Social Inequalities in Early Antidepressant Discontinuation
Abstract
Objective
This study aimed to test for social disparities in early discontinuation of antidepressant treatment and to explore associations with type of drug and composition of prescriber’s clientele.
Methods
The cohort was 14,518 Marseille residents (ages 18–64 years) covered by the National Health Insurance Fund who had a new episode of antidepressant treatment (specifically, no prescription claim in the six months before the index claim) prescribed by a private general practitioner in 2008 or 2009. Factors associated with early discontinuation (prescription filled or refilled fewer than four times in the six months after the index claim) were analyzed with multilevel models that were adjusted for patient morbidity and number of consultations with private general practitioners and psychiatrists. Sensitivity analyses were conducted with different definitions of new treatment and early discontinuation.
Results
Low income, type of antidepressant (tricyclics versus selective serotonin reuptake inhibitors), and prescribers’ clientele composition (specifically, a high proportion of socioeconomically disadvantaged patients) were independently associated with an increased risk of early antidepressant discontinuation. A significant interaction was found between low income and gender. Low-income patients were more likely than other patients to receive tricyclic antidepressants.
Conclusions
These results add further evidence of inequalities in care for major depression and suggest that women are at greater disadvantage than men. Educational programs for general practitioners should focus on the risks of antidepressant discontinuation among disadvantaged patients. Enhancing therapeutic education of low-income patients may improve their treatment adherence.
International clinical practice guidelines (including guidelines in France) recommend that antidepressant treatment of major depression be continued for several months after recovery to reduce the risk of relapse and recurrence (1–4). However, there is growing evidence that patients discontinue new antidepressant prescriptions at high rates during the first several weeks of treatment (usually 50%−60% within ten weeks) (5–8). Diverse factors may account for this discontinuation: the drug’s characteristics and side effects, personal characteristics related to adherence (such as gender and age and self-reported severity of illness, fear of drug dependence, and perception of treatment efficacy) (5,7), and factors related to the prescribers (including specialty, skill in diagnosing depression and prescribing appropriate antidepressant treatments, and amount of time spent in patient education) (6,8–10).
Although compelling evidence indicates that low socioeconomic status (SES) increases the duration of new episodes of major depression (11), several studies have shown the duration of antidepressant treatment to be shorter among people of low SES (6–8,12,13). This fact suggests that the inverse-care law—that is, “the availability of good medical care tends to vary inversely with the need for it in the population served” (14,15)—applies to continuation of antidepressant treatment, especially among highly disadvantaged populations. Nonetheless, some studies have reached different conclusions (10,16–19).
Few studies have sought to study the reasons for social inequalities specifically in the duration of antidepressant treatment (18). In a previous study (13), we suggested that these inequalities exist both in access to care (availability and use of mental health services) and in processes of care (the technical and interpersonal care provided to the patient, such as diagnostic and therapeutic procedures and coordination and continuity of care) (20). For one aspect of the processes of care, previous studies have suggested that people of low SES, compared with those who are better off, are more likely to receive tricyclic antidepressants (21), which are not tolerated as well as newer classes of antidepressants (22). Differences in type of treatment might modify patient adherence to treatment (23,24) but are rarely taken into account in examining social inequalities in antidepressant treatment discontinuation (6,25). Similarly, the characteristics of prescribers (specialty excepted) and their clientele have been taken into account only rarely in studies analyzing social inequalities in antidepressant treatment (8,10,25), although these characteristics have been shown to be associated with prescriptions for and duration of antidepressant treatment (6,8–10,12,25) and to vary according to patients’ SES (26–28).
We conducted a cohort study to obtain further insights into the reasons for social inequalities in early discontinuation of antidepressant treatment with a focus on the processes of care. This study aimed to verify whether early antidepressant discontinuation occurs more frequently with low-income patients than with others, independently of other factors, and whether early discontinuation depends on prescribers’ choices of specific antidepressant drug treatment or on the composition of their clientele. We used multilevel models to take the hierarchical structure of the data (patient-level data embedded in physician-level data) into account and adjusted for proxies for patient morbidity.
Methods
Study design, population, and period
The target population consisted of the inhabitants of Marseille, the second largest city in France. Residents were ages 18–64 years and were covered by the French National Health Insurance Fund (NHIF). The fund’s beneficiaries are mainly salaried workers, including workers who have become unemployed. The cohort included all beneficiaries with a new episode of antidepressant treatment, defined as no antidepressant prescription claims in the six months before the index claim (8,29–31), between July 1, 2008, and June 30, 2009. We included all new treatments prescribed by private general practitioners. In France, general practitioners manage more than 80% of patients with depression and prescribe 75% of new antidepressant treatments (3,13).
Data sources and variables
Individual data.
