Major depressive disorder is an important mental health problem that is associated with significant role impairment, mortality, and substantial costs to society (1). Antidepressants are the most frequently used therapy to treat depression because of their easy administration and relatively high effectiveness. Clinical practice guidelines from the Agency for Healthcare Research and Quality (AHRQ), formerly the Agency for Health Care Policy and Research, and the American Psychiatric Association (APA) recommend frequent follow-up visits during the initiation of antidepressant treatment to provide patient support, adjust dosage, and monitor side effects and clinical response (2,3,4). Visits also should be frequent enough to promote treatment adherence and to reduce communication gaps between the treating physicians and patients about the anticipated duration of treatment. The APA guidelines recommend patients to be seen on a weekly basis during the first 90 days of treatment (2). The AHRQ guidelines recommend patients to be seen every ten to 14 days for the first six to eight weeks, or more frequently if they have more severe depression (4).
Most patients with depression are treated in primary care settings (5). Three prior studies, which identified provider specialty from clinic types or service claims, found substantial differences in rates of follow-up visits between primary care and mental health specialty care settings (6,7,8). Identifying provider specialty by clinic type (for example, general internal medicine clinic versus mental health clinic) may be misleading because providers with various specialties may practice in a given clinic. Using medical claims as the source of provider specialty may not identify the physician providing antidepressant management.
The objectives of this study were to examine the rate of receiving guideline-concordant follow-up visits among patients with major depressive disorder initiating antidepressant treatment and to evaluate whether this rate differs between patients initially given a prescription for antidepressants by primary care providers, by psychiatrists, or by nonpsychiatrist specialists. We identified provider specialty from the initial antidepressant prescription, because providers writing the index prescription were most likely to be the same provider in subsequent outpatient visits managing the antidepressant treatment. We also conducted sensitivity analyses to provide robustness of our overall results.
The study used 2000–2004 medical and prescription claims from a large national health plan affiliated with i3 Innovus. Patients were included if they were diagnosed as having major depressive disorder between July 1, 2000, and December 31, 2002, by using the ICD-9-CM diagnosis codes 296.20–296.24 in any diagnosis field; filled a second-generation antidepressant prescription within 45 days of the index diagnosis; and were continuously enrolled for at least six months before and two years after the index diagnosis. Patients were excluded if they were younger than 18 years, had service claims with an ICD-9-CM diagnosis code of bipolar disorder or schizophrenia during the study period, or had antidepressant claims during the six-month preindex period. Individuals with an overlapping supply of another antidepressant indicating augmented therapy were excluded, because these patients may require a different frequency of monitoring. Finally, patients with an unknown provider specialty in their index antidepressant prescription claim were excluded because of missing observations for the explanatory variable of interest.
Follow-up visits were identified on the basis of outpatient service claims with mental health diagnosis codes (ICD-9-CM codes 290–319) in any diagnosis field. Based on National Committee on Quality Assurance's Healthcare Effectiveness Data and Information Set (HEDIS) measures, patients were considered to have had minimum guideline-concordant follow-up visits if they received at least three follow-up visits during the first 90 days after the index antidepressant prescription and least one of the follow-up visits was with a provider with prescribing privileges (9).
The main explanatory variable of interest was provider specialty. The index antidepressant prescription claim was identified, and the provider specialty code from that claim was used to identify the initial provider specialty. We used the prescription as the source of provider specialty because we assumed that patients received the initial antidepressant prescription from the providers who initiated and subsequently managed their antidepressant therapy. Provider specialty was categorized into three mutually exclusive groups: primary care providers, psychiatrists, and all other nonpsychiatrist specialists (for example, cardiologist, gastroenterologist, neurologist, and oncologist).
To adjust for differences between patients seen by providers of different specialties, we included several covariates in the regression analyses. Demographic variables included age, gender, and region of residence. Median household income and percentage of population 25 years or older with a high school degree in the state of residence were included as proxies for socioeconomic status. A chronic disease score based on prescription claims during the six-month preindex period was calculated to adjust risk of comorbid health conditions (10). Preexisting comorbid psychiatric conditions, including anxiety and alcohol or drug abuse, were identified at baseline. An indicator for receiving pregnancy-related outpatient visits during the 90-day period was included in the model. Type of initial antidepressant was included to control for potential differences in side effect profiles that may result in later outpatient visits. Quarter of year in which treatment was initiated was controlled for as a proxy for seasonality. Health plan characteristics included in the model were copayment for primary care visits and whether the patient saw an initial prescriber who was paid on a capitated basis.
