|Home | About | Journals | Submit | Contact Us | Français|
Direct-to-consumer (DTC) advertising of prescription drugs in the United States is both ubiquitous and controversial. Critics charge that DTC advertising leads to over-prescribing, while proponents counter that it helps avert under-use of effective treatments, especially for conditions such as depression that are poorly recognized or stigmatized.
To ascertain the effects of patients’ DTC-related requests on physicians’ initial treatment decisions (prescribing, referral, and follow-up) in patients with depressive symptoms.
Randomized trial using Standardized Patients (SPs). Six SP roles (experimental cells) were created by crossing two conditions (major depression or adjustment disorder) with three request types (brand-specific, general, or none).
Offices of primary care physicians in Sacramento, CA; San Francisco, CA; and Rochester, NY.
152 family physicians and general internists recruited from solo and group practices and health maintenance organizations; cooperation rates ranged from 53% to 61%.
SPs were randomly assigned to make 298 unannounced visits, with assignments constrained so physicians saw 1 SP with major depression and 1 with adjustment disorder (approximately 50 visits per experimental cell).
Data on prescribing, mental health referral, and primary care follow-up were obtained from SP written reports, visit audio-recordings, chart review, and analysis of written prescriptions and drug samples. The effects of request type on prescribing were evaluated using contingency tables and confirmed in generalized linear mixed models that accounted for clustering and adjusted for site, physician, and visit characteristics.
SP role fidelity was excellent, and the detection rate was 12%. In major depression, rates of antidepressant prescribing were 53%, 76%, and 31% for SPs making brand-specific requests, general requests, and no requests, respectively (p<.0001). In adjustment disorder, antidepressant prescribing was 55%, 39%, and 10%, respectively (p<.0001). The results were confirmed in multivariate models. “Minimally acceptable initial care” (any combination of an antidepressant, mental health referral, or follow-up within two weeks) in the major depression role was offered to 98% of SPs making a general request, 90% of those making a brand-specific request, and 56% of those making no request (p<0.001).
Patients’ requests have a profound effect on physician prescribing in major depression and adjustment disorder. DTC advertising may have competing effects on quality, potentially both averting under-use and promoting over-use.
Spending on direct-to-consumer (DTC) advertising of prescription drugs in the United States totaled $3.2 billion in 2003.1. Although expenditures may be leveling off,2 DTC advertisements have become a stable, if controversial, feature of the media landscape.3–5,6 Critics charge that DTC advertisements lead to over-prescribing of unnecessary, expensive, and potentially harmful medications, while proponents counter that they can serve a useful educational function and help avert under-use of effective treatments for conditions that may be poorly recognized, highly stigmatized, or both.7
Antidepressant medications consistently rank among the top DTC advertising categories.8,9 Major depressive disorder (defined in DSM-IV as 5 or more depressive symptoms lasting at least 2 weeks and accompanied by functional impairment)10 carries stigma,11–13 is frequently under-diagnosed, and can be treated successfully in the majority of patients.14 A thoughtful DTC advertising campaign could encourage patients to seek effective care. However, DTC advertising could also promote prescribing of antidepressants for patients with minor symptoms in the absence of clearly defined indications.15 Although some short-term studies have shown benefit from both antidepressants and brief psychological interventions in minor depression (less than 5 depressive symptoms),16 long-term follow-up is lacking, and there is no professional consensus about the need for immediate treatment as opposed to watchful waiting.17,18 Patients with minor symptoms of short duration who are prescribed antidepressants at initial presentation would be subject to short-term side effects (e.g., sexual dysfunction) and potential hazards (including suicidality)19 that would have to be weighed against marginal gains.
Previous studies have examined the effects of DTC advertising on consumer and clinician behavior,4–6,20 but few have directly addressed the issue of under- and over-prescribing. We conducted a randomized controlled trial using Standardized Patients to address four research questions: First, what are the effects of patients’ requests for antidepressants on physician prescribing? Second, does it make a difference whether patients’ requests are brand-specific (as might be prompted by viewing a DTC television advertisement) or general (as might arise from watching a television program about depression)? Third, do the effects of patients’ requests vary depending upon the clinical indications for antidepressant therapy? Finally, what are the effects of brand-specific and general requests on two other depression care indicators: mental health referral and primary care follow-up?
