General practitioners’ drug treatment for depression by patients’ educational level: registry-based study

BACKGROUND
Antidepressant drugs are often prescribed in general practice. Evidence is conflicting on how patient education influences antidepressant treatment.


AIM
To investigate the association between educational attainment and drug treatment in adult patients with a new depression diagnosis, and how gender and age influence the association.


DESIGN AND SETTING
Nationwide registry-based cohort study, Norway, 2014-2016.


METHOD
The study comprised all residents of Norway born before 1996 and alive in 2015. We obtained information on all new depression diagnoses in general practice in 2015 (Primary Care Database) and data on all dispensed depression medication (Norwegian Prescription Database) 12-months after the date of diagnosis. Independent variables were education, gender, and age. Associations with drug treatment were estimated using a Cox proportional hazard model, for genders separately.


RESULTS
Out of 49,967 patients with new depression (61.6% women), 15,678 were dispensed drugs (30.4% women, 33.0% men). Highly educated women were less likely to receive medication (Hazard Ratio (HR) =0.93, 95% CI (0.88 - 0.98)) than women with low education. No such differences appeared among men. Women aged 20-29 were more likely to be treated with drugs than those aged 30-59, and women aged 70+ were more likely to receive drugs (HR=1.65, (1.54 - 1.77)) than those aged 20-29. The pattern was similar but less pronounced for men.


CONCLUSION
Educational differences in antidepressant therapy among women may reflect different treatment approaches that clinicians should be aware of to avoid unintended variation. Reasons for this variation and consequences for quality of treatment should be.


INTRODUCTION
The use of antidepressant drugs has increased substantially during the 1990s and early 2000s. [1][2][3] With respect to prevalence of depression, studies indicate that women, 4,5 older people, [6][7][8] and lower socioeconomic groups 9 are at increased risk. Moreover, women are more likely to consult their general practitioner (GP) for depression compared to men 10 . Whether drug use is equally distributed across population groups, such as higher and lower educated, is poorly examined and findings so far are inconsistent. [11][12][13] With respect to gender and age, some studies indicate that GPs prescribe more depression drug therapy to women and older people than to men and younger people. 14,15 However, there is a lack of studies that use a new depression diagnosis to confirm this trend.
GPs play a key role in providing health services to patients with depression. According to guidelines, talking therapy by GP is the first choice of treatment for mild depression. 16,17 With increased severity, talking therapy may be combined with medication. 16 In Norway, about 80% of antidepressant prescriptions are issued by a GP. 14 Large registry-based studies with complete data on diagnoses, depression medication and population demography may increase the knowledge and awareness about variation in health care provision to patients with depression.
The aim of this study was to investigate the association between educational attainment and drug treatment in adult patients with a new diagnosis of depression, and to what extent gender and age influence the association.

Design
We conducted a nationwide registry-based cohort study with data from the Norwegian GP-DEP Study, which investigates pathways of depression care in general practice. Our cohort comprised all individuals with a new depression diagnosis in general practice in 2015. The cohort was examined regarding dispensing of medication for depression in the 12 months after the first date of depression diagnosis (index date) in 2015.

Data sources
A c c e p t e d M a n u s c r i p t -B J G P O p e n -B J G P O . 2 0 2 0 . 0 1 2 2 4 Information from national registries was linked at the individual level, using the unique personal identity number (encrypted) assigned to all residents of Norway. All data was stored and analysed at a safe server at the University of Bergen.
The Control and reimbursement of health care claims (KUHR) database stores data on all feefor-service claims from GPs. For each encounter, the claims contain a GP-and patientidentifier, date of contact, and one or more diagnoses according to the International Classification of Primary Care 2 nd version (ICPC-2).
The Norwegian prescription database (NorPD) stores information on all prescription drugs dispensed to patients treated in ambulatory care. 18 For each prescription, NorPD contains an encrypted prescriber-and patient-identifier, date of dispensing, generic drug information (Anatomical Therapeutic Chemical (ATC) code), and any reimbursement code. NorPD lacks information at individual level on medication dispensed to people staying in hospitals or nursing homes.
The Norwegian Patient Registry (NPR) comprises information on all patient contacts with secondary health care, with diagnoses according to the International Classification of Disease 10 th revision (ICD-10).
The National Education Database stores information on the highest level of completed education.
The Population Registry contains information on gender, year of birth, death and emigration.

Study population
The source population comprised all inhabitants of Norway born before 01.01.1996 and alive in NPR) and/or dispensed drug treatment for depression (NorPD) during 12-months prior to index date. The resulting study population comprised 49,967 individuals ( Figure 1).

