin from 2 of the insurers to help cover the expenses of providing this service to our patients.

Total projected annual cost savings ranged from $150 to $250 for each patient because of lower copayments and from $670,000 to $1.14 million for the payers because of lower drug costs. The range in cost savings for the payers reflects the savings that might accrue if all payers engaged in a passive approach such as that by the University of Michigan Benefits Office or if none did. Allowing for some variation in drug pricing and assuming that patients get monthly refills, the cost difference between atorvastatin and simvastatin translates to an annual savings of about $1000 per person switched for the payer.

Our results demonstrate that physicians in a large academic medical center that serves a diverse patient population can successfully conduct a centralized statin switch program. In contrast to some of the other medication switch programs reported, the voluntary and flexible nature of our program allowed physicians and patients to be in control of the switch process, from approval to dosing to prescription processing.

Our security-enabled computer-aided telephone interview program facilitated the electronic management of our database and allowed us to track the status of our patients from initial data load through every step of the switch process. Dynamic call and print queues were set up based on prespecified algorithms and were automatically updated in real to facilitate administrative operations and coordinate staffing needs.

One problem we encountered was that intensive manual data lookup using our electronic medical record system was needed to fill in missing patient information, dosing regimens, and interacting drugs for some patients. Inaccurate patient contact information prevented us from reaching 143 (3.9%) patients. In addition, some patients were assigned to physicians who were no longer seeing them. Some of these problems would not occur with more robust electronic medical records systems (eg, those with coded medication lists) and accurate identification of each patient’s physician.

Articles

MANAGERIAL

This report shows that a successful, cost-effective statin switch program can be implemented by a large physician group via a centralized, collaborative process.

Moving From A to Z: Successful Implementation of a Statin Switch Program by a Large Physician Group

To use the administrative capacity of a health maintenance organization to enroll uninsured patients at an academic health center into a coordinated care program in which patients were assigned to community primary care physicians over 3 years.

Observational case study of a cohort of 2389 patients enrolled for at least 1 year and cross-sectional observations for all enrollees.

Among 18,336 eligible patients enrolled in the program between January 1, 2001, and December 31, 2003, a total of 2389 patients were continuously enrolled before and after the inauguration of the program.

Over the 3-year study, most of the eligible uninsured patients were successfully enrolled in community-based practices. For the cohort studied, reductions were observed in the proportions of enrollees with inpatient hospitalizations (17.6% vs 13.8%) and with emergency department visits (73.9% vs 42.9%) (

 <.001 for both). Although the rates of emergency department visits and hospitalizations were not reduced for the before-and-after cohort, utilization rates per 1000 enrollees declined for the overall program.

Contractual arrangements with community physicians were used to augment primary care capacity for uninsured patients. Although per-member hospitalizations and use of the emergency department did not improve for the cohort studied, declines were observed for the program overall when examined by study year. Using community primary care physicians to coordinate care for the uninsured seems to reduce emergency department use and hospitalizations.

This article describes a program to coordinate the care of an inner-city uninsured population at an academic health center. Using the administrative capacity of a provider-sponsored health maintenance organization, the academic health center contracted with community primary care physicians to better manage uninsured patients. Key findings include the following:

The expanded primary care capacity through the contractual arrangements allowed the academic health center to reduce emergency department overcrowding with uninsured patients.

The improved access to primary care seemed to reduce emergency department visits and hospitalizations over a 3-year period.

Community primary care physicians were enthusiastic about participation, and most remained with the program from its conception.

While academic health centers (AHCs) represent only about 6% of the nation’s hospitals, they provide almost 40% of charity care.

 Approximately 26% of Medicaid discharges are from AHCs,

 and the volume of uninsured patients at AHCs is expanding.

Although some uninsured patients are referred to AHCs for tertiary care, many seek primary care services at AHCs located in the inner city. This can lead to overcrowded emergency departments (EDs) and some resource-intensive hospitalizations that could be avoided with adequate access to primary care.

As AHCs struggle to manage the care of large numbers of uninsured patients, options include expanding primary care services, decompressing EDs, and avoiding unnecessary hospitalizations. Several AHCs have demonstrated that expansion of primary care coverage can reduce ED use.

 When primary care access is enhanced under Medicaid managed care, unnecessary ED visits are reduced by one-third.

 However, increases in primary care coverage can be difficult for AHCs because the physician workforce at many AHCs deemphasizes primary care. This deficit in the academic medical workforce was temporarily altered in the 1990s, when managed care forced many AHCs to expand their primary care capacity.

 However, more recent trends have resulted in a reduction in primary care,

 further exacerbating the capacity of AHCs to effectively manage uninsured patients.

Many urban AHCs are bordered by neighborhoods that include community primary care physicians (PCPs). Therefore, an alternative to enhance primary care capacity is for AHCs to use community-based primary care through contractual arrangements. Herein, we describe a program implemented and the data patterns of patient assignment during a 3-year transition from AHC to community-based settings. The data represent the healthcare use patterns of patients who were transferred from AHC primary care practices to community-based practices during this interval.

