Mapping and Harmonising Outcomes in Post-Stroke Cognitive Impairment Intervention Trials to Inform Core Outcome Set Development: A Scoping Review Protocol

Eligibility criteria

For the purposes of this review, PSCI was pragmatically defined as any stroke-attributed cognitive impairment identified by study authors as post-stroke cognitive impairment and confirmed at baseline through study eligibility criteria, clinical diagnosis, or author report. Given the absence of internationally agreed diagnostic criteria for PSCI, no fixed threshold for cognitive severity, functional impairment, or time since stroke was imposed.^2,8 The boundary between PSCI and post-stroke dementia remains uncertain, with considerable overlap in definitions and diagnostic approaches across the literature.² Consistent with the registered review scope, studies primarily targeting post-stroke dementia populations were excluded, while recognising existing core outcome set initiatives in dementia and aphasia populations.^11,12

Eligibility is restricted to randomised and quasi-randomised trials and their protocols to align with the COS-PSCI programme focus on effectiveness trials; trade-offs are addressed in the limitations. We acknowledge that a methodologically distinct body of PSCI intervention trials is indexed in Chinese-language databases not covered by the English-language sources searched here; the implications for outcome and OMI capture are addressed in the limitations. Full eligibility criteria are outlined in Table 1.

Search strategy

Databases and Interfaces

Searches will be conducted in CINAHL (via EBSCOhost); PsycINFO (via EBSCOhost); PubMed (via the National Library of Medicine interface); Scopus (via Elsevier); Cochrane Central Register of Controlled Trials (CENTRAL, via the Cochrane Library); and ClinicalTrials.gov. These sources were selected to provide broad coverage of published, unpublished, and ongoing randomised and quasi-randomised controlled trials relevant to PSCI, balancing retrieval breadth against feasibility for a Phase 1 scoping review whose aim is to characterise outcome and OMI practice.

A publication date limit of 2020–2025 was applied as a deliberate design choice rather than solely a pragmatic constraint. As a scoping review, the objective is to map the extent and nature of current outcome and outcome measurement instrument (OMI) practice,¹⁴ rather than to list every outcome ever used in PSCI trials. The volume of eligible records published since 2020 is sufficient to characterise contemporary measurement practice and the outcome heterogeneity that motivates COS development. This scoping review represents the trial-literature strand of the COS-PSCI evidence base; overall comprehensiveness is achieved through triangulation with qualitative research, PPI input, and pre-Delphi consolidation. Restricting to recent literature therefore characterises current practice without compromising the completeness of the final output.

Search strategies were developed collaboratively by the research team and iteratively refined in consultation with an experienced health sciences librarian and reviewed for comprehensiveness. Searches will be constructed using a combination of controlled vocabulary (e.g., MeSH terms and database-specific subject headings) and free-text keywords. Boolean operators (AND, OR), truncation, phrase searching, and proximity operators will be applied as appropriate for each database platform. Strategies will be adapted to reflect differences in indexing systems, syntax, and interface functionality across databases. The core search concepts and terms for CINAHL are presented in Table 2; the complete search strategies for all databases will be made available on publication.

Trial Protocols and Registrations

Where trial protocols or registry records are identified, supplementary searches will determine whether a corresponding full trial publication exists. If a full results publication is available, the published article will be treated as the primary study record. The protocol and/or registry entry will be excluded from included-study counts and documented in the PRISMA flow diagram with the exclusion reason recorded as ‘Protocol – results publication available’. Formal comparison of prespecified versus reported outcomes will not be undertaken, given the scoping nature of this review.

Screening

Search results will be imported to Covidence and deduplicated. Title and abstract screening will be performed independently by two reviewers in Covidence, followed by full-text review. Reasons for exclusion at full-text screening will be recorded and reported in the final review. Any inconsistencies will be resolved through discussion amongst screeners. If consensus cannot be reached, a third reviewer (SH) will act as arbitrator. The selection process will be illustrated in a PRISMA-ScR flowchart.¹⁵

Data extraction

Data extraction will be conducted using a study-specific extraction form developed in Covidence. The form is informed by Core Outcome Set–STAndards for Development (COS-STAD) recommendations,¹⁶ COMET Initiative methodological guidance,⁹ and the Template for Intervention Description and Replication (TIDieR) framework.¹⁷ Data will be extracted independently by two reviewers, with disagreements resolved by discussion within the research team until consensus is reached. The extraction form will be piloted on a random sample of five included studies prior to full data extraction; any amendments made during piloting will be documented.

