Development of core outcome sets for vision screening and assessment in stroke: a Delphi and consensus study

Objectives Visual impairment following stroke is common with a reported incidence of visual impairment in 60% of stroke survivors. Screening for visual impairment is neither routine nor standardised. This results in a health inequality where some stroke survivors receive comprehensive vision assessment to identify any existent visual problems while others receive no vision assessment leaving them with unmet needs from undiagnosed visual problems. The aim of this study was to define two core outcome sets (COS), one for vision screening and one for full visual assessment of stroke survivors. Design A list of potentially relevant visual assessments was created from a review of the literature. The consensus process consisted of an online 3-round Delphi survey followed by a consensus meeting of the key stakeholders. Participants Stakeholders included orthoptists, occupational therapists, ophthalmologists, stroke survivors and COS users such as researchers, journal editors and guideline developers. Setting University. Outcome measures COS. Results Following the consensus process we recommend the following nine assessments for vision screening: case history, clinical observations of visual signs, visual acuity, eye alignment position, eye movement assessment, visual field assessment, visual neglect assessment, functional vision assessment and reading assessment. We recommend the following 11 assessments for full vision assessment: case history, observations, visual acuity, eye alignment position, eye movement assessment, binocular vision assessment, eye position measurement, visual field assessment, visual neglect assessment, functional vision assessment, reading assessment and quality of life questionnaires. Conclusions COS are defined for vision screening and full vision assessment for stroke survivors. There is potential for their use in reducing heterogeneity in routine clinical practice and for improving standardisation and accuracy of vision assessment. Future research is required to evaluate the use of these COS and for further exploration of core outcome measures.

Strengths and limitations of this study ► This study followed robust methodologies in accordance with Core Outcome Measures in Effectiveness Trials-initiative guidelines. ► Two core outcome sets are produced for vision screening and for full assessment of stroke survivors with visual impairment. ► There is potential for their use with other types of acquired brain injury causing visual impairment. ► Attrition rates were moderate but similar to other Delphi surveys. ► Larger response numbers, including international participants, would be of benefit.

AbStrACt
Objectives Visual impairment following stroke is common with a reported incidence of visual impairment in 60% of stroke survivors. Screening for visual impairment is neither routine nor standardised. This results in a health inequality where some stroke survivors receive comprehensive vision assessment to identify any existent visual problems while others receive no vision assessment leaving them with unmet needs from undiagnosed visual problems. The aim of this study was to define two core outcome sets (COS), one for vision screening and one for full visual assessment of stroke survivors. Design A list of potentially relevant visual assessments was created from a review of the literature. The consensus process consisted of an online 3-round Delphi survey followed by a consensus meeting of the key stakeholders.
Participants Stakeholders included orthoptists, occupational therapists, ophthalmologists, stroke survivors and COS users such as researchers, journal editors and guideline developers. Setting University. Outcome measures COS. results Following the consensus process we recommend the following nine assessments for vision screening: case history, clinical observations of visual signs, visual acuity, eye alignment position, eye movement assessment, visual field assessment, visual neglect assessment, functional vision assessment and reading assessment. We recommend the following 11 assessments for full vision assessment: case history, observations, visual acuity, eye alignment position, eye movement assessment, binocular vision assessment, eye position measurement, visual field assessment, visual neglect assessment, functional vision assessment, reading assessment and quality of life questionnaires.
Conclusions COS are defined for vision screening and full vision assessment for stroke survivors. There is potential for their use in reducing heterogeneity in routine clinical practice and for improving standardisation and accuracy of vision assessment. Future research is required to evaluate the use of these COS and for further exploration of core outcome measures.