The Virginia Commonwealth University Medical Center (VCUMC) is the only comprehensive AHC in central Virginia. It comprises a licensed hospital containing 779 beds, outpatient clinics, and a 600-physician practice. The VCUMC is the region’s only level I trauma center and offers more than 200 specialty areas, including organ transplantation, spinal cord trauma, burn, and cancer treatment. It has more than 30,000 inpatient admissions per year, more than 500,000 annual outpatient visits, and approximately 80,000 ED visits annually. The VCUMC also owns and operates a Medicaid health maintenance organization (HMO), Virginia Premier Health Plan, which began operations in August 1995. This HMO was originally a joint venture with a public forprofit company. Eventually, the HMO was purchased by the VCUMC. The HMO operates in more than 80 counties in central, eastern, western, and southwestern Virginia and has a total enrollment of approximately 130,000 Medicaid beneficiaries. The HMO uses IDX (IDX System Technology, Inc, Torrance, CA) systems for billing and claims processing, referrals, and enrollment tracking.

About 15% of the population of central Virginia is without health insurance, about the same as statewide figures.

 Approximately one-half of uninsured individuals statewide live in households below 200% of the federal poverty level, and the Virginia Medicaid program covers only specific populations within this income level (ie, pregnant women, children, and adults who are aged, blind, or disabled). Adults with income levels above 24% of the federal poverty level are ineligible for Medicaid. The VCUMC provides healthcare for a  disproportionate number of uninsured individuals in Virginia compared with other hospitals and medical centers. The VCUMC is responsible for approximately 30% of total inpatient admissions for indigent patients in Virginia.

The Indigent Care Program in the Commonwealth of Virginia provides assistance to patients who are US citizens and whose household incomes are below 200% of the federal poverty level. All individuals must also meet a financial means asset test. Approximately 30,000 individual patients who qualified for the Indigent Care Program were seen in a VCUMC outpatient setting in 2000. Within this population, approximately 6200 indigent care patients were hospitalized during the same year. 

In November 2000, using the managed care infrastructure from its Medicaid HMO, the VCUMC established the Virginia Coordinated Care (VCC) program to deliver healthcare to eligible uninsured individuals. The VCC program was designed as a care coordination program for patients who qualified for the state’s Indigent Care Program, lived in the Greater Richmond area (	≤50-mile radius of the VCUMC), and had no other coverage options (ie, Medicaid, Medicare, or another sponsoring health plan). This population initially represented about 30% of the uninsured indigent population seeking care at the VCUMC. The VCC was designed to improve access to primary care for uninsured patients who frequently sought care through the ED. The strategy was to use managed care principles to contain the costs of caring for this population. The program has been previously described in a study8 of metropolitan programs that used managed care principles for the care of uninsured patients.

In addition to a monthly management fee, community PCPs were paid fee-for-service rates comparable to approximately 110% of the Medicaid fee schedule in Virginia. The $5 per member per month was comparable to the management fee paid by the AHC’s Medicaid HMO.

In all, there were 50 PCPs at 32 different community practice sites. Among the community providers, there were 26 family medicine physicians at 16 community practices, 5 pediatricians at 5 community practices, 2 family medicine physicians at 2 free clinics, 12 family medicine physicians and 2 pediatric physicians at 8 federally qualified health centers, and 3 internists at a community-based practice affiliated with the AHC. Since the beginning of the program in 2000, few providers have discontinued their participation with the VCC program. All of the community-based providers in the VCC also participated in Virginia’s Medicaid programs. About onehalf of the community providers received graduate or postgraduate degrees from the AHC. Approximately 60% of the community providers have been in practice for more than 15 years. The VCUMC physicians remained the predominant source of specialty care during the study.

The VCC members shared the cost of services based on a sliding fee scale. Enrollees with family income below 100% of the federal poverty level had no cost-sharing responsibilities. More than two-thirds of the VCC members fell into this category. For patients between 101% and 200% of the federal poverty level, cost sharing was based on a sliding scale that ranged from 5% to 70% patient responsibility.

 In addition to the sliding scale payment, patients were responsible for paying $4 per prescription for a 1-month supply. In addition to the systematic assignment of uninsured patients to community-based PCPs, the VCC included a care coordination component. The concept for care coordination services was developed through a private foundation grant in 1998. The model was developed to coordinate services across the AHC’s healthcare delivery system for its uninsured population and included interventions by outreach workers and by nurse coordinators. The outreach workers were individuals with experience in working with the uninsured population in community-based organizations or in outpatient clinic settings. Outreach workers were located in the AHC’s ED and identified patients who were eligible for participation in the VCC, educated them regarding primary care options, scheduled appointments for needed services, provided information regarding community social resources, and arranged for transportation as needed. These outreach workers also facilitated communication between community providers and VCUMC specialists to ensure that patients were linked with appropriate resources and received optimal services. Nurse care coordinators responded to healthcare questions, reinforced treatment plans, and assisted VCC members to navigate through the VCUMC.

