EQUITY-MS: evaluating the quality of investigation and treatment of multiple sclerosis using data from five specialist neurology centres in the UK—study protocol Authors: Elizabeth O’Nions, Dan Lewer, Rachael W Cheung, Ellen Da Silva, Joyutpal Das, Ruth Dobson, Radoš Keravica, Faith Moyo, David Rog, Melanie Russell, Cord Spilker, Emma Tallantyre, Helen L Ford Published: 2026 ABSTRACT AbstractIntroductionThe social determinants of health are the conditions in which people live and the systems around them that shape these conditions. Social determinants of health may affect diagnosis and treatment for people with multiple sclerosis (MS) even in countries with universal healthcare systems. This study will investigate whether a person’s age, sex and/or gender, living in a more socioeconomically deprived area and being of a minoritised ethnicity impact access to a diagnosis of MS and disease-modifying treatments (DMTs) for people living in the catchments of specialist neurology centres with a total estimated catchment population of approximately 7 million people in England and Wales.Methods and analysisEQUITY-MS is a multicentre retrospective cohort study using routinely collected healthcare data. Study participants are people aged 16 and above who received a new diagnosis of MS between 1 January 2018 and 31 December 2024 while living within the catchment of five specialist neurology centres (located in Leeds, Bradford, Greater Manchester, Cardiff and East London). Data will be collected from patients’ hospital records by clinical teams. The primary outcome is the length of time between MS diagnosis and prescription of any DMT. Secondary outcomes are prescriptions of a high-efficacy DMT and the duration between patient-reported symptom onset and MS diagnosis. We plan to adjust for clinical and demographic factors that could impact prescribing decisions.Ethics and disseminationThe study was approved by the North West–Greater Manchester East Research Ethics Committee (National Health Service) (25/NW/0184) on 14 August 2025. Results will be published in peer-reviewed journals, and summaries will be provided to local MS societies and disseminated via the study website (https://bradfordresearch.nhs.uk/bradford-centre-for-health-data-science/equity-ms/). STRENGTHS AND LIMITATIONS OF THIS STUDY The study will include comprehensive measures, including baseline disability and disease activity (multiple sclerosis relapse rate), which will improve control of potentially confounding variables. The study will include comprehensive measures, including baseline disability and disease activity (multiple sclerosis relapse rate), which will improve control of potentially confounding variables. Inclusion of consecutively diagnosed participants will minimise selection bias. Inclusion of consecutively diagnosed participants will minimise selection bias. The findings may not generalise to other geographical regions, partly because the source populations and clinical teams participating in this study may be different from those elsewhere. The findings may not generalise to other geographical regions, partly because the source populations and clinical teams participating in this study may be different from those elsewhere. Analyses aiming to estimate inequalities experienced by people of minoritised ethnicities will group multiple different ethnic groups together due to sample size constraints, so quantitative estimates will not be generalisable to specific ethnicity groups. Analyses aiming to estimate inequalities experienced by people of minoritised ethnicities will group multiple different ethnic groups together due to sample size constraints, so quantitative estimates will not be generalisable to specific ethnicity groups. The study will account for the COVID-19 pandemic period during which services operated unusually, and disease-modifying treatment prescribing patterns may have changed based on guidance. The study will account for the COVID-19 pandemic period during which services operated unusually, and disease-modifying treatment prescribing patterns may have changed based on guidance. Introduction Multiple sclerosis (MS) is a chronic immune-mediated inflammatory neurodegenerative disorder of the central nervous system. Globally, it is the most common cause of non-traumatic disability in young adults.1 Analysis of routinely collected health data suggests there are currently over 150 000 people with MS in the UK—around 1 in 400.2 Around 7100 people are diagnosed annually in the UK. Most people are diagnosed with MS in their 30s and 40s, and women are around 2.5 times as likely to be diagnosed as men.2 As a precipitant to diagnosis, people often experience symptoms associated with a ‘relapse’: an acute attack of the disease producing symptoms lasting days to months.3 4 Common symptoms include new-onset visual impairments (eg, unilateral blurred vision or double vision), weakness, muscle spasms, sensory problems, pain, poor coordination, gait problems and bowel and bladder dysfunction.5 6 Although there is no cure for MS, disease-modifying treatments (DMTs) developed over the past 25 years reduce relapses and accumulation of disability.7 8 Earlier initiation of DMTs can reduce accumulation of disability and all-cause mortality.9 In guidance published in 2018 and updated in 2023, National Health Service (NHS) England recommends that, after a single clinical episode meeting the McDonald criteria for MS, a moderate-efficacy medication is prescribed unless ‘clinical or radiological markers indicate a poor prognosis for rapidly developing permanent disability’.10 However, evidence indicates that initiation of treatment with newer high-efficacy DMTs delays accumulation of disability.11–13 Therefore, there is an evolving trend towards earlier use of high-efficacy DMTs, although they can have more burdensome monitoring requirements or complex safety profiles, which require careful consideration by healthcare professionals and patients. Growing evidence