Exploring the Use of Washington Group Questions to Identify People with Clinical Impairments Who Need Services including Assistive Products: Results from Five Population-Based Surveys

This study analyses the use of the self-reported Washington Group (WG) question sets as a first stage screening to identify people with clinical impairments, service and assistive product (AP) referral needs using different cut-off levels in four functional domains (vision, hearing, mobility and cognition). Secondary data analysis was undertaken using population-based survey data from five countries, including one national survey (The Gambia) and four regional/district surveys (Cameroon, Chile, India and Turkey). In total 19,951 participants were sampled (range 538–9188 in individual studies). The WG question sets on functioning were completed for all participants alongside clinical impairment assessments/questionnaires. Using the WG “some/worse difficulty” cut-off identified people with mild/worse impairments with variable sensitivity (44–79%) and specificity (73–92%) in three of the domains. At least 64% and 60% of people with mild/worse impairments who required referral for surgical/medical and rehabilitation/AP services, respectively, self-reported “some/worse difficulty”, and much fewer reported “a lot/worse difficulty.” For moderate/worse impairment, both screening cut-offs improved identification of service/AP need, but a smaller proportion of people with need were identified. In conclusion, WG questions could be used as a first-stage screening option to identify people with impairment and referral needs, but only with moderate sensitivity and specificity.


Introduction
In global health, alongside mortality and morbidity data, there is an increasing emphasis on addressing a third health indicator, "functioning" [1,2]. The International Classification of Functioning, Disability and Health (ICF) defines functioning as an umbrella term for body functions, structures, activities and participation; it denotes the interaction between an individual (with a health condition) and the environmental and personal Specifically, we assess: (1) the sensitivity and specificity of the WG questions (at different cut-off levels) to identify people with clinically assessed impairments (vision, hearing, mobility and cognitive domains); (2) the proportion of people identified by the WG questions (at different cut-off levels) who are in need of surgical/medical and/or rehabilitation/AP services.

Children and Adults
Clinical mobility assessment -Clinical mobility assessment -Clinical mobility assessment

Definitions
See Box 1 for definitions of vision, hearing, mobility, cognitive and WG terms used in this paper. Box 1. Definition of vision, hearing, mobility, cognitive and Washington Group terms used in this paper.

Washington Group Question Sets for Four Domains
Participants (or representatives if unable to self-report) were asked a series of questions from the WG sets. Adults > 17 years were asked either the WG Short Set, the Short Set-Enhanced or the Extended Set on functioning, and children aged 2-17 years were asked the Child Functioning Module [11]. See Box 2 for the WG questions analysed in this paper. The question sets also ask about use of glasses, hearing aids and mobility APs/assistance.
For the purpose of our analyses, we used the screening cut-offs of "some difficulty or worse" (herein referred to as some+) and "a lot of difficulty or worse" (a lot+) either without or with the corresponding AP (if the participant used them) in each of the four functional domains.

Vision
Distance vision: Presenting visual acuity (VA) (i.e., with correction, if available) was assessed using a tumbling E single optotype, on cards in India and Cameroon and on Peek Acuity mobile application in The Gambia [33]. In India and The Gambia mild or worse vision impairment (VI) was defined as presenting VA < 6/12 in the better eye; and in India, The Gambia and Cameroon, moderate or worse VI was defined as presenting VA < 6/18 in the better eye [34]. Pinhole vision was assessed for all participants with VI to identify individuals with uncorrected refractive error (URE), and in The Gambia a refraction was performed to record best corrected visual acuity (BCVA). In all three countries (Cameroon, India and The Gambia), participants with VI were examined by a trained eye care worker to determine the cause using the WHO protocol for the condition that is "easiest to treat" [35]. Those identified with URE were reported as needing distance glasses.
Near vision: Presenting (i.e., with near correction if available) binocular near vision was assessed in The Gambia survey only. A binary outcome of can or cannot see N8 at 40 cm (correctly identifies 4 out of 5 E optotypes) was recorded. Participants unable to see N8 were re-tested using an age-appropriate correction for near and recorded as needing near glasses (unmet need) or needing a change in prescription of existing near glasses (undermet need).

