Item generation and content validation of domains with item pool for the individual with knee osteoarthritis : a mixed-method study

| INTRODUCTION: There are several scales to evaluate subjective perceptions and individual components in individuals with knee osteoarthritis (IKOA). Till date, no scale is available to measure the combined balance, mobility, ADL and QoL in IKOA based on the International Classification of Functioning, Disability and Health (ICF). OBJECTIVE: The purpose of the study was to generate items and domains related to problems faced by IKOA and to validate the content by experts. METHODS: The domains and items were generated through extensive literature search (ELS) to extract items related to symptoms, balance, mobility, ADL and QoL in IKOA based on the International Classification of Functioning, Disability and Health (ICF) and through in-depth direct interview (IDDI) from 13 IKOA and three experts. The content validation of domains and items generated were validated by 10 experts through online Delphi survey. Minimum itemlevel content validation index (I-CVI) of 0.80 was considered to validate the identified items and the overall scale-level content validation index (S-CVI) of 0.90 was fixed to validate the generated items to use in scale development process. RESULTS: 117 items generated by IDDI and ELS were grouped under 18 domains initially. Content validation by Delphi method resulted in reduction with 56 item pool being grouped under the 14 domains with SCVI is 0.93. CONCLUSION: The comprehensive impairment, activity limitation and participation restriction item pool for IKOA under the proposed domains, have been developed and content validated. These items are recommended for their use in development of new comprehensive knee osteoarthritis index scale (CKOAI).


Introduction
One of the worldwide leading causes of disability and pain is osteoarthritis 1 . 22% to 39% of the 1.252 billion population suffer from osteoarthritis (OA). Hip and knee OA are the most prevalent forms of OA with the overall prevalence of knee OA, 28.7% 2 . This will increase by 33.5% in 2030 due to the alarming increase in the aging population 3 . Individuals with knee OA (IKOA) are seen with deficits in static and dynamic balance which comprises impaired proprioception, muscle strength, disturbed postural control, and decreased range of motion at knee joint 4 . Furthermore, IKOA has pain and increase physical limitation and functional limitation 5,6 . Eventually decreasing their quality of life (QoL) 7 .
Static balance in IKOA is assessed using several outcome measures such as timed single-leg stance 8 , functional reach test 9 , and variation of postural sway in unipedal or bipedal stance 10 . For more than two decades, the Berg Balance Scale and Tinetti Performance-Oriented Mobility Assessment (balance subscale) were used to assess dynamic balance 11,12 . Recently, the Community Balance and Mobility Scale (CB&M) have been validated for the purpose 13 . The isokinetic dynamometer has been in use to estimate muscle strength 14,15 . Proprioception was measured by joint repositioning test 16,17 . Several researchers explored QoL in IKOA 7,18-20 and combined it with functional independence 21 .
Various measures of knee function adopted by Rating Scale (ARS), and Tegner Activity Score (TAS) are subjective in nature. The patient reported a problembased rating scale in IKOA with objective scoring is still lacking. This way, the purpose of this research project is to develop items for the rating scale that combines the assessment of balance, mobility, ADL, and QoL in IKOA.

Protocol approval
The study protocol was approved by the institutional research committee (IRC) on 28th February 2017 (MMIPT/2017/5180) and then submitted and approved by the Institutional Ethics Committee (IEC) of Maharishi Markandeshwar Deemed-to-be University on 8th December 2017 (MMU/IEC/1021). After obtaining the approval from the Research Advisory Committee (RAC) and Institutional Ethics Committee (IEC), the study protocol was registered in the open-access public domain, ClinicalTrials.gov, on 5th April 2018 (NCT03498833). The study is composed of two main phases. First, item generation related to comprehensive impairment, activity limitation, and participation restriction in IKOA and grouping of item pool to relevant domains, and second, validation of generated item pool for the content validity. The first phase is composed of three sub-phases, item pool generation through extensive literature search (ELS), item pool generation through the in-depth direct interview (IDDI) method, and grouping of identified item pool under relevant domain related to impairment, activity limitation, and participation restriction domain. The first phase of the study was qualitative in nature, while the second phase used the Delphi survey method to validate the identified item pool. Hence, overall, the study was a mixed-method study. The details of the study phases were displayed in Figure 1. The study strictly adhered to the ethical principles for medical research involving human subjects, Helsinki declaration, 2013 adopted by the World Medical Association, the International ethical guidelines for health-related research involving humans (Revised, 2016) adopted by the Council for International Organizations of Medical Sciences (CIOMS) and also adopted the National ethical guidelines for biomedical and health research involving human participants by Indian Council of Medical Research (ICMR), 2017. Before the recruitment, all IKOA signed an informed consent form for their participation in an in-depth direct interview. The online informed consent form was obtained before the expert begins scale validation through the Delphi process.