indicates that social determinants of health, the conditions in which people live, work and age, and the systems that surround them14 impact healthcare experiences and outcomes for patients with MS.15 Some previous work suggests that people with MS from more deprived socioeconomic backgrounds may experience diagnostic delays,16 poorer access to DMTs or second-line treatments,17 18 faster disease progression19–21 and premature mortality.22 23 Regarding DMT initiation, a 2022 UK study based on data from 1449 participants from the UK MS Register initially diagnosed with relapsing-remitting MS between 2010 and 2017 found that patients from the most socioeconomically deprived areas had 0.69 times the odds of receiving a DMT relative to the other groups combined, adjusting for age, duration of disease, educational attainment, employment status and occupation.24 Another study in the UK MS Register including patients diagnosed between 2008 and 2021 found that university education was associated with an approximately 1.5-fold increase in the odds of receiving high-efficacy DMTs adjusting for year of diagnosis, age at diagnosis, health-related quality of life and sex.25 A single-centre study based in Greater Manchester, UK, including 1648 people newly diagnosed with relapsing-remitting MS between 2010 and 2020 found that patients living in areas in the most socioeconomically deprived quintile nationally had 0.66 times the odds of being offered any DMT adjusted for ethnicity, sex, age at diagnosis and year of diagnosis.17 The probability of being offered a DMT did not vary significantly by ethnicity in analyses adjusted for deprivation, but people with MS of minoritised ethnicities had 0.60 times the odds of starting a DMT.17 This analysis did not adjust for differences in disease activity and baseline disability at the point of diagnosis. Patients from less advantaged sociodemographic groups may have more severe disability or more active MS at the point of diagnosis, potentially due to delayed diagnosis or more severe disease due to greater exposure to risk factors (eg, smoking). If this is not taken into account, the impact of inequalities may be underestimated. A further aim of the planned study is to identify whether diagnostic delays (time between disease onset and diagnosis) are disproportionately experienced by males, patients in age groups where MS onset is less common, patients living in socioeconomically disadvantaged areas and patients of minoritised ethnicities. Previous work suggests that both younger26–30 and older patients,26 31 32 those with more comorbidities,27 33 those experiencing socioeconomic disadvantage,31 those with lower educational levels,31 34 or whose parents had lower education levels,16 and those living in rural areas34 are more likely to experience diagnostic delays, with inconsistent results for sex and/or gender.27 31 34 Previous work using data from the UK MS Register found evidence for both earlier age at onset and shorter diagnostic delays among minoritised patients.35 In EQUITY-MS, we will use data from a cohort of patients newly diagnosed with MS between 2018 and 2024 to examine whether patients from different sociodemographic groups (a) are less likely to receive any DMT (vs no DMT); (b) are less likely to receive a high-efficacy DMT (vs a moderate-efficacy DMT or no DMT); and (c) experience longer diagnostic delays, meaning that a larger number of patients within a clinic’s catchment may be undiagnosed. We chose to start the study in 2018 because changes to the diagnostic algorithm for MS were made in 2017, which allowed for a diagnosis to be made after a single clinical episode (ie, relapse),36 provided that investigations showed dissemination in time and space. Our primary exposures of interest will be socioeconomic deprivation, measured using indices of income and employment available for both England and Wales,37 and ethnicity. Secondary exposures will be age and sex. This study forms part of a larger research programme funded by the Horne Family Foundation that will also include qualitative studies investigating diagnostic journeys of people with MS, their experiences of MS care and treatment and healthcare professionals’ views on barriers to accessing treatments, plus investigation of health inequalities related to sociodemographic factors in patients with MS using national primary care records data. Methods and analysis Study design EQUITY-MS is a retrospective cohort study of routinely collected healthcare data, which aims to quantify the extent to which patients who live in more socioeconomically deprived areas and patients of minoritised ethnicities experience inequalities in access to DMTs. The analysis plan is based on the following principles: Clinical factors occurring before diagnosis, such as differences in symptom severity, diagnostic results, comorbidities and patient age, are considered as confounders in our primary analysis because they are not within the control of MS treatment services. We will adjust for these clinical factors statistically in our models. This means that we will estimate inequalities in access to treatment within MS services rather than across the broader healthcare system. Clinical factors occurring before diagnosis, such as differences in symptom severity, diagnostic results, comorbidities and patient age, are considered as confounders in our primary analysis because they are not within the control of MS treatment services. We will adjust for these clinical factors statistically in our models. This means that we will estimate inequalities in access to treatment within MS services rather than across the broader healthcare system. Mechanisms that cause inequalities such as the ability to travel to a physical clinic are potentially mediators of inequalities related to local area deprivation, that is, people from more deprived areas might be less likely to get DMTs because they live further from a clinic or in an area with poor public transport. We plan to collect