Hearing
In India and Cameroon, all-age participants were screened using Otoacoustic Emissions (OEA) Testing, and participants aged ≥ 4 years old who failed this underwent Pure Tone Audiometry (PTA) at 0.5, 1, 2 and 4 kHz to assess the presence and severity of hearing loss. The definition of moderate or worse hearing impairment (HI) was a pure tone average (at 0.5, 1, 2 and 4 kHz) of >31 dB for children (4 to 17 years) and >41 dB for adults (≥18 years) in the better ear [36]. Individuals with HI underwent examination by an ENT specialist to assess the cause and likely service needs, including hearing aids.
In Chile [14], PTA was tested using a mobile-based audiometry system HearTest [37] at 0.5, 1, 2 and 4 kHz in each ear. According to WHO's definition, mild or worse HI was defined as >25 dB in the better ear. All participants had their ears examined by an ENT resident or consultant.

Mobility
In Cameroon, India and Turkey, participants were asked six validated screening questions from the Rapid Assessment of Musculoskeletal Impairment survey tool (RAM) [38]. Anyone who screened positive underwent a standardised examination by a physiotherapist using the RAM protocol which includes head/neck, upper limb, lower limb/pelvis, trunk and spine assessment. The presence, severity (mild/moderate/severe) and cause/diagnosis of MSI, as well as the need for services and APs including wheelchairs and prosthetics (both upper and lower limb) was recorded. In Turkey an updated version of RAM was used [27].

Cognition
The Mini Mental State Examination (MMSE) is a brief neurophysiological test [39]. A short validated Chilean Mini Mental State Examination (SCh-MMSE) was developed by an audiologist in Chile [40] to include populations with low levels of literacy [41]. This includes six questions selected from the original 11 question version. The six questions evaluate: spatial and temporal orientation (day, month, year); -short-and long-term memory (3 word retention); -attention (inverse repetition of 5 numbers); -executive capacity (verbal order with 3 steps); -visual constructive capacities (copy of two circles).
Each of the questions has a score, with a possible maximum of 19 points; a total score <13 is considered "suspected cognitive impairment."

Data Analysis
Stata version 16.0 (StataCorp LP, College Station, TX, USA) was used to manage and analyse the data. The cluster design was accounted for in the analysis using the "svy" command.
To test whether the WG self-reporting questions, as a first-stage screen, are able to identify people assessed as having clinical impairments, we calculated sensitivity, specificity, positive predictive value and negative predictive values, with clinical impairment assessment being the reference.
To understand the extent to which the WG questions identify people who could benefit from referral for a specific intervention, we calculated the proportion of individuals who by clinical impairment assessment were found to need surgical/medical or rehabilitation/AP interventions who self-reported "some+ difficulty" or "a lot+ difficulty" for both mild+ and moderate+ impairment levels. For the purposes of our analyses, we restricted clinically assessed service and AP need to only those participants who responded to the corresponding WG question in each domain. Surgical/medical and rehabilitation/AP service needs, and need for five individual APs classified as "priority APs" by ATScale [42] (distance glasses, near glasses, hearing aids, wheelchairs and prosthetics), were clinically assessed according to cause, diagnosis and severity. Domain-specific details, used for our analysis, are provided below.
-Vision: For participants with vision loss due to cataract, surgical intervention was assigned. For participants with URE, distance glasses were assigned as the intervention. For participants with other causes of visual loss, e.g., glaucoma, both medical and rehabilitation services were recorded, and, for causes with no medical or surgical treatment possibilities, only rehabilitation services were assigned. -Hearing: Following the protocol used in RAHL [14], for participants with hearing loss due to chronic otitis media (dry/wet/possible Cholesteatoma), acute otitis media, otitis media with effusion, otitis externa, impacted wax and foreign body, surgical/medical intervention was assigned. Participants with sensorineural/mixed hearing loss in both ears, or unknown cause, were categorised as needing "referral to audiological rehabilitation services and likely hearing aids". In Cameroon and India, clinician-assessed hearing aid referrals were used. -Mobility/MSI: According to the RAM [38], surgical/medical and rehabilitation/AP interventions were clinically assessed based upon the examination with specific referral recommendations recorded by the clinician. For example, rehabilitation services included referrals to physiotherapy and environmental modifications, and APs including up to 11 mobility APs, such as wheelchairs, prosthetics, sticks/canes and orthotics.
In each of the functional domains, some participants were assessed to need both medical/surgical interventions and rehabilitation/AP services. Data on intervention need were not available for the cognition domain.