Data for the study were from several sources. The VCUMC information systems created monthly patient demographic and utilization data files. These data included information about inpatient discharges and outpatient encounters, including ED visits and prescription drug use. Virtually all inpatient, ED, and outpatient laboratory and radiology services were also provided at the VCUMC. All prescription medications to this population were available only through the VCUMC outpatient pharmacy. Claims for community-based primary care were generated by using the IDX systems processing capabilities of the VCUMC’s Medicaid HMO. A monthly data transfer from the VCUMC’s Medicaid HMO provided claims data for community-based services, patient demographics, and a member enrollment file. The latter file was used to generate data on per-member-per-month utilization and costs. The common unique identifier across these various files was the patient’s medical record number. The study was approved by the VCUMC’s Committee on Human Subjects.

The time frame chosen for this analysis was from January 1, 2001, through December 31, 2003. The analysis focused on patients 18 years or older at enrollment who had a community PCP assignment for at least 1 month during the study. The various data files were combined and summarized using SAS (version 9.1; SAS Institute, Cary, NC). Patient demographic characteristics were determined at the time of initial enrollment. Months of enrollment were summarized at the patient level by PCP type. Healthcare utilization records of patients were aggregated by service type (eg, ED visit or primary care visit). Two different analyses were performed. 

First, to study the effects of the program on patients with prior experience at the AHC, a cohort of 2389 enrollees who received care at the VCUMC before and after the implementation of the VCC program was identified. This cohort included men and women with age at enrollment ranging from 18 years to younger than 64 years who had sought care at the VCUMC before enrolling in the program and who were subsequently continuously enrolled with the VCC for at least 1 year. Before-and-after comparison of service utilization for the cohort was performed. Only enrollees transferred to community PCPs were evaluated in the before-and-after design. Before-and-after comparison of service use was conducted. The null hypothesis of equal proportions was evaluated at the 5% level of significance for these analyses.  

Second, a separate descriptive analysis was performed for all enrollees in each study year. Approximately one-half of the population is new to the program each year. Therefore, many of these enrollees had no prior experience with the AHC or with PCPs.

Microsoft (Redmond, WA) Access 2000 and Excel 2000 were used to calculate the percentages of enrollees with claims and per-member-per-month use for inpatient discharges, ED visits, primary care visits, and specialty care visits. Although most enrollees were transferred to community practice PCPs during the study, some patients stayed with VCUMC practices longer than others, mostly based on physician advocacy. Therefore, the program was not randomized, and the study was nonexperimental in design. Demographic and enrollment characteristics, as well as binary measures of service use, were determined using patient-level data.

Between January 1, 2001, and December 31, 2003, a total of 18,336 unique individuals enrolled in the program for at least 1 month. Some disenrolled from the program during the study because they became eligible for Medicaid, acquired employer-based coverage, or had other reasons to leave the program at least temporarily. Therefore, the analysis was restricted to those individuals who were enrolled for at least 75% of the study. There were 11,834 individuals who met this eligibility requirement. Of these, 9383 patients were assigned to a community physician practice during the study. As patients were enrolled and assigned to community practices, the number of patients with a community PCP steadily increased during the study (

). Eventually, most patients had their primary care transferred to community PCPs. The initial results represent the subset of patients who were continually paneled with a community PCP practice for at least 1 year and had used clinical services at the AHC before enrollment. We also present results from the overall enrollment on a crosssectional basis for each study year.

 gives the characteristics of 2389 continuously enrolled participants. As shown, among 2389 patients, 71% were black, 24% were white, and the rest were of other races/ethnicities. Sixty-two percent of all cohort members were female, and 38% were male. Seventy-six percent reported residence in the City of Richmond. Among 1489 patients with identified sex, 98.8% were female, and 1.2% were male. Seventy-six percent of all cohort members reported residence in the City of Richmond.

 gives proportionate utilization data with administrative claims from the cohort studied. There were statistically significant reductions in the proportions of enrollees with any inpatient hospitalization (17.6% vs 13.8%) and with any ED visit (73.9% vs 42.9%) (

 <.001 for both). Conversely, there were increases in the numbers of enrollees with any primary care visit (23.3% vs 33.0%) and with any specialty care visit (72.4% vs 79.3%) (

 summarizes per-member-per-month utilization data. Enrollees in the program had significant increases in permember-per-month primary care, but per-member-per-month specialty care visits and rates of inpatient hospitalizations remained similar between the 2 periods. Although there was a trend for decreases in ED visits, this reduction was not statistically significant.

gives cross-sectional data by year for the entire program, including members of the before-and-after cohort and patients who were enrolled as the program grew during the 3-year study. Therefore, these samples were substantially larger than the continuously enrolled cohort. The data show that the first 2 years of the program represented a steady state with respect to inpatient hospitalizations. In the third year of the initiative, there was an 8.3% decline in inpatient admissions. There was a steady decrease in the rate of ED visits per 1000 enrollees over the 3 years studied.

