Introducing a core dataset for real-world data in multiple sclerosis registries and cohorts: Recommendations from a global task force

Abstract

Background:

As of September 2022, there was no globally recommended set of core data elements for use in multiple sclerosis (MS) healthcare and research. As a result, data harmonisation across observational data sources and scientific collaboration is limited.

Objectives:

To define and agree upon a core dataset for real-world data (RWD) in MS from observational registries and cohorts.

Methods:

A three-phase process approach was conducted combining a landscaping exercise with dedicated discussions within a global multi-stakeholder task force consisting of 20 experts in the field of MS and its RWD to define the Core Dataset.

Results:

A core dataset for MS consisting of 44 variables in eight categories was translated into a data dictionary that has been published and disseminated for emerging and existing registries and cohorts to use. Categories include variables on demographics and comorbidities (patient-specific data), disease history, disease status, relapses, magnetic resonance imaging (MRI) and treatment data (disease-specific data).

Conclusion:

The MS Data Alliance Core Dataset guides emerging registries in their dataset definitions and speeds up and supports harmonisation across registries and initiatives. The straight-forward, time-efficient process using a dedicated global multi-stakeholder task force has proven to be effective to define a concise core dataset.

Introduction

There is an increased awareness of the importance of utilising real-world data (RWD) in multiple sclerosis (MS) with the number of RWD collection efforts growing around the world.^1–3 However, RWD collection efforts lack standardisation across sources especially due to heterogeneous content and the semantic and syntactic representation (the ‘what’ and ‘how’ of data collection). ⁴ Different stakeholders, such as clinicians, researchers, or regulators, have a particular interest in and need for alignment (‘harmonisation’) in RWD collection in MS. This would help facilitate collaboration between MS registries, especially when striving towards large-scale global collaborative efforts.^1,3

Several initiatives in the field of MS have already faced the challenge of harmonising datasets for their research purposes, which successfully enabled collaborative RWD-driven insights.^5–9 For example, the Big MS Data (BMSD) network agreed upon a minimal dataset for their network and research questions dealing with post-authorisation safety studies in MS. ¹⁰ In July 2017, the European Medicines Agency (EMA) held an MS workshop in the framework of their patient registries initiative for promoting the use of RWD for regulatory purposes. Different stakeholders discussed three aspects of real-world MS data, one being the establishment of core data elements in the data collection of MS registries. Although these initiatives working towards harmonisation in RWD in MS were a big step in the right direction, they failed to communicate the developed and used minimal or core datasets (see Figure 1 for differentiation) effectively. Also it remained unclear whether there was any alignment with more generic initiatives (not MS-specific) or inter-stakeholder validation for the development of used datasets. The published core data elements from EMA ¹¹ could serve as a ‘core dataset’ for MS. However, as it was not inside the scope of the workshop, the proposal lacked detailed information (e.g. the operational definition of the format of the specific data elements).

Figure 1.

For a better understanding, we defined ‘core dataset’, ‘minimal dataset’, and ‘common data model’ for the use in our paper and process of defining the MSDA core dataset.

The MS Data Alliance (MSDA) is a global multi-stakeholder collaboration aiming to develop tools to reduce the level of heterogeneity among real-world MS data sources and promote timely alignment for data harmonisation across data sources. ¹² The MSDA has successfully led global data harmonisation activities throughout the COVID-19 pandemic with the MS Global Data Sharing Initiative (GDSI) and its corresponding global recommendations for data collection on COVID-19 in people with MS. ¹³ Within the GDSI, the MSDA, MS International Federation and a global data task force established recommendations for a ‘COVID-19 in MS core dataset’ to enable a harmonised approach to the challenge of building a global data network. These recommendations were published ¹³ and provided on the MSDA website ¹⁴ for prospective and retrospective alignment of participating registries and databases to promote joint analyses. This inspired the MSDA to stimulate the definition and agreement on a standardised core dataset in MS (called ‘MSDA Core Dataset’, ‘Core Dataset’) to further strengthen and speed up collaboration to generate evidence from RWD.

The aim of this paper is to present the process and outcome of defining a core dataset for RWD in MS using a global multi-stakeholder task force.

Materials and methods

Workflow of the Core Dataset definition and agreement

The definition and approval of this core dataset in MS was guided by a multi-stakeholder global task force. The task force consisted of 20 experts and key opinion leaders in the field of RWD in MS, including clinicians, data custodians, and leads of MS registries and cohorts, a patient organisation, pharmaceutical industry representatives, and regulatory agencies. The range of different stakeholders enabled a rich diversity of opinions and experiences contributing to in-depth discussions on the core variables in MS. The experts, with representatives from North and South America, North-Eastern Africa, Europe, and Oceania, were involved in MSDA activities and were chosen to participate in the task force. The process of landscaping, discussing, reviewing, and agreeing upon a core dataset was orchestrated and coordinated by the MSDA. A workflow was defined prior to initiating the task force to ensure the process was as efficient as possible, while keeping the time and resources requested from the task force members feasible (see Figure 2). With regards to communication, we informed the members of the task force about steps or results. The key messages from the meetings were circulated after each meeting via mail. The Core Dataset was transparently shared with the task force together with the manuscript. The final agreement was based on the absence of contrary views of the circulated versions of the Core Dataset.