travel time to the clinic but we will not adjust for it in our primary analysis. Mechanisms that cause inequalities such as the ability to travel to a physical clinic are potentially mediators of inequalities related to local area deprivation, that is, people from more deprived areas might be less likely to get DMTs because they live further from a clinic or in an area with poor public transport. We plan to collect travel time to the clinic but we will not adjust for it in our primary analysis. Local area deprivation is a potential mediator of inequalities related to minoritised ethnicity, given that structural racism increases the chances that minoritised groups live in more deprived areas. Socioeconomic deprivation might present access barriers38 such as not having access to a car; greater difficulty taking time off work or away from caring responsibilities to access infusions; or not being able to receive medication deliveries at home due to long working hours. We will provide descriptive data on travel time and local deprivation by ethnicity to support our discussion of inequalities. Local area deprivation is a potential mediator of inequalities related to minoritised ethnicity, given that structural racism increases the chances that minoritised groups live in more deprived areas. Socioeconomic deprivation might present access barriers38 such as not having access to a car; greater difficulty taking time off work or away from caring responsibilities to access infusions; or not being able to receive medication deliveries at home due to long working hours. We will provide descriptive data on travel time and local deprivation by ethnicity to support our discussion of inequalities. Our analysis assumes that there are common mechanisms across geographical areas through which personal characteristics affect healthcare access, including structural racism, sexism, classism and ageism. There are plausible reasons why inequalities might act differently in different geographical subregions. We are estimating the average effect across all regions sampled within our centre catchments (Leeds, Bradford, Cardiff, East London, Greater Manchester). Our analysis assumes that there are common mechanisms across geographical areas through which personal characteristics affect healthcare access, including structural racism, sexism, classism and ageism. There are plausible reasons why inequalities might act differently in different geographical subregions. We are estimating the average effect across all regions sampled within our centre catchments (Leeds, Bradford, Cardiff, East London, Greater Manchester). The probability of DMT prescription irrespective of patient characteristics could vary across treatment centres, as could demographic characteristics of the population living within catchments; hence, the treatment centre is a potential confounder. The treatment centre that patients attend to could also be a driver of inequality, for example, if people from less advantaged backgrounds are more likely to live in areas with lower rates of DMT provision. This type of inequality is not measurable in our study because of (a) the relatively small number of centres included and (b) that we cannot fully account for prediagnostic factors. For example, delayed diagnosis increases the chances that patients are no longer eligible for DMTs due to disability accumulation. The probability of DMT prescription irrespective of patient characteristics could vary across treatment centres, as could demographic characteristics of the population living within catchments; hence, the treatment centre is a potential confounder. The treatment centre that patients attend to could also be a driver of inequality, for example, if people from less advantaged backgrounds are more likely to live in areas with lower rates of DMT provision. This type of inequality is not measurable in our study because of (a) the relatively small number of centres included and (b) that we cannot fully account for prediagnostic factors. For example, delayed diagnosis increases the chances that patients are no longer eligible for DMTs due to disability accumulation. Selection of sample We will use patient-level data from five specialist MS services (Leeds General Infirmary, Bradford Royal Infirmary, University Hospital of Wales Cardiff, Barts Health London, Manchester Centre for Clinical Neurosciences), collated for the purposes of this study and deidentified prior to use. Each MS centre is based at an NHS hospital and provides treatment free of charge following referral from another healthcare setting. The rationale for collecting a new dataset is that currently there is no centralised repository for diagnostic, clinical and prescribing information across neurology clinics in the UK for consecutively referred or randomly sampled patients. Since this study focuses on inequalities, a representative sample is crucial, and existing datasets are often biased towards certain patient groups or only include patients prescribed DMTs. Eligibility and exclusion criteria We will include consecutively diagnosed patients who: Were newly diagnosed with MS between 1 January 2018 and 31 December 2024. Were newly diagnosed with MS between 1 January 2018 and 31 December 2024. Lived within the catchment area of one of the five MS centres at the time of their MS diagnosis. Lived within the catchment area of one of the five MS centres at the time of their MS diagnosis. Were aged 16 years or older at the time of their diagnosis. Were aged 16 years or older at the time of their diagnosis. Had no clear contraindications to receiving a DMT at the time of diagnosis. Had no clear contraindications to receiving a DMT at the time of diagnosis. We will include patients who were diagnosed privately while residing within clinic catchment but were then referred to the participating MS clinic to receive treatment. We will exclude patients who: Lived outside the clinic’s catchment area at the time of their diagnosis (eg, if they were