Independent variables
The National Education Database is based on the International Standard Classification of Education. 19 We recoded 11 levels into three categories: low (primary school (Grades 1-7) A c c e p t e d M a n u s c r i p t -B J G P O p e n -B J G P O . 2 0 2 0 . 0 1 2 2 5 and lower secondary school (Grades 8-10) or less), medium (13 years, upper-secondary school) and high (>13 years, university and higher education). We recoded patient age into decennial categories.

Outcome
From NorPD we included all medications reimbursed for the treatment of depression: antidepressants (ATC code N06A), selected antiepileptic drugs (N03A) and selected antipsychotic drugs (N05A), dispensed during 12-months follow-up after index date (yes, no).
Number of days from index date to first drug dispensing, was categorised as 0-7, 8-31, 32-183, and 184-365 days. Since drug treatment was initiated by a GP in 86% of the cases, we use the term "GP drug treatment".

Statistical analysis
Descriptive statistics was used to examine the distribution of antidepressants, antiepileptics and antipsychotics among the patients treated with drugs, given by numbers and percentages.
Dispensing of medication and time interval from index date to first drug dispensing was provided by numbers and percentages, by educational level, gender, and age category. The associations between drug dispensing and education, gender and age were examined by chi square test. Further, Cox proportional hazard models were used to estimate the likelihood of being dispensed medication for the independent variables education, gender and age.
Interactions between education and gender, and between education and age, in the association with drug dispensing were tested in separate Cox proportional hazard models. Follow-up was defined in days from index date to first drug dispensing, and individuals were censored at the time of death, emigration, or end of follow-up, whichever occurred first. The results from the final model were presented stratified by gender (due to interaction), both crude and adjusted, as hazard ratios (HR) with 95% confidence intervals (CI). Reference groups were low education and age group 20-29 year.
The association between educational level and time to first drug dispensing was illustrated by Kaplan-Meier survival curves, and a pairwise log rank test was used to test the equality of the distribution of the survival curves between different education levels, for gender separately.

RESULTS
The study population comprised 49,967 individuals with a new depression diagnosis in 2015, with a mean age of 44.4 (standard deviation, 16 Due to a significant interaction between education and gender (P=0.010), the Cox proportional hazard model was performed for men and women separately. Women with high and medium education were less likely to receive drugs (crude HR=0.94, (0.90 -0.99) and HR=0.86, (0.81 -0.90), respectively) compared to women with low education (reference), Table 3. The age adjusted estimates were less pronounced but still significant for highly educated women. Among men there was no association between drug treatment and education. There was no interaction between gender and age. Women aged 20-29 were more likely to be treated with medication compared to those aged 30 to 59, and women aged 70+ were even more likely to receive medication (HR=1.65, 95% CI (1.54 -1.77)) than those aged 20-29. The pattern was similar but less pronounced for men. Figure 2 illustrates the distribution of survival curves for time to first drug dispensing.
Women (Figure 2a) with high education were dispensed depression drugs to a lesser extent 7 and later after index date, compared to those with low or medium education (P ≤ 0.001 for both groups). There was also a different distribution between medium and low educated women (P=0.018). Highly educated men (Figure 2b) had a different distribution of drug dispensing than men with medium education (P=0.013).

Summary
In a nationwide cohort of patients with a new diagnosis of depression, we found a gendered pattern in the occurrence of depression and in the likelihood of receiving medication by educational level. While GP-diagnosed depression was considerably more prevalent among women, the proportion being treated with drugs was higher among men. Gender modified the relation between education and medication; hence all analyses were performed separately for men and women. A novel finding was that highly educated women with new depression were significantly less likely to receive medication than lower educated women, even after adjusting for age. No educational differences were found for men. Finally, the youngest and the oldest patients were the age-groups most likely to receive depression drugs.