Figure 2.

The task force meetings represented three stages of the definition and agreement workflow: The process, estimated workload, and expected engagement were shared with the task force in advance, to ensure transparency and effectiveness. The workflow of the definition and consensus process of the task force was divided into three consecutive meetings held within 3 months (June to September 2022). (1) Pre-stage of the landscaping (consensus on aim, purpose and limitations of the core dataset), (2) Consensus building on the landscaping-based proposed categories and variables, and (3) Addressing and decision-making focussed on open questions regarding single variables and values. After each stage, the experiences and input shared from the task force members were incorporated into the design of the Core Dataset. In the end, the Core Dataset data dictionary was finished and shared with the task force for final approval.

Landscaping

The landscaping process for Core Dataset development aimed at identifying relevant initiatives and material around the topic of minimal and core data elements in MS. Three pillars were identified: (1) an exploratory literature scan, (2) recommended material from the task force, and (3) metadata from the MSDA Catalogue, ⁴ a web-based platform which was launched in 2019 and allows end-users to browse metadata profiles of MS RWD cohorts.

The literature scan and recommended material focussed on two categories: (a) MS-specific recommendations and publications on common data elements and (b) examples of MS registries with available information on their datasets. Literature was identified using PubMed and Google Scholar with specific search terms ([‘core dataset’ OR ‘minimum dataset’ OR ‘minimal dataset’] + ‘multiple sclerosis’).

The recommended material from the task force members was provided electronically and included additional publications on common data elements as well as the MS registry dataset examples.

Metadata from the MSDA Catalogue was accessed internally. It was exported and further processed to serve as a look-up and validation source.

Comparison of available sources and the drafting process

The suggested core data elements and wish list items of the referred EMA workshop were used for comparison for the mapping of available datasets. Although this list has not been validated or approved by experts beyond the workshop itself to be able to act as a ‘gold standard’, it was chosen to enable a pragmatic and feasible approach for an initial Core Dataset definition. The available reference recommendations, publications, or dataset descriptions were checked for EMA variables. Variables that were collected by references but were not part of the EMA core data elements were not included for comparison. The resulting comparison table led to the selection of category and variable suggestions for the first draft of the Core Dataset. Some variables were added based on proposals during the task force meetings and thereon-based discussions. This added to the eventual final Core Dataset v2022 over the course of the three task force meetings and the subsequent finishing (see Figure 2). Determination of the variables’ values followed the definition of the categories and variables. As it is the aim of the Core Dataset to be as standardised and harmonised as possible, the values of the identified and defined variables had to fulfil certain requirements. If available and applicable, standards were used, for example, for date (ISO-8601) or educational level (International Standard Classification of Education (ISCED)). If no standard existed, values were chosen for discussion that were already established or used by one or more registries, with the data dictionary of MSBase ¹⁵ as a leading coding model due to its wide global application.

Results

Scope of the Core Dataset: consensus on aim, purpose, and limitations

The consensus of the task force was that the Core Dataset should serve as a guideline for existing and emerging registries and databases to reduce heterogeneity, provide coverage of the key data elements for care in MS and ensure feasibility of data collection. Twelve of the 20 experts of the task force were clinicians and data custodians with year- and even decade-long expertise in collection of such data and a proven scientific track record which enabled a solid feasibility assessment. The Core Dataset represents the common denominator across initiatives and experts’ input as a list of recurring variables and aims to be research question agnostic. Indeed, the Core Dataset does not aspire to enable every possible research question since a have-it-all core dataset for MS would be impossible to achieve (as for e.g. sustainability and feasibility). The decision was made to focus on a core dataset that can be augmented in future revisions and (local) adaptations regarding what (variables) and how (values, formats) to collect. Application guidelines, for example, focussing on frequency of data collection, are out of scope for the current version of the Core Dataset, but conceivable for future work.

Landscaping: identify existing (and available) sources on minimal/core datasets

Our literature scan performed in July 2022 found little published information in the field of MS-specific recommendations and publications on data elements. The literature scan on PubMed did not deliver any relevant results. Google Scholar delivered a broader picture for the chosen search terms but no publication that dealt exactly with the topic of any minimal or core dataset in MS could be found. Broadening the search term to ‘real-world data’ and ‘multiple sclerosis’ delivered a larger amount but mostly single study publications with a specific use case (mostly a certain treatment) where RWD was used. Neither dataset definition processes nor the datasets themselves were included in these publications. The scan for relevant material based on prior knowledge of MS initiatives that have published their datasets resulted only in the publication about the EMA MS patient registry workshop. ¹¹ The additional material from task force members helped identifying initiatives and datasets that were not publicly available.

The National Institute of Neurological Disorders and Stroke (NINDS), ¹⁶ as part of the National Institutes of Health (NIH), has defined ‘Common Data Elements’ (CDE) for neuroscientific clinical research, with MS being one of the neurological disorders. This very extensive list of variables aims to define data collection standards for neuroscientific clinical research across NINDS-funded studies in the United States. This list was included in the comparison with the EMA core data elements to identify common data elements for MS RWD.

