Using the ICF Framework to Assess Communicative Competence in Dyadic Communication among Children and Adolescents Who Use Augmentative and Alternative Communication Devices in Taiwan

Augmentative and alternative communication (AAC) devices enable children and adolescents (CAD) with communication disorders to communicate competently and develop friendships through communicative competence (CC). Existing assessment tools are unable to indicate whether CAD aged 0 to 18 years would competently use the subsidized AAC devices provided by the Ministry of Health and Welfare in Taiwan. This study, thus, aimed to develop an assessment tool by using the International Classification of Functioning, Disability and Health (ICF) to measure CC in dyadic communication among CAD using AAC devices. Five speech-language pathologists (SLPs), five special education teachers, and four AAC experts (14 in total) selected codes relevant to the four domains of CC via the Delphi method. Next, they categorized the selected codes into one of the four domains of CC through a face-to-face expert panel. A total of 112 codes were listed in the tool and fully classified into the four domains of CC. Among these, seven codes were concurrently placed under two domains of CC. Consequently, this study developed an assessment tool by employing the ICF