IntrODuCtIOn
Visual impairment is common in stroke occurring in up to 73% of stroke survivors. 1 Visual impairment is typically categorised into impairments of central vision, eye movements, visual fields and visual perception. 2 Vision is arguably our most important sense. Visual impairment results in impaired activities of daily living with reduced quality of life through loss of independence, greater risk of trips and falls and accidents. [3][4][5][6] This leads to loss of independence and potentially results in social isolation and depression. 5 6 The primary focus of stroke rehabilitation is often occupational therapy and physiotherapy to mobilise patients, improve limb function and balance and engage in activities of daily living plus speech and language therapy for communication difficulties. 7 Many rehabilitation strategies require visual input, for example to safely mobilise around potential obstacles, recognise depth and position of objects and recognise visual cue cards. Given so many stroke survivors have visual impairment it is important to screen for this at an early time point post-stroke onset with the aim to optimise the rehabilitation process.
The recent Impact of Visual Impairment after Stroke (IVIS) study reported specialist orthoptist vision screening is possible at a median of 3 days post-stroke onset with the majority of stroke survivors being assessed within 1 week of stroke onset. 1 This study used standardised visual assessment methods with portable equipment to be used at the Open access patient's bedside. There is, however, no standardised visual screening assessment for post-stroke visual impairment. In one UK survey it was found that 45% of stroke services provided no formal vision assessment for stroke patients. 8 A further survey of practice identified that only 7% of stroke units had a policy relating to vision assessment and management. 9 Both surveys showed lack of standardisation for vision assessment and treatment for stroke survivors. The National Stroke Strategy argues that vision and visual perceptual difficulties are components requiring multifaceted stroke specific rehabilitation and support. 10 The Royal College of Physicians recommend that every patient with stroke has a practical assessment of vision and examination of the visual field. 11 On the basis that there is no consensus on how to adequately screen for visual impairment after brain injury, the aim of this study was to achieve consensus on the content of vision screening and full vision assessment for stroke survivors in order to better identify visual impairment. Screening and/or full vision assessments are to be undertaken at any time point post-stroke onset with the intention that identification of visual impairment enables prompt access to earlier visual rehabilitation options. One approved process to reach consensus on screening and assessment for specific conditions is through the development of core outcome sets (COS). 12 COS indicate the minimum that should be measured and reported in all studies of a specific condition. The overall purpose of a vision screening and full assessment COS is to improve routine care in the UK National Health Service (NHS) through standardisation of assessments.
In this study we report the results of a Delphi process and consensus meetings in the development of a COS for vision screening and a COS for full vision assessment of stroke survivors. Vision screening was defined as the assessments considered important for use by clinicians not working in eye care settings and without formal experience or training in performing eye tests. Full vision assessment was defined as the assessments considered important for use by clinicians who had formal eye care training and were principally based in eye clinics and providing more detailed assessment than possible with screening assessment of basic levels of visual function.

MethODS ethical approval
Informed consent was obtained for the Delphi surveythe participant checked a consent tick box on the opening page of the survey. Informed written consent was obtained from all participants in the consensus meeting.
Development of the COS involved three phases: (1) the generation of a comprehensive list of outcomes; (2) a Delphi survey involving three rounds to gain consensus as to which outcomes are most important; and (3) patient and professional consensus meetings to agree a final COS. A protocol for the development of this COS project was written by the steering committee, registered in the Core Outcome Measures in Effectiveness Trials (COMET) initiative website (http://www. comet-initiative. org/ studies/ details/ 275? result= true) and available as open access (http:// pcwww. liv. ac. uk/~ rowef/ index_ files/ Page356. html). When developing this COS, we followed the minimum set of development standards set out by Core Outcome Set-STAndards for Development Statement, and report the results against the Core Outcome Set-STAndards for Reporting guideline. 13 14 Steering group A steering group was set up to inform the development of the various stages of this study and to discuss the results at each phase of the study. It comprised a representative from clinical professions involved in the vision screening of patients with acute stroke brain injury (orthoptist, occupational therapist and neuro-ophthalmologist), two stroke survivors and COS users (clinical and academic researchers, systematic reviewer and clinical guideline developer); eight members in total. Clinicians were identified from a national orthoptist stroke clinical advisory group, a regional allied health profession research network and regional ophthalmology research network. Stroke survivors were identified from a national stroke/ vision patient and public involvement panel. COS users were identified from local university staff with links to the COMET initiative and/or those involved in national clinical guideline panels. The COMET initiative brings together people interested in the development and application of agreed standardised sets of outcomes (http:// www. comet-initiative. org/).

Patient and public involvement
The development of this research question was informed by patients' priorities, experience and preferences outlined by a National Institute for Health Research James Lind Alliance priority setting partnership. Patients were involved in the design of this study as members of the steering group set up specifically for the study and outlined previously. Further, patients were involved in the recruitment to and conduct of the study through the VISable patient and public involvement panel which is described below in the next section. Results will be disseminated to study participants by emailing a copy of the full publication of this research along with a lay summary of the results.