A few publications were identified that discuss the importance of using standardised/harmonised datasets for RWD use in MS^1,3 or attempt to find a ‘common language of MS data elements’. ¹⁷ The latter study compared two registry data collections with regards to the used data elements in both data collection software and gave a list of recommendations based on these two comparators. This list of recommendations was also compared to the EMA list.

Similar to the effort to find recommendations on CDE in MS, the search for publicly accessible data dictionaries from MS registries was challenging. Some registries have published their dataset description or sent them to the MSDA for the purpose of the landscaping for the Core Dataset. For example, MSBase, ¹⁵ NARCOMS,^18,19 and the German MS registry ²⁰ have published their dataset descriptions online. The dataset descriptions of the Egyptian Registry, iConquerMS,^21,22 and the Canadian Multiple Sclerosis Monitoring System (not active anymore) were provided to the MSDA.

Given that a lot of analyses in MS require derived variables, there’s also a need to have data specification documents (‘data dictionaries’) of observational MS data sources available. The publication of established datasets would be, beyond the definition of a minimal or core dataset, an important step to promote a common understanding across MS RWD sources and facilitate harmonisation.

The Core Dataset: categories, variables, values, and data dictionary

The task force discussions led to the agreement on 8 categories and 44 (plus date of visit) variables for the Core Dataset v2022 (see Figure 3). The discussions consisted of a presentation of a certain preparatory work (landscaping results, draft of Core Dataset variables, draft of data dictionary) and an inter-group debate from which the key reflections and decisions were recorded and shared with the task force afterwards. The inclusion and exclusion of variables, as well as the approach to appoint their values, were the main topics of the discussions. Excluded variables and value decisions are provided in the ‘Discussion’ section. After agreeing upon the categories, variables, and values, the data dictionary of the Core Dataset was finalised. A detailed data dictionary (see Table 1) was developed to facilitate the implementation of the dataset and to promote the harmonisation across data sources not only through content but also through the specification of the structure and syntax. To ensure alignment between the MSDA Core Dataset and important non-MS-specific initiatives, it was decided to add a disease or data source independent standard into the data dictionary of the Core Dataset: SNOMED CT (Systematised Nomenclature of Medicine Clinical Terms). SNOMED CT is, as an international validated multilingual healthcare terminology, broadly used across the globe. With its more than 358 000 standardised concepts (as of 16 May 2023), ²³ it enables a harmonised and consistent representation of clinical content. SNOMED CT is mapped to other healthcare standards as well as used as a leading standard in common data models (CDM, see Figure 1). It can be an additional benefit for registries, for example, by being used for health practitioner databases or patient-reported registries, where variables can be adapted to questions to suit the needs of the circumstances.

Figure 3.

Core Dataset categories and variables: The 8 categories of the core dataset (demographic data, disease history, disease status, relapses, MRI, comorbidities, disease-modifying treatments, non-pharmaceutical treatments) with their respective 44 variables as they were agreed-upon during the task force meetings.

Table 1.

Core Dataset data dictionary to specify the content, structure, and syntax of the collected variables.