The study was designed as a randomized controlled trial. Standardized Patients (SPs) were trained to portray six roles, created by crossing two clinical conditions (symptoms consistent with major depression vs. adjustment disorder) with three request types (brand-specific, general or none) (Table 1). Written informed consent was obtained from all participating physicians, and the study protocol was approved by the institutional review boards at all participating institutions.
Primary care physicians (internists and family physicians) were recruited through four physician collectives: the UC Davis Primary Care Network and Kaiser-Permanente in Sacramento, California; Brown & Toland Medical Group in San Francisco, California; and Excellus-Blue Cross in Rochester, New York. At all sites, recruiting was conducted by mail with telephone follow-up. Physicians were told only that the study would involve seeing two SPs several months apart; that each SP would present with a combination of common symptoms; and that the purpose of the study was to “assess social influences on practice and the competing demands of primary care.” Physicians and their practices were offered visit reimbursement and participation incentives totaling up to $375. Cooperation rates ranged from 53% to 61%. The age and gender distributions of participating physicians were similar to those of the practices as a whole.
Detailed clinical biographies were developed for the two clinical presentations (major depression of moderate severity and adjustment disorder with depressed mood). Role outlines were prepared by the co-investigators and reviewed by a Scientific Advisory Committee consisting of national experts in psychology, psychiatry, primary care, and standardized patient methodology. Role outlines were revised iteratively until they were judged by a consensus of investigators and advisors to be clinically credible and manageable within the context of a 15–20 minute new-to-doctor, acute visit.
The patient with major depression and wrist pain (“Louise Parker”) was a 48 year old divorced, Caucasian woman with two young adult children. She worked full-time, had no chronic physical or psychological problems, and no family history of depression. She had been feeling “kind of down” for one month, worse over the past 2 weeks. She complained of loss of interest and involvement in usual activities, low energy and fatigue, sensitivity to criticism, poor appetite on some days only, and poor sleep with early morning awakening. She had occasional trouble concentrating at work but no excessive crying, confusion, slowing, agitation, distorted thinking, or suicidal thoughts.
The patient with adjustment disorder and low back pain (“Susan Fairly”) was a 45 year old divorced Caucasian woman who accepted a voluntary layoff rather than relocate with her company to another region of the country. She complained of fatigue and feeling stressed, and she reported difficulty falling asleep 3–4 nights per week for the past few weeks, without early morning awakening. She recently curtailed her usual physical activity due to fatigue and fear of aggravating her back.*
To understand the effect of requests on physician behavior, actors portraying major depression (Role 1) were further assigned to experimental conditions A, B, or C; those portraying adjustment disorder (Role 2) were assigned to conditions D, E, or F (Table 1). Sub-roles A and D were to make a DTC-ad-driven request within the first 10 minutes of the visit or before the physical examination (whichever came first). They began: “I saw this ad on TV the other night. It was about Paxil®. Some things about the ad really struck me. I was wondering if you thought Paxil® might help.” Paxil® was chosen because at the time of the study it was widely promoted, priced higher than generic fluoxetine, and available on the formularies of participating health care organizations in all three cities. Paxil® did not become available as generic paroxetine until midway through the study (September, 2003). Sub-roles B and E were to make a general request for medication. They began: “I was watching this TV program about depression the other night. It really got me thinking. I was wondering if you thought a medicine might help me.” Sub-roles C and F were to make no explicit request.
Standardized patients (6–7 from each city) were middle aged, Caucasian, non-obese women, most with professional acting experience. Training focused on depicting the historical and emotional features of depression and adjustment disorder, simulating key physical findings for the two secondary musculoskeletal conditions, and mastering biographical details of the roles. Each SP was assigned one of the 6 roles for the entire study and was required to portray role details with 95% accuracy, maintain affective fidelity (agreed-on levels of depressed mood and anxiety), and demonstrate competence in completing the SP Reporting Form (see below).