The costs for the program over the first 3 years escalated as patients were transitioned from the AHC to the care of community PCPs (data not shown). In the first full year of operations, payments were $320,128 to community PCPs. Thereafter, payments rose to $551,878 in the second year and to $1,323,808 in the third year. In 2007, payments and management fees to PCPs in the community amounted to $2,485,238.The monthly costs of primary care were approximately $5 per member per month in 2003 and grew to approximately $7 per member per month in 2007. These costs exclude the administrative overhead of processing claims and care coordination for the program.

With a disproportionately large responsibility for delivering care to the uninsured, as well as the potential for avoidable resource-intensive utilization, some AHCs may seek to expand primary care capacity. Although many AHCs made substantial investments in expanding primary care capacity in the 1990s, the emerging dominance of inpatient capacity concerns and reversal of consumer popularity for gatekeeping models of care led ultimately to a reduced emphasis on primary care.6 With this diminished primary care capability, it has been difficult to implement effective coordinated care for patient populations such as the inner-city uninsured. Therefore, the AHC described herein contracted with community-based practices to supplement its primary care needs for managing this population. The program used the administrative infrastructure from a Medicaid HMO owned by the AHC to utilize managed care principles for the care of an uninsured inner-city population.

In a series of 5 case studies conducted by the National Association of Public Hospitals and Health Systems8 in 2005, including the program described herein, there were several notable common attributes that were shared among the programs that sought to use managed care principles for uninsured populations. First, all of the programs were careful to avoid insurance product structures to circumvent unnecessary state insurance regulation. Therefore, none of the programs, including the VCC, had a defined benefit package. Second, all of the programs sought to more effectively manage their uninsured populations by establishing a medical home and by controlling avoidable costs (eg, unnecessary ED visits, specialty care, and hospitalizations). Third, each of the programs used inpatient care management and coordinated discharge planning to ensure a seamless post—acute care process further aimed at averting unnecessary resource utilization. There were other programmatic attributes among the 5 case studies such as copayment strategies, but these were not consistently applied.

The findings from this study indicate that expanding the primary care capacity of an AHC for the care of an uninsured inner-city population resulted in a decrease in the proportion of enrollees who used the ED and in a reduction in hospitalizations. The results demonstrate no significant reductions in utilization rates in hospitalizations or ED visits for a continuously enrolled before-and-after cohort. Nonetheless, evaluation of patterns of care for all enrollees by study year revealed declining rates of use per 1000 enrollees for ED visits and for inpatient hospitalizations. Because the turnover for the population is approximately 50% each year, it is unclear whether this decrease in utilization is more reflective of use patterns for those who remained in the program or for those who disenrolled. Further research will be needed to evaluate the dynamics of how enrollment influences utilization and outcomes.

Our program included care management through nurse coordinators who were based in the ED and identified patients eligible for the program and provided services. Other models may use community-based case managers who could be more effective at linking indigent patients to needed services versus in an ED setting. This approach would have the benefits of optimizing preventive strategies and of intervening much earlier for chronic illnesses. Although our model was efficient because of the large population base served by the AHC (ie, >1 million), targeted community-based efforts (eg, health fairs and other safety net providers) may also be efficient. Nonetheless, while community-based care coordination models may have appeal, limited resources may compel AHCs to begin with models that target ED populations to provide immediate returns on their investments.

Because of the substantive burden of the uninsured that is concentrated at AHCs, some centers are restricting access. For instance, formal AHC policies limiting care to uninsured patients were reported by up to 13% of AHC faculty in a 2003 survey.

 Other AHCs have used community-based practices to support the care of underserved populations, with varying degrees of success. At the University of New Mexico, reduced hospitalizations were estimated to have saved the AHC substantial resources.

 The primary care capacity of the community practice sites from this program enabled the AHC to accommodate a large portion of its uninsured population from the inner-city region.

Virtually all of the community-based practices that participated in the program described in this study were also part of a Medicaid HMO provider network. Medicaid HMOs may be uniquely suited for expanding administrative services to uninsured populations, as beneficiaries often alternate between Medicaid eligibility and uninsured status. Unlike commercial HMOs, Medicaid health plans have continued to use many traditional methods of managed care, including primary care gatekeeping and utilization review.