Ethics and Consent
This secondary analysis study received approval from the London School of Hygiene & Tropical Medicine. Each survey received separate approval from the London School of Hygiene & Tropical Medicine and the relevant ethics committees in each study country [10,27,28,30,31]. Written (signed or fingerprinted) informed consent was obtained from all participants or their proxies. Table 1 presents the survey details for each country. The sample size ranged from 538-9188 participants with response rates of 47% to 88%. In all five surveys at least half of the study population were female.

WG Questions to Screen for Clinical Impairment
The association between clinical impairment assessment and self-reported difficulty in functioning for each domain is presented in Table 2 with additional analyses in Supplemental Table S1.
Across the different impairments and study settings, using the WG category "some or worse" difficulty identified people with clinical impairments with a sensitivity range of 44% to 85%, and a specificity range of 65% to 92%. There was one exception of very low specificity (18%) for mild+ cognitive impairment (Chile). Using the more restrictive "a lot or worse" difficulty consistently, across impairment types and studies, reduced sensitivity (range 9-62%) and improved specificity (range 86-99.7%). "Near VI" was only measured in The Gambia and had low/very low sensitivity (39% and 3%) and high specificity (85% and 99.5%) using both WG cut-offs of some+ and a lot+ difficulty, respectively. Specific ranges in each domain were as follows. For distance vision, self-reported WG "some+ difficulty" seeing had good/high sensitivity (67-85%) and specificity (77-80%) when compared to clinical VI. Moving to a cut-off of "a lot+ difficulty" increased the sensitivity (99%) but radically reduced the specificity (10-39%). For hearing, "some+ difficulty" hearing had moderate/high sensitivity (44-83%) and good/high specificity (65-92%) when compared to clinical HI. Moving to a cut-off of "a lot+ difficulty" increased the sensitivity (98-99.7%) but reduced the specificity (9-50%). For mobility, some+ difficulty" walking had good/high sensitivity (64-84%) and specificity (76-90%) when compared to clinical MSI. Moving to a cut-off of "a lot+ difficulty" increased the sensitivity (97.6-99.7%) but reduced the specificity (16-62%). Table 3 shows the proportion of participants with identified clinical impairment who were assessed to need either medical/surgical interventions (e.g., cataract surgery) and/or rehabilitation/AP services (e.g., hearing aids) who were identified by the WG self-reported questions. Over three-fifths of participants (range 62-96%) with impairments who needed a surgical/medical intervention self-reported "some+ difficulty", whereas much fewer (range 13-64%) reported "a lot+ difficulty" across the studies. Of those who needed rehabilitation services and/or APs, 60-86% of persons with impairments self-reported "some+ difficulty" and much fewer (5-62%) reported "a lot+ difficulty."

WG Questions to Screen for Service/Intervention Needs
The detailed results for each country and service/intervention need are shown in Supplemental Table S2. Additionally, only 39% of people who were clinically assessed to need functional near vision services reported "some+ difficulty".
Across all domains and countries, the overall population-level need for rehabilitation/AP services (2-43%) was approximately equal to or greater than the need for surgical/medical services (2-10%), except for moderate+ VI in India and The Gambia (see Supplemental Table S3).