We obtained cohort members’ deidentified data from the databases of the NHIF in southeastern France after approval from the National Data Protection Authority (Commission Nationale de l’Informatique et des Libertés). The NHIF provides health insurance for all residents of France and about 87% of the French general population—the same proportion it insures in southeastern France. It covers most health care expenses (including doctors’ visits, medication, hospitalizations, laboratory tests, and imaging) and reimburses about 75% of overall individual health care costs (65% of costs for both older and newer antidepressant medications) (32,33). Only medications prescribed by a physician (private or public), purchased in commercial pharmacies in France, and considered effective by the Ministry of Health are reimbursed by the NHIF. The fund does not record prescriptions dispensed by public or private hospital pharmacies during hospital stays.
We collected information on the following drugs dispensed between January 1, 2008, and December 31, 2009: antidepressants (ATC classification code N06A), anxiolytics (ATC code N05B), hypnotics (ATC code N05C), antipsychotics, and lithium (ATC code N05AN01). [A list of the antipsychotics and their ATC codes is provided in appendix 1 of the online data supplement to this article.] For each prescription claim, we retrieved the dates it was prescribed and dispensed and the anonymous identifier of the prescriber.
Using the clinical guidelines as a reference framework and the same methodology used by Melfi and collegues (30), we defined early discontinuation as fewer than four claims (including refills) for the antidepressant within six months after the index claim or any number of claims when the last purchase was made fewer than 75 days after the index claim. Because prescriptions are dispensed in France for a maximum of 28 days (31), this definition refers to episodes of treatment lasting three months or less. We classified new treatments into four categories according to the type of antidepressant first dispensed during the study period: selective serotonin reuptake inhibitors (SSRIs), tricyclics, other antidepressants (including venlafaxine, a serotonin-norepinephrine reuptake inhibitor), and combined antidepressants (such as two or more types of antidepressant). We built a coprescription variable for the dispensing of anxiolytics or hypnotics at least once within the six months after the index claim, and we used the coprescription of antipsychotics or lithium (at least one claim during the same period) as a marker of severe mental illness (31).
We used beneficiaries’ participation, as indicated in NHIF records, in the public CMUC (Couverture Maladie Universelle Complémentaire, or the Complementary Universal Health Insurance) as an indicator of very low income. This program exempts from any out-of-pocket costs individuals throughout France who are younger than 65 and have annual incomes below €9,000 (32). Physicians can prescribe the medication of their choice to their CMUC patients.
The following data were also available for all NHIF beneficiaries: gender, age, and number of consultations with general practitioners (including house calls) and with private psychiatrists in 2008–2009. We used visits to psychiatrists as a proxy for depression severity (34). We had no access to beneficiaries’ hospital consultations or admissions. Chronic disease status, recorded by physicians according to the tenth revision of the International Classification of Diseases (35), was obtained for each beneficiary. Chronic status is attributed to persons with specific and chronic diseases that are costly to treat (defined by NHIF) and that make patients eligible for 100% reimbursement for treatment. We used this information as a proxy for the presence of severe somatic (yes-no) or chronic psychiatric illnesses (none, major depression, severe anxiety disorder, or other severe psychiatric illness, such as psychosis) (31).
Characteristics of prescribers and their clientele.
For private physicians, the NHIF databases contain characteristics of the prescribers and their clientele. We matched the cohort members to the anonymous private general practitioner who prescribed their new treatment. We obtained the following data for deidentified private physicians for 2008: gender, age, and total volume of medication prescribed and dispensed (in euros). For the physician’s 2008 patient list (all patients were seen at least once during 2008), we obtained data concerning the proportion of patients age ≥60 years or covered by CMUC (as an indicator of a very low SES or disadvantaged clientele) or having a chronic disease. We calculated the annual mean of total medications prescribed and dispensed per patient as a proxy for the physicians’ general prescription practices.
Statistical analysis
We used chi square tests to compare CMUC beneficiaries with others. Then we analyzed factors associated with the dependent variable of early discontinuation. We used unadjusted logistic regressions and then a multivariate multilevel logistic regression (36) to consider the hierarchical structure of the data where individuals’ data (level 1) were nested among the physician’s clientele (level 2: the prescriber responsible for the new treatment). We first tested a null model (with no variables) to evaluate the significance of interprescriber variability and to assess whether the multilevel approach was justified (37); we used the median odds ratio (MOR) to quantify this variability (38). To focus the analysis on the processes of care, we adjusted the multivariate model for the number of visits each beneficiary had with private general practitioners (13). To assess whether socioeconomic differences in early discontinuation varied according to patients’ gender and age, we tested interactions between CMUC coverage and these variables. Because of the high frequency of early discontinuation, we corrected estimated odds ratios with the method of Zhang and Yu (39) to estimate relative risks (RRs) with their 95% confidence intervals (CIs).
We reassessed the full model with an interval defining a new treatment as one year without antidepressant treatment (versus six months), with a less stringent definition of early discontinuation (no antidepressant refill at all within the six months after the index claim) and with a more stringent one (fewer than six claims, including refills, for antidepressants within the six months after the index dispensing).