referred into the service for a second opinion or moved to the area after their diagnosis). Lived outside the clinic’s catchment area at the time of their diagnosis (eg, if they were referred into the service for a second opinion or moved to the area after their diagnosis). Had missing data on the date of diagnosis or whether they received a DMT (centres will indicate whether exclusion criteria are met). Had missing data on the date of diagnosis or whether they received a DMT (centres will indicate whether exclusion criteria are met). Each participating MS centre will identify eligible participants seen within their specialist MS clinics according to local practice. Data collation will be performed by local teams, and data will be entered into an Excel spreadsheet. Steps are as follows: Identify eligible patients: clinical teams will identify all eligible patients meeting study inclusion criteria. They will generate a list of patients retaining a unique patient identifier, for example, NHS number or local patient number. Identify eligible patients: clinical teams will identify all eligible patients meeting study inclusion criteria. They will generate a list of patients retaining a unique patient identifier, for example, NHS number or local patient number. Extract research variables: clinical teams will extract research variables from clinical records or local databases. Extract research variables: clinical teams will extract research variables from clinical records or local databases. Create derived variables: clinical teams will generate derived variables—pseudonymous study ID; whether the patient lived in the clinic’s catchment area at the time of diagnosis (to determine eligibility); the decile of the English Index of Multiple Deprivation 2019 (IMD)39 or Welsh Index of Multiple Deprivation (WIMD)40 based on home address at the time of diagnosis; deciles based on income and employment domain scores available for both England and Wales based on home address at the time of diagnosis37; distance between home address and clinic by straight line and transport duration (car and public transport); and month/year of birth, month/year of death. Create derived variables: clinical teams will generate derived variables—pseudonymous study ID; whether the patient lived in the clinic’s catchment area at the time of diagnosis (to determine eligibility); the decile of the English Index of Multiple Deprivation 2019 (IMD)39 or Welsh Index of Multiple Deprivation (WIMD)40 based on home address at the time of diagnosis; deciles based on income and employment domain scores available for both England and Wales based on home address at the time of diagnosis37; distance between home address and clinic by straight line and transport duration (car and public transport); and month/year of birth, month/year of death. Create a deidentified dataset: clinicians at each site will delete the unique patient identifier and other identifiers, including local patient number, name, address, full date of birth and full date of death. They will only share a minimised dataset with the central research team. Create a deidentified dataset: clinicians at each site will delete the unique patient identifier and other identifiers, including local patient number, name, address, full date of birth and full date of death. They will only share a minimised dataset with the central research team. Securely share data with the research team: data will be sent to the central research team at Bradford Institute for Health Research (Bradford Teaching Hospitals NHS Foundation Trust). Securely share data with the research team: data will be sent to the central research team at Bradford Institute for Health Research (Bradford Teaching Hospitals NHS Foundation Trust). Delete research datasets at the end of the study: analyses of data will begin during Autumn 2026. Data will be deleted when analyses are complete, which is anticipated by 31 December 2028. Delete research datasets at the end of the study: analyses of data will begin during Autumn 2026. Data will be deleted when analyses are complete, which is anticipated by 31 December 2028. Variables Online supplemental eTable 1 lists variables that will be collected for the purposes of the study. Where the format of variables or level of granularity differ between centres, we will clean the data to ensure consistency. Exposure Our primary exposure variables are ethnicity (White British vs minoritised ethnicities) and local area deprivation proxied using income and employment decile.37 When extracting data from clinical records, where possible, centres will categorise ethnicity based on the Office for National Statistics Ethnic group classification 8a, resulting in seven ethnicity subgroups.41 Where this granularity is not available based on clinical records, we will instead adopt classification 6a, resulting in five ethnic groups.41 Classification 8a is preferred as this will allow us to disaggregate White ethnic groups into White British and other White ethnicities. Our primary analysis in relation to ethnicity will compare patients of White British ethnicity to those of other or minoritised ethnicities, which will measure the effect of being in a minoritised group. We will also report results for more detailed ethnicity categories with the limitation that the precision of these results is likely to be low. To quantify local area deprivation, we will use local area income employment decile, which reflects average income and employment rates for Lower layer Super Output Areas (LSOAs), which are small, consistent geographic units across England and Wales.42 LSOAs comprise between 400 and 1200 households and approximately 1000–3000 residents. We will use income employment decile rather than raw score. Income employment decile is used rather than IMD decile because the latter is constructed separately for England and Wales and therefore cannot be harmonised. IMD and WIMD will also be collected and reported descriptively. Our secondary