Phase 1: Domain and item generating
This phase of the study was aimed at generating the item pool related to impairment, activity limitation, and participation restriction in IKOA using ELS and IDDI.
Extensive literature (ELS) in the English language were searched in PubMed, ProQuest, MD Consult, SCOPUS, Cochrane Library, and EbscoHost databases in the time frame between January 1980 and February 2018. Also, the reference part of the filtered articles was searched manually to confirm that no articles would be missed by any error in the electronic search. The primary author conducted the electronic search using the following medical subject headings (MeSH) terms: "knee", "knee joint", "knee osteoarthritis", "osteoarthritis", "outcome", "outcome assessment", "outcome studies", "outcome research", "pain", "Physical therapy modalities", "Physical therapy techniques", "physiotherapy", and "exercise". These MeSH terms were used in association with Boolean operators such as "AND", "OR" and "NOT". From a total of 1954 articles, 874 duplicates were removed, 312 screened, 562 excluded, and 46 full-text articles were assessed for eligibility. After the removal of 19 non-relevant articles, only 27 scales were included in the qualitative analysis. The flowchart describing the details of the study included was displayed in Figure 2. For IDDI, thirteen IKOA who comprised all four grades of on knee radiograph (Kellgren and Lawrence) grading system were approached in person at home or at the outpatient department (OPD) by the principal investigator to collect the data. We have recruited thirteen patients that were asked to generate the items that are relevant to assess their impairment, activity limitation, and participation restriction. First, the patients were asked to mention different items related to impairment that they feel important to be included in the pool. Second, they were asked to report items related to activity limitation Patients, and third, items related to participation restriction. Fourth, they were motivated to fill a maximum number of items which they feel appropriate based on their experience concerning functional activity in daily life. Fifth, the patients were given the item pool identified from the literature and were asked to add more items that were not mentioned in the literature. This procedure was chosen to yield the maximum number of items the patients felt that needed to be on the scale. After the generation of items from the literature and interviews, the items were pooled together and corrected for duplicates.

Sub-phase 2b: In-depth direct interview from experts
Three physiotherapists who had a minimum clinical experience of 10 years (experts) in the field of physiotherapy were approached to report about the functional activities, which IKOA reported to be difficult to perform. This approach was considered to obtain the maximum number of items to be included under the functional task, which is unique to this scale.

Sub-phase 3: Grouping item in domain
After generating items by ELS and IDDI, the items were grouped under the domains identified. First, the items that were closely related in function were identified and grouped. This was followed by placing the grouped items under the relevant domain generated. It was ensured that all the items were included under any of the domains identified.

Phase 2: Content validation by Delphi methodology
The content validation of identified items was executed by the online Delphi method. The two round Delphi survey was carried out to achieve a consensus of 80% agreement among the identified panel of 10 experts. These experts were physiotherapists with at least 10 years of clinical experience in treating IKOA from different geographical locations within India, to identify the needs for IKOA in each round of Delphi survey. As more than 10 experts were deemed to be unnecessary, we have included not more than a panel of 10 experts in each Delphi survey. The panel of experts who participated in the first round Delphi survey was not selected in the second round. The identified items were added to the Google Forms with each item bearing three options, "agree", "disagree", and "neutral". As acknowledged and advised by the early writers 23-25 , we have used a 3-point rating scale for the item validation. The experts were asked to exercise their opinion about each item with the above options. The "neutral" option was included to avoid skipping the items.
Eighty percent of agreement between the experts was fixed in a recommendation by Lynn to yield the item-level content validation index (I-CVI) of 0.80. According to Lynn's (1986) criteria, a minimum I-CVI of 0.78 for 6 to 10 experts is required to validate each item on a scale 24,25 . The item pool was selected by the experts based on the relevance to impairment, activity limitation and participation restriction, simplicity, and clarity. Feedback and comments provided by the experts regarding the inclusion of additional items which was not included earlier in scale validation were also considered. The grouping of items under the domains was sent to the panel of experts (sample of 10). The first round Delphi method of scale validation resulted in scale-level content validation index (SCVI) less than the recommended level (SCVI/Ave = 0.90)26, hence the second round Delphi method of scale validation was carried out after the elimination of individual item less than 0.78 (I-CVI) 25 . Each round of Delphi method of scale validation was executed by emailing the Google Forms to the identified panel of experts (n=10 in each round). Twenty experts (2n) who are expert in treating IKOA were identified created Google Form link were emailed. That were sent email reminders, phone calls, and messages on WhatsApp® to the non-responded experts after five working days, and the Google Form link was closed once the required response was obtained. The filled online content validation forms received from the experts were analyzed.

Data analysis
Description of the articles screened, excluded, and included were reported in the whole number. Demographic dimensions of IKOA, along with the grade of OA, were tabulated. Item pool generated through ELS and IDDI were also tabulated. The duplicate item generated was highlighted and retained under the tabulated items of either ELS or IDDI. Each generated item was validated and reported in terms of I-CVI. The overall validation of the proposed scale with item pool was reported with S-CVI after the end of each Delphi method of scale validation. S-CVI was computed by both approaches, the universal agreement calculation method (S-CVI/ UA) and the averaging calculation method (S-CVI/ Ave) 24 . Lynn recommended that minimum I-CVI of .78, in case of 6 to 10 experts, and overall, the scale should an SCVI/Ave of .90 or higher for considered to have excellent content validity 25 .