Documented by	Name	Variable_name	Collection time	Data format	SNOMED term	SNOMED code	Value(s)	Value_name	SNOMED term	SNOMED code
B	Date of visit/reporting date (for PRO)	date_visit	FV, FU	YYYY-MM-DD	Date of visit (observable entity)	406543005	–	–	–	–
Demographic data
B	Date of birth	date_birth	FV	YYYY-MM-DD	Date of birth (observable entity)	184099003	(if DD or MM is not allowed, 15 is used for DD or MM, respectively)	–	–	–
B	(Biological) sex	sex	FV	Radio/select	Biological sex (property) (qualifier value)	734000001	Female	female	Female (finding)	248152002
							Male	male	Male (finding)	248153007
B	Country of residence	residence	FV, FU	List	Country of residence (observable entity)	416647007	ISO 3166-1 (alpha-2)	residence_xx (xx = country per ISO standard)	–	–
B	Race/Ethnicity	race_ethnicity	FV	List	Race (observable entity) Ethnic group finding (finding)	103579009 397731000	(there are no value recommendations given because of the global differences in race/ethnicity as well as the corresponding legislation on data collection; a collection of local ethnic/racial data where possible is strongly encouraged though)	–	–	–
B	Educational level	education	FV	List	Educational achievement (observable entity)	105421008	ISCED 0 = Early childhood education ISCED 1 = Primary Education ISCED 2 = Lower Secondary Education ISCED 3 = Upper Secondary Education ISCED 4 = Post-secondary non-Tertiary Education ISCED 5 = Short-cycle tertiary education ISCED 6 = Bachelors degree or equivalent tertiary education level ISCED 7 = Masters degree or equivalent tertiary education level ISCED 8 = Doctoral degree or equivalent tertiary education level	isced_0 isced_1 isced_2 isced_3 isced_4 isced_5 isced_6 isced_7 isced_8	–	–
B	Employment status	employment	FV, FU	List	Employment status (observable entity)	224362002	Employed with unknown status (full-/part-time)	employed	Employed (finding)	224363007
							Full-time employed	ft_employed	Full-time employment (finding)	160903007
							Part-time employed	pt_employed	Part-time employment (finding)	160904001
							Student	student	Student in full time education (occupation)	413327003
							Retired	retired	Retired, life event (finding)	105493001
							Medically retired (due to MS)	med_retired_ms	Medically retired, life event (finding)	160898008
							Medically retired (due to other diseases)	med_retired_other	Medically retired, life event (finding)	160898008
							Unemployed	unemployed	Unemployed (finding)	73438004
							Maternal/ parental leave	parental_leave	On maternity leave (finding)/On parental leave (finding)	224457004/700149001
							Homemaker (housewife/houseman)	homemaker	Homemaker (person)	444168002
B	Smoking status	smoking	FV, FU	Radio/select	–	–	Never smoked	never_smoked	Never smoked tobacco (finding)	266919005
							Current smoker	current_smoker	Smoker (finding)	77176002
							Former smoker	former_smoker	Ex-smoker (finding)	8517006
							Unknown	smoking_unknown	Tobacco smoking consumption unknown (finding)	266927001
							If current_smoker: number of cigarettes (daily)	smoking_count	Cigarette consumption (observable entity)	230056004
B	Family history of MS	ms_family	FV	Boolean	Family history: Multiple sclerosis (situation)	160337009	Yes No Unknown/not sure	yes no unknown/not_sure	Yes (qualifier value) No (qualifier value) Unknown (qualifier value)	373066001 373067005 261665006
Disease-specific information: disease history
B	Date of diagnosis	date_diagnosis	FV	YYYY-MM-DD	Date of diagnosis (observable entity)	432213005	(if DD or MM is not available, 15 is used for DD or MM, respectively)	–	–	–
B	Date of onset	date_onset	FV	YYYY-MM-DD	Date of onset (observable entity)	298059007	(if DD or MM is not available, 15 is used for DD or MM, respectively) or unknown/unsure	–	- Unknown (qualifier value)	261665006
C	Cerebrospinal Fluid: Oligoclonal IgG Bands present (for diagnosis)	csf_olib	FV	Boolean	Cerebrospinal fluid oligoclonal bands (procedure)	113073005	Yes No Unknown	yes no unknown	Positive (qualifier value) Negative (qualifier value) Unknown (qualifier value)	10828004 260385009 261665006
B	MS course	ms_course	FV, FU	List	Multiple sclerosis (disorder)	24700007	Radiologically isolated syndrome	ris	Radiologically isolated syndrome (disorder)	16415361000119105
							Clinically isolated syndrome	cis	Clinically isolated syndrome (disorder)	445967004
							Relapsing remitting	rrms	Relapsing remitting multiple sclerosis (disorder)	426373005
							Secondary progressive	spms	Secondary progressive multiple sclerosis (disorder)	425500002
							Primary progressive	ppms	Primary progressive multiple sclerosis (disorder)	428700003
Disease-specific information: disease status
C	MS current status (clinician)	ms_status_clin	FV, FU	List	Chronic disease–follow-up assessment (finding)	170558000	Not active and without progression (stable disease)	statusC_stable	Patient’s condition stable (finding)	359746009
							Active (MRI activity or relapses) and with progression	statusC_act_progr	Active (qualifier value)/Progressive (qualifier value)	55561003/255314001
							Active but without progression	statusC_act_noprogr	Active (qualifier value)/non-progressive (qualifier value)	55561003/702322003
							Not active but with progression	statusC_inact_progr	Inactive (qualifier value)/Progressive (qualifier value)	73425007/255314001
P	MS current status (patient), e.g. “Compare your overall MS status now with what you experienced since last visit: Is your MS. . .”	ms_status_pat	FV, FU	List	Chronic disease–follow-up assessment (finding)	170558000	Much worse	statusP_worse2	Patient status determination, much worse (finding)	45112000
							Worse	statusP_worse1	Patient’s condition deteriorating (finding)	275723000
							A little worse	statusP_worse	Patient status determination, slightly worse (finding)	6718007
							No change	statusP_stable	Patient’s condition stable (finding)	359746009
							A little better	statusP_better	Patient status determination, slightly improved (finding)	67405006
							Better	statusP_better1	Patient’s condition improved (finding)	268910001
							Much better	statusP_better2	Patient status determination, greatly improved (finding)	3286006