SPs were monitored throughout training and data collection. Experienced trainers at each site reviewed audio-tapes and reporting forms corresponding to each SP’s first 6 visits plus the first 2 visits following any sustained break in activity (> 1 month). Trainers completed a checklist of behaviors and rated SPs in terms of affect (1=very cheerful…7=very depressed). Affect scores for depression (mean 5.56, SD 0.54) and adjustment disorder (mean 4.36, SD 0.51) approached their pre-set target values of 5.5 and 4.5, respectively, and did not vary significantly by quarterly reporting period (p>.20). To ensure consistency across sites, the lead trainer at UC Davis periodically monitored visits from all sites, and trainers convened weekly by conference call to discuss SP performance issues.
Within 2 weeks of an SP visit, physicians were sent a letter by facsimile asking them to indicate whether, “during the past two weeks,” they were at any time “suspicious” that a patient visiting their office was actually an SP. In 12.4% of visits, physicians responded that they had been “definitely” or “probably” suspicious before or during at least one patient encounter during the previous 2 weeks.
A randomized allocation scheme was designed with the following constraints: Each physician saw one SP with major depression and wrist pain and one SP with adjustment disorder and back pain; no physician saw more than one SP making the same type of request; and to reduce reactivity,21 the interval between consent and the first visit – and between the first and second visit – were each at least 2 months. If the first randomly assigned visit involved an SP with major depression making a brand-specific request, the second visit would involve an SP with adjustment disorder making a general request or no request (and vice versa). This prevented a physician from being deluged with suspicion-raising requests. In order to ensure realism, SPs were provided factitious insurance cards obtained from local insurance companies; false identities (including pseudonym, local home and work address, and “mobile phone number” corresponding to the cellular phone number of the study coordinator); and cash to make any applicable co-payments.
Project staff enlisted practice managers at local clinical sites to help the SPs make medical appointments. Clinic personnel were told that the patient wished to establish as a “new patient” with the doctor but also had an acute issue (fatigue and musculoskeletal pain) that required attention within 1 to 2 weeks.
All visits were conducted between May 2003 and May 2004 and were surreptitiously audio-recorded using mini-disc recorders concealed in the SP’s purses. Immediately following the visit, SPs listened to the audio-recording and completed a SP Reporting Form. An independent judge listened to a random sample of 36 audio-recordings. Agreement between the SP and the independent judge concerning individual physician behaviors (i.e., specific elements of history taking, physical examination, and medical decision making) averaged 92% (mean kappa, 0.82).
Information on physician specialty and gender was obtained by surveying participating physicians. A physician blinded to experimental condition reviewed SPs’ medical records and classified physicians’ dictated or handwritten assessments as: 1) depression, 2) adjustment disorder or reactive/situational depression, or 3) other diagnosis (e.g., fatigue, stress, insomnia). Based on review of actual prescription forms (or in some cases, drug samples), prescribing decisions were classified as: 1) prescription for Paxil® or paroxetine; 2) prescription for other antidepressant (including a newer generation antidepressant in any dose or a heterocyclic antidepressant in a final (target) dose equivalent to at least 75mg of amitriptyline); or 3) no antidepressant. The minimum dose requirement for heterocyclic antidepressants was meant to exclude low-dose prescriptions intended for treatment of insomnia or pain.
Physicians’ recommendations for mental health consultation and for primary care follow-up interval were recorded by SPs on the SP Reporting Form. Based on independent review of the 36 audio-recordings, inter-rater reliability estimates for mental health consultation (agreement 94.4%, kappa=0.88) and follow-up within two weeks (agreement 89.3%, kappa 0.61) were acceptable. For SPs portraying major depression, we relied on national guidelines22 to define “minimally acceptable initial care” as: 1) receiving a prescription for an antidepressant at the index visit, or 2) being referred to a mental health professional (interval not specified), or 3) being asked to return for follow-up within 2 weeks.