Identification of Persons Needing Specific Assistive Products
The proportion of people who were assessed as having a clinical impairment who needed glasses (distance and near), hearing aids, wheelchairs and protheses that were identified by WG question "some+ difficulty" is presented in Table 4, with all categories presented in Supplemental Table S4. Abbreviations: UL/LL = upper limb/lower limb; * "Need" includes both "unmet need" and "undermet need" for each assistive product;ˆDenominator includes participants who needed specific assistive products and who completed Washington Group questions in the respective functional domain; ** Hearing aid need includes all participants who needed a referral for audiological services and likely hearing aid need.
Of people with mild+ VI who were clinically assessed to need distance glasses, 59-76% reported having "some+ difficulty" seeing, and of those with moderate+ VI, it was 60% to 81%. Of those who were clinically assessed to need near glasses, only 34% of people with near VI reported having "some+ difficulty" seeing. Of the people with mild+ HI who were clinically assessed to likely need hearing aids, 60% reported having "some+ difficulty" hearing, and, of those with moderate+ HI, 72% to 81% reported having "some+ difficulty" hearing. Overall in three countries, 14 of 15 (93%) people clinically assessed as needing wheelchairs, and 4 of 5 (80%) of people who needed a prosthesis reported having "some+ difficulty" walking.

Use of Washington Group Questions for Initial Screening in Population-Based Clinical Assessment Surveys
Overall, using the "some or worse" difficulty cut-off for WG questions demonstrates better agreement with the presence of clinical impairments and service/AP referral needs than using "a lot or worse" difficulty. This pattern remained true for both mild+ and moderate+ impairments in each of the three functional domains of vision, hearing and mobility. Use of "some or worse" WG screening cut-off would identify at least 60% of people with mild+ impairment who could potentially benefit from a service/intervention, but with many false negatives. In contrast, using the cut-off "a lot or worse" difficulty would miss the vast majority of people with service needs. Similarly, the proportion of eligible people identified through WG screening increased when using moderate+ impairment threshold, but a smaller proportion of people with need were identified.
Specifically, our study explored Mactaggart et al.'s recommendation to use a WG cutoff of "some difficulty or worse" as first-stage screening followed by clinical impairment assessment in the same functional domain to identify people with disabilities, based upon a moderate+ impairment threshold [10]. Though our overall findings were congruent with the general recommendation to use "some difficulty or worse" cut-off, Mactaggart et al.'s research anticipated that at least 80% of people with disabilities would be identified using this method, whereas our study found much fewer people with impairment (44-79%) and people with service/AP needs (60-82%) would be identified using updated recommended mild+ impairment thresholds [2,[12][13][14]27,43,44]. Therefore, it appears use of this screening recommendation might not be transferrable to a mild+ impairment threshold.
There are few population-based prevalence studies that allow comparison with our findings to ascertain what might be a recommended "minimum" identification screening threshold. A few hearing impairment studies exist, and one study similarly found a selfreport screen identified 80% of people with clinically assessed hearing loss [45]. However, regardless of the threshold, it could be argued that using a two-stage screening might indicate the proportion of "service demand" in a given population. Though literature is limited, the rationale for this statement could be that people who report a difficulty in functioning may be more likely to consider that they need services, and therefore uptake related services, creating a "service demand." In contrast, people who report having no difficulty may be unlikely to uptake referrals for services. For example, a study in New Zealand found that measuring unserviced health needs through a patient-initiated general practitioner consultation was directly relevant to service planning because the gaps identified reflected clinically indicated services that patients want and need [46]. Similarly, this relationship has been evidenced for mental health, where perceived mental health need has been shown to be predictive of seeking services [47,48]. However, research has also recognised that demand-based health needs planning could increase access and utilisation service gaps and inequities between social groups in populations; therefore, it is recommended that demand-based health needs planning should also be coupled with need-based allocation of resources and a focus on the empowerment of groups who have greater needs [49]. There is a need to further explore this relationship in the context of collecting population-based data to inform service planning.
Our findings were closely aligned with Sprunt et al.'s findings which found variable sensitivity and specificity overall and by impairment severity (none to severe) when exploring the use of the WG question set CFM "some difficulty" as a screening for schoolaged children in Fiji with impairments [16][17][18][19][20]. Following their analysis, Sprunt et al. recommended to use the WG first-stage screening of "some difficulty" in a minimum of one functional domain, and that subsequently additional wide-ranging clinical assessments should be administered by the school system in Fiji in order to pick up unidentified and unexpected impairments [16][17][18][19]. Therefore, it is recommended that future research explore this additional analysis as an option for population-based multi-domain survey two-stage screening, whilst parallel research should also consider the feasibility, affordability and acceptability of administration. Further, the Fiji study also specifically highlighted the importance of including environmental factors specific for learning and support needs [19]. Therefore, other potential screening and clinical assessment tools incorporating more environmental factors should be explored.
Our study has shown that the proposed use of the WG "some+ difficulty" as a first stage screening could be a practical and feasible option to reduce the survey duration, cost and response burden compared to conducting multi-domain impairment assessments. However, our findings have also shown that this approach will not capture everyone with impairments and service/AP needs in each domain so it will not be appropriate for surveys that aim to estimate prevalence of impairment and service/AP need.

