MACCS is derived entirely from routinely collected administrative healthcare data and integrates information from multiple existing national datasets within NHS Scotland. Individuals are included in MACCS on the basis of medicine exposure: the eligibility criteria for inclusion are being aged 18 years or older; and having received at least one prescription medicine within NHS Scotland – dispensed through a community pharmacy since 04.2009; administered in hospital since 07.2022; and/or delivered directly to their home through the Medicines Home Care Services since 01.2019. Extensive clinical, demographic, and laboratory information is available for included individuals through record linkage via the Community Health index (CHI) number, a unique person identifier which enables deterministic linkage across the datasets [12]. Record linkage is performed by the Electronic Data Research and Innovation Service (eDRIS) [15] within PHS and follows established information governance procedures.

MACCS currently combines information from 14 different sources, thereby providing data on medicines use; medical conditions including laboratory data; hospitalisations; deaths; and socio-demographics. National medicines datasets include Prescription Information System (PIS), holding information on medicines supplied in community pharmacies [4]; Hospital Electronic Prescribing and Medicines Administration (HEPMA), covering inpatient wards in hospitals (excluding systemic anti-cancer therapy and intensive care units) [5]; and Home Care Medicines (HCM), i.e., data on medicines prescribed in secondary care and delivered directly to a patient’s home (typically highly specialist medicines such as biologics) [16]. Disease registries such as the Cancer Registry (SMR06)[17] and the Scottish Renal Registry (SRR)[18] provide data on medical conditions; key biochemistry and haematology test results (e.g., kidney and liver function tests) are held in the Scottish Care Information (SCI) Store [19] while microbiology surveillance data stem from the Electronic Communication of Surveillance Scotland (ECOSS) [20]. Data on hospitalisations is provided through Scottish Morbidity Records (SMR) covering outpatient, inpatient, maternity, and mental health care (SMR00/01/02/04) [21–24]; and the Scottish Intensive Care Society Audit Group (SICSAG) [19] covering critical care. Death records from the National Records of Scotland (NRS) are available to provide data on mortality [25]. Patient demographics – including age, sex, health board and urban-rural classification of residence, and the Scottish Index of Multiple Deprivation (SIMD) [26] are available across the datasets; see Table 1 for further details on the individual datasets. Work is in progress to add further data; in the first instance, Unscheduled Care Data comprising data from Emergency Departments, GP out-of-hours systems, and the Scottish Ambulance Service. Discussions are ongoing to also include primary care data (data from GP records).

Source extracts from PHS datasets are ingested into a controlled staging layer in which original dataset structures are preserved, and initial data quality checks are performed (Figure 1). Within the National Safe Haven (NSH), a secure environment used to enable researcher access to pseudonymised individual-level data, MACCS is organised using a structured database design (or schema) that defines how different types of information are stored, related, and queried. The core schema includes person-level tables that hold basic demographic information for each individual, alongside harmonised medicines exposure tables derived from the PIS, HEPMA and HCM, which record medicines prescribing and supply (where available) in a consistent format. Additional event-based tables capture linked information on hospital activity, laboratory results, disease registries, and mortality data. Data is updated at regular intervals through repeat data extractions from the source datasets; this approach preserves a consistent data structure while allowing new records to be incorporated over time. All records include dates of relevant events, allowing researchers to follow individuals over time, subject to data availability, and to support longitudinal analyses, the use of look-back periods, and time-to-event study designs.

Variables are harmonised across datasets as far as possible (e.g., consistent variable naming and formatting); standardised coding systems are used where available and appropriate. These include, e.g., British National Formulary (BNF) coding [29] or medicines dispensed in community pharmacies and Dictionary of Medicines & Devices (dm+d) codes [30] for medicines prescribed and administered in hospital settings, both of which can be mapped to the Anatomical Therapeutical Chemical (ATC) classification system [12]; and International Classification of Diseases 10^th Revision (ICD-10) codes [31] to capture reasons for admission to hospital and causes of death. Dates are recorded using standard ISO formats, and categorical variables use clearly defined code lists.