 A trend that has developed is the emergence of Medicaid-focused health plans, and more than one-half of these are provider sponsored. Some AHCs have responded to this trend in Medicaid managed care by purchasing, developing, or coventuring in HMOs. The AHC described herein was able to take advantage of its provider-sponsored HMO to establish needed administrative functions for managing the primary care needs of a disadvantaged vulnerable population. Therefore, the existing infrastructure from a provider-sponsored HMO enabled the AHC to support functions such as member services, claims processing, and provider relations without significantly increasing administrative overhead for the program. In addition, the HMO’s existing infrastructure enabled the program to access information systems to support needed services for enrollment

and data analyses. Finally, the HMO staff provided expert advice in areas such as policy development, case management services, medical management, and performance improvement. The decision to develop this program based on managed care principles to accommodate the primary care needs of the uninsured in a community setting had many benefits for the AHC.

There were several lessons learned from the initiative. The positive response of community PCPs to the program was widespread; they exhibited a shared interest in the coordination of care for the population. Because the program contracted with physicians who were underrepresented in the inner city, an opportunity was provided to bridge town-and-gown relationships. However, community PCPs expressed concerns that payments would be inadequate, and they raised questions regarding the fate of patients in the event that contracts ended. They also had a strong desire to maintain existing referral relationships with specialty colleagues, and this led to some conflict in management decisions, especially when access to AHC specialists was constrained by capacity. For the AHC, faculty leaders were worried about funding support that would be shifted to community providers. However, few faculty expressed apprehension over quality differences, and the referral relationships between community PCPs and medical center faculty were strengthened by the program. Most faculty acknowledge that the VCC has decompressed the medical campus and has enabled a primary care network that is more convenient for disadvantaged patients, many of whom had longstanding difficulties with obtaining transportation and with navigating the complexities of an urban medical center.

This observational study has several limitations. First, because the evaluation was conducted in a single AHC, this case study of enrollment and utilization characteristics of an uninsured population in a managed care model may not be generalizable to other settings. Similar studies need to be performed at other centers to confirm these results. However, our study involved a large sample size at an AHC located in an inner-city teaching hospital, and many AHCs are located in urban environments with high rates of uninsured individuals. Second, secular changes in utilization and patterns of care seeking could have influenced the results. Third, the study represents data collected over a 3-year period from January 1, 2001, to December 31, 2003, and recent economic flux may affect the generalizability of the data.

Nevertheless, the program remains strong. In 2008, there were more than 20,000 enrollees. However, because the AHC receives disproportionate share of hospital (ie, DSH) payments for indigent care, the costs of furnishing primary care in the community are funded through its operating margin, not through federal or state funding. The AHC realizes that the cost savings through reduced ED visits and hospitalizations may take several years to be realized, when more effective care management of chronic illnesses may lead to improved outcomes. Nevertheless, this model of contracting with community PCPs has had the added advantage of providing a new flexible capacity for primary care among the uninsured population. While further studies are planned to document the continuance of observed trends in reduced ED visits and inpatient admissions, the sustainability of the community PCP network may be the most lasting contribution for this AHC.

This study reported on an AHC that contracted with PCPs in community settings by using the administrative and networking capacity of a Medicaid HMO. The study sought to evaluate the benefits of a program that coordinated primary care access in the community and attempted to avert resource-intensive utilization using managed care principles. The results indicate that enrollees in the program had reduced rates of ED visits and lower rates of hospitalization. However, the declining rates for hospitalizations only came after 2 years in the program and in fact rose in the second year (Table 4). Further research is needed to determine if these findings are sustainable and are generalizable to other settings.

This article describes a program to coordinate the care of an inner-city uninsured population at an academic health center.

The Transfer of Uninsured Patients From Academic to Community Primary Care Settings

To describe the taxonomy of methods used by pharmaceutical companies to influence physicians’ nonsteroidal anti-inflammatory drug (NSAID) prescribing behaviors and to elicit physicians’ perceptions of and counterbalances to these influences.

In-depth interviews analyzed using the constant comparative method of qualitative data analysis.

 Qualitative interviews were conducted with physicians representing various clinical specialties. Interviews were transcribed and coded inductively using grounded theory. Recruitment was stopped at 25 participants after the attainment of thematic saturation, when no new concepts emerged from ongoing analysis of consecutive interviews.

 Physicians described a variety of influences that shaped their NSAID prescribing behaviors, including detailing and direct contact with pharmaceutical representatives, requests from patients inspired by direct-to-consumer advertisements, and marketing during medical school and residency training. Physicians described practice guidelines, peer-reviewed evidence, and opinions of local physician experts as important counterweights to pharmaceutical company influence. Local physician experts interpreted and provided context for new clinical evidence, practice guidelines, and NSAID-related marketing.

 The social and communicative strategies used by pharmaceutical companies can be adapted to improve physicians’ adoption of guidelines for safer NSAID prescribing. Communicative interactions between local experts and other physicians who prescribe NSAIDs may be the critical target for future interventions to promote safer NSAID prescribing.

	Adherence to practice guidelines for safer nonsteroidal anti-inflammatory drug (NSAID)  prescribing remains low, despite their acceptance as markers of high-quality care.

