Discussing Out-of-Pocket Expenses During Clinical Appointments: An Observational Study of Patient-Psychiatrist Interactions

Patient visits were sampled from the Verilogue Point-of-Practice database of audio recorded and transcribed clinical encounters (www.verilogue.com). Visits took place from May 2010 to January 2014 for medication management in outpatient, private-practice offices across the United States. Details of the sample have been discussed in previous publications (23,24). Briefly, Verilogue recruits physicians randomly from available lists of active, board certified physicians and pays them to record patient visits for the purpose of marketing or health services research. No other specific inclusion or exclusion for the physicians exist; the recruitment rate is 1:1,200. Physicians and patients are unaware of the research questions for which their recorded visits will be used. Patients are approached and asked to provide consent to participate prior to initiation of the visit, with approximately 80% of patients agreeing to participate. In the sample, only one visit per patient is included, although multiple visits per provider may exist. The sampled visits can occur at any time within the patient-physician relationship. All protected health information is removed during the transcription process; no other information is censored. This study was approved by the Duke University Institutional Review Board.

We defined “cost conversation” as any mention of the patient’s out-of-pocket expenses or insurance coverage for a past, present, or potential health care service. The development and validity of this definition has been discussed in detail elsewhere (23). To mitigate individual coder biases and errors, encounters were analyzed independently by at least two team members. The following cost-related measures were evaluated. First, we determined the presence or absence of a cost conversation. If a cost conversation was present, we then determined which interventions were related to the cost conversation, the duration of the cost conversation, and the discussion of cost-saving strategies. The duration of the cost conversation was calculated by multiplying the total number of cost-related words by the average rate of words spoken, as described elsewhere (23). A team member experienced in conversation analysis assessed coder agreement on each of the above measures. In cases of agreement, the corresponding decision was assigned as “final.” When there were discrepancies, the final coding decision was decided by group consensus. Team members with clinical experience (WGH and PAU) supervised coder training and discrepancy resolution to ensure proper interpretation of clinical matters. All coding was applied using NVivo software, version 10.

We compared cost discussion code frequencies by using Pearson’s chi-square tests and Fisher’s exact test, and we calculated 95% confidence intervals by using Clopper and Pearson’s exact method (25). We used nonparametric Kruskal-Wallis and Mann-Whitney U tests to compare duration of cost discussions, because the distribution of these durations was nonnormal. Statistical analyses were conducted using R software (R Core Team, version 3.1; www.R-project.org). All authors had full access to the data and take responsibility for the data’s integrity.

Our final sample included 422 clinic visits. Characteristics of patients and psychiatrists are provided in Table 1. A total of 422 patients with major depressive disorder were included in our study. They were predominantly female (66%), 35 to 54 years of age (44%), and white (83%). Eight percent of patients were uninsured, which is similar to the national average of 9.8% for patients with depression (26). A total of 36 psychiatrists were included in this study; they recorded a median of 13 visits each (interquartile range [IQR]=4–17 visits), were all male, and most commonly had 21 to 30 years of experience in practice (44%). Data on physicians’ race were not available. The clinic visits occurred in 20 states: Arizona, Georgia, Hawaii, Illinois, Indiana, Kansas, Kentucky, Massachusetts, Michigan, Minnesota, Mississippi, North Carolina, New Jersey, New York, Ohio, Pennsylvania, South Carolina, Texas, Virginia, and Wisconsin.

Thirty-eight percent of visits (N=161 of 422) contained a cost conversation (95% confidence interval [CI]=33%–43%). Incidence of cost conversations varied significantly by patients’ insurance status and was highest among uninsured patients (N=21 of 33, 64%), followed by patients enrolled in private insurance plans (for example, Aetna and Blue Cross–Blue Shield) (N=84 of 190, 44%) and then by patients enrolled in public insurance plans (that is, Medicare and Medicaid) (N=50 of 168, 30%) (p<.001 for comparison across groups). Additionally, incidence of cost conversations was higher in visits with white, non-Hispanic patients, compared with visits with nonwhite or Hispanic patients (N=142 of 352, 40% versus N=19 of 70, 27%; p=.05). Finally, cost conversations were more common in visits with middle-aged patients (35 to 54 years old; N=80 of 184, 43%) than with younger patients (19 to 34 years old; N=37 of 96, 39%) or older patients (≥55 years old; N=44 of 142, 31%) (p=.06 for comparison across age groups).

Twenty-two percent of cost conversations (N=36 of 161) addressed costs for more than one clinical topic: 16% (N=26 of 161) addressed two different topics, and 6% (N=10 of 161) addressed three different topics. In sum, 208 different cost topics were discussed in our sample. Cost conversations most commonly addressed medications (N=136 of 208, 65% of all cost conversation topics), and more than half dealt with costs for psychotropic drugs (N=107 of 208, 51%). The most frequently discussed classes of psychotropic medications were serotonin norepinephrine reuptake inhibitors (SNRIs; N=22 of 208, 11% of all cost conversation topics) and antipsychotics (N=22 of 208, 11%). Selective serotonin reuptake inhibitors (SSRIs), sleep aids, and stimulants were discussed less frequently, constituting 7% (N=15 of 208), 7% (N=15 of 208), and 5% (N=10 of 208) of cost conversation topics, respectively. The costs of over 50 different medications were discussed in our sample; the most frequently mentioned were duloxetine (Cymbalta), aripiprazole (Abilify), quetiapine (Seroquel), zolpidem (Ambien), bupropion (Wellbutrin, Aplenzin), and escitalopram (Lexapro). Examples of medications among the different classes as well as a single example representative of the seven conversations that discussed cost of medications in general are available as an online supplement to this article.