exposures are age band and sex. Outcome The primary outcome is receipt of a DMT (one of the following: alemtuzumab (Lemtrada), Avonex (interferon beta 1a), cladribine (Mavenclad), dimethyl fumarate (Tecfidera), diroximel fumarate (Vumerity), fingolimod (Gilenya), glatiramer acetate (Copaxone), natalizumab (Tysabri), ocrelizumab (Ocrevus), ofatumumab (Kesimpta), ozanimod (Zeposia), Plegridy (peginterferon beta 1a), ponesimod (Ponvory), Rebif (beta interferon-1a), siponimod (Mayzent), teriflunomide (Aubagio), autologous haematopoietic stem cell transplantation (AHSCT), ublituximab (Briumvi)). The secondary outcome is receipt of a high-efficacy DMT (one of the following: alemtuzumab (Lemtrada), cladribine (Mavenclad), natalizumab (Tysabri), ocrelizumab (Ocrevus), ofatumumab (Kesimpta), AHSCT, ublituximab (Briumvi)). Variables to control potential confounding Covariates will be the other sociodemographic variables of interest, MS centre of treatment, type of MS diagnosis, baseline disability measured by the Expanded Disability Status Scale43 (if available), relapse rate over the 2 years prior to diagnosis (if available), presence of relapses requiring hospitalisation over the 2 years prior to diagnosis (if available), time from onset to diagnosis (if available) and calendar year. These variables have been identified as potential confounders of the relationship between sociodemographic factors and receipt of a DMT, and therefore all will be included as covariates. Missing data Variables where we anticipate potentially high levels of missingness include MS relapse rate, baseline disability and initial disease onset. There could be non-random factors that affect missingness (eg, a person with milder relapses may have less relapse data). Similarly, the date of disease onset may be missing where MS had a more gradual or less clearly defined onset. We plan to examine patterns of missingness between centres and consult clinical leads about likely determinants of missingness. We plan to use multiple imputation to manage missing data and will evaluate the plausibility of the missing-at-random assumption. Sample size calculation Our expected sample size is 2595 patients (table 1). We conducted a power analysis by simulating our primary study, which will be a survival analysis of the time between diagnosis and receipt of a DMT (https://github.com/equityms/power-analysis-MS-centres). We used the following assumptions: (1) numbers of new patients seen in each of the five clinics between 2018 and 2024 based on estimates from clinical leads and (2) the proportion of patients newly diagnosed with MS during the time period who are prescribed DMTs. Expected sample sizes stratified by socioeconomic status (SES) quintile (based on income and employment domain scores37) and ethnicity (White British and minoritised ethnicities) are provided in table 1. Our simulation suggested that we have 80% power to detect a risk ratio of 0.91 (ie, a 9% lower probability of DMT receipt) among the lowest versus the highest SES quintile, and a risk ratio of 0.92 (ie, an 8% lower probability of DMT receipt) among patients of minoritised ethnicities versus White British patients. Assuming 15% missing data (uniformly across groups), the detectable risk ratio reduced to 0.90 for SES and 0.91 for ethnicity. We will also estimate the effect of ethnicity defined using the more detailed variable—mixed/multiple/other ethnicities; Asian/Asian British; Black/African/Caribbean/Black British—to identify larger inequalities that may disproportionately affect particular ethnic groups when compared with White British patients. Table 1Expected numbers of participants per groupBradfordn (%)Leedsn (%)East Londonn (%)Cardiffn (%)Manchestern (%)SES income employment quintiles 1 (most deprived)49 (41.1)126 (25.1)80 (15.9)183 (34.9)347 (36.5) 220 (16.5)77 (15.3)173 (34.5)134 (25.6)173 (18.2) 324 (19.6)83 (16.5)96 (19.1)67 (12.8)137 (14.4) 414 (11.5)102 (20.3)21 (4.2)67 (12.7)134 (14.1) 5 (least deprived)13 (11.2)114 (22.8)132 (26.3)74 (14.0)160 (16.8)Ethnicity White British80 (66)400 (80)350 (70)500 (95)850 (89) Minoritised40 (33)100 (20)150 (30)25 (5)100 (11)Total120500500525950SES estimated from distribution of postcodes within catchment, assuming equal prevalence of MS across groups.MS, multiple sclerosis; SES, socioeconomic status. Expected numbers of participants per group SES estimated from distribution of postcodes within catchment, assuming equal prevalence of MS across groups. MS, multiple sclerosis; SES, socioeconomic status. Statistical analysis plan Differences in time to receiving a DMT We will use regression models to estimate the effect of sociodemographic variables on the time to receiving a DMT using Cox proportional hazards regression. In each analysis, we will estimate the effect of each sociodemographic variable while adjusting for potential confounding variables, including a fixed effect for the treatment centre. There are various possible analytical approaches to estimating this effect. In this section, we will discuss the relative strengths of each approach and the reasons for our chosen approach. A two-stage approach, where fixed effect models are estimated for each centre and the results are pooled across centres using meta-analysis. This approach means that all model parameters are estimated separately for each centre and their effects on the outcome can vary. The key limitation of this approach is that there is no reason to assume that potential confounders such as calendar year impact DMT prescribing differently between centres. This method would therefore reduce the power to adjust for these confounders. A two-stage approach, where fixed effect models are estimated for each centre and the results are pooled across centres using meta-analysis. This approach means that all model parameters are estimated separately for each centre and their effects on the outcome can vary. The key limitation of this approach is that there is no reason to assume that potential confounders