Strengths and limitations
The main strength of this study is the use of complete registry data from the publicly subsidised primary care services in Norway. Linkage of data from five national registries at the individual level provides a unique source of information, eliminating recall bias and selection bias.
Information on GP-diagnosed depression is another strength. A new depression diagnosis was defined as a GP-consultation with the ICPC-2 code P76, after a 1-year washout period.
However, the individual GPs set the diagnosis and the KUHR database has no formal control on diagnostic categories. Differing coding behavior may therefore challenge the internal validity. However, potential misclassification by the GP would be non-differential and distributed randomly across population groups. Another limitation is lack of information on severity of depression, as ICPC-2 does not allow for such grading. Severity probably influences GPs´ decisions to initiate drug treatment. On the other hand, variation in severity is most likely also distributed evenly across patient educational level, gender and age. The NorPD contains complete data on all prescription drugs dispensed. Although we may have slightly underestimated the prevalence of prescribed medication for depression, the use of drug dispensing data is recognised as an acceptable proxy in epidemiological studies. 20 Low out-of-pocket payment in Norway makes medication for depression easily available, and we thus believe that primary non-compliance is low and evenly distributed across population groups. To strengthen the internal validity, we have considered drugs reimbursed for the treatment of depression only, eliminating e.g. SSRIs for anxiety disorder and tricyclic antidepressants for adjuvant pain therapy.

Comparison with existing literature
The prevalence of drug therapy for new depression found in our study (one of three patients) is considerably lower than antidepressant prescription rates of 45%-75% reported in studies from European countries. [21][22][23] This discrepancy may be due to use of a new diagnosis of depression, a strict definition of depression medication, and the use of drug dispensing data in our study. Half of the patients who started on drug treatment collected medication within one week after index date. This finding is in line with studies in the Netherlands and Sweden examining the time interval from depression diagnosis to initiation of drug treatment. 22,23 Turning to gender, a Swedish registry study also including only newly diagnosed patients with depression found a slightly greater proportion of men than women receiving antidepressant drugs, 23 in line with our findings. The lower proportion of women using depression drugs may be related to gendered preferences; women preferring talking therapy, men drug treatment. [24][25][26] On the other hand, GPs may initiate talking therapy to women due to a preconception that they are more inclined to conversation.
A study in the UK among people aged 55 years and older showed that treatment rates with antidepressants were high for those recorded with a new depression diagnosis but varied little by age. 8 The present study indicates that the youngest and the oldest patients were most commonly prescribed medication. Higher prescription rates for older people may be an expression of unwarranted variation, since studies suggest that elderly judge talking therapy more favorably than medication. 27,28 GPs should be aware of older patients' increased risk of polypharmacy and adverse side effects. 29 This practice may nevertheless be due to brief GP encounters with focus on somatic conditions, 27 or to limited access to secondary mental health care. 31 9 Depression is more prevalent among people with low socioeconomic status (SES) compared to those better off. 9 Previous studies investigating the relationship between education and antidepressant use have demonstrated divergent findings. In line with our results, a registrybased study in Sweden showed that poorly educated people received more prescriptions of antidepressants than those with higher education; the educational gradient being somewhat stronger among women than among men. 12 Accordingly, Packness and colleagues in Denmark found that higher educated groups with no/few self-reported symptoms of depression were less likely to use medication, however, no associations were found among people with more pronounced symptoms. 13 In contrast, Kivimäki and colleagues found lesser antidepressant use among men with lower education compared to men with higher education in Finland, while such differences were not seen in women. 11 Finally, two studies conducted in Denmark and one in Australia showed no association between education and antidepressant use. [32][33][34] The conflicting results in these studies could be due to different study populations and measures of educational status, or to cross-national differences in access to treatment among disadvantaged people. In contrast to the current study, none of these studies comprised information on new depression diagnoses made by a GP.
Socioeconomic variation in antidepressant use found in this study may reflect an unintended bias by the GP letting SES influence prescription of medication, 34 differential patient preferences supported by the GP, or a combination. One may speculate whether GPs prescribe drugs rather than provide talking therapy to patients they perceive as less educated, or whether better educated patients are more assertive about having non-pharmacological therapy. A survey among patients in GP waiting rooms in Norway showed that lower educational level was associated with greater preference for medication. 35 The association between educational level and drug treatment found in our study among women only, may suggest that highly educated women are more skeptical to medication in general. This may apply particularly to younger women since older women in the study population were less educated.

Implications for research and/or practice
The results of our study support an association between drug treatment for new depression and patient gender, education and age. Highly educated women received less medication than women with low education, and no such differences appeared among men. Educational differences among women may reflect different treatment approaches that clinicians should be   e p t e d M a n u s c r i p t -B J G P O p e n -B J G P O . 2 0 2 0 . 0 1 2   Educational level: Low (primary school (Grades 1-7) and lower secondary school (Grades 8-10) or less), medium (13 years, upper-secondary school) and high (>13 years, university and higher education) *Adjusted for age and education, respectively. ** Missing data on educational level for 321 women and 227 men.