Article Text
PDF
Abstract
Background Socioeconomic deprivation is associated with health inequality. Previous studies have described associations between primary care prescribing rates and deprivation for individual drugs or drug classes. We explore the correlation between socioeconomic deprivation and the rate of prescribing of individual pharmaceutical drugs, and drug classes, in primary care in England, to identify prescribing inequalities that would require further investigation.
Methods In this cross-sectional study, national primary care prescribing data, by primary care practice, were retrieved for the calendar year 2019 in England. Socioeconomic deprivation was quantified using the Index of Multiple Deprivation (IMD) score. Correlations were calculated using Spearman’s rank correlation coefficient (ρ), adjusting for practice list size and demographics, with a Bonferroni-corrected p value threshold of 5×10–5.
Results We included 1.05 billion prescription items dispensed from 6896 England practices. 142/206 (69%) drug classes and 505/774 (65%) drugs were significantly correlated with IMD score (p<5×10−5). Of the 774 included drugs, 31 (4%) were moderately positively associated with IMD score (ρ>0.4). Only one was moderately negatively correlated with IMD score (ρ<−0.4), suggesting higher prescribing rates in more affluent areas. The drug classes most strongly associated with IMD score included opioid and non-opioid analgesics, antipsychotics and reflux medications. Drug classes most strongly associated with affluence included epinephrine, combined oral contraceptives and hormone replacement therapy.
Conclusion We identify novel associations of prescribing with deprivation. Further work is required to identify the underlying reasons for these associations so that appropriate interventions can be formulated to address drivers of inequality.
- primary care
- public health
Data availability statement
Data are available in a public, open access repository. The data sources used in this study are publicly available and have been referenced within the manuscript. The analysis code is available on the project GitHub page: https://github.com/sirsazofduck/2020MooneyJ.
Statistics from Altmetric.com
Introduction
Health inequalities may be reflected in prescribing rates across selected drug classes. For example, higher volumes of statins are prescribed in more socioeconomically deprived parts of England, even after adjusting for the increased prevalence of cardiovascular diseases and diabetes in these communities.1 Although prescriptions for antibiotics are generally falling, there is 20% more antibiotic prescribing in the most deprived centile of England compared with the most affluent centile.2 Benzodiazepines and opioids are also prescribed in higher rates in English primary care practices serving more deprived populations.3 4 While this may, in part, reflect a higher prevalence of chronic pain and anxiety disorders in these areas, other hypotheses are offered. For example, general practitioners in more socioeconomically deprived areas may be more likely to be under significant time pressures because of the higher level of healthcare demand in the communities they are serving, in turn having less capacity for medication reviews. Furthermore, alternatives to medications, such as social prescribing, may not be as effective in areas where both resources and uptake are low.5 Regardless of the underlying reasons, areas with increased prescribing are likely to have a higher incidence of medication-related side-effects and dependency. This iatrogenic morbidity could result in increased costs and even greater pressures on resources in these communities. It is also possible that some drugs are prescribed with higher frequency in areas of affluence.
Understanding the relationship between deprivation and prescribing is the first step in exploring the underlying reasons for differential prescribing and identifying where prescribing may be optimised. The aim of this investigation was to explore the correlation between socioeconomic deprivation and the rate of prescribing of individual pharmaceutical drugs, and drug classes, in primary care in England, to identify prescribing inequalities that would require further investigation.