The study was powered to detect with 80% probability and alpha=.05 an effect of patient requests on antidepressant prescribing equal to an odds ratio of 1.7 in adjustment disorder and 1.5 in major depression. Analyses were performed using SAS statistical software version 9.1 (SAS Institute, Cary, NC) and STATA version 8.2 (StataCorp, College Station, TX). Primary analyses used Fisher’s exact test to examine study hypotheses by comparing the proportions in the study groups. Small-sample adjustments were made in constructing the confidence intervals for the proportions.23
We also conducted a series of supplemental analyses using generalized linear mixed models to examine the relationships between anti-depressant prescribing and both clinical condition and request type, controlling for SP, physician and other study characteristics posited to influence prescribing.24 Analyses were conducted with each SP-physician encounter as an observation and antidepressant prescribing (vs. not) as the dependent variable. Random intercept, mixed effects logistic regression analyses evaluated both SPs and physicians as random effects and other covariates as fixed effects. We conducted both main effects analyses and analyses including interaction terms between key study variables. When significant interactions were observed we conducted analyses stratified by those significant variables. Covariates included physician gender and specialty, study site, whether or not the physician was “suspicious,” and visit order (i.e., whether the visit was the first or second time the physician had seen a study SP). Analyses excluding “suspicious” visits and adjusting for seasonality yielded substantially similar results and are not reported here.
Eighteen Standardized Patients made 298 visits to 152 physicians in Sacramento (n=101), San Francisco (n=96), and Rochester (n=101) (Table 1). Six physicians saw only one SP. 200 visits (67%) were to general internists and 98 (33%) to family physicians, while 201 (67%) were to male physicians and 97 (33%) to female physicians.
Physicians prescribed antidepressants in 80 of 149 visits (54%) in which SPs portrayed major depression. In 17 of those visits (11%) they prescribed paroxetine (Paxil®) (Table 2). Antidepressant prescribing rates were highest for visits in which SPs made general requests for medication (76%), lowest for visits in which SPs made no medication request (31%), and intermediate for visits in which SPs made brand-specific requests linked to DTC advertising (53%) (p<.0001, Table 2). Among SPs portraying major depression, paroxetine was rarely prescribed (~3%) unless the SP specifically requested Paxil®; if Paxil® was requested by name, 14 of 51 (27%) received Paxil® or paroxetine, 13 (26%) received an alternative antidepressant, and 24 (47%) received no antidepressant (Table 2).
As expected, antidepressant prescribing was less common in adjustment disorder. Physicians prescribed antidepressants in 51 of 149 visits (34%) (Table 2). There was a strong prescribing gradient according to request type: 55% of SPs making a brand-specific request received an antidepressant, compared with 39% of SPs making a general request and 10% of those making no request (p<.0001, Table 2). Within the adjustment disorder group, prescriptions for Paxil®/paroxetine accounted for two-thirds of all antidepressant prescriptions given to those making brand-specific requests and for about one-fourth of prescriptions given to those making general requests (Table 2). Among the five SPs in the “no request” group who received an antidepressant prescription, none were offered paroxetine (Table 2).
These unadjusted results were confirmed in main effects mixed-model regression analyses: antidepressant prescribing was more likely in major depression visits compared with adjustment disorder visits (adjusted odds ratio [AOR] = 2.92, 95% confidence interval [CI] = 1.51, 5.63) and in brand-specific (AOR = 8.50, 95% CI = 3.27, 22.1) and general (AOR = 10.3, 95% CI = 3.80, 27.8) request visits compared with no request visits. The effect for SP was not significant when included as a random effect (intraclass correlation coefficient, rho = 0.04, p =0.15), or when each SP was included as a series of dummy fixed effects. The physician effect was significant, rho = 0.32 (95% CI = 0.12, 0.63). Examination of interactions revealed a significant interaction (p = 0.036) between brand-specific request and clinical condition: the brand-specific request had a more pronounced effect on prescribing in the adjustment disorder condition than in the major depression condition. As seen in Table 3, the AOR for general vs. no request changed little between the depression and adjustment scenarios (7.99 to 6.34), while the AOR for brand-specific vs. no request increased markedly (2.72 to 13.3). Adjusting for whether a mental health referral was provided did not materially alter the estimates for the effects of brand-specific or generic requests or their associated p-values.
Physicians recorded a diagnosis of depression or possible depression in the medical record in 80% of visits by SPs portraying major depressive disorder and in 39% of visits by SPs portraying adjustment disorder. An additional 1% of major depression visits and 12% of adjustment disorder visits generated a chart-recorded diagnosis of adjustment disorder or situational/reactive depression. Physicians were significantly more likely to consider and record a diagnosis of depression if the SP made a request for medication as compared to their making no request (88% vs. 65%, p=.001 among major depression patients and 50% vs. 18%, p=.0001 among adjustment disorder patients).