Analyses were performed with SAS version 9.2 and the GLIMMIX procedure.
Results
Individual characteristics and type of drug treatment
The study population included 14,518 individuals with a new treatment prescribed by a private general practitioner between July 1, 2008, and June 30, 2009. Mean age at inclusion was 44.0±11.4 years; 31.9% of the sample were men, 22.0% were covered by CMUC, 14.9% had a severe chronic somatic illness, and 5.7% a chronic psychiatric disorder (Table 1).
Full cohort(N=14,518) | Not covered by CMUC(N=11,329)a | Covered by CMUC(N=3,189)a | |||||
---|---|---|---|---|---|---|---|
Characteristic | N | % | N | % | N | % | pb |
Gender | <.001 | ||||||
Female | 9,892 | 68.1 | 7,602 | 67.1 | 2,290 | 71.8 | |
Male | 4,626 | 31.9 | 3,727 | 32.9 | 899 | 28.2 | |
Age (years) | <.001 | ||||||
18–34 | 3,265 | 22.5 | 2,416 | 21.3 | 849 | 26.6 | |
35–44 | 4,047 | 27.9 | 3,045 | 26.9 | 1,002 | 31.4 | |
45–54 | 4,032 | 27.7 | 3,166 | 28.0 | 866 | 27.2 | |
55–64 | 3,174 | 21.9 | 2,702 | 23.9 | 472 | 14.8 | |
Consultations with private general practitioners, 2008–2009 | <.001 | ||||||
0–8 | 4,004 | 27.6 | 3,637 | 32.1 | 367 | 11.5 | |
9–23 | 6,957 | 47.9 | 5,670 | 50.1 | 1,287 | 40.4 | |
>23 | 3,557 | 24.5 | 2,022 | 17.9 | 1,535 | 48.1 | |
Consultations with private psychiatrists, 2008–2009 | <.001 | ||||||
0 | 10,975 | 75.6 | 8,389 | 74.1 | 2,586 | 81.1 | |
≥1 | 3,543 | 24.4 | 2,940 | 26.0 | 603 | 18.9 | |
Chronic somatic illnessc | <.001 | ||||||
No | 12,358 | 85.1 | 9,577 | 84.5 | 2,781 | 87.2 | |
Yes | 2,160 | 14.9 | 1,752 | 15.5 | 408 | 12.8 | |
Chronic psychiatric illnessc | <.001 | ||||||
No | 13,691 | 94.3 | 10,638 | 93.9 | 3,053 | 95.7 | |
Major depression | 336 | 2.3 | 292 | 2.6 | 44 | 1.4 | |
Severe anxiety disorder | 57 | .4 | 45 | .4 | 12 | .4 | |
Other severe psychiatric disorder | 434 | 3.0 | 354 | 3.1 | 80 | 2.5 | |
Coprescription of antipsychotics, lithium, or both | <.001 | ||||||
No | 13,785 | 95.0 | 10,824 | 95.5 | 2,961 | 92.9 | |
Yes | 733 | 5.1 | 505 | 4.5 | 228 | 7.2 | |
Coprescription of anxiolytics, hypnotics, or both | .74 | ||||||
No | 4,232 | 29.2 | 3,310 | 29.2 | 922 | 28.9 | |
Yes | 10,286 | 70.9 | 8,019 | 70.8 | 2,267 | 71.1 | |
Type of antidepressant treatment | <.001 | ||||||
SSRIsd | 9,522 | 65.6 | 7,546 | 66.6 | 1,976 | 62.0 | |
Tricyclics and monoamine oxidase inhibitors | 1,165 | 8.0 | 797 | 7.0 | 368 | 11.5 | |
Miscellaneous other antidepressants | 3,755 | 25.9 | 2,927 | 25.8 | 828 | 25.9 | |
Combination (≥2 types of antidepressant) | 76 | .5 | 59 | .5 | 17 | .5 |
SSRIs were prescribed for 65.6% of the cohort members, tricyclics for 8.0%, other antidepressants for 25.9%, and a combination of two or more types of antidepressant for .5%; anxiolytics or hypnotics or both were prescribed in the six months after the index claim for 70.9% of the sample. The type of antidepressant prescribed differed significantly (p<.001) according to CMUC coverage (Table 1). These differences remained significant after adjustment for gender, age, and number of consultations with private general practitioners and with private psychiatrists: CMUC coverage was associated with a higher probability of treatment with tricyclics versus SSRIs (results not shown).
Prescriber characteristics
Prescribers for CMUC-covered cohort members were more frequently male and younger than 45. Their prescriptions resulted in lower annual mean drug expenditures than those prescribed by providers for new treatments for people not covered by CMUC (Table 2). Their clientele had higher rates of chronic illness and low income than patients of the other providers (Table 2).