Nonpharmacologic therapies and services accounted for 35% of cost conversation topics (N=72 of 208) and most commonly addressed costs for psychotherapy (N=19 of 208, 9%), provider visits for medical management (N=19 of 208, 9%), procedures (N=9 of 208, 4%), and medical equipment (N=5 of 208, 2%). Two cost conversations discussed the cost of neuropsychiatric testing. Of the conversations pertaining to procedures, five involved electroconvulsive therapy (ECT) or transcranial magnetic stimulation (TMS) and one discussed a vagus nerve stimulator.

Nearly equal percentages of cost conversations were initiated by physicians and patients (N=81 of 161, 50%; N=78 of 161, 48%; respectively). Rarely, cost conversations were initiated by patients’ caregivers or companions (N=2 of 161, 1%). The breakdown of initial cost topics and who initiated them is presented in Figure 1. A significantly greater percentage of physician-initiated cost conversations initially addressed costs for psychotropic medications (N=50 of 81, 62%) compared with those cost conversations initiated by patients (N=30 of 78, 38%; p=.009 for comparison of percentages of physician versus patient initial cost topics which addressed psychotropic medications). By contrast, a significantly greater percentage of patient-initiated cost conversations initially addressed costs for provider visits (N=21 of 78, 27%) compared with physician-initiated cost conversations (N=8 of 81, 10%; p=.008 for comparison of percentages of physician versus patient initial cost topics which addressed provider visits).

Overall, 45% (N=72 of 161) of cost conversations involved at least one cost-reducing strategy (CI=37%–53%). Sixteen percent of cost conversations (N=26 of 161) involved discussion of two or more unique cost-reducing strategies. In sum, physicians and patients discussed cost-reducing strategies for 110 of the 208 total cost topics addressed in our sample (53%).

Cost-reducing strategies most commonly dealt with costs for psychotropic medications (N=86 of 110; 78%) (Table 2). Specifically, psychiatrists and patients most often mentioned cost-reducing strategies for antipsychotics (N=22 of 110, 20%) and SNRIs (N=22 of 110, 20%, also). Fifteen percent (N=16 of 110) of cost-reducing strategies pertained to nonpharmacologic topics, such as provider visits (N=7 of 110, 6%) and other topics, including procedures, diagnostic testing, and medical equipment (N=9 of 110, 8%).

The majority of cost-reducing strategies (N=75 of 110, 68%) did not require changes to the care plan. Instead, they involved discussions of lowering cost by changing the source or timing of an intervention (N=25 of 110, 23%), providing free samples for the purpose of lowering the patient’s costs (N=25 of 110, 23%), or facilitating use of copay assistance programs (CAPs) or cost-saving coupons (N=22 of 110, 20%). The remaining 32% of cost-reducing strategies (N=35 of 110) required changes to the care plan, such as switching to a lower-cost therapy or diagnostic (N=24 of 110, 22%) and changing the dose or frequency of an intervention (N=11 of 110, 10%).

Frequently, psychiatrists and patients sought to lower costs for psychotropic medications by providing free samples or facilitating use of CAPs or coupons. These often involved a variety of interrelated scenarios in which the doctor offered a medication option because he or she had samples, offered a trial of the medication before writing a prescription, or offered to start the patient on a sample of medication that also is associated with a coupon program (Table 3). Sometimes psychiatrists discussed coupon programs for nonmedication expenses (for example, a conversation around gender affirmation surgery that utilized a grant program).

Sometimes psychiatrists recommended reducing expenses by switching to a different formulation of the same medication. This was particularly true with bupropion (example in Table 3). Other times, psychiatrists lowered patients’ expenses by prescribing generic medications that were not the equivalent of the brand-name medications that patients had taken previously (Table 3).

Sometimes psychiatrists lowered patients’ expenses by changing the source or timing of medications (for example, discussing an alternative location for obtaining the medication, such as using a different pharmacy to obtain a lower price). Other times, psychiatrists lowered patients’ medication expenses by prescribing a different quantity of medications (for example, a 90-day versus 30-day prescription).

Psychiatrists also reduced patients’ expenses by addressing the source or timing around nonmedication expenses (for example, by having a patient come to another office location that would accept the new insurance plan, or by spacing out the intervention, such as the appointment follow-up time). Psychiatrists also suggested logistical changes to reduce therapy expenses (for example, by recommending that patients seek counselors through churches) (Table 3). Discussions around ECT and TMS often focused on obtaining insurance approval.

The median visit duration in our sample was 12 minutes and 15 seconds (IQR=9 minutes and 13 seconds to 18 minutes and 36 seconds). In visits containing cost conversations, the median time spent discussing patients’ health care costs was 36 seconds (IQR=16–81 seconds). The majority of cost conversations lasted less than one minute (69%), and 8% lasted three minutes or longer (Figure 2). There was no significant difference in the duration of cost conversations initiated by physicians versus patients. However, cost conversations containing discussion of cost-reducing strategies lasted significantly longer than those without such discussion (median=71 seconds and 22 seconds, respectively, p<.001).