		Physician participants described a variety of social and communicative strategies, including those used by pharmaceutical companies, that shaped their NSAID prescribing behaviors.

		Participants detailed how relationships with local physician experts provide interpretation and context to new clinical evidence, practice guidelines, and pharmaceutical marketing related to NSAID prescribing behaviors.

		An approach that models these social and communicative strategies and targets communication

		among physicians may significantly enhance adoption of practice guidelines for safer NSAID prescribing.

	Nonsteroidal anti-inflammatory drugs (NSAIDs) are among the most commonly prescribed medications,

 despite their well-documented adverse effects, including gastrointestinal bleed and renal dysfunction.

 Recommendations for improving benefits while reducing adverse events have focused on strategies to improve physicians’ NSAID prescribing behaviors.

 Safer prescribing strategies include modifications of type and dosing of NSAIDs and concomitant use of protective agents such as misoprostol and proton pump inhibitors in high-risk patients.

 Despite their acceptance as markers of highquality care,

 adherence to safer NSAID prescribing strategies remains between 27% and 42%.

 Publication of guidelines in high-impact journals and endorsement by medical societies have not translated into higher rates of safer NSAID prescribing behaviors. Ineffective dissemination and implementation strategies at the individual physician level may explain the poor adoption of guidelines for safer NSAID prescribing given the contrasting success of pharmaceutical marketing.

	NSAID prescribing behaviors changed dramatically from 1999 to 2004, with a significant rise in total prescriptions and a proportionate increase in cyclooxygenase-2 (COX-2)—selective NSAIDs compared with nonselective types.

 The significant change has been attributed to pharmaceutical companies’ marketing efforts after approval of novel COX-2—selective NSAIDs in 1999.

 Healthcare organizations and investigators might benefit from a greater understanding of how pharmaceutical companies influence physicians’ prescribing behaviors to harness these strategies in interventions designed to improve the quality and safety of NSAID prescribing behaviors. Using in-depth qualitative interviews of physicians who routinely prescribe NSAIDs, this study explores the methods that pharmaceutical companies use to influence physicians’ NSAID prescribing behaviors and physicians’ perceptions of and counterweights to these influences.

	The Institutional Review Board at Baylor College of Medicine approved this study. Qualitative methods

 were used to explore physicians’ perceptions of the ways pharmaceutical companies influence their NSAID prescribing behavior. Study subjects participated in semistructured indepth interviews from February to April 2006 to allow us to describe the methods of pharmaceutical company influence on NSAID prescribing behaviors.

	Study personnel identified individuals who met the eligibility criteria of being a licensed (nontrainee) physician who routinely prescribed NSAID therapy. Participants were purposefully recruited from a variety of practice settings, including a veterans hospital, publicly funded health system, private hospital and academic ambulatory clinic, and a diversity of medical and surgical specialties. Purposive sampling identifies cases that will generate new hypotheses, reveal novel modifying or moderating variables, or describe the complexity of multiple interrelations simultaneously among variables.

 Participant recruitment was stopped at the point of thematic saturation,

 when no new concepts emerged from ongoing analysis of at least 5 interviews from participants of different specialties.

	Interviews were conducted by one of us (ALW) and consisted of open-ended questions to elicit perceptions about the role of pharmaceutical representatives and marketing on NSAID prescribing practices. Interviews began with the following statement and question: “We know that NSAIDs are used very commonly for pain relief and for control of acute and chronic inflammation. When prescribing NSAIDs to your patients, how do you decide which NSAID to prescribe?” Responses to this question were explored using probes focused on the influence of pharmaceutical companies, representatives, drug samples, and marketing, including the variety of marketing techniques and participants’ perceptions of those techniques.

	Interviews lasted approximately 1 hour and were audiotaped and transcribed for analysis. Data were analyzed using the constant comparative method.

 Three of us (ADN, ALW, NSA) independently reviewed each transcript line by line to identify and sort segments of data with similar concepts into distinct categories. Sorted categories evolved inductively into a coding system based on the grounded theory of qualitative analysis.

 Investigators applied the coding system to each transcript independently. After additional rounds of independent coding, investigators convened as a group for careful review, negotiation, and consensus building to resolve discrepancies in coding. With coding of successive transcripts, the coding system was expanded, refined, and applied to previously coded data.

 Resulting categories comprise the final coding scheme for elaborating study findings and themes.

	Twenty-five participants were recruited. All were affiliated with an academic medical center, and they spent most of their time in direct clinical care. Characteristics of participants are given in the 

. One-half were women, and one-third were of nonwhite race/ethnicity. Specialties commonly associated with a high volume of NSAID prescriptions were well represented, including internal medicine, rheumatology, general surgery, physical medicine and rehabilitation, and geriatrics. On average, participants reported 16 years of postresidency clinical experience. Encounters with pharmaceutical company representatives varied from daily (12%) to no reported interactions (12%).