Stakeholders
For participation in this study, we sought the input of clinicians (eligibility=those involved in the vision screening of patients with acute brain injury), patients (eligibili-ty=those with brain injury causing visual impairment), caregivers (eligibility=those caring for people with brain injury causing visual impairment), editors of journals and COS users. Clinicians, journal editors and COS users were invited to participate through advertisements circulated via national professional societies (eg, British and Irish Orthoptic Society, Royal College of Occupational

Open access
Therapists), via international orthoptic editorial groups (eg, Strabismus and British and Irish Orthoptic Journal) and via regional allied health profession research networks (Council for Allied Health Professions Research). Stroke survivors and caregivers were invited through advertisements circulated via a national stroke/vision patient and public involvement panel (VISable) and through national charities (eg, the Stroke Association, Royal National Institute for the Blind).

Phase 1 Literature review
In order to develop a preliminary list of outcomes for a Delphi survey we undertook an overview of seven systematic reviews of studies/trials reporting vision screening, assessment and treatment of post-stroke visual impairment. 2 15-20 We extracted 119 outcomes, many of which were variations on test choices for specific visual functions. Therefore these were combined into one outcome domain (table 1).
COS do not address how tests should be defined or measured. Thus this combination decision was made by the steering group who decided through discussion that test choice would vary on a case-by-case basis dependent on the patient's ability to undertake the test and that COS should consider only the general assessment needed and not the specific methods of measurement. For example, logMAR, Snellen, fixation/following, grating cards and fixation distance were outcomes combined under one outcome domain of 'visual acuity' assessment.
This process produced a list of 22 domains for vision screening and a list of 24 domains for full visual assessment. These lists were then circulated to the VISable patient and public involvement panel for approval and checking of writing for lay terms as the basis of the online survey development. VISable advised that patient-important outcomes were covered by the outcome domains and did not add any further outcomes to those already identified.

Phase 2 Delphi survey
We undertook a prospective consensus study using a Delphi process. Delphi is a structured process that is widely used in developing core outcomes sets. 21 The process aims to achieve consensus through the collection of stakeholder opinions.
SurveyMonkey (SurveyMonkey, 2015) was used as the online platform to administer the Delphi process. The survey was piloted for checks of ambiguity and appropriate use of lay language by the steering group and then released live.
An email outlining the project and survey was sent out via stakeholder networks (as outlined earlier in the stakeholder section) with a snowballing technique of onward roll-out from individual stakeholders. The Delphi survey consisted of three rounds ( figure 1). Rounds remained open for 10 weeks, with regular 2-week reminders sent to those that had partially completed or not completed to maximise response rates.
In round 1, the Delphi started with an introductory page, which outlined the purpose of the study and how to complete each section, as follows: "We are not asking you to provide detailed answers to these items, but to judge how important they are when assessing visual impairment in brain injury. Please view the items listed below and score their importance with 1 being not important and 9 being critical. After a brain injury, patients may first have visual screening to detect if a visual problem is present. Once a visual problem is found, patients may then receive a full visual assessment. In part 1a, please rate the items below as to how important you believe they are to a screening assessment. In part 1b please rate the items below as to how important you believe they are to a full visual assessment." All terms had explanatory notes to aid interpretation. All participants were asked to score each assessment in terms of importance and asked to identify any additional outcomes of importance that did not appear in the list of assessments. In the acknowledgement that definition of importance varies between individuals and between stakeholder groups, the Delphi survey was circulated widely to a variety of stakeholder groups to obtain greater consistency in responses. Additional outcomes added in round 1 were reviewed by the steering group to consider their relevance to the survey and to identify and remove duplicates.
In round 2, all data from round 1 were analysed and compiled by stakeholder groups; (1) stroke survivors/ carers, (2) stroke team professionals and (3) eye team professionals to allow different perspectives to be considered prior to re-rating. 12 The summarised results for each stakeholder group were provided to all participants who fully completed the survey in the first round. Each participant was also shown their personal original score for each domain. Participants had the opportunity to re-score based on the summary scores from the three stakeholder groups versus their previous personal scores. All participants were also asked to score on additional outcomes that were identified and added in round 1.
In round 3, the results from round 2 were analysed. Following analysis, the results from each stakeholder group were reviewed by the steering group and judged to be similar in terms of percentage spread across the responses of 1-9. Thus they were presented as one compilation of all responses. Each participant was shown their personal score from round 2 for each domain along with the summary scores from all other participants, and asked to score each domain again in terms of importance.
Scoring method: In each round, participants were asked to score the importance of each domain listed on a 9-point scale (1-3: not important; 4-6: important but not critical; 7-9: critical, as well as an 'unable to score' option). The scale was devised by the Grading of Recommendations Assessment, Development and Evaluation (GRADE) working group 22 to score the quality of Open access  Open access Figure 1 Flow chart of Delphi process across three survey rounds.
evidence for outcomes in systematic reviews and has been adopted in other COS development work research using Delphi methods. Methods of analysis: For each round of Delphi, descriptive statistics were used to summarise the results for each domain, including the percentage of participants scoring the domain at each possible response from 1 to 9.
Consensus was defined a priori; however, this information was not provided on the Delphi survey. Participants were aware of these cut-off values at the time of attending the consensus meeting: 'Consensus in' (ie, consensus that the domain should be included in the core set) was defined as greater that 70% of participants scoring as 7-9 and less than 15% of participants scoring 1-3. 'Consensus out' (ie, consensus that the domain should not be included in the core set) was be defined as greater than 70% of participants scoring as 1-3 and less than 15% of participants scoring as 7-9. All other combinations were seen as equivocal. The domains that were designated as 'consensus in' or seen as 'equivocal' were taken forward and discussed in more detail at the consensus meeting for inclusion into the final COS, one COS for screening and one COS for full assessment. Participants were reminded of all domains not reaching consensus as part of the Delphi process.