B	Current symptoms	current_symp	FV, FU, RI	Checkboxes/list	Symptom checklist (assessment scale)	273859002	Walking/Mobility	symp_walking	Ability to walk (observable entity)	282097004
							Fatigue	symp_fatigue	Fatigue (finding)	84229001
							Pain	symp_pain	Pain (finding)	22253000
							Anxiety	symp_anxiety	Anxiety (finding)	48694002
							Depression	symp_depression	Depressive disorder (disorder)	35489007
							Vision	symp_vision	Disorder of vision (disorder)	95677002
							Dizziness	symp_dizziness	Dizziness (finding)	404640003
							Bladder control	symp_bladder	Impaired urinary system function (finding)	1156448003
							Bowel control	symp_bowel	Bowel control, function (observable entity)	129008009
							Spasticity	symp_spasticity	Spasticity (finding)	221360009
							Sensory symptoms	symp_sensory	Sensory disorder (disorder)	85972008
							Cognition	symp_cognition	Impaired cognition (finding)	386806002
							Dexterity	symp_dexterity	Poor manual dexterity (finding)	306741005
P	Symptom severity (for each symptom that was collected)	sever_symp	FV, FU, RI	Scale	Symptom severity level (observable entity)	405162009	0–10 (0 being the lowest severity/impact on life and 10 the highest severity/impact on life)	–	–	–
B	Symptoms treated	treat_symp	FV, FU	Boolean			Yes No	yes no	Yes (qualifier value) No (qualifier value)	373066001 373067005
C	Expanded Disability Status Scale (EDSS) Score	edss_score	FV, FU, RI	Float	Kurtzke multiple sclerosis rating scale (assessment scale)	273554001	0, 1.0, 1.5, . . ., 10	–	–	–
P	Patient determined disease steps (PDDS)	pdds_score	FV, FU	Integer			0 (“Normal”), 1 (“Mild disability”), 2 (“Moderate disability”), 3 (“Gait disability”), 4 (“early cane”), 5 (“late cane”), 6 (“bilateral support”), 7 (“wheelchair/scooter”), 8 (“bedridden”)	–	–	–
C	Timed 25 Foot Walk (T25FW)	t25fw	FV, FU	Float	–	–	(time in seconds)	–	–	–
C	9-Hole Peg Test (9HPT)	9hpt	FV, FU, RI	Float	Nine hole peg test score (observable entity)	446602000	(time in seconds, dominant hand)	–	Dominant hand (attribute)	263557007
							(time in seconds, non-dominant hand)	–	Non-dominant side (qualifier value)	262458006
C	Vibratory sense	vib_sense	FV, FU, RI		Vibratory sense, function (observable entity)	397656001	Normal (right foot)	vib_normal_right	Normal vibration sensation of right foot (finding)	830005003
							Decreased (right foot)	vib_decreased_right	Impaired vibration sense of right foot (finding)	418711005
							Absent (right foot)	vib_absent_right	Absence of vibration sense of right foot (finding)	830006002
							Normal (left foot)	vib_normal_left	Normal vibration sensation of left foot (finding)	830004004
							Decreased (left foot)	vib_decreased_left	Impaired vibration sense of left foot (finding)	130980003
							Absent (left foot)	vib_absent_left	Absence of vibration sense of left foot (finding)	417892007
C	Symbol Digit Modalities Test (SDMT)	sdmt	FV, FU	Int	Symbol Digit Modalities Test score (observable entity)	718387005	(number of correct answers/substitutions)	–	–	–
Disease-specific information: relapses
B	Relapse event	relapse	RI	Boolean	Exacerbation of multiple sclerosis (disorder)	192929006	Yes No Unknown/not sure	yes no unknown/not_sure	Yes (qualifier value) No evidence of (qualifier value) Unknown (qualifier value)	373066001 41647002 261665006
B	Date of relapse	date_relapse	RI	YYYY-MM-DD	–	–	(if DD or MM is not available, 15 is used for DD or MM, respectively)	–	–	–
B	Steroid treatment for relapse	relapse_treat	RI	Boolean	Corticosteroid and/or corticosteroid derivative therapy (procedure)	788751009	Yes No	yes no	Yes (qualifier value) No (qualifier value)	373066001 373067005
B	Recovery	relapse_recovery	RI, FU	Select/radio	Convalescence (finding)	105499002	Complete	compl_recovery	–	–
							Partial	partial_recovery	–	–
							No	no_recovery	–	–
							Unknown/not sure	unknown/not_sure	Unknown (qualifier value)	261665006
Para-clinical investigations: MRI
B	MRI done	mri	FU	Boolean	Magnetic resonance imaging (procedure)	113091000	Yes No	yes no	Yes (qualifier value) No (qualifier value)	373066001 373067005
B	MRI region	mri_region	FU	Radio/select	–	–	Brain	mri_brain	Magnetic resonance imaging of brain (procedure)	816077007
							Cervical cord	mri_cervical	Magnetic resonance imaging of cervical spine (procedure)	241646009
							Thoracic cord	mri_thoracic	Magnetic resonance imaging of thoracic spine (procedure)	241647000
							Whole spinal cord	mri_myelon	Magnetic resonance imaging of spine (procedure)	241645008
B	Date of MRI	mri_date	FU	YYYY-MM-DD	–	–	–	–	–	–
C	Number of Gd enhanced lesions	mri_gd_les	FU	Integer	–	–	number or “Unknown”	–	-Unknown (qualifier value)	261665006
C	Number of new or enlarged T1 lesions	mri_new_les_T1	FU	Integer	–	–	number or “Unknown”	–	-Unknown (qualifier value)	261665006
C	Number of new or enlarged T2 lesions	mri_new_les_T2	FU	Integer	–	–	number or “Unknown”	–	-Unknown (qualifier value)	261665006
Comorbidities
B	Comorbidities	comorbidity	FV, FU	Boolean	Co-morbid conditions (finding)	398192003	Yes No Unknown/not sure	yes no unknown/not_sure	Yes (qualifier value) No (qualifier value) Unknown (qualifier value)	373066001 373067005 261665006
B	Type of comorbidity	com_type	FV, FU	List	–	–	Alcohol abuse	com_alc_abuse	Alcohol abuse (disorder)	15167005
							Anxiety	com_anxiety	Anxiety disorder (disorder)	197480006
							Autoimmune Disorder	com_autoimmune	Autoimmune disease (disorder)	85828009
							Bipolar disorder	com_bipolar	Bipolar disorder (disorder)	13746004
							Cancer	com_cancer	Malignant neoplastic disease (disorder)	363346000
							Cardiac arrhythmia	com_arrhythmia	Cardiac arrhythmia (disorder)	698247007
							Chronic lung disease	com_chronic_lung	Chronic lung disease (disorder)	413839001
							Congestive heart disease	com_congestive_heart	Congestive heart failure (disorder)	42343007
							Depression	com_depression	Depressive disorder (disorder)	35489007
							Diabetes mellitus	com_diabetes	Diabetes mellitus (disorder)	73211009
							Drug abuse	com_drug_abuse	Drug abuse (disorder)	26416006
							Epilepsy	com_epilepsy	Epilepsy (disorder)	84757009