The following data items will be extracted:

Data synthesis

Data synthesis will proceed in sequential stages, aligned with the outcome harmonisation workflow.

Overview

Data will be synthesised descriptively, in accordance with JBI methodology for scoping reviews.^14,18 Although construct-level outcome harmonisation extends beyond purely descriptive mapping, this approach is explicitly recommended within core outcome set methodology to support transparent and reproducible consensus development.⁹

No statistical meta-analysis or quantitative transformation is planned. Outcome data will be exported from Covidence in CSV format and imported into a purpose-built COS-PSCI master Excel dataset designed to support transparent synthesis. This master dataset will contain linked worksheets for study characteristics, intervention details, and a master outcome log. Each outcome instance will be recorded as a single row linked to study authors, publication year, associated OMI, primary or secondary designation, and assessment timepoints.

The following terms in Table 3 are used throughout this review to distinguish different levels of outcome reporting:

Given the substantial variation in outcome terminology, scope, and measurement in PSCI intervention research, it is important to define the level at which outcomes will be grouped early in COS development.^9,19 In this review, outcomes will be grouped at the level of clinically meaningful constructs rather than broad domains or individual tasks. This approach is consistent with COS guidance, which emphasises that grouping decisions should be clearly defined and applied consistently throughout the review and consensus process.^16,19 Focusing on conceptual meaning, rather than study-specific wording or measurement tools, aims to reduce unnecessary variation while retaining distinctions meaningful to patients, clinicians, and researchers.

Outcome organisation and harmonisation workflow

Stage 1: Verbatim Outcome Extraction

All outcomes will be extracted exactly as reported in the included studies using Covidence and exported into a structured Excel master dataset for analysis. Retaining original wording at this stage ensures that subsequent categorisation, construct grouping, and long-list development remain transparent and fully traceable to source publications. Outcomes will then be harmonised through the six-stage workflow outlined in Table 4.

Where a study reports an OMI without specifying the outcome construct (e.g., Trail Making Test Part B without stating “executive function”), the entry will be marked with an asterisk (*). Because an outcome is defined by its underlying construct rather than the instrument used to measure it,^9,13 this asterisk distinguishes author-defined outcomes from instrument-only citations, ensuring analyses reflect intended outcome constructs rather than measurement practices. Starred entries will be excluded from outcome-label frequency analyses but retained in the OMI summaries. This pragmatic approach reduces misclassification where outcome constructs are not explicitly reported, consistent with prior stroke-cognition reviews that classified neuropsychological tests into domains using established frameworks and team adjudication.^5,20 Assignment of the asterisk (*) will be completed independently by two authors at data extraction stage.

Stage 2: Broad Outcome Categorisation

Seventeen broad outcome categories will be defined a priori to provide an organisational structure for the anticipated volume and heterogeneity of outcomes in PSCI intervention research. The categories were informed by established neuropsychological classification frameworks,^21,22 outcome groupings used in previous stroke and cognition intervention reviews,^7,20,23 the International Classification of Functioning, Disability and Health (ICF),²⁴ and the COMET taxonomy of outcome domains.¹³ They are listed in Table 5. During harmonisation, separate Excel workbooks will be created for each broad outcome category, each containing worksheets to support outcome organisation, construct grouping, and traceability.

Where extracted outcomes do not clearly fit a predefined category, they will be assigned temporarily to a provisional ‘unclassified’ category and resolved through discussion. OMI-only outcomes (see above) will be assigned to the most appropriate category independently by two researchers based on the primary measurement construct of the instrument, with disagreements resolved through discussion.

These categories serve solely as an organisational structure for managing and navigating the extracted data. They do not define, constrain, or predetermine the structure of the COS-PSCI consensus stages; outcomes will not be merged, removed, or prioritised at this stage.