Full cohort(N=14,518) | Not covered by CMUC(N=11,329)a | Covered by CMUC(N=3,189)a | ||||
---|---|---|---|---|---|---|
Characteristic | N | % | N | % | N | % |
Prescribers | ||||||
Gender | ||||||
Female | 2,535 | 17.5 | 2,097 | 18.5 | 438 | 13.7 |
Male | 11,983 | 82.5 | 9,232 | 81.5 | 2,751 | 86.3 |
Age (years) | ||||||
<45 | 2,060 | 14.2 | 1,559 | 13.8 | 501 | 15.7 |
45–55 | 8,515 | 58.7 | 6,727 | 59.4 | 1,788 | 56.1 |
>55 | 3,943 | 27.2 | 3,043 | 26.9 | 900 | 28.2 |
Annual mean expenditure on prescriptions in 2008–2009 (euros per patient) | ||||||
<129 | 2,204 | 15.2 | 1,392 | 12.3 | 812 | 25.5 |
129–266 | 7,611 | 52.4 | 6,057 | 53.5 | 1,554 | 48.7 |
>266 | 4,703 | 32.4 | 3,880 | 34.3 | 823 | 25.8 |
Prescribers’ clientele | ||||||
Patients age ≥60 (%) | ||||||
<38 | 3,825 | 26.4 | 2,887 | 25.5 | 938 | 29.4 |
38–47 | 8,188 | 56.4 | 6,426 | 56.7 | 1,762 | 55.3 |
>47 | 2,505 | 17.3 | 2,016 | 17.8 | 489 | 15.3 |
Patients with chronic illness (%)b | ||||||
<27 | 2,081 | 14.3 | 1,946 | 17.2 | 135 | 4.2 |
27–46 | 7,398 | 51.0 | 6,356 | 56.1 | 1,042 | 32.7 |
>46 | 5,039 | 34.7 | 3,027 | 26.7 | 2,012 | 63.1 |
CMUC-covered patients (%)a | ||||||
<5 | 2,006 | 13.8 | 1,910 | 16.9 | 96 | 3.1 |
5–23 | 7,164 | 49.4 | 6,212 | 54.8 | 952 | 29.9 |
>23 | 5,348 | 36.8 | 3,207 | 28.3 | 2,141 | 67.1 |
Early discontinuation of antidepressant treatment
Overall, 71.6% of the cohort discontinued their new antidepressant treatment early; this percentage was 44.7% when discontinuation was defined as no refills at all, and 84.2% when defined as fewer than six claims within the six months after the index claim. [Details are illustrated in appendix 2 of the online data supplement.]
In the unadjusted analyses, the risk of early discontinuation was higher among patients covered by CMUC and among those receiving tricyclics or other antidepressants, compared with those receiving SSRIs (Table 3). Risk of early discontinuation also was associated with characteristics of prescribers and their patients; in particular, it was most likely when the proportion of CMUC-covered patients was highest (Table 3).
Factor | N | UnadjustedRR | 95% CI | AdjustedRR | 95% CI |
---|---|---|---|---|---|
Individual level | |||||
Male (reference: female) | 4,626 | 1.03 | 1.01–1.05 | 1.02 | .99–1.04 |
Age (reference: 18–34) | |||||
35–44 | 4,047 | .90 | .88–.92 | .90 | .87–.93 |
45–54 | 4,032 | .87 | .84–.89 | .87 | .84–.90 |
55–64 | 3,174 | .84 | .82–.87 | .85 | .82–.89 |
Consultations with private general practitioners, 2008–2009 (reference: 0–8) | |||||
9–23 | 6,957 | .86 | .84–.88 | .86 | .83–.89 |
>23 | 3,557 | .87 | .85–.90 | .84 | .81–.87 |
Consultations with private psychiatrists, 2008–2009 (reference: 0) | 3,543 | .72 | .69–.74 | .76 | .74–.79 |
Chronic somatic illness (reference: no)b | 2,160 | .92 | .89–.95 | .99 | .96–1.02 |
Chronic psychiatric illness (reference: no)b | |||||
Major depression | 336 | .56 | .50–.64 | .70 | .62–.78 |
Severe anxiety disorder | 57 | .77 | .61–.97 | .87 | .68–1.03 |
Other severe psychiatric disorder | 434 | .78 | .72–.85 | .90 | .83–.96 |
Coprescription of antipsychotics, lithium, or both (reference: no) | 733 | .82 | .78–.88 | .92 | .86–.97 |
Coprescription of anxiolytics, hypnotics, or both (reference: no) | 10,286 | .84 | .82–.85 | .89 | .86–.91 |
Type of antidepressant treatment (reference: SSRIsc) | |||||
Tricyclics and monoamine oxidase inhibitors | 1,165 | 1.19 | 1.16–1.22 | 1.14 | 1.10–1.18 |
Miscellaneous other antidepressants. | 3,755 | 1.04 | 1.01–1.06 | 1.04 | 1.01–1.06 |
Combination (≥2 types of antidepressant) | 76 | .98 | .84–1.14 | .98 | .80–1.13 |
CMUC coveraged (reference: no) | 3,189 | 1.16 | 1.14–1.18 | 1.12 | 1.09–1.15 |
Prescriber level | |||||
Male (reference: female) | 11,983 | 1.05 | 1.02–1.08 | 1.02 | .98–1.06 |
Age (reference: <45) | |||||
45–55 | 8,515 | 1.00 | .97–1.03 | .99 | .95–1.04 |
>55 | 3,943 | 1.06 | 1.03–1.10 | 1.05 | 1.00–1.09 |
Annual mean expenditure on prescriptions in 2008–2009 (euros per patient) (reference: <€129) | |||||
129–266 | 7,611 | .96 | .94–.99 | 1.01 | .98–1.05 |
>266 | 4,703 | .95 | .93–.98 | 1.01 | .97–1.05 |
Patients age ≥60 (%) (reference: <38%) | |||||
38–47 | 8,188 | .97 | .95–.99 | .98 | .94–1.01 |
>47 | 2,505 | .95 | .92–.98 | .96 | .92–1.01 |
Patients with chronic illnesses (%) (reference: <27%)b | |||||
27–46 | 7,398 | 1.07 | 1.03–1.10 | 1.06 | 1.01–1.11 |