such as calendar year impact DMT prescribing differently between centres. This method would therefore reduce the power to adjust for these confounders. A one-stage approach where data from all centres are pooled and included in a single regression model, and the effect of patient-level variables (including the primary exposures) is assumed to be fixed across centres. Although we could allow the baseline time to DMT receipt to vary by centre, the impact of sociodemographic variables on the rate of DMT receipt would be estimated as a single value across centres. This could be implemented in two ways:A hierarchical (random intercept) or shared frailty model where separate intercepts (reflecting baseline prescribing rates) are modelled for each centre.Including the MS centre as a categorical fixed effect. A one-stage approach where data from all centres are pooled and included in a single regression model, and the effect of patient-level variables (including the primary exposures) is assumed to be fixed across centres. Although we could allow the baseline time to DMT receipt to vary by centre, the impact of sociodemographic variables on the rate of DMT receipt would be estimated as a single value across centres. This could be implemented in two ways: A hierarchical (random intercept) or shared frailty model where separate intercepts (reflecting baseline prescribing rates) are modelled for each centre. A hierarchical (random intercept) or shared frailty model where separate intercepts (reflecting baseline prescribing rates) are modelled for each centre. Including the MS centre as a categorical fixed effect. Including the MS centre as a categorical fixed effect. A one-stage approach allowing the effect of exposure variables to vary across centres. This means that the estimated time taken to receive a DMT and the effect of sociodemographic variables can both vary by centre, but the effects of confounders are fixed. This could be implemented by:Estimating both random intercepts (for MS centre) and random slopes (for sociodemographic variables on MS centre) in a hierarchical model.Using a non-hierarchical model that includes an interaction between centre and sociodemographic variables. A one-stage approach allowing the effect of exposure variables to vary across centres. This means that the estimated time taken to receive a DMT and the effect of sociodemographic variables can both vary by centre, but the effects of confounders are fixed. This could be implemented by: Estimating both random intercepts (for MS centre) and random slopes (for sociodemographic variables on MS centre) in a hierarchical model. Estimating both random intercepts (for MS centre) and random slopes (for sociodemographic variables on MS centre) in a hierarchical model. Using a non-hierarchical model that includes an interaction between centre and sociodemographic variables. Using a non-hierarchical model that includes an interaction between centre and sociodemographic variables. A final option is to pool data for all participants without modelling any centre-related baseline differences or centre-specific effects of key exposure variables. Models 1–3 include centre as a confounder, meaning that if people of one sociodemographic group are more likely to live in an area/centre where prescribing of DMTs is generally less common, associated differences in the time to DMT receipt will be ‘adjusted out’ in our analysis. Not including centre as a confounder would mean that centre-level differences could drive apparent sociodemographic differences. Models 3a and 3b would allow the effect of the exposure (eg, minoritised ethnicity) to vary across centres. There are plausible reasons why the effect of personal characteristics might differ across centres; for example, (a) in areas of greater ethnic density (such as the Pakistani community in Bradford), minority ethnic groups might have improved healthcare access due to mechanisms within these communities to mitigate the impacts of structural racism,38 and (b) some services may have greater investments in accessibility for underserved patient groups. However, there are also many common processes across centres, such as structural racism, material deprivation, social capital, etc, and our aim is to estimate the overall average effect of personal sociodemographic characteristics on healthcare access, rather than understand if/how these effects differ across centres. We have therefore chosen to estimate the fixed effect of personal sociodemographic characteristics across the study. Therefore, we have decided to use model 2b only to estimate our effects. We will use survival analysis to estimate the time to receipt of a DMT. Outcomes are (1) receipt of any DMT (vs no DMT) and (2) receipt of a high-efficacy DMT (vs any other DMT or no DMT). Estimating the potential number of undiagnosed patients with MS within clinic catchments We will use microsimulation to estimate the number of undiagnosed people in different community subgroups within clinic catchments, inferred from the length of diagnostic delays that patients report. This will involve the following: Describing the distribution of ‘delays’ between onset and diagnosis (ie, how long patients report between their initial MS symptom and their MS diagnosis) by ethnicity and deprivation. Describing the distribution of ‘delays’ between onset and diagnosis (ie, how long patients report between their initial MS symptom and their MS diagnosis) by ethnicity and deprivation. Based on this distribution, use a microsimulation to estimate the total number of patients ‘awaiting’ diagnosis, assuming that the delays are representative of delays experienced by everyone who has MS. Other assumptions would be that a similar number of referrals had been made into clinics, and clinics had not expanded their capacity to take on referrals, which would impact the distribution of delays over the time period. We will then repeat this by each of our sociodemographic variables of interest. Based on