Phase 3 Consensus meeting
Representatives from the Delphi survey were invited to a face-to-face consensus meeting. All round 3 survey completers (n=51) were emailed an invitation. All stakeholders were reasonably represented. The format of the meeting included a short study overview, a presentation containing a summary of the results and number of domains reaching consensus from the survey. Each consensus domain was considered in turn, in order of their presentation in the Delphi survey, to ratify these results. Each remaining domain was then considered, in turn, with full discussion. Similar to the survey, each domain was considered as reaching consensus with 70% of participants voting in favour of its inclusion.
Each participant voted on every domain being asked to vote 'Yes' (this domain should be included in the COS), 'No' (this domain should not be included) or 'Unsure' using voting slips. After voting for all domains was completed, the results were collated during a rest break for the participants. When the meeting resumed, the results were presented to the group. Domains were retained or dropped when consensus was reached. Discussion and further rounds of voting, restricting the options to 'Yes' or 'No', were undertaken until consensus was reached on all domains. All domains retained were included in the final COS.
reSultS Figure 2 outlines a flowchart of the results for number of participants and number of domains.

Phase 1
Outcome identification An overview of systematic reviews produced a list of 119 outcomes. Outcomes that were variations on test choices for specific visual functions were combined into outcome domains. This produced 22 domains for vision screening and a list of 24 domains for full visual assessment.

Phase 2 Delphi survey
In total 123 participants registered for round 1 of the Delphi survey. Six did not complete the survey. The remaining participants comprised 79 orthoptists, 20 occupational therapists, 17 stroke survivors and 1 ophthalmologist. There were 20 males and 97 females with ages ranging from 18 to 84 years (figure 3). In round 2, 65 participants completed the survey-an attrition of 44.5% from round 1. These participants comprised 47 orthoptists, 10 occupational therapists, 7 stroke survivors and 1 ophthalmologist. In round 3, 51 participants completed the survey-an attrition of 56.4% from round 1 (78.5% from round 2). These participants comprised 39 orthoptists, 6 occupational therapists, 5 stroke survivors and 1 ophthalmologist.
Following completion of the survey of 46 domains in round 1, 12 additional outcomes were put forward by the participants for inclusion in round 2. These comprised five screening outcomes and seven full assessment outcomes. After round 2 of 58 domains, three outcomes were excluded; these were test variations already included within the existent domains. Four outcomes were combined; these were variations on test choices and could be combined into one domain-as per the process outlined in phase 1. No new outcomes were introduced. Following round 3 of 52 domains, consensus was achieved for 14 screening domains leaving ten domains for discussion, and consensus was achieved for 22 full assessment domains leaving six domains for further discussion.