Influence of Pharmaceutical Company Practices on NSAID Prescribing Behaviors

	Participants described several themes regarding methods used by pharmaceutical companies to influence their NSAID prescribing behaviors. These can be categorized into the following 3 broad themes: detailing and direct marketing, patients’ requests for medications, and shaping prescribing habits during formative training. Participants related that these influences have been present over most of their careers but were heightened during the period when COX-2—selective NSAIDs were introduced and marketed.

. Participants described a variety of methods used by pharmaceutical companies to directly influence their NSAID prescribing behaviors. Visits to physicians’ offices by pharmaceutical representatives to provide details of their products (ie, detailing) were frequently described:

f you can get the drug reps, of course, to promote it if it’s their drug, because those people are actually out in the physicians’ offices and they quite often can have 5, 10, 15 minutes of, you know, one-on-one time to actually show

	Another participant put it more bluntly:

he drug reps for the COX-2s…They hammered us with all the statistics….

	Participants were savvy about the influence of detailing visits. One described the approaches thusly:

I guess we all get somewhat influenced by the drug representatives. You know…I appreciate the ones that bring me in some literature. I don’t like the ones that come in and just say, well, you know, use my drug, it’s better, but they have nothing to back it up, but if I have some literature that backs up the efficacy or safety or some overall benefit of the drug, then I think I’m more likely to believe that.

	Participants also described how provision of medication samples by pharmaceutical representatives was frequently associated with direct marketing efforts. They had mixed feelings about the availability of medication samples, stating that samples assist patients who have difficulty paying for medications:

I do reserve my drug samples for people that I believe cannot afford the more expensive medicines, and so I do give out, sometimes I give a patient a full course of medications from my sample cabinet. The patient may try to fill the prescription and then find out the cost and then will call me back and say, well, I’m just going to take Advil…it opens a bunch of risk factors.

	Participants also suggested that providing samples could influence their NSAID prescribing behaviors:

 usually, instead of following guidelines, they usually end up writing a prescription they were provided a sample 

 and once they give the patient a sample and the medication works, then they tend to continue prescribing the same medication. So whatever is available from some pharmaceutical company with free drugs, that’s what they would use.

Theme 2: Patients’ Requests for Medications.

 Participants frequently cited the effect of patients’ requests for specific brands of NSAIDs. Invariably, these were the products of direct-to-consumer advertising that influenced the patient or a close friend or relative:

I’ve had in the past people bring it up, because it was so heavily advertised.

	Oh, absolutely. Oh, yeah, yeah. I think that direct consumer marketing is huge with patients.

	Several described how patient requests influenced their NSAID prescribing behaviors:

he direct consumer advertising is the bane of every working physician in this country…Ask your doctor if you need…Of course…Absolutely. If you’re asking, you bet you do. Right. And, you know, it’s how much resistance can I put up?

Theme 3: Shaping Prescribing Habits During Formative Training.

 Participants described how NSAID prescribing behaviors were established during their formative training as physicians, especially in residency:

I think people’s patterns in prescribing get set pretty early when they’re in residency, and they tend to fall back on those patterns quite a bit….

Sure. When you do it as a resident, you’re inclined to keep doing it….

	Many described that pharmaceutical companies influenced how their medications were used during formative training and how this shaped inchoate prescribing behaviors:

Pfizer pretty much ruled the formulary. So I think they did a very good job of getting their point across and educating on their medications, and I can say easily in my first several years out in practice, those medicines that I knew very well from residency, and really understood their side-effect profile and everything, that I continued to use for several years after that….

	Participants, from various specialties and years in practice, described the influence of pharmaceutical companies during formative training. Participants related how their residency programs established policies barring such contacts with residents during work hours.

hat hospital had a total anti—drug rep mindset. I remember one time I was precepting residents in the clinic, and I said, “Hey, where’s the drug reps?” And a resident looked at me like I had 3 heads and said, “Why would we want them around here? They just influence our practice the wrong way.”

Counterbalances to Pharmaceutical Company Influence on NSAID Prescribing Behaviors

	Participants were well aware of the role of pharmaceutical marketing, direct advertising, and representatives in shaping NSAID prescribing behaviors. Most participants described how they attenuated this influence using alternative sources of information, including practice guidelines or peer-reviewed evidence and opinions of trusted local experts.

Theme 4: Alternative Information Sources for Shaping NSAID Prescribing Behaviors.

 Most participants described routine use of journals, electronic peer-reviewed literature, and professional meetings as their primary sources of new evidence and practice guidelines. One physician summarized these sources:

I try to keep up with reading in the journals, like the

…New things that come out, I read up on…And is probably my other primary source for finding quick facts. And then I’ll go into PubMed and do searches if I need to know a little more detail.

	Participants also described guidelines as sources of unbiased information:

It’s better to have a big conglomerate of doctors form some sort of committee without any outside bias or interests…to come up with what they feel is the most appropriate way to treat patients…it bears more weight than just one person who did one study or was funded by one company to do one study, to publish a set of guidelines.