Methods
Data sources
This is a cross-sectional study using publicly available, nationally collected primary care data from England in 2019. Monthly primary care dispensing data between January and December 2019 were downloaded from NHS digital,6 along with the age and sex distribution of registered patients at eachprimary care practice.7 These datasets include, for eachprimary care practice in England, the number of items dispensed for each unique drug preparation, as defined by British National Formulary (BNF) codes. Socioeconomic deprivation was quantified using the national Index of Multiple Deprivation (IMD) score for 2019—a relative measure of deprivation. IMD scores for each practice were obtained from Public Health England’s National General Practice Profiles.8
Data processing
BNF codes are 15-character codes that specify a unique drug product, dose and formulation. A unique drug, irrespective of dose, formulation or route is identified from the first nine digits of the code, and a drug ‘class’ was defined to be a BNF paragraph (eg, antibacterials, fibrinolytic drugs), as identified by the first seven digits of the BNF code. Monthly prescribing data for 2019 were aggregated by practice to create a single dataset for the year. Separate aggregations were then made by (1) individual drug and (2) individual drug class. Prescribing data were linked, by a uniqueprimary care practice identifier, to practice list size, the proportion of over-65s and males in the practice and practice IMD scores. Only drugs and drug classes from the first 14 chapters of the BNF were analysed. This excluded chapters on anaesthesia, preparations used in diagnosis (eg, X-ray contrast media), ‘other drugs and preparations’ (eg, poison antidotes and homeopathic preparations), dressings and appliances (eg, stoma bags). Rarely prescribed items, defined as fewer than 1000 prescriptions being dispensed over the year across England, were also excluded.
Data analysis
Prescribing rates were calculated as the number of items dispensed per 1000 registered patients in a practice. The association between drug, or drug class, prescribing rate and practice IMD score was calculated using Spearman’s rank correlation coefficient. Adjusted (partial) correlation coefficients were calculated to account for practice list size, the proportion of males and the proportion of over-65s in each practice. All correlation coefficients presented here are adjusted. To account for multiple testing, a Bonferroni-adjusted significance threshold was used. We anticipated that there would be a maximum of 1000 tests performed, and thus set an a priori threshold of p<5×10−5. Data were visually presented using bubble charts, with bubble size proportional to the total number of items prescribed. Adjusted correlation coefficient cut-offs of ±0.2 and ±0.4 represent weak and moderate correlations. All data processing and analysis was performed, and all plots generated, using the software R (V.3.6).9 Interactive versions of the included figures, and the R code written to perform the analysis and generate the plots, are available on the GitHub page https://github.com/sirsazofduck/2020MooneyJ. No ethical approval was required as all datasets used are publicly available and contain non-identifiable aggregate data.
Results
In total, 774 drugs and 206 drug classes were included in the analysis, with 1.05 billion prescription items dispensed from 6896 practices. The prescribing rate of most analysed drugs has a significant association with practice IMD score, using the Bonferroni-adjusted p value threshold (online supplemental figure S1). Medications are more likely to be positively correlated with IMD score than negatively correlated (see below and figure 1), that is, it is more common for drugs to be prescribed at higher rates in practices with more socioeconomically deprived patients, rather than at higher rates in more affluent practices.
Supplemental material
Correlation coefficients of prescribing rates and practice Index of Multiple Deprivation score relationships by individual drug. Each bubble represents one drug, with the area of the bubble proportional to the number of items dispensed in 2019 in England. Drugs are ordered along the x-axis by their position in the BNF and colour-coded by chapter. Drug names and p values are not shown here but can be viewed on the interactive online version. A similar plot showing p values can be found in online supplemental figure S1. BNF, British National Formulary.
Individual drugs and drug classes
Of 206 included drug classes, 64 (31%) had an adjusted ρ between 0.2 and 0.4, and a further 30 (15%) had an adjusted ρ greater than 0.4. Only four drug classes (2%) had an adjusted correlation coefficient less than −0.2, with just one being moderately correlated with affluence (ρ<−0.4 online supplemental figures S2 and S3). Similar results were seen with individual drugs: of the 774 included drugs, 137 (18%) had an adjusted ρ between 0.2 and 0.4, and a further 31 (4%) had an adjusted ρ greater than 0.4, suggesting a moderate correlation between prescribing rate and IMD score. Conversely, only 10 drugs (1.2%) had an adjusted correlation coefficient less than −0.2, with only one drug having a correlation coefficient less than −0.4 (ie, with prescribing rates that were higher in more affluent areas).