							Eye disease	com_eye	Disorder of eye proper (disorder)	371405004
							Fibromyalgia	com_fibromyalgia	Fibromyalgia (disorder)	203082005
							Irritable bowel syndrome	com_ibs	Irritable bowel syndrome (disorder)	10743008
							Viral hepatitis	com_hepatitis	Viral hepatitis (disorder)	3738000
							Hyperlipidaemia	com_hyperlipidemia	Hyperlipidaemia (disorder)	55822004
							Hypertension	com_hypertension	Hypertensive disorder, systemic arterial (disorder)	38341003
							Ischemic heart disease	com_ischemic_heart	Ischemic heart disease (disorder)	414545008
							Peripheral vascular disease	com_peripheral_vasc	Peripheral vascular disease (disorder)	400047006
							Psychosis	com_psychosis	Psychotic disorder (disorder)	69322001
							Stroke (any)	com_stroke	Cerebrovascular accident (disorder)	230690007
C	Comorbidity system (alternative to comorbidity type)	com_system	FV, FU	List	–	–	skeletal	sys_skel	Disorder of skeletal system (disorder)	88230002
							muscular	sys_musc	Disorder of musculoskeletal system (disorder)	928000
							nervous	sys_nerv	Disorder of nervous system (disorder)	118940003
							endocrine	sys_endo	Disorder of endocrine system (disorder)	362969004
							cardiovascular	sys_cardio	Disorder of cardiovascular system (disorder)	49601007
							lymphatic	sys_lymph	Disorder of lymphatic system (disorder)	362971004
							respiratory	sys_resp	Disorder of respiratory system (disorder)	50043002
							digestive	sys_digest	Disorder of digestive system (disorder)	53619000
							urinary	sys_urin	Disorder of the urinary system (disorder)	128606002
							reproductive	sys_reprod	Disorder of reproductive system (disorder)	362968007
DMT: for current and past treatments
B	Current or past DMT	dmt_status	FV, FU	Select/radio	–	–	Yes No (but in the past) DMT naive (Never)	yes no dmt_naive	Yes (qualifier value) No (qualifier value) Treatment naive (finding)	373066001 373067005 844585000
B	DMT type	dmt_type	FV, FU	List	–	–	Alemtuzumab	alemtuzumab	Alemtuzumab (substance)	129472003
							Cladribine	cladribine	Cladribine (substance)	386916009
							Dimethyl Fumarate	dimethyl_fumarate	Dimethyl fumarate (substance)	724035008
							Diroximel fumarate	diroximel_fumarate	Diroximel fumarate (substance)	1268407006
							Fingolimod	fingolimod	Fingolimod (substance)	449000008
							Glatiramer acetate	glatiramer_acetate	Glatiramer acetate (substance)	108755008
							Interferon-beta	interferon	Interferon beta (substance)	420710006
							Peg-Interferon-beta 1a	peg_interferone	Interferon beta (substance)	420710006
							Monomethyl fumarate	monomethyl_fumarate	Monomethyl fumarate (substance)	1010557002
							Natalizumab	natalizumab	Natalizumab (substance)	414805007
							Ocrelizumab	ocrelizumab	Ocrelizumab (substance)	733464008
							Ofatumumab	ofatumumab	Ofatumumab (substance)	444607009
							Ozanimod	ozanimod	Ozanimod (substance)	870525003
							Ponesimod	ponesimod	Ponesimod (substance)	1179253008
							Siponimod	siponimod	Siponimod (substance)	786997005
							Teriflunomide	teriflunomide	Teriflunomide (substance)	703785006
							Ublituximab	ublituximab
							Azathioprine	azathioprine	Azathioprine (substance)	372574004
							Cyclophosphamide	cyclophosphamide	Cyclophosphamide (substance)	387420009
							Fludarabine	fludarabine	Fludarabine (substance)	386907005
							IV Immunoglobulin	immunoglobulin	Immunoglobulin (substance)	112133008
							Methotrexate	methotrexate	Methotrexate (substance)	387381009
							Minocycline	minocycline	Minocycline (substance)	372653009
							Mitoxantrone	mitoxantrone	Mitoxantrone (substance)	386913001
							Mycophenolate mofetil	mycophenolate_mofetil	Mycophenolate mofetil (substance)	386976000
							Rituximab	rituximab	Rituximab (substance)	386919002
							Corticosteroids	corticosteroids	Corticosteroid series (substance)	255877006
							aHSCT	ahsct	Transplantation of autologous hematopoietic stem cell (procedure)	709115004
B	Start date	dmt_start	FV, FU	YYYY-MM-DD	Date treatment started (observable entity)	413946009	–	–	–	–
B	Stop date	dmt_stop	FV, FU	YYYY-MM-DD	Date treatment stopped (observable entity)	413947000	–	–	–	–
B	Stop reason	dmt_stop_reas	FV, FU	List	Reason for (attribute)/drug therapy discontinued (situation)	410666004/274512008	Developed contraindications	stop_contraindication	Medication stopped–contra-indication (situation)	395008009
							Inefficacy	stop_inefficacy	Medication stopped–ineffective (situation)	395007004
							Intolerance/treatment-related adverse event	stop_intolerance	Medication stopped–side effect (situation)	395009001
							Non-adherence	stop_non_adherence	Non-compliance of drug therapy (finding)	702565001
							Patient Choice/Convenience	stop_patient_choice	Requested by patient (contextual qualifier) (qualifier value)	15635006
							Persistent disease activity (MRI lesions/relapses)	stop_disease_activity	Active disease following therapy (finding)	110278006
							Progression of disease/disability	stop_progression	–	–
							Financial/insurance coverage issues	stop_financial	Finding related to health insurance issues (finding)	419808006
							Other	stop_other	Other category (qualifier value)	394841004
							Unknown	stop_unknown	Unknown (qualifier value)	261665006
Non-pharmaceutical treatments
B	Non-pharmaceutical treatment type	np_treat_type	FV, FU	List	Professional/ancillary services care (regime/therapy)	409023009	Occupational therapy	occupational_therapy	Occupational therapy (regime/therapy)	84478008
							Physiotherapy	physiotherapy	Physical therapy procedure (regime/therapy)	91251008
							Psychotherapy	psychotherapy	Psychotherapy (regime/therapy)	75516001
							Rehabilitation	rehabilitation	Rehabilitation therapy (regime/therapy)	52052004
							Speech therapy	speech_therapy	Speech therapy (regime/therapy)	5154007
							Other	np_other	Other category (qualifier value)	394841004
B	Start date	np_treat_start	FV, FU	YYYY-MM-DD	Date treatment started (observable entity)	413946009	date, if available/applicable	–	–	–
B	Stop date	np_treat_stop	FV, FU	YYYY-MM-DD	Date treatment stopped (observable entity)	413947000	date, if available/applicable	–	–	–