Further Gaps in the Survey Measurement Approaches
Our analysis highlighted gaps and recommendations to be considered in the collection of data on functional service needs.
First, rehabilitation/AP needs are often neglected, but this paper highlights that need was at least equal to or higher than the surgical/medical service need across all countries and domains. This further highlights the importance of increased data collection efforts, using robust methodology, to assess need in different settings.
Second, adjustments to the first stage screening questions might be needed. For example, in the vision domain, there was poor identification of near VI service/AP needs using both "some+ difficulty" and "a lot+ difficulty" WG cut-off levels (39% and 3%, respectively). This may be expected as the specific WG question asked in general about difficulty seeing (see Box 2). Therefore, it is recommended that a specific near vision screening question is included for surveys that intend to assess near VI and the need for services and AP to improve the sensitivity. The WG extended question set provides an optional vision question that asks about a functional activity related to use of near vision-difficulty clearly seeing the picture on a coin-so this should be incorporated at a minimum [11]. In the mobility domain, the extended WG questions for difficulty walking over certain distances and/or climbing stairs were compared to MSI (Box 2). Though walking could be one activity limitation for a person with MSI, other possible activity limitations also assessed in WG questions include self-care, upper body, pain and fatigue. Future analysis should explore whether combinations of these questions, in addition to environmental questions, increase identification of people with MSI and service needs by improving the sensitivity and specificity. Furthermore, consideration should be given as to whether additional first-stage screening options might provide better prevalence estimates of need, such as the validated RAM's six self-reported first-stage screening questions [38].
Third, the development of a multi-domain modular survey tool would allow flexibility, depending on aim of the data collection and time and resources available. For example, this could include options to (i) include or not include the first stage WG screen in the survey and (ii) select which functional domains to include.
Fourth, this paper used secondary analysis of datasets and the analyses were therefore constrained by data that were collected. For example, in the cognitive domain, there was very low specificity (18%) for mild+ cognitive impairment without service recommendations. Though the SCh-MMSE was contextually developed for low literacy populations which is a strength, using this screening tool as a "gold standard" in our study has limitations. Future studies are needed to explore and compare additional cognitive assessment and screening tools which include assessments of cognitive service/AP needs as well.
Fifth, when using the WG questions in service/AP need surveys, consideration could be given to ask about the presence of functional difficulties without the use of assistance or APs. For example, Danemayer et al.'s systematic review recommended AP indicators of total need and met need, as well as unmet and undermet need for service/AP need, are collected in population-based surveys [50]. To collect these data, a first-stage screening would also need to capture people who are using services/APs who could then undergo impairment assessment. Future research could consider asking participants about reported functional difficulties without assistance/APs as well which collects important quality of services data. The WHO Rapid Assessment of Assistive Technology [51,52] uses the WG Short Set with this modification and includes an additional survey section which asks about broader AP use; however, this sole modification generates non-comparable WG functioning data so is not a viable approach.
Finally, when exploring options for the second stage of a two-stage survey to estimate participants' functional service and AP needs, it is important not to rely on clinical assessments solely measuring impairments since this more "medical" model of assessment is only estimating one ICF component of functioning. It is key that second-stage assessments integrate broader functioning components when developing survey tools for assessing need, including consideration of environmental factors as highlighted by Sprunt et al. Therefore, alongside clinical impairment assessments, more hybrid clinical assessments measuring broader functional needs should be incorporated through structured observation and demonstration of tasks/activities, in addition to self-reported measures on activities, participation and environmental factors. This integration would ensure enhanced alignment with the ICF's broader definition of functioning and also would provide more detailed data about specific rehabilitation/AP service and human resource needs for evidence-based health and social policy and planning beyond solely surgical and medical needs.