Stage 3: Identification of Unique Outcome Labels

Within each category, spreadsheet filtering will be used to identify unique outcome labels, defined as each distinct phrasing used across the included studies. Two labels will be retained as separate unique labels where they differ in phrasing, specificity, or implied construct, even if they appear conceptually related.

Labels will be collapsed only where the difference is purely surface-level and does not alter lexical content, specifically: (i) spelling or spacing variants (e.g., “visuo-spatial” vs “visuospatial”); (ii) capitalisation or punctuation; (iii) grammatical inflection of the same root word (e.g., “cognitive function” vs “cognitive functioning”); and (iv) obvious typographic errors or duplicated words. Labels will be retained as separate where they differ in any substantive word, qualifier, or specifier (e.g., “global cognition” vs “global cognitive function”; “memory” vs “verbal memory”), even where the underlying meaning may appear related, because such differences are resolved at construct grouping (Stage 4) rather than here. Where a collapse decision is ambiguous, the label is retained as separate and the decision recorded in the traceability log.

This approach differs from definitions of “unique outcomes” that merge outcomes sharing the same underlying meaning and context²⁵ and is intended to preserve label-level variation in outcome terminology prior to construct-level grouping. These unique outcome labels represent distinct units of outcome terminology rather than conceptually distinct outcomes and are used to characterise heterogeneity in outcome reporting prior to construct-level grouping.

Unique-label decisions will be independently verified by a second researcher on a minimum 20% sample, purposively weighted towards high-volume and conceptually complex categories. All discrepancies will be recorded in the traceability log and resolved through discussion. A verification sample is appropriate at this stage because label identification is largely mechanical (collapsing only superficial variation) and every source-to-label link remains documented.

Stage 4: Grouping of Conceptual Outcome Constructs

Unique outcome labels will be organised according to their underlying constructs, in line with COMET guidance, which emphasises clinically meaningful groupings over measurement instruments or superficial terminological differences.^9,13 Labels will be merged where they meet the rules for determining whether two outcome labels refer to the same construct ( Table 6) (e.g., “delayed recall” and “recall ability”), whereas conceptually distinct constructs (e.g., verbal versus visuospatial memory) will be retained separately. For closely related but theoretically distinct constructs, particularly within higher-order domains such as executive function, separation will be maintained to minimise inappropriate aggregation. All verbatim outcome labels, assigned constructs, and the rationale for each decision will be recorded in a traceability log within each category workbook.

Construct assignment will follow pre-specified classification decision rules applied across all outcome categories. These rules prioritise author-defined outcome intent, as reflected in the verbatim label, with outcome measurement instruments (OMIs) used only as a secondary reference where the label is ambiguous or maps to multiple constructs. This preserves the distinction between the intended outcome and its measurement and reduces the risk of OMI-driven reclassification.

The rules in Table 6 and Table 7 were developed for this review to operationalise COMET guidance on construct-level grouping decisions.^9,13

These examples are provided for illustrative purposes only and do not predetermine classification, which will be applied systematically across all extracted outcomes using the predefined criteria.

Initial grouping will be undertaken by the lead researcher and independently reviewed by a second researcher on a minimum 20% sample, purposively weighted towards high-volume, conceptually complex categories (e.g., executive function), where construct boundaries are least clearly defined, with all decisions and rationale recorded in the traceability log. Classification will be informed by established frameworks and the broader stroke and cognition literature.^20–22,26 Disagreements will be documented and resolved through team discussion, and the proportion requiring adjudication will be reported. The application of predefined rules alongside a full audit trail of label-to-construct decisions is intended to support transparency and auditability. Although independent duplicate coding is undertaken on a minimum 20% sample, every label-to-construct assignment in the complete preliminary long list is subsequently examined and verified by the full research team at Stage 5, so that no construct assignment enters the long list without team-level review.

Stage 5: Internal Review & PPI Consultation

The research team will review the complete preliminary long list, examining every construct assignment, to verify construct-level grouping decisions, refine terminology, and adjudicate any misclassifications identified during harmonisation. All decisions and rationale will be documented in the traceability log.