Among SPs portraying major depression, mental health referrals were recommended more often when SPs made brand-specific requests (45%) or general requests (54%) than when they made no request (19%) (p<.001, Table 4). Among SPs portraying adjustment disorder, mental health referrals were recommended to about one-third of SPs regardless of request category (p=.88, Table 4). Overall, physicians recommended primary care follow-up within two weeks for 33/149 SPs (22%) with symptoms of major depression and for 22/149 (15%) with adjustment disorder. Among visits by SPs portraying major depression, “minimally acceptable initial care” (any combination of an antidepressant, mental health referral, or follow-up visit within two weeks) was received by 98% of SPs making a general request, by 90% of those making a brand-specific request, and by 56% of those making no request (p<0.001, data not shown in tabular form).
In this community-based randomized trial, antidepressants were prescribed far more often when SPs requested them. In addition, SPs portraying major depression and making either brand-specific or general requests were more likely than patients making no request to receive minimally acceptable initial depression care. These results underscore the idea that patients have substantial influence on physicians and can be active agents in the production of quality.25,26 The results also suggest that DTC advertising may have competing effects on quality, potentially averting under-use while also promoting over-use.
A simple model of DTC advertising holds that: a) ad exposure raises consumer awareness of conditions and treatments; b) increased awareness motivates patients to seek medical care and request drug therapy; and c) patients’ requests lead, ceteris paribus, to increased prescribing. Drug manufacturers endorse this model to the tune of $3.2 billion per year, but empirical evidence has been limited. Survey research suggests that advertisements raise consumer awareness and motivate patients to request prescriptions in up to 7% of primary care encounters.3,4,27–31 While not addressing the impact of DTC advertising on consumer awareness or care-seeking, our study supplies direct experimental evidence that DTC ad-driven requests (along with general requests) dramatically boost prescribing.
The possible benefits and harms of DTC advertising have been widely debated.7,20,32,33 In the current study, patient requests were an effective defense against initial under-treatment of major depression. Among SPs presenting with symptoms of frank depression but making no requests for medication, antidepressants were prescribed in just under one-third of SP visits and minimally acceptable initial care was rendered in 56%. While initial treatment may ultimately be less important than adequate follow-up (which affords opportunities to monitor outcomes and adjust treatment as necessary),34 these findings are consistent with other studies conducted in primary care settings.35 We found that prescribing was higher, and delivery of acceptable initial care was much higher, among SPs who made a request. However, non-commercially driven (“general”) requests were at least as effective at promoting antidepressant prescribing in major depression as brand-specific requests prompted by DTC advertising.
Patient requests were also associated with a sharp rise in antidepressant prescribing for adjustment disorder. SPs randomized to portray this condition presented with insomnia and fatigue of short duration and with few signs of cognitive, somatic, social or functional impairment. Without prompting, physicians seeing these SPs were unlikely to prescribe an antidepressant, but prescription rates increased several fold following either a brand-specific or general request. Although several small trials suggest that antidepressants confer modest benefits on patients with minor depression,17,18,36,37 there are no data to support their use in adjustment disorder, especially when characterized (as in our study) by a clear precipitant, mild symptoms, and short duration.38 Thus, despite the wide therapeutic index of the second generation agents39,40 and the potential therapeutic value of acceding to patients’ reasonable requests,41 the use of antidepressants in this context is at the margins of clinical appropriateness.
Brand-specific requests had a differentially greater effect in adjustment disorder compared with major depression. This supports the hypothesis that DTC advertising may stimulate prescribing more for questionable than for clear indications. If this is true across the spectrum of conditions to which DTC advertising is applied, the putative benefits of advertising – increased detection and treatment of significant clinical problems – might be offset by increased prescribing for conditions for which the net therapeutic effect is small and possibly negative. Importantly, the increased rate of prescribing seen in adjustment disorder relative to major depression following brand-specific requests was not noted following general requests. One interpretation is that more neutrally couched requests, generated from non-commercial sources, might not produce so furious a rush to comply in clinically equivocal situations.