>46 | 5,039 | 1.22 | 1.17–1.26 | 1.13 | 1.06–1.20 |
CMUC-covered patients (%) (reference: <5%)d | |||||
5–23 | 7,164 | 1.07 | 1.03–1.10 | 1.02 | .97–1.07 |
>23 | 5,348 | 1.23 | 1.18–1.27 | 1.11 | 1.03–1.18 |
The MOR of the null model was 1.58 (p<.001) and thus justified the multilevel approach. It decreased to 1.52 after adding individual-level covariates and remained significant in the full hierarchical model (MOR=1.40, p<.001) (Table 4). The full model, adjusted for individual and prescriber characteristics, confirmed the above-mentioned associations (Table 3). We found a significant interaction between SES and gender: the relative risk of early discontinuation associated with CMUC was higher for women (RRa=1.15, CI=1.12–1.18) than for men (RRa=1.05, CI=1.00–1.10). There was no significant interaction between SES and age.
Prescriber random effect | |||||
---|---|---|---|---|---|
Model | Variance | SE | MORa | p | Variation of the interprescriber varianceb |
Null model | .233 | .029 | 1.58 | <.001 | — |
Individual level only | .193 | .029 | 1.52 | <.001 | –17.0% |
Full model | .126 | .024 | 1.40 | <.001 | –34.7% |
Extending the criterion defining new treatment to one year or using different definitions of early discontinuation produced results similar to those presented above for the factors associated with discontinuation. [Appendixes 3 and 4 of the online data supplement provide details.]
Discussion
This cohort study of French NHIF beneficiaries residing in Marseille found early discontinuation of 71.6% of the new antidepressant treatments. Low income (CMUC coverage), type of antidepressant prescribed (tricyclics versus SSRIs), and the composition of the prescriber’s clientele (in particular a high proportion of patients with CMUC coverage) were independently and significantly associated with an increased risk of early discontinuation of antidepressant treatment.
Limitations and strengths
To the best of our knowledge, our study is the first to analyze the association between prescribers’ personal and professional characteristics and the risk of clients’ early discontinuation of antidepressant treatment while adjusting for personal and drug treatment characteristics and using multilevel regression models. This method allowed us to compare the risk of discontinuation according to prescribers’ characteristics while controlling for the systematic biases related to the composition of their clientele.
Our results cannot be extrapolated to population categories that are not covered by the NHIF (such as students, farmers, trades people, and the self-employed). Hence, the generalizability of these results to rural areas or to countries without universal health insurance may be limited.
As in most observational studies of administrative databases (6,21,31,40–42), information about diagnoses and disease severity was not available, except for a subset of patients with chronic or severe cases of depression. To improve our consideration of case severity, we adjusted the multilevel model for claims for antipsychotics or lithium and for visits with private psychiatrists, both of which can be considered markers of psychiatric disease severity given that general practitioners refer patients to psychiatrists more frequently in severe versus mild cases of depression (13,34). No information on disease duration or age at onset was available, although these variables may be associated with prescribers’ choices of the type of antidepressant.
Because NHIF databases did not contain information on the actual treatment duration, we defined early discontinuation by the number of claims, that is, reimbursements for pharmacy expenses (and we note that reimbursement usually happens automatically at purchase, with no action required by the patient). We used the same cutoff as Melfi and colleagues (30) (fewer than four claims within the six months after the index claim), which may seem to indicate poor adherence to guidelines and underestimate rates of discontinuation (30). However, changing the cutoff did not alter our results (appendix 4 of online data supplement).
We evaluated individual SES with the proxy variable of CMUC coverage because the NHIF database does not record either household income or education level. CMUC coverage is a strong and reliable indicator of low income and social deprivation in France (32,43) but does not allow analysis of socioeconomic gradients.