The reviewed long list will then be presented to people with stroke and carers from the established UL Stroke PPI group with lived experience of PSCI. PPI contributors will assess the list for completeness from a lived experience perspective, flagging any outcomes that matter to stroke survivors or carers but are absent from the trial literature; such outcomes will be carried forward for consideration during the qualitative phase and/or long-list consolidation.²⁷ It should be noted that the preliminary long list at this stage represents the trial-literature strand of the COS-PSCI evidence base only and will subsequently be supplemented by outcomes identified through qualitative research with stroke survivors and carers. An evidence synthesis process will then lead to the final consolidated list for Delphi. PPI contributions will be reported in the final programme output in accordance with the GRIPP2 short-form checklist.²⁸

Stage 6: COMET Taxonomy Mapping and Final Output

The consolidated preliminary long list of outcomes will be mapped to the COMET taxonomy of outcome domains¹³ through research team discussion. The taxonomy classifies outcomes across five core areas: physiological/clinical, life impact, resource use/economic, adverse events, and mortality. Where a construct spans more than one core area, primary classification will reflect dominant construct intent, with secondary classifications noted. All mapping decisions and rationale will be recorded in the traceability log.

The final output will be a fully traceable, COMET-mapped preliminary long list of COS-PSCI outcomes constituting the trial-literature evidence base for consensus. This will include: all verbatim outcome labels linked to their harmonised constructs; outcome instance counts and reporting frequencies; COMET taxonomy classifications; an overview of included studies and key study characteristics; descriptive summaries of assessment timepoints and follow-up; and a structured catalogue of outcome measurement instruments.

Outcome measurement instruments

Each OMI reported in the included studies will be extracted verbatim, linked to the COS-PSCI long-list outcome it is intended to measure and standardised in three steps: (1) spelling, abbreviation, and formatting variants of the same instrument will be merged under a single label (e.g., “MMSE” and “Mini-Mental State Examination”); (2) subtests will be kept separate where they measure different constructs (e.g., Trail Making Test Parts A and B), or where study authors report and analyse them as distinct outcomes; otherwise, they will be merged under the parent instrument; and (3) translated or adapted versions will be retained separately only where they are formally validated national versions with distinct normative data.

For each instrument variant, citation frequency (i.e., the number of times it is paired with an outcome across studies) will be calculated. In addition, distinct instrument variants per construct will be recorded to quantify measurement fragmentation. Instrument-use frequencies will be calculated across all extracted rows, including OMI-only (starred) entries.

Assessment timepoints

Assessment timepoints will be extracted for all included studies and classified as: baseline; post-intervention; short-term follow-up (≤3 months post-intervention); medium-term follow-up (>3–6 months); or long-term follow-up (>6 months). Where no follow-up beyond immediate post-intervention assessment is reported, the study will be classified as having no defined follow-up. These data will inform discussions among stakeholders about current norms in follow-up practice and provide an empirical basis for discussions about appropriate assessment timepoints during Delphi consensus.

Quality assessment

No critical appraisal of individual sources of evidence will be undertaken, consistent with standard scoping review methodology, which aims to map the extent and nature of evidence rather than evaluate its quality.

Limitations

Several limitations reflect deliberate design choices appropriate to a Phase 1 scoping review focused on characterising contemporary outcome and outcome measurement instrument (OMI) use.

First, searches are restricted to English-language publications, which may under-represent PSCI trials indexed in non-English databases, particularly for Chinese-language literature, which is indexed in databases not covered by the sources searched here. As the review focuses on outcome selection and measurement practices, this limitation may influence the range and frequency of outcomes and OMIs identified and should be considered when interpreting findings.

Second, the search is limited to studies published between 2020 and 2025 to capture current outcome measurement practice. While this timeframe is sufficient to characterise contemporary outcome selection and heterogeneity, it does not capture all outcomes historically used in PSCI research.

Third, eligibility is restricted to randomised and quasi-randomised trials and their protocols, reflecting the programme’s focus on outcome standardisation in effectiveness trials. This enhances comparability across studies but excludes non-randomised intervention designs, including single-case and early-phase rehabilitation studies, which may report additional outcomes.