Phase 3 Consensus meeting
The consensus meeting was a 1 day event held in Liverpool, UK with 12 participants comprising five occupational therapists (2 with research roles), 3 orthoptists (1 with a research role), 2 patients, 1 Cochrane editor and 1 facilitator. The facilitator did not take part in the voting. The objective of the meeting was to discuss and vote on the Delphi domains-the results of the Delphi survey had

DISCuSSIOn
A COS is an agreed minimum set of outcome measures that should be reported. By reporting a minimum set of measures, this reduces the heterogeneity of outcomes across studies, which, in turn, supports future evidence synthesis. [12][13][14] Once a COS is defined, it is important to achieve consensus on how the outcomes should be measured. 23 Delphi and consensus process methods have been used extensively for research-oriented COS but has also been reported in the development of a test battery for visual perception screening 24 and Traumatic Brain Injury (TBI)-related visual impairment. 25 To our knowledge this is the first study to use these methods to develop test batteries for general vision screening and assessment for post-stroke visual impairment.
The goal of this study was to agree a COS to be used in the vision screening and full assessment of visual impairment following stroke. After round 3 and the consensus meeting, stakeholders agreed with the proposed test battery for vision screening consisting of 9 domains and for full vision assessment consisting of 11 domains. The vision screening COS is intended for clinicians who have no formal training or experience in providing eye tests. However, it is acknowledged by expert consensus that the specific sections of visual inattention and functional vision assessments are key areas where assessments by members of the stroke team (eg, by occupational therapists) do not require formal eye training. The full vision assessment COS is intended for specialist vision assessment by professions such as orthoptists and ophthalmologists. Of importance, these COS are the core identified assessments to be made wherever possible when screening or fully assessing the visual functions of stroke survivors. These COS do not exclude further visual assessments which should be added as appropriate and relevant for the individual stroke survivor. Equally, some COS measures may not be achievable due to patient cognitive, communication and physical impairments which prevent assessment.
There are a number of strengths for this study. Two COS have been produced, one for vision screening and one for full vision assessment which addresses a gap in evidence-based practice for post-stroke visual impairment.
Each COS is composed of domains which were ranked by Delphi and consensus opinion from multidisciplinary experts and stroke survivors. As these COS are derived from clinical and research stakeholders, and patients and carers, they are potentially suitable for research in addition to clinical practice. Furthermore they are based on existent outcomes that are easily accessible for implementation, and accepted as validated clinical measures. Although these COS were developed for stroke, there is potential for their use with other types of acquired brain injury causing visual impairment.
There are limitations to this study. Participants were from UK and Ireland and largely represented views and practices within the NHS. Thus, assessments are those used clinically in these countries. It would be valuable to seek expert opinion from international participants regarding the key assessments and their specific outcome measures. However, we recognise that the assessment domains selected for the COS are those that are widely used internationally in clinical and research settings as evidenced from our initial literature review. 2 19 21 We had a limited response by ophthalmologists in the Delphi process, with none subsequently involved in the consensus meeting. However, within the UK and Ireland, orthoptists are the clinical professionals predominantly responsible for the diagnosis and management of stroke-related visual impairment. Thus the greater response from orthoptists in the Delphi survey and consensus process was expected and, we believe, provided robust information required for this study. In other countries, decisions should be taken as to whether involvement of other clinical professionals in the care of these patients would alter the COS lists. We also experienced attrition bias across the Delphi rounds. The attrition rate at round 2 was 44.5% and, by round 3, was 56.4% from the initial sample. High attrition rates however are common using Delphi methods and our rates can be considered within an acceptable range based on those reported for other COS developments. 26 Nevertheless a larger sample would have been preferable. Researcher bias is also a potential limitation. We aimed to limit this by providing a summary of results across all rounds of the Delphi survey and with final decisions left to the consensus meeting with experts. There can be risks from using Delphi methods in which participants can have very disparate views of each outcome. However, we sought a wide variety of participants across a number of stakeholder groups to achieve greater consistency in responses and balance potential outlier responses. Further, this core outcome development process included a final stage of consensus meeting such that decisions were not purely made from the Delphi responses.

COnCluSIOnS
This study reports the use of Delphi and consensus methods in the development of COS for vision screening and full vision assessment of stroke survivors. Vision screening comprises 9 assessment domains and full vision Open access assessment comprises 11 domains. These COS will facilitate standardisation of screening and assessment of poststroke visual impairment while also having the potential to reduce heterogeneity in assessment in future research. Further research is now required to evaluate the use of these COS and outcome measures.