We used a six-month interval without any antidepressant purchases to distinguish between new and prevalent antidepressant treatments (8,29–31); this criterion is commonly considered to distinguish between a relapse (same episode) of major depression and a new episode. Some authors have used longer intervals (6,40). The choice of this interval is important because it might imply misclassification of prevalent cases as new cases (31). However, using an interval of one year instead of six months gave similar results (appendix 3 of online data supplement).
Finally, our study focused on ambulatory care and did not consider antidepressants prescribed during hospital stays. Given that there is evidence of large socioeconomic differences in hospitalization for psychiatric reasons (44), social inequalities in antidepressant use may be wider than estimated in our study.
Social inequalities in early discontinuation of antidepressants
The rate of antidepressant discontinuation in our study was higher than that observed in most previous studies in other countries (6,12,25,29) but is in line with previous results in France for the period 2004–2007 (81.8% of antidepressant treatments were discontinued within six months in a nationally representative sample of NHIF beneficiaries: 87.2% for low-income people versus 81.2% for others) (8). As previously reported in France and elsewhere (6–8,12,13), this discontinuation rate was higher for low-income patients; even after adjustment for characteristics of the drug treatment and the prescriber; low income can sometimes be a marker for severe mental illness (11). Our results suggest that the inverse-care law persists in this field and is more pronounced among women.
Higher rates of off-label antidepressant prescription among low-SES patients may partly explain this result. General practitioners may find it more difficult to diagnose depression and assess its severity among these patients (45–49). Moreover, low-SES patients may manage their illness and adhere to treatment poorly; they may have a greater expectation, for example, that the drug will work immediately, although it takes several weeks for antidepressants to become effective (50). Social inequalities in patient-physician communication, as shown in other health fields (26,51,52), may worsen this situation. Physicians may exhibit a pessimistic view of the possible outcomes of depression management for low-SES patients and thus may be reluctant to respond with their full attention (53).
Further evidence of the inverse-care law is our finding that low-income patients were more likely to receive tricyclics and less likely to receive SSRIs than other patients, as suggested elsewhere (21,41,54). Moreover, treatment with tricyclics was associated with a higher risk of antidepressant discontinuation compared with treatment with SSRIs, independent of other factors. This finding is consistent with previous studies (6,7,24,25,55) and may reflect that tricyclics are tolerated less well than SSRIs (24). In the absence of specific indication, physicians should thus prescribe SSRIs or other antidepressants rather than tricyclics (4). In France, as elsewhere, tricyclics are less expensive than newer antidepressants (21,33,41), and the out-of-pocket cost is accordingly lower, but these facts should not explain our findings because the CMUC program is designed to eliminate out-of-pocket costs for filling prescriptions from dispensing pharmacies, and the program does not limit doctors’ prescription choices. Nonetheless, drug cost might influence prescribers’ choices for highly disadvantaged patients (52).
Our results are also consistent with previous studies showing that CMUC beneficiaries tend to be concentrated among certain practitioners (28,56), perhaps because of the reluctance or refusal of some physicians to treat CMUC patients for financial or attitudinal reasons, as well as because of patient preferences and geographic location (56). This concentration in the clientele of some prescribers was independently associated with a higher risk of early discontinuation. This finding suggests differences in depression management between general practitioners with a more disadvantaged clientele and those whose patients are better off. Videau and colleagues (27) showed that the mean consultation duration of general practitioners with a low-SES clientele was shorter than among those whose patients had a higher SES, probably because the density of general practitioners is much lower in disadvantaged areas than in more affluent ones; as a consequence, physicians practicing in these areas have higher workloads and must shorten their consultations.
Conclusions
Initial training and continuing medical education of general practitioners in the area of major depression should include specific information and focus on the acquisition of specific skills to diagnose and treat low-SES populations and minimize the early discontinuation of antidepressant treatment. Specific actions should also target attitudes and knowledge of low-SES patients to improve their treatment adherence. For example, programs of therapeutic education especially tailored for this population should be developed, as they have been for treatment of other illnesses (57,58). Previous American studies have shown that depression care management programs (involving a care manager who conducts patient assessment, education, follow-ups, and care coordination) can be effective in reducing social inequalities in depression treatment and outcomes (50,59); such programs should be tested elsewhere to reduce social inequalities in depression management.