The drug classes most strongly positively correlated with socioeconomic deprivation are shown in table 1. These included both non-opioid and opioid analgesics (adjusted ρ=0.67 and 0.57, respectively), adrenoreceptor agonists (ρ=0.64, such as salbutamol) and drugs used to treat epilepsy (ρ=0.56). Listing drugs individually, tramadol (ρ=0.59) and gabapentin (ρ=0.57) were most correlated with deprivation (online supplemental table S1). Indeed, five of the top ten drugs most associated with deprivation were analgesics.
The 10 drug classes with prescribing rates most positively correlated with Index of Multiple Deprivation score
Drug classes that were most strongly negatively correlated with deprivation (ie, that were prescribed at higher rates in more affluent areas) were those relating to allergic emergencies (ρ=−0.49), combined hormonal contraceptives (ρ=−0.33) and female sex hormones (hormone replacement therapy (HRT), ρ=−0.24, table 2). At the individual drug level, epinephrine prescribing was most strongly associated with affluence (ρ=−0.48, online supplemental table S2). Of the top 10 drugs most correlated with affluence, six were female reproductive hormones used for contraception, menstrual bleeding disorders and HRT, and three were travel-related vaccines.
The 10 drug classes with prescribing rates most negatively correlated with Index of Multiple Deprivation score
A complete list of all studied drugs and drug classes, along with the unadjusted and adjusted correlation coefficients and p values, is given in online supplemental tables S3 and S4, respectively.
British National Formulary chapters
The BNF chapter with the drug classes having the greatest association with deprivation was that relating to the central nervous system (figure 2). Other than opioid analgesics and drugs used to control epilepsy (which include gabapentin and pregabalin, both of which are also used in the management of chronic pain), this section includes antipsychotics (ρ=0.49), selective serotonin reuptake inhibitors (such as fluoxetine, ρ=0.42) and tricyclic antidepressants (ρ=0.47).
Correlation coefficients of central nervous system drug classes and their association with Index of Multiple Deprivation score. Drug classes are given along the x-axis in their order in the British National Formulary (chapter 4). Bubble area is proportional to the number of prescriptions dispensed in primary care in 2019. Statistically significant associations are signified by red bubbles. A positive correlation coefficient reflects a tendency towards greater prescribing in more socioeconomically deprived practices. Some bubbles are annotated. All drug class names, number of items, correlation coefficients and p values can be viewed on the interactive online version. (SSRI: selective serotonin reuptake inhibitor)
Another BNF chapter with a number of drug classes strongly correlated with deprivation was that relating to the respiratory system (figure 3). All but one of these classes are prescribed in greater amounts in more deprived practices, including mucolytics (ρ=0.47), theophylline (ρ=0.42) and antihistamines (ρ=0.38).
Correlation coefficients of respiratory system drug classes and their association with Index of Multiple Deprivation score. Drug classes are given along the x-axis in their order in the British National Formulary (chapter 3). Bubble area is proportional to the number of prescriptions dispensed in primary care in 2019. Statistically significant associations are signified by red bubbles. A positive correlation coefficient reflects a tendency towards greater prescribing in more socioeconomically deprived practices. Some bubbles are annotated. All drug class names, number of items, correlation coefficients and p values can be viewed on the interactive online version.
In addition to the top ten list given in table 1, other drug classes moderately positively correlated with deprivation include proton pump inhibitors (ρ=0.44), insulin (ρ=0.45), other antidiabetic drugs (ρ=0.44), and antiplatelet agents (ρ=0.40).
A complete set of figures showing the correlation coefficients and significance of drug classes for each BNF chapter is given in online supplemental figure S4–S17.
