Influences on the use of antidepressants in primary care: All England general practice‐level analysis of demographic, practice‐level and prescriber factors

General practice (GP) antidepressants (ADs) prescribing in England has almost doubled in the past decade: how does location, GP characteristics, and prescribing selection influence antidepressant prescribing rate (ADPR) and growth.


| INTRODUCTION
The number of prescriptions for antidepressants (ADs) in England has almost doubled in the past decade. Data from NHS Digital show that 70.9 million prescriptions for ADs were given out in 2018, compared with 36 million in 2008 (Iacobucci, 2019).
Evidence suggests that medication prescribing for many chronic health conditions, particularly in older persons is often inappropriate (Spinewine et al., 2007) with associated increases in morbidity and economic burden (Simonson & Feinberg, 2005).
In 2017, one in six adults in England was prescribed ADs. The United Kingdom figures, covering the NHS as a whole, saw a total of 7.3 million people given at least one AD prescription in 2017. This included more than 70,000 people under the age of 18 years. Those aged over 60 were twice as likely as those in their twenties to be on ADs. One in five people in towns such as Blackpool and Great Yarmouth was prescribed ADs in 2017, while in London the figure was less than 1 in 10 (www.pulsetoday.co.uk).
We have previously applied multivariate regression analysis on publicly available NHS data at the general practice level to identify how general practice factors relate to outcome in terms of glycosylated haemoglobin (HbA1c) (Heald et al., 2017(Heald et al., , 2018. This approach can be generalised to other areas of medicine, including prescribing in psychiatry and has proved informative in terms of appreciating the drivers of prescribing year-on-year in several long-term conditions. This exploratory study using national-level data aimed to look at how a range of quantifiable and nationally audited factors at a general practice (family doctor practice) relate to general the practice variation in antidepressant prescribing rate (ADPR) across England.

| METHODS
We collected the England national public published population demographic, practice characteristics and AD prescribing behaviour in each general practice and year and used multivariate regression analysis to establish their link to the practice ADPR. Only general practices with more than 2,000 registered patients (i.e. requiring ≥ one full-time general practitioner) were included in the analysis. The population demographic, general practice processes and prescribing behaviour in each practice and year were analysed.
We examined three different classes of possible factors that could influence the ADPR.

| Statistical analysis
Stepwise multivariate regression analysis was used to establish the link between these factors to the ADPR at a general practice level.
Only factors that had a p-value < .05 were retained within the analysis. As many factors are not independent of each other, this analysis was carried out both for each class (location, characteristics and prescribing behaviour) and across all classes and factors.

| Ethical approval
As we used publicly available general practice-level data, with no individual patient data, it was not considered necessary to seek Ethics Approval for this study. there was a 37% rise in the number of people being recorded on the depression register and 22% rise in total doses of ADs.
F I G U R E 2 Cross-sectional analysis of the link between practice-level factors and AD prescribing. AD, antidepressant Total costs of ADs fell 15%, a reduction in the unit cost of 35%.
The total number of different unique ADs at different dose levels increased from 94 to 107 in 2017-2018, with 2.1 billion doses of AD being prescribed into a total population of 52 million people. Average ADPR, Defined Daily Doses of AD/head population/day, was 0.096 and 80% of practices lay between 50 and 150% of this value. This highlights the wide variation in the use of antidepressants by local practices (Figure 1).

| Multiple regression analysis
Location and demographics including age, gender, ethnicity, social deprivation, population density and latitude accounted for 62% of the variation (Figure 2). The results for each of the factors that were included in the final model are shown in Table S1.
Practice characteristics on their own including levels of comorbidities including depression could account for 62% of the variation.
It is worth noting that the univariant analysis for % of patients on the depression register accounted for 30% of the variation in overall ADPR.
The prescribing behaviours accounted for 51% of prescribing variation. The remaining explained variation came from practice prescribing behaviour including the number and mix and costs of different ADs being prescribed. Practices with higher cost/dose had lower ADPR, those using a higher number of different ADs had higher ADPR.
As many factors were codependent, that is, age, social disadvantage and BME ethnicity could impact on comorbidities and prescribing behaviour so when all the factors were included, the model could account for 81% of the variation in GP practice ADPR.
Factors cross-sectionally linked with relatively more AD prescribing at general practice level included: • Higher proportion of people with COPD and diabetes as major

| Cost and AD prescribing rate
The multivariate regression highlighted that practices with higher cost/dose had lower ADPR. Also, those using a higher number of different ADs had higher ADPR, and there was a significant reduction in The association of comorbidities-COPD and diabetes with increased AD prescribing highlights the importance of holistically addressing long-term health concerning the impact of long-term physical conditions on mental health. The influence of GP practice size and location on AD prescribing has not been reported before. The finding that a higher overall social disadvantage level is associated with greater prescribing of ADs is not surprising. Conversely, the link between higher proportion BAME ethnicity in the GP practice and lower AD prescribing may be a marker for profound cultural influences on the way that individuals perceive symptoms of depression and the implications of those symptoms.
We have recently described using national GP practice-level data how the empowerment of individuals in managing long-term conditions, has the potential to reduce GP practice level prescribing of ADs (Heald et al., 2020). Practices more effective in empowering their patients as assessed by "How confident are you that you can manage any issues arising from your condition (or conditions)", was non-linear with less antidepressants prescribed for both high and low responses.
The difference between the lowest and highest decile of prescribing for this response was over 10% and potentially modified by changing practice approach. Therefore measures that facilitate patient empowerment can potentially decrease the level of antidepressant prescribing.
We have shown that demographic factors, socioeconomic deprivation, population density and location are also important factors associated with increasing prescriptions. Further research is needed to examine whether prescriptions are being effectively prescribed and whether areas of lower prescribing have lower rates of depression, higher unmet need or better use of non-pharmacological strategies.
We nevertheless accept that we are applying at an individual level, conclusions drawn from general practice-level analysis.
The limitations of our analysis are that it does not look at individual patient data and only includes data that is recorded on national registries. However, the data covers all GP surgeries in England and is, therefore, representative of the determinants of antidepressant prescribing across these nations.

| CONCLUSION
The results represent a benchmark against which general practices can establish their baseline ADPR, incorporating their local demographic and practice profile-and then consider mix and relevance of the various ADs to enhance the patient benefit of their prescribing protocols.
We hope that our findings can inform local clinical behaviour, medicines management recommendations and provide insight that is helpful to general practices.

DATA AVAILABILITY STATEMENT
Any requests for data extracts will be considered by Dr Adrian H. Heald as the corresponding author.

ETHICS STATEMENT
As we used publicly available and GP level data, with no individual patient data, it was not necessary to seek Ethics Approval for this study.

SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of this article.