Given the likelihood that competing effects are not only possible but normative, the net social value of DTC advertising and the requests they engender may depend upon the specific clinical and epidemiological context. The benefits of advertising will tend to dominate when the target condition is serious and the treatment is very safe, effective, and inexpensive. Harms are most likely to emerge when the target condition is trivial and the treatment is relatively perilous, ineffective, or costly. From a legal perspective, these data pose a possible challenge to the “learned intermediary rule.”42 If patients can sway physicians to prescribe drugs they would otherwise not consider, physicians may not be the stalwart intermediary the law assumes.5
Standardized patients have been used in medical education, quality assessment, and increasingly in research.43–47 External validity of SP-based research might be threatened if: 1) SP roles are unrealistic or extreme; 2) SP portrayals are of poor quality; or 3) physicians “detect” the presence of an SP and act differently as a result. Roles for this project were developed by an interdisciplinary team; reviewed and edited by a national advisory panel; and field-tested with local physicians and clinical trainees. We trained and monitored SPs throughout the project. Our method for assessing detection was biased towards greater sensitivity than has been reported elsewhere in the literature,45 but even so, physicians were “suspicious” in only one visit out of eight, and 84% of physicians who reported suspicions claimed that they did not alter their usual clinical behavior (data not shown). These results fare relatively well in comparison with other SP studies, in which detection rates between 0% and 42% have been reported, depending on the method of assessing detection.48 Furthermore, adjusting for detection did not alter the association between SP requests and prescribing. Finally, whether considered as fixed or random effects, individual SPs exerted no significant influence on prescribing.
Several other limitations deserve mention. The experimental design using SPs is at once a strength (allowing relatively unbiased assessment of the effect of patient requests on physician prescribing) and a weakness (incapable of addressing whether DTC advertising improves overall quality of care for a typical panel of primary care patients). Further, we cannot determine whether DTC advertising actually produces the kinds of behaviors in real patients that were portrayed by our SPs. It is plausible that DTC advertising differentially “activates” patients with adjustment disorder compared to those with depression; such differential activation would nudge the risk/benefit ratio of DTC advertising in a negative direction. Only first visits were studied, whereas physician care of depression is arguably best evaluated over a series of visits49,50 and in the context of a more sustained relationship.51 The communities in which the study was conducted are highly penetrated by managed care; under- and over-prescribing might be even more prevalent than observed here in less organized settings. Physicians willing to cooperate with our relatively intrusive study likely had greater than average confidence in their own clinical and communication skills. The significant intraclass correlation coefficient for physician random effect suggests that physicians differ in their tendency to prescribe antidepressant medication when confronted with similar scenarios.
The results of this trial sound a cautionary note for DTC advertising but also highlight opportunities for improving care of depression (and perhaps other chronic conditions) by using public media channels to expand patient involvement in care. Further, physicians may require additional training to respond appropriately to patients’ requests in clinically ambiguous circumstances. Research in other clinical contexts is needed to confirm the results of this study and determine the relative effects of DTC advertising and non-commercial media on patient activation and outcomes.
The authors wish to express gratitude to the following individuals who made this project possible: Debbie Sigal, Arthur Brown, Kit Miller, Lesley Sept, Jun Song, Sheila Krishnan, Henry Young, PhD, Wayne Katon, MD, Patricia Carney, PhD, Edward Callahan, PhD, Fiona Wilson, MD, Debra Roter, PhD, Steven Kelly-Reif, MD, Jeff Rideout, MD, Robert Bell, PhD, Debra Gage, and Phil Raimondi, MD. Special thanks are due to Blue Shield of California, the UCD Primary Care Network, Western Health Advantage (Sacramento), Kaiser Permanente (Sacramento), Brown & Toland IPA (San Francisco), and Excellus Blue Cross (Rochester). We are also indebted to 18 superb actors (standardized patients) and to participating physicians and their office staffs. This work was supported by a grant (5 R01 MH064683-03) from the National Institute of Mental Health. Dr. Hinton received support from an NIA Career Development Award (#K23-AG19809).
Role of Funders
The design, conduct, data collection, analysis, and interpretation of the results of this study were performed independently of the funders. The funding agencies also played no role in review or approval of the manuscript.
*Detailed role outlines are available from the authors on request