1 : Evidence-based guidelines for treating depressive disorders with antidepressants: a revision of the 2000 British Association for Psychopharmacology guidelines. Journal of Psychopharmacology 22:343–396, 2008Crossref, Medline, Google Scholar
2 : World Federation of Societies of Biological Psychiatry (WFSBP) Guidelines for Biological Treatment of Unipolar Depressive Disorders in Primary Care. World Journal of Biological Psychiatry 8:67–104, 2007Crossref, Medline, Google Scholar
3
4 Clinical Guidelines: Outpatient Management of Isolated Depressive Episodes in Adults [in French]. Paris,
5 : Risk factors and predictors of compliance in depression. European Neuropsychopharmacology 13(suppl 3):S69–S75, 2003Crossref, Medline, Google Scholar
6 : Early discontinuation of antidepressants in general practice: association with patient and prescriber characteristics. Family Practice 21:623–629, 2004Crossref, Medline, Google Scholar
7 : Continuity of antidepressant treatment for adults with depression in the United States. American Journal of Psychiatry 163:101–108, 2006Link, Google Scholar
8 : Duration of antidepressant drug treatment and its determinants in France [in French]. L’Encéphale 37(suppl 1):S36–S41, 2011Crossref, Medline, Google Scholar
9 : Suboptimal duration of antidepressant treatments in the older ambulatory population of Quebec: association with selected physician characteristics. Journal of the American Geriatrics Society 50:1365–1371, 2002Crossref, Medline, Google Scholar
10 : A nationwide prospective study on prescribing pattern of antidepressant drugs in Italian primary care. European Journal of Clinical Pharmacology 69:227–236, 2013Crossref, Medline, Google Scholar
11 : Socioeconomic inequalities in depression: a meta-analysis. American Journal of Epidemiology 157:98–112, 2003Crossref, Medline, Google Scholar
12 : Factors associated with duration of new antidepressant treatment: analysis of a large primary care database. British Journal of General Practice 62:e104–e112, 2012Crossref, Medline, Google Scholar
13 : Social inequalities in new antidepressant treatment: a study at the individual and neighborhood levels. Annals of Epidemiology 23:99–105, 2013Crossref, Medline, Google Scholar
14 : The inverse care law today. Lancet 360:252–254, 2002Crossref, Medline, Google Scholar
15 : The inverse care law. Lancet 1:405–412, 1971Crossref, Medline, Google Scholar
16 : Discontinuation of use and switching of antidepressants: influence of patient-physician communication. JAMA 288:1403–1409, 2002Crossref, Medline, Google Scholar
17 : The role of the primary care physician in patients’ adherence to antidepressant therapy. Medical Care 33:67–74, 1995Crossref, Medline, Google Scholar
18 : Treatment of depression in primary care: socio-economic status, clinical need and receipt of treatment. British Journal of Psychiatry 191:164–169, 2007Crossref, Medline, Google Scholar
19 : New users of antidepressant medications: first episode duration and predictors of discontinuation. European Journal of Clinical Pharmacology 68:65–71, 2012Crossref, Medline, Google Scholar
20 : Evaluating the quality of medical care: 1966. Milbank Quarterly 83:691–729, 2005Crossref, Medline, Google Scholar
21 : Social inequalities in antidepressant treatment and mortality: a longitudinal register study. Psychological Medicine 37:373–382, 2007Crossref, Medline, Google Scholar
22 : Selective serotonin reuptake inhibitors versus tricyclic antidepressants: a meta-analysis of efficacy and tolerability. Journal of Affective Disorders 58:19–36, 2000Crossref, Medline, Google Scholar
23 : Is antidepressant-benzodiazepine combination therapy clinically more useful? A meta-analytic study. Journal of Affective Disorders 65:173–177, 2001Crossref, Medline, Google Scholar
24 : Efficacy and tolerability of selective serotonin reuptake inhibitors compared with tricyclic antidepressants in depression treated in primary care: systematic review and meta-analysis. BMJ 326:1014–1017, 2003Crossref, Medline, Google Scholar
25 : Socio-economic determinants of early discontinuation of anti-depressant treatment in young adults. European Journal of Public Health 23:433–440, 2013Crossref, Medline, Google Scholar
26 : Socioeconomic and racial/ethnic differences in the discussion of cancer screening: “between-” versus “within-” physician differences. Health Services Research 42:950–970, 2007Crossref, Medline, Google Scholar
27 : Why patients of low socioeconomic status with mental health problems have shorter consultations with general practitioners. Journal of Health Services Research and Policy 15:76–81, 2010Crossref, Medline, Google Scholar
28 : An overview of CMU-C beneficiaries’ access to ambulatory care in 2006 [in French]. Etudes et Résultats 629:1–8, 2008Google Scholar
29 : Duration of antidepressant drug treatment and its influence on risk of relapse/recurrence: immortal and neglected time bias. American Journal of Epidemiology 170:280–285, 2009Crossref, Medline, Google Scholar
30 : The effects of adherence to antidepressant treatment guidelines on relapse and recurrence of depression. Archives of General Psychiatry 55:1128–1132, 1998Crossref, Medline, Google Scholar
31 : Impact of duration of antidepressant treatment on the risk of occurrence of a new sequence of antidepressant treatment. Pharmacopsychiatry 44:96–101, 2011Crossref, Medline, Google Scholar