SNOMED: Systematised Nomenclature of Medicine Clinical Terms; Documented by: B: both; C: clinician-reported specific; P: patient-reported specific; Collection time: FV: first visit, FU: follow-up visit; RI: relapse incidence; ISO: International Organisation for Standardisation; ISCED: International Standard Classification of Education; MS: multiple sclerosis; MRI: magnetic resonance imaging; DMT: disease-modifying treatment; aHSCT: autologous haematopoietic stem cell transplantation.

Discussion

We propose a Core Dataset for MS, consisting of 44 variables (plus date of visit) in eight categories, which can greatly benefit the use of RWD and the generation of real-world evidence (RWE) in MS.

It is important to note that when designing a new collection of clinical data in the form of a disease registry, a common experience is that initial ambitions are higher than what is eventually possible to collect longitudinally with a high degree of data completeness. Even if the proposed Core Dataset has been limited to contain what was considered a reasonable minimum of variables, it indeed exceeds the current data collections and the completeness of the collected variables of most, if not all, existing registries.^4,24 Thus, the value of the Core Dataset in this context is to provide a structured and formatted list of variables to be considered for a new registry, rather than as a mandatory template.

The use and implementation of the Core Dataset is recommended for prospective (i.e. improving data collection efforts moving forward) and retrospective (i.e. coping with heterogeneity of existing data sources) harmonisation efforts. The Core Dataset may benefit prospective harmonisation in:

• (1) Adapting RWD collection efforts by using the Core Dataset as a blueprint for the data acquisition tool to shape or alter the source dataset of the registry/cohort. MS registries, especially those with patient-reported data, may tailor the formulation of a question for a certain variable to incorporate local differences, for example, regarding language or the level of detail needed to ask for a certain variable.