	Another participant described the role of local physician experts as counterbalances to pharmaceutical company influence:

Well, thinking about myself, the biggest disseminator of information around here is really word of mouth from other physicians. You know, we don’t have drug reps coming in to tell us things, and certainly I don’t want to listen to what the drug rep has to say. I don’t want that to be my source of information….

Theme 5: Local Physician Experts Interpret and Provide Context for New Clinical Evidence, Practice Guidelines, and Pharmaceutical Marketing.

 Despite their affirmation of evidence-based literature and familiarity with practice guidelines, participants reported having trouble integrating practice guidelines into their NSAID prescribing behaviors. One physician described his difficulty with guidelines as follows:

The guidelines that you commonly use, I think you continue to practice by, but other than that, I mean they all tend to blend together for me, because they tend to all be so similar.

	Another physician described her apprehension regarding a practice guideline after she had a previous adverse experience:

We’ve all had circumstances where we give a patient a particular drug and they have such a serious adverse effect…we become skittish about ever giving another patient that

even though you know intellectually that maybe 0.1% of all the people who ever took this drug…I think that had a large effect on our practice patterns and probably [is] one of the reasons why we don’t follow guidelines.

	Participants described how local experts and trusted physicians serve as interpreters of evidence and guidelines into a practice context that influenced their prescribing behaviors:

Do I use guidelines?…Really, really not. I would say the things that have most shaped my recent NSAID prescribing practices (knowing that I don’t have contact with drug reps anymore)…is conversations with colleagues.

I think that my own impression is and I think it’s based on our rheumatology database…the guy that runs it is a doctor whom I’ve heard speak before…he has a lot of data on the traditional NSAIDs, and I think that those pan out to be a little bit safer, so I try to use that if I can.

	Physicians described several social and communicative strategies used by pharmaceutical companies to influence their NSAID prescribing behaviors, including detailing and direct contact with pharmaceutical representatives, requests from patients inspired by direct-to-consumer advertisements, and marketing during formative medical school and residency training. Practice guidelines and peer-reviewed evidence, as well as local physician experts, were viewed as important counterweights to the influence of pharmaceutical companies on prescribing behaviors. All study participants described routine experiences with pharmaceutical marketing and their use of counterweights to this influence. These multifaceted influences that affect physicians’ NSAID prescribing behaviors are shown in the 

	The marketing strategies of pharmaceutical companies are shown as ovals in the Figure. Detailing and direct promotion to physicians were frequently cited as direct influences on prescribing behaviors and as indirect influences mediated through the experiences of formative training, as well as by suspicion that guidelines were tainted by pharmaceutical promotion. Critics of industry influence often decry the more egregious use of free meals, honoraria, and medication samples, but the influence of direct communication with physicians using in-office presentations and other detailing of products may be the most effective.21 Pharmaceutical detailing is characterized by persuasive use of epidemiologic and statistical data that is often biased with only the presentation of favorable results.

 reporting the characteristics and influence of detailing by the manufacturers of gabapentin found that most visits to physicians lasted 5 minutes or less and had high informational value. The most effective visits were accompanied by delivery or a promise of samples and occurred in small groups. Detailing visits were surprisingly effective at influencing prescribing behaviors, even for offlabel use of medications,

 as most participants reported some change in their prescribing behaviors after marketing.

	Consumer advertising that instructs patients to “tell their physician” about particular symptoms and specific medications was described as having an indirect influence on physicians’ NSAID prescribing behaviors (Figure). Directto-consumer advertising has been strongly associated with changes in physician prescribing behavior when advertising is closely associated with requests by patients for specific drugs.

 For example, advertising of rofecoxib and celecoxib to consumers increased the number of patients seen by physicians each month and the likelihood that patients received both NSAIDs,

 especially when patients specifically requested a COX-2—selective NSAID.

 Participants also described how the influence of pharmaceutical companies on prescribing behavior begins early in a physician’s education and training (formative training box in the Figure). Influences during formative training are effective in anchoring subsequent behavior within a fixed set of prescribing patterns that can become resistant to change. This phenomenon, often described as clinical inertia,

This study describes the social and communicative strategies pharmaceutical companies use to influence NSAID prescribing behaviors and elicits physicians' perceptions and counterbalances to these strategies.

April 2009

Moving From A to Z: Successful Implementation of a Statin Switch Program by a Large Physician Group

The Transfer of Uninsured Patients From Academic to Community Primary Care Settings

Pharmaceutical Company Influence on Nonsteroidal Anti-Inflammatory Drug Prescribing Behaviors

Primary Care Strength Linked to Prevention Programs for Cardiovascular Disease

Does Opioid Therapy Affect Quality of Care for Diabetes Mellitus?

Cost-Effectiveness of Salmeterol, Fluticasone, and Combination Therapy for COPD

Assessing Outcomes in Child Psychiatry