this distribution, use a microsimulation to estimate the total number of patients ‘awaiting’ diagnosis, assuming that the delays are representative of delays experienced by everyone who has MS. Other assumptions would be that a similar number of referrals had been made into clinics, and clinics had not expanded their capacity to take on referrals, which would impact the distribution of delays over the time period. We will then repeat this by each of our sociodemographic variables of interest. We will then test the effect of relaxing assumptions regarding delays between onset and diagnosis, for example, the effect of some patients never being diagnosed. We will then test the effect of relaxing assumptions regarding delays between onset and diagnosis, for example, the effect of some patients never being diagnosed. The modelling uses the following assumptions: (1) patients and clinicians can accurately report the time between onset of symptoms and diagnosis; (2) these reported durations are not differentially biased between patient subgroups and between those with longer or shorter durations between onset and diagnosis; and (3) the length of the ‘queue’ (ie, the number of patients waiting to be diagnosed) does not change substantially over time. An example of the microsimulation approach is available here: https://github.com/danlewer/equity-ms/blob/main/example_microsim.R Limitations Limitations of the planned analysis include the following: The study will include the COVID-19 period, which is known to have impacted DMT prescribing due to guidance44 and evidence45 that certain drugs might impact susceptibility to severe COVID-19 infections. COVID-19 altered how services operate, and some patients may have been reluctant to enter hospitals to receive infusions. We cannot be sure that any observed differences in relation to COVID-19 would have held across ethnicities and socioeconomic groups, given that people from minoritised ethnicities were more severely impacted,46 so there may have been greater concerns around them continuing immunosuppressant treatments. COVID-19-related differences affect the generalisability of these data to other time periods. To address this, we will perform sensitivity analysis omitting data from March 2020 until January 2021, when COVID-19 vaccines became widely available. The study will include the COVID-19 period, which is known to have impacted DMT prescribing due to guidance44 and evidence45 that certain drugs might impact susceptibility to severe COVID-19 infections. COVID-19 altered how services operate, and some patients may have been reluctant to enter hospitals to receive infusions. We cannot be sure that any observed differences in relation to COVID-19 would have held across ethnicities and socioeconomic groups, given that people from minoritised ethnicities were more severely impacted,46 so there may have been greater concerns around them continuing immunosuppressant treatments. COVID-19-related differences affect the generalisability of these data to other time periods. To address this, we will perform sensitivity analysis omitting data from March 2020 until January 2021, when COVID-19 vaccines became widely available. Possible limited generalisability to other areas and times. The findings from this study will be based on patients across five neurology clinics with an estimated total catchment population of 7 million. Given that these clinics are research active and have a particular interest in inequalities, they may not be representative of specialist centres for MS in the UK nationally. Some rural areas may not have access to a specialist MS service at all. Findings may also not generalise beyond the time period studied. Possible limited generalisability to other areas and times. The findings from this study will be based on patients across five neurology clinics with an estimated total catchment population of 7 million. Given that these clinics are research active and have a particular interest in inequalities, they may not be representative of specialist centres for MS in the UK nationally. Some rural areas may not have access to a specialist MS service at all. Findings may also not generalise beyond the time period studied. Will not estimate geographical inequalities. With five contributing clinics, the present study will not generate sufficient data to study geographical inequalities, though we will describe travel time within our centre catchments. Estimation of geographical inequalities would require data from a larger number of centres nationally, which could identify regions where services for MS are very limited or inequalities in access to DMTs are concentrated. Such data would be challenging to obtain as it would require the cooperation of a larger number of centres. Patients in regions not in the catchment of a specialist MS centre may be linked to a general neurology service, making it very difficult to capture data for those patients and thus estimate the true magnitude of geographical inequalities. Will not estimate geographical inequalities. With five contributing clinics, the present study will not generate sufficient data to study geographical inequalities, though we will describe travel time within our centre catchments. Estimation of geographical inequalities would require data from a larger number of centres nationally, which could identify regions where services for MS are very limited or inequalities in access to DMTs are concentrated. Such data would be challenging to obtain as it would require the cooperation of a larger number of centres. Patients in regions not in the catchment of a specialist MS centre may be linked to a general neurology service, making it very difficult to capture data for those patients and thus estimate the true magnitude of geographical inequalities. Missing data. As noted above, we anticipate that there could be substantial missingness in our data for certain data fields (eg, relapses, disease onset, baseline disability). We cannot be sure