32 : Complementary health insurance in France. Who pays? Why? Who will suffer from public disengagement? Health Policy 81:166–182, 2007Crossref, Medline, Google Scholar
33 Base Drugs and Tariff Information. Paris, National Health Insurance Fund of Employees, 2013. Available at www.codage.ext.cnamts.fr/codif/bdm_it/index_presentation.phpGoogle Scholar
34 : General practitioners’ choices and their determinants when starting treatment for major depression: a cross sectional, randomized case-vignette survey. PLoS One 7:e52429, 2012Crossref, Medline, Google Scholar
35 International Classification of Diseases, 10th rev. Geneva, World Health Organization, 1993Google Scholar
36 : A brief conceptual tutorial of multilevel analysis in social epidemiology: linking the statistical concept of clustering to the idea of contextual phenomenon. Journal of Epidemiology and Community Health 59:443–449, 2005Crossref, Medline, Google Scholar
37 : The contribution of multilevel models in contextual analysis in the field of social epidemiology: a review of the literature [in French]. Revue d’Epidemiologie et de Sante Publique 50:489–499, 2002Medline, Google Scholar
38 : Appropriate assessment of neighborhood effects on individual health: integrating random and fixed effects in multilevel logistic regression. American Journal of Epidemiology 161:81–88, 2005Crossref, Medline, Google Scholar
39 : What’s the relative risk? A method of correcting the odds ratio in cohort studies of common outcomes. JAMA 280:1690–1691, 1998Crossref, Medline, Google Scholar
40 : Socio-economic inequalities in first-time use of antidepressants: a population-based study. European Journal of Clinical Pharmacology 60:51–55, 2004Crossref, Medline, Google Scholar
41 : The use of antidepressants and introduction of new types in different socio-economic groups: a Danish registry-based cross-sectional study. Nordic Journal of Psychiatry 64:268–272, 2010Crossref, Medline, Google Scholar
42 : Neighborhood deprivation and psychiatric medication prescription: a Swedish national multilevel study. Annals of Epidemiology 21:231–237, 2011Crossref, Medline, Google Scholar
43 : Universal health insurance coverage: contribution of complementary protection [in French]. Médecine Sciences 20:105–108, 2004Crossref, Medline, Google Scholar
44 : Does hospital admission risk for depression vary across social groups? A population-based register study of 231,629 middle-aged Finns. Social Psychiatry and Psychiatric Epidemiology (Epub May 28, 2013)Medline, Google Scholar
45 : Clinical diagnosis of depression in primary care: a meta-analysis. Lancet 374:609–619, 2009Crossref, Medline, Google Scholar
46 : GP treatment decisions for patients with depression: an observational study. British Journal of General Practice 55:280–286, 2005Medline, Google Scholar
47 : Determinants of early identification of suicidal ideation in patients treated with antidepressants or anxiolytics in general practice: a multilevel analysis. Journal of Affective Disorders 99:253–257, 2007Crossref, Medline, Google Scholar
48 : Psychotropic drug use and mental psychiatric disorders in France; results of the general population ESEMeD/MHEDEA 2000 epidemiological study [in French]. L’Encéphale 31:195–206, 2005Crossref, Medline, Google Scholar
49 : A qualitative study exploring how GPs decide to prescribe antidepressants. British Journal of General Practice 55:755–762, 2005Medline, Google Scholar
50 : Collaborative depression care management and disparities in depression treatment and outcomes. Archives of General Psychiatry 68:627–636, 2011Crossref, Medline, Google Scholar
51 : Patient-physician interaction in general practice and health inequalities in a multidisciplinary study: design, methods and feasibility in the French INTERMEDE study. BMC Health Services Research 9:66, 2009Crossref, Medline, Google Scholar
52 : Influence of patients’ socioeconomic status on clinical management decisions: a qualitative study. Annals of Family Medicine 6:53–59, 2008Crossref, Medline, Google Scholar
53 : Managing depression in primary care: another example of the inverse care law? Family Practice 19:632–637, 2002Crossref, Medline, Google Scholar
54 : Prescribing patterns of antidepressants in Europe: results from the Factors Influencing Depression Endpoints Research (FINDER) study. European Psychiatry 23:66–73, 2008Crossref, Medline, Google Scholar
55 : Course of antidepressant treatment drug type, and prescriber’s specialty. Psychiatric Services 49:1180–1186, 1998Link, Google Scholar
56 : Explaining the strong disparities observed in the CMUC clientele of independent practitioners? Questions d’économie de la santé 130:1–6, 2008Google Scholar
57 : Outcomes of a Latino community-based intervention for the prevention of diabetes: the Lawrence Latino Diabetes Prevention Project. American Journal of Public Health 102:336–342, 2012Crossref, Medline, Google Scholar
58 : Comprehensive diabetes management program for poorly controlled Hispanic type 2 patients at a community health center. Diabetes Educator 37:680–688, 2011Crossref, Medline, Google Scholar
59 : The impact of nonclinical factors on care use for patients with depression: a STAR*D report. CNS Neuroscience and Therapeutics 15:320–332, 2009Crossref, Medline, Google Scholar