Discussion
We explored the correlation between socioeconomic deprivation and the rate of prescribing of individual pharmaceutical drugs, and drug classes, in primary care in England, to identify prescribing inequalities that would require further investigation. Medication prescriptions for most individual drugs and drug classes increased with increasing deprivation. After adjusting for practice list size and age and sex make-up, this positive correlation was at least moderate for 31/774 (4%) of individual drugs and 30/206 (15%) of drug classes. Only one drug and one drug class were moderately negatively correlated with deprivation (prescribed more frequently in more affluent settings), although there were many other drugs with weaker statistically significant negative correlations. In seeking an explanation for these findings, further work would require an exploration of a combination of societal factors (including the wider determinants of health, such as education and employment), healthcare system factors (such as prescribing habits and staffing adequacy) and patient-level factors (such as differences in the prevalence of chronic disease).
Comparison to previous literature
Some of the associations found in this study align with previous research, and other findings are novel. The drug categories that are most strongly correlated with deprivation include analgesics, both opioids—such as tramadol and co-codamol—and non-opioids, including non-steroidal anti-inflammatory drugs, gabapentin and pregabalin. An association between opioid prescribing and deprivation in England has previously been described,3 although that study did not adjust for covariates. A recent review into medications associated with dependence by Public Health England also found that gabapentinoid and opioid use was associated with deprivation and that coprescribing of both drug classes was 1.4 times more common in the most deprived practices.10 The prevalence of chronic pain has been shown to be correlated with low socioeconomic status, for both musculoskeletal11 and neuropathic pain.12 Furthermore, deprivation is associated with poor outcomes following pain management.13 Taken together, it is possible that the increased prescribing rate of analgesics in deprived practices reflects the underlying prevalence and duration of chronic pain. However, further work is needed to exclude other potential contributors, such as lack of access to drug dependency services and the prevalence of complex mental health problems.
There are many other health conditions whose prevalence varies with deprivation. This underlying variation may partially explain differential levels of prescribing. For example, epilepsy drugs (including those other than gabapentinoids) were found to have one of the strongest correlations with deprivation in this study. Epilepsy is known to have a higher incidence in more deprived areas.14 Cardiovascular disease, asthma, chronic obstructive pulmonary disease (COPD) and diabetes have a well-characterised association with deprivation in terms of prevalence and severity.15–20 Social and behavioural factors play a key role in the pathogenesis and severity of these conditions.21 We found that drugs used to manage these chronic diseases are prescribed in higher rates in the most deprived areas, including nitrates, antiplatelets, lipid-lowering medications, inhalers, theophylline and antidiabetic drugs such as metformin and insulin.
A simple explanation for these findings would be that prescribing rates are purely a reflection of underlying disease burden and thus of healthcare need. However, there is a possibility that inequalities in prescribing or wider management exist in addition to inequalities in health status. Indeed, we previously found that higher rates of antipsychotics were prescribed in more deprived areas after adjusting for the prevalence of psychosis, dementia and depression, as well as for demographic factors.22 In that case, the higher prescribing may reflect a lack of alternative non-pharmacological management options in areas of socioeconomic deprivation, such as psychological therapy and social support, although this is yet to be investigated. In another study of asthma, we found that rates of salbutamol inhaler prescribing were higher in more deprived areas after adjusting for the prevalence of asthma, COPD and smoking.23 High salbutamol use is related to inadequate day-to-day management of asthma, and further investigation found that practices with higher salbutamol prescribing were also less likely to provide patients who had asthma with regular medication reviews. These examples suggest that prescribing inequalities require investigation as they may highlight wider inequalities that need to be addressed beyond the mere prevalence or severity of a disease. Possibilities include issues with health literacy, inadequate provision for those using languages other than English, subcultural preferences for alternative therapies and difficulties in attending appointments.
Interestingly, prescribing in areas of women’s health is significantly correlated with deprivation, with both HRT and combined oral contraception being prescribed more in affluent areas. A study from 2009 suggested that women from lower and middle social classes were less likely to use contraception in general24 and this may still be the case. We recently described the finding of increased HRT prescribing in areas of affluence, where this association remained after adjusting for additional factors that may influence prescribing, such as smoking, obesity and cardiovascular disease.25 While contraception can be obtained from alternative services, such as community genitourinary medicine clinics and pharmacies (which are not represented in the analysed data), our findings suggest the possibility of a gap of provision, or in accessing provision, of women’s healthcare services.