Results
Extensive literature search resulted in total of 19856 articles. After the removal of duplicates (18742), screening (1114), excluded (962), assessed for eligibility (152) and non-relevant articles (125). 27 studies were included in qualitative synthesis. From 27 studies, 13 articles were included in the item pool development. From 13 articles, seven scales' (AIMS -Arthritis Impact Measurement Scales; KOOS -Knee Injury and Osteoarthritis Outcome Score; KSKSS -Knee Society Knee Scoring System; LISOHK -Lequesne Indexes of Severity for Osteoarthritis of the Hip and Knee; NKSKSS -The New Knee Society Knee Scoring System; OKS -Oxford Knee Score; TLKSS -Tegner Lysholm Knee Scoring Scale) were used in generating item pool and domain. Total 51 items were identified by ELS, 48 items by IDDI. After removing two duplicates with ELS, IDDI resulted in 46 items. The details of item pool generated by extensive literature search with source of literature was displayed in Chart 1. The demographic characteristic of IKOA included in-depth review with their qualitative report related to their problem due to knee osteoarthritis were reported in Chart 2 and the combined 97 items pool generated by both ELS and IDDI were tabulated in Chart 3.
Ten performance-based functional tasks developed through ELS and IDDI are displayed in Chart 4. That way, 117 identified item pool were grouped under 18 domains according to their relevance was displayed in Chart 5. The first round Delphi survey results in the removal of 43 items and yields a total of 64 items (in Chart 6) with SCVI/Ave is 0.77 and mean expert proportion is also 0.77. Hence, the item pool entered the second round Delphi survey. At the end of the second round, the Delphi survey resulted in 56 items with SCVI/Ave is 0.93, and mean expert proportion is 0.93. As SCVI /Ave of 0.90 or higher is considered to have excellent content validity, the third round of the Delphi survey was not performed. Thus, the content validation by a panel of experts resulted in 56 item pool being grouped under the 14 domains. The I-CVI for each item and domain are tabulated in Chart 7. Chart 7. Level of agreement between experts expressed in terms of item-level content validation index for selected domains and items after content validation after second round Delphi survey (conclusion)

Discussion
From the identified 117 item pool under 18 domains, 56 item pools were validated under 14 domains. About 50% of the identified items were excluded. The majority of the excluded items were from the activity limitation domain. The reason might be the difference in the level of agreement among the expert panel of a physiotherapist from various geographical zones of India. 10 expert panel was used in each Delphi survey as it was advised by Lynn that more than 10 was probably unnecessary 25 . The larger sample size would probably lead to issues of data handling and analysis 27 . In the Delphi survey, a panel of identified experts in a particular field is asked to complete a set of questions to identify the panel consensus on specific issue 28 . We have used the Delphi survey to validating the item pool because it has advantages over questionnaires and panel discussions. It is an efficient method over others as the members do not need to interact, which making their response possible even by distance. The consensus developed is without interaction among respondents, and thereby potential bias of one dominant person influencing other's thoughts in delivering their opinion could be eliminated 28 . But the disadvantage of this method were the reminder emails, phone calls and messages on WhatsApp® required to attain the required sample size. Another disadvantage is that the members of the expert panel should be computer or smartphone literate, which made us exclude a handful of potential expert members.
We have set 80% agreement 29 among the members of the expert panel to include in the item pool to be used in scale, CKOI. Green et al. 27 recommended 80% consider that the particular item has attained consensus and also, if 80% of the expert agree with the particular item then it would yield the item content validation index (I-CVI) of 0.80 25 . The findings of the current study provide some preliminary information about the range of items required in a scale representing mobility disability in the community. The identified item pool and domains highlight that the patient has reported problems with relation to their knee OA under various components of the ICF model. The validated item pool has input from both patients with knee OA and physiotherapist experts in treating knee osteoarthritis. The strength of this study lies in the qualitative development of items from IDDI and ELS. The process of using both methods in item pool generation resulted in the overlapping of a few items and the generation of unique items. This would minimize the missed out items. Functional task domain was also added which is unique in this report. This study had a few limitations. The participant recruited for IDDI in this study represented a convenience sample of IKOA that may have led to some degree of selection bias and judging fit of items under the domains were not conducted by Confirmatory factor analysis. Nevertheless, this was the first study to develop and validate the item pool under the ICF framework model. Future studies should consider the use of the statistical method, Confirmatory factor analysis to judge the fit of the item pool under the domains, and random sampling technique in recruiting IKOA.
The developed and validated items should be tested for their psychometric and clinimetric properties, for their effective use among IKOA.

Conclusion
The comprehensive impairment, activity limitation and participation restriction item pool for IKOA under the proposed domains have been developed and content validated. These items are recommended for their use in development of new comprehensive knee osteoarthritis index scale (CKOAI).