• (2) Guiding new RWD initiatives. In order to reach emerging registries and initiatives, the Core Dataset will be part of the educational programme of the MSDA called ‘How to set up a registry?’ ²⁵ as a leading example in the topic dealing with the harmonisation and alignment for data collection. The educational activities will be updated along with the revisions and (local) adoptions of the Core Dataset. In addition, the Core Dataset is made available on the website and the social media account of the MSDA and partner organisations for easy access and to promote dissemination.

When it comes to retrospective harmonisation efforts, the Core Dataset aims to serve as the target schema definition for harmonisation activities. A target schema definition is the desired output to which (an extract of) the registry’s source data is transformed. The source dataset is not altered, and the routine prospective data collection within the registry/cohort is not changed. An example of this is the transformation of a birth date in the original format (e.g. MM/DD/YYYY) to the target format (e.g. YYYY-MM-DD). Retrospective harmonisation of existing RWD has proven successful within the GDSI, where RWD from different data sources was transformed into and verified against the COVID-19 core dataset before uploading the patient-level data into the central platform.

A challenge for this version of the Core Dataset data dictionary was the value definition of some variables that had a variety of options and no straight-forward value set was identifiable. For example: (1) For the topic of ‘Comorbidities’ it was decided to use Marrie et al. ²⁶ as the leading published list of comorbidities in MS although it may not include all the relevant comorbidities (as they are either dependent on the specific use case/research question or part of the detailed PASS data collections). As an optional variable, body systems were added for a broader collection of comorbidities. This may serve RWD collections where the granular collection of comorbidities is not possible; (2) The variable ‘race/ethnicity’ was added because of the influence of ancestry on disease prevalence, course or treatment effectiveness.^27–29 Due to the large variety of local racial and ethnic categories across the globe, it was decided not to include recommendations for values but to include the variable into the Core Dataset with the note to adhere to local ethnicity/race data values;

• (3) For ‘current symptoms’, SymptoMScreen, ³⁰ a validated and published list of symptoms, was chosen for the first version of the Core Dataset to be the guide for the symptom list to address both patient-driven and clinician-driven data collection needs. It is conceivable that it will be extended in the future or adapted to further validations like Zhang et al. ³¹ or other (arising) symptom questionnaires. A recommendation for a severity score for those symptoms was also added to enhance granularity in symptom follow-up/management, especially in patient-driven data collections.

A regular revision of the current Core Dataset (v2022) is anticipated, especially in regards to the currently excluded variables or pragmatic choices of values. Variables excluded for the 2022 version were those considered future-oriented but not yet widely collected (e.g. biomarkers, including neurofilament, optical coherence tomography, or new progression assessments). Variables on magnetic resonance imaging (MRI) measurements and outcomes focussing not only on relapsing-remitting MS (RRMS) but progressive forms of MS as well are also excluded for now due to feasibility and data coverage limitations. They are also potentially exposed to noise in the data, for example, through the use of different scanners, MRI protocols or interobserver variety for brain volume measurements. Dataset variables needing a dedicated set of data elements (e.g. in the area of patient-reported outcomes (cf. PROMS initiative)^32,33 or pharmacovigilance) are also not included. The latter is anticipated to be driven by leading networks like BMSD or PROMS initiative focusing on these specific topics. For example, the BMSD has already established a ‘core BMSD post-authorisation safety studies (PASS) protocol’ for disease-modifying treatments within their network that provides a high-level concept for future PASS although further customization is required for a specific PASS depending on its research questions. The BMSD is working on an updated core PASS protocol and the registries involved in the BMSD aim to become qualified registries for PASS by the EMA. ³⁴ For that reason, data elements on serious adverse events related to disease-modifying treatments were not included in the MSDA Core Dataset.

The Core Dataset was developed within a predefined time frame using a pragmatic approach by including clinical MS variables commonly observed in real-world datasets and feasible to collect. This pragmatic approach naturally has limitations. The task force consisted of a limited number of experts in the field of MS and RWD and only indirectly involved patients (represented by the patient registries and organisation). An extension of the task force, including patients or more experts from other countries, could be considered for future work. This is also helpful in minimising bias in the selection of the core variables. The personal and professional background and interests of the participating experts of the task force may also have influenced the selection of the core variables.

When considering future revisions of the Core Dataset, the cost/effort is another aspect to consider. Getting a task force together requires time and resources spent by involved parties. Compared to consistently new definitions of minimal or core datasets for each new collaboration, this seems to be a benevolent calculation of cost–value ratio though.

Conclusion

The Core Dataset aims to reduce heterogeneity and to promote harmonisation across MS data sources, by incorporating elements that promote faster collaboration within and outside of the MS community. This will help answer questions and solve challenges related to clinical care more quickly–benefitting people with MS. The MSDA Core Dataset v2022 can act as a consolidated step towards a more effective and efficient use of RWD in MS. It is foreseen that the future development of the current Core Dataset and the establishment of core datasets for special interest topics will further support the prospective and retrospective harmonisation in and across MS registries and cohorts. The detailed Core Dataset is presented in Table 1 and freely downloadable on the MSDA website (https://www.msdataalliance.org/).