of the extent of this until data capture has been completed, and we have discussed the possible reasons for missingness with clinical leads. These discussions will determine our analytical approach. Missing data. As noted above, we anticipate that there could be substantial missingness in our data for certain data fields (eg, relapses, disease onset, baseline disability). We cannot be sure of the extent of this until data capture has been completed, and we have discussed the possible reasons for missingness with clinical leads. These discussions will determine our analytical approach. Power for smaller ethnicity groups. Given that we anticipate relatively small samples of patients from minoritised ethnicities, we will not have the statistical power to look at these groups separately in our primary analysis. This may lead to an underestimation of inequalities disproportionately affecting some minoritised groups. We will present rates stratified by specific minoritised groups (mixed/multiple/other ethnicities; Asian/Asian British; lack/African/Caribbean/Black British) in a funnel plot. Power for smaller ethnicity groups. Given that we anticipate relatively small samples of patients from minoritised ethnicities, we will not have the statistical power to look at these groups separately in our primary analysis. This may lead to an underestimation of inequalities disproportionately affecting some minoritised groups. We will present rates stratified by specific minoritised groups (mixed/multiple/other ethnicities; Asian/Asian British; lack/African/Caribbean/Black British) in a funnel plot. Inherent limitations of studying ethnicity as we plan to. We acknowledge the limitations of the construct of ethnicity as we are applying it here, focusing on arbitrary groupings that do not reflect the diversity of people’s varied ethnic backgrounds, and potentially underestimating the extent of inequalities that certain groups face. Our approach reflects the inherent limitations of data recorded in the course of usual clinical care. Future work collecting data prospectively could include more granular indicators of potential structural determinants of inequalities disproportionately experienced by minoritised groups, and thus better quantify the extent of and mechanisms underpinning potential inequalities related to racial discrimination. Inherent limitations of studying ethnicity as we plan to. We acknowledge the limitations of the construct of ethnicity as we are applying it here, focusing on arbitrary groupings that do not reflect the diversity of people’s varied ethnic backgrounds, and potentially underestimating the extent of inequalities that certain groups face. Our approach reflects the inherent limitations of data recorded in the course of usual clinical care. Future work collecting data prospectively could include more granular indicators of potential structural determinants of inequalities disproportionately experienced by minoritised groups, and thus better quantify the extent of and mechanisms underpinning potential inequalities related to racial discrimination. Coproduction The patient and public involvement (PPI) group consists of 14 patients with MS from four of the five contributing centres who are invited to attend six monthly PPI meetings. The PPI group has provided consultation and advice on study design and will consult on how best to make results accessible to lay audiences. Patients unanimously supported the use of non-consented individual-level data for the proposed research. Patients also informed development of the analysis plan, including a need for a greater focus on potential age and sex-related inequalities in diagnosis of MS, which we plan to examine in this study and in a complementary line of research that will use national primary and secondary care records. Our patient group also highlighted numerous barriers to MS diagnosis, such as misdiagnosis of symptoms and/or lack of recognition that symptoms could have a neurological origin on the part of primary care physicians, which had led some to experience significant diagnostic delays. This influenced our decision to make statistical adjustments to our models to account for clinical factors, such as fixed disability, at the point of diagnosis. Ethics and dissemination Ethical and safety considerations The planned studies involve analysis of deidentified data that are routinely collected and recorded as part of standard NHS care. Therefore, there will be no impact on the services received by participants. The planned research will require Health Research Authority approval and NHS Research Ethics Committee approval. Although the study involves non-consented processing of individual-level data, a Clinical Advisory Group section 251 approval is not required because data will be anonymised by the usual care team prior to transfer to the central research team, following advice from the Health Research Authority. All data will be handled in accordance with the Data Protection Act 2018 and the General Data Processing Regulation. Plans for data transfer procedures are outlined in the data flow diagram (figure 1). Data flow diagram. DOB, date of birth; DOD, date of death; NHS, National Health Service. Dissemination plan We plan to disseminate our findings via publications in peer-reviewed journals and lay summaries disseminated via local MS societies. We will also present findings at national and international neurology conferences. We will translate the results into policy-relevant numbers, such as the number of patients with MS who may have ‘missed out’ on or experienced a delay in accessing DMTs due to their demographic characteristics. We aim to inform national policy and clinical guidance around diagnosis, treatment and care of people with MS from different sociodemographic groups. Information about the study results will be made available via the study website (https://bradfordresearch.nhs.uk/bradford-centre-for-health-data-science/equity-ms/). Supplementary material