The average number of follow-up visits was compared between provider specialty groups with two-tailed Student's t tests, and the proportion of patients with guideline-concordant follow-up visits was compared among provider specialty groups with Pearson's chi square statistics. Logistic regression models were used to evaluate the association between provider specialty and receipt of guideline-concordant follow-up visits, with adjustment for patient and health plan characteristics. To control for confounding from imbalance in observed covariates, we implemented propensity score matching techniques to create comparable patient cohorts between primary care and mental health specialty settings (11). The results were similar between the original model and the propensity score matched model, so we present results from only the original model.
Because the accuracy of the results may be contingent on how providers coded diagnoses, we conducted two sensitivity analyses to investigate how results may have differed by varying the approach used to identify follow-up visits from administrative claims data. The HEDIS measure requires that one of the three follow-up visits be with a prescribing provider, but patients are likely to have visits with nonprescribing providers (for example, psychologist or social worker) who would not be able to change a patient's antidepressant therapy. In the first sensitivity analysis, we tested whether considering only follow-up visits with a mental health diagnosis made to providers with prescribing privileges would change the results of the main analysis. We expected that by focusing only on the visits made to prescribing providers, the results would more closely represent the differences in visits to manage antidepressant therapy.
Also, we thought that depression diagnosis could be underreported on the claims for various reasons. To understand how robust our results were in regard to this issue, we conducted a second sensitivity analysis. Our data allowed us to link the provider from antidepressant prescription claims to outpatient service claims based on encrypted provider identification. Visits, regardless of diagnosis code, that were made to the same provider who wrote the initial antidepressant prescription were identified. We counted these visits and the visits with a mental health diagnosis made to providers with prescribing privileges (as used in the first sensitivity analysis) as the outcome in the second sensitivity analysis. On one hand, we relaxed the requirement for mental health diagnosis; on the other hand, we restricted the relaxation of the requirement to only the visits made to the provider who initially prescribed the antidepressants in order to ascertain some level of confidence that the issues about antidepressant treatment would likely be addressed during these visits. We expected this change in the definition of follow-up visits to be most evident for patients seen by primary care providers because psychiatric diagnoses are commonly undercoded in primary care to accommodate patients who may not want it recorded because of stigma (12).
Institutional review board approval of this study was obtained from University of North Carolina at Chapel Hill.
The final sample included 4,102 patients aged 18 years or older with major depressive disorder who initiated treatment with second-generation antidepressants. There were 2,441 patients (60%) whose index antidepressant prescription was written by a primary care provider, 1,443 (35%) whose index prescription was written by a psychiatrist, and 218 (5%) whose index prescription was written by a nonpsychiatrist specialist. Compared with patients who were initially given a prescription by primary care providers, patients with an initial prescription from psychiatrists were younger (mean±SD of 38±12 years versus 41±12 years) (p<.05) and patients with an initial prescription from nonpsychiatrist specialists had a higher chronic disease score (mean score of 1.95±1.00 versus 1.77±.79) (p<.05). (Chronic disease score was measured by prescription claims during six months prior. Scores ranged from .16 to 7.5 for the study sample, with higher scores representing a higher level of comorbidities.)
The mean±SD number of follow-up visits in the entire sample was 2.64±3.06 during the first 90 days after the index prescription. When stratified by provider specialty, the mean number of follow-up visits was higher for patients with an initial prescription from psychiatrists (mean of 3.8 ±3.5 visits; median of two visits) and for patients with an initial prescription from nonpsychiatrist specialists (mean of 2.8±3.3 visits; median of one visit) than it was for patients with an initial prescription from primary care providers (mean of 2.0±2.6 visits; median of one visit) (p<.05 for both). Overall, 1,265 patients (31%) received guideline-concordant follow-up visits. When stratified by provider specialty, the proportion was significantly higher for the psychiatrist group (N=745, 52%) and the nonpsychiatrist specialists group (N=58, 27%) than it was for the primary care providers group (N=464, 19%) (p<.05 for both).
In adjusted analyses, patients were more likely to receive guideline-concordant follow-up visits if they received their initial prescription for an antidepressant from psychiatrists (odds ratio [OR]=4.6, 95% confidence interval [CI]=3.9–5.4) or from nonpsychiatrist specialists (OR=1.5, CI=1.1–2.0) than they were if they received their initial prescription from primary care providers (Table 1). Patients were less likely to receive guideline-concordant follow-up visits if they were older than 50 years (50–64 years, OR=.77, CI=.62–.96; 65 years or older, OR=.3, CI=.1–.6) and if their copayment for primary care visits was more than $20 (OR=.6, CI=.4–.9). Patients who had preexisting comorbid anxiety disorders were more likely than those without them to receive guideline-concordant follow-up visits (OR=1.8, CI=1.4–2.3).