Other novel associations found in this study include the higher rate of epinephrine and travel-related vaccine prescribing in more affluent practices, and the higher rate of antihistamine, antacid and topical dermatological agents and of vitamin supplementation, in more deprived practices. In the case of epinephrine, which is often prescribed in the form of auto-injectors for those at risk of a severe allergic reaction, it is possible that individuals in more affluent areas are more likely to seek medical attention for allergies. The higher prescribing rates of certain upper gastrointestinal medications in areas of deprivation, along with vitamin supplements, may be influenced by differences in diet or in health-seeking behaviours or may reflect an inability for individuals to afford over-the-counter options. Such hypotheses will need to be the focus of further investigation.
Strengths, limitations and further work
This work examined prescribing in all NHS primary care practices in England, as captured by NHS Digital. As such, it has used a robust and unbiased data source for reviewing prescribing associations. The IMD score was recently updated (in 2019), so the deprivation scores assigned to individual practices are an up-to-date marker of their population deprivation status. A hypothesis-free approach was taken to examining the associations between deprivation and prescribing, and a conservative p value threshold to account for multiple testing was used to minimise the risk of reporting falsely significant correlations.
There are limitations to the current study. First, data are from 1 year only, and prescription rates and deprivation were aggregated at the primary care practice level, so it is not known which individuals were given specific prescriptions, nor how many or for what reasons. For example, some epilepsy drugs may be prescribed for epilepsy or the management of pain, and it is impossible to identify the indications from aggregate data. Second, the IMD score is used as a surrogate marker of deprivation but is not a direct measure. Finally, this study aimed to identify which drugs are potentially significantly associated with deprivation. It is important to acknowledge these as preliminary findings that need to be used as the foundation for further work, specifically examining prescribing at the individual patient level, adjusting for underlying clinical and demographic factors to confirm associations, and they may not apply to healthcare settings outside of England. Where these exist, investigation is required into the reasons for prescribing inequalities, to identify where corrective intervention is required in order to improve outcomes for patients.
Conclusion
Socioeconomic deprivation is correlated with higher rates of prescriptions for a large number of drugs, though a few are also correlated with affluency. The reason behind many of these associations is unclear and may be explained through a combination of differences in disease prevalence rates, in the wider determinants of health including behavioural factors, in access to health and social care, and in the prescribing and management practices in different settings. More research is required into each of these associations to confirm the findings, to identify the underlying reasons for the observed inequality between communities and to formulate appropriate interventions to ensure that every individual receives the level of care that they require.
Main messages
We identify novel associations between prescribing and deprivation.
Drugs prescribed at higher rates in more deprived practices include non-opioid analgesics and salbutamol.
Contraceptives and travel vaccination prescriptions are associated with affluence
Current research question
To identify drugs which are prescribed in higher or lower amounts in more socioeconomically deprived primary care practices
Prescribing rates and deprivation measures were analysed for primary care practices across England
What is already known on the subject
Prescribing rates of many medications are associated with socioeconomic deprivation.
Benzodiazepines and opioids have been found to be prescribed in greater amounts in areas of deprivation
Data availability statement
Data are available in a public, open access repository. The data sources used in this study are publicly available and have been referenced within the manuscript. The analysis code is available on the project GitHub page: https://github.com/sirsazofduck/2020MooneyJ.
Ethics statements
Patient consent for publication
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Twitter @sirsazofduck
Contributors JM, FK, SH, DT and SS conceptualised the study and finalised the protocol. JM, RY, AE and SS undertook data curation and analysis. JM and SS drafted the manuscript. RY, HM, FK, AE, SH and DT critically revised the manuscript. SS is the guarantor.
Funding FK is supported by a National Institute of Health Research (NIHR) Academic Clinical Fellowship, and SS by an NIHR Clinical Lectureship. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.