Strengths and Limitations
The biggest strength is that approximately 20,000 survey participants from five countries are included in these secondary analyses presenting important data on the potential use of WG questions as first-stage screening questions in population-based surveys. We compared two cut-off levels of self-reported WG data to two cut-off levels of clinically assessed impairment data, as well as comparing clinically assessed need for services and select APs with the two methodologies. However, there were limitations. First, the authors acknowledge that the comparison in methodologies is based upon two separate ICF components of functioning-(i) impairment or body structure/function, and (ii) activity limitation-using clinically assessed impairment as the "gold standard." It is possible that some of the variation between the methodologies was due to measuring two separate ICF components. Second, our analyses in this paper were limited to "unmet" and "undermet" need comparing only a proportion of "total need" in the two methodologies to identify those who needed services. Therefore, we did not consider those who "use" services and APs that might actually have/had "met," "overmet" or "undermet" clinically assessed service and AP needs. Third, for the vision and hearing domains service groupings, we allocated participants to medical/surgical and/or rehabilitation/AP services, and often both, using a list of possible diagnoses. This retrospective allocation is likely to have over-estimated the need for both types of services given certain clinical diagnoses were assigned to both categories. Additionally, bilateral moderate+ impairment might be used as the referral threshold in some countries, such as in Chile for government financed hearing aids for people > 65 years. Fourth, the WG may not be the best possible screening tool, but it was used in the five surveys because it is widely endorsed and utilised. It could be interesting for future research to compare other self-reported functioning survey tools, such as the WHO's Disability Assessment Schedule [53] and Brief Model Disability Survey [54], with the WG questions for potential screening questions in multi-domain population-based surveys. Finally, all five surveys were supported by the same research group, the International Centre of Evidence in Disability at the London School of Hygiene & Tropical Medicine, using similar methodologies for clinically assessing vision, hearing and mobility impairments and service/AP need. While this is a strength in terms of comparability of methods, it is also possible that the use of alternate and/or additional methodologies and/or tools incorporating broader functioning components might have provided different results and should be considered in future research.

Conclusions
This paper explores the use of self-reported WG questions as a first-stage screening in population-based surveys. Our analyses found the WG questions could be used as a first-stage screening option to identify people with impairment and referral needs, but only with moderate sensitivity and specificity. If developing a multi-domain hybrid assessment survey tool, it therefore would be important to include options to (i) include or not include the first stage WG screen in the survey and (ii) select which functional domains to include. It is also recommended to explore additional first-stage screening cut-offs and options to provide better prevalence estimates of need, and incorporate assessments for other ICF components, especially personal and environmental factors, for more holistic hybrid methodology assessment of functional needs. Overall, our findings are important for the ongoing development and feasibility testing of population-based survey methodology and survey implementation considerations.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/ijerph19074304/s1, Table S1: Agreement between Washington Group question responses and impairment severity level in each functional domain. Table S2: Proportion of participants assessed to have a clinical impairment who need interventions as identified by Washington Group questions. Table S3: Overall proportion of survey participants who were assessed to have a clinical impairment and need interventions. Table S4: Relationship between clinical impairment assessed need * for four priority assistive products (glasses, hearing aids, wheelchairs and protheses) and Washington Group responses in three functional domains.  Informed Consent Statement: Informed consent was obtained from all participants involved in the study.

Data Availability Statement:
The Gambia data presented in this study are still undergoing analysis and will be made openly available once complete. For the other datasets, we are unable to make the databases publicly available as we do not have participant consent for this. We can, however, share the databases with researchers upon request.