In the first sensitivity analysis that limited the visits coded with a mental health diagnosis only to those made to providers with prescribing privileges, we found that results were nearly identical to those in the main model (Table 1). In the second sensitivity analysis that added visits made to the same provider who wrote the initial antidepressant prescription, regardless of diagnosis, to the visits in the first sensitivity analysis, the odds ratio for receiving guideline-concordant follow-up visits among psychiatrists became much smaller but was still significant (OR=2.3, CI=1.9–2.7). The odds ratio for nonpsychiatrist specialists became nonsignificant.
The study used a large claims database to examine whether patients with major depressive disorder treated with antidepressants received guideline-concordant care (according to the HEDIS measure of three or more follow-up visits within 90 days) and whether these visit rates differed by provider specialty. The findings from this study suggest that follow-up visits are underutilized by all providers—31% of all patients received guideline-concordant follow-up care. Patients who were given an initial prescription for an antidepressant by a psychiatrist were more likely than those who were given such a prescription by a primary care provider to have guideline-concordant follow-up (52% versus 19%, p<.05). However, it should be noted that nearly half of the patients who were seen initially by psychiatrists did not receive guideline-concordant care. This unadjusted difference was confirmed in adjusted analyses that controlled for patient, provider, and health plan factors, which is consistent with prior studies that used different definitions for provider specialty (6,7).
In light of the differences between these types of providers, this study provides additional evidence highlighting the need for new care models that promote close management to maximize antidepressant effectiveness (13). Previous studies have highlighted patient, provider, and health system barriers to providing guideline-concordant care (14), but further investigation is needed to fully understand these barriers, including assessment of provider differences in attitudes toward and knowledge of depression treatment and knowledge about guidelines. The findings also underscore the need to improve follow-up for patients with depression managed in primary care settings, because it is neither efficient nor realistic to channel all patients with depression into specialty care. Rather, the goal should be to improve management of depression in primary care. Indeed, when guideline-concordant care is provided, depression outcomes in primary care and psychiatric settings appear equivalent (15).
As with prior studies, this analysis identified follow-up visits by service claims that had a mental health diagnosis code. Depression is a common disease that is underdiagnosed in primary care, and patients might not want it to be recorded because of stigma (12). It is likely that some visits that involved antidepressant management were actually coded without mental illness diagnoses, and this is the first study to address this issue in sensitivity analyses. Our data allowed us to identify visits made to the same provider who wrote the initial antidepressant prescription, which provided an opportunity for patients to talk about antidepressant treatment, although it might not be the main reason for the visit. Results from the sensitivity analyses that accounted for these types of visits still found that initiating antidepressant treatment with a psychiatrist was associated with greater odds of receipt of guideline-concordant follow-up visits. This sensitivity analysis provides greater confidence that there is a systematic difference between primary care providers and psychiatrists who initiate antidepressant treatment.
Our study has several limitations. The study was based on a commercially insured population. We also included a continuously enrolled population, who might be more stable in the plan and have a greater likelihood of having guideline-concordant follow-up visits. Therefore, the results might not generalize to the publicly insured population. The study utilized data before the black-box warning for antidepressants, and practice patterns might have changed after this warning was implemented. Many plan-specific variables (for example, benefit structure) and provider-specific variables (for example, years of practice, practice setting, and patient volume) were not available in the data, which may explain variation in follow-up rates between provider specialties. Unobservable confounders may exist. Finally, the observational design prohibits attribution of causality.
Our study highlights the existence of suboptimal care and differences between provider specialties. Follow-up visits are likely driven by both providers and patients. Future research should systematically evaluate the differences in attitudes toward, knowledge of, and barriers to depression care between mental health specialists and primary care providers. Studies should also examine how patient behavioral factors (such as attitude, knowledge, and stigma) and clinical factors (such as depression severity) may affect patient demand for follow-up visits. These studies could assist policy makers in understanding how to support primary care to provide more guideline-concordant care to patients with depression and to simulate effective organizational intervention to improve quality of care. Finally, future research should focus on linking follow-up visits to patient outcome, such as antidepressant response, to understand their effectiveness over managing antidepressant therapy.
Our findings suggest that only a small proportion of patients initiating antidepressant treatment for major depressive disorder received guideline-concordant follow-up visits. Large differences were found between primary care and mental health specialty settings in follow-up visits, but reasons for such differences require further investigation. There is still much room for improvement if primary care settings are to be the preferred location for depression treatment.
Dr. Chen is employed by Abt Bio-Pharma Solutions. Dr. Hansen has received consulting fees from Takeda Pharmaceuticals. Dr. Farley has received grant funding from the Pfizer Foundation and consulting fees from Takeda Pharmaceuticals. Dr. Gaynes has received grants and research support from M-3 Information, Bristol-Myers Squibb, and Novartis, and he has served as an advisor for Bristol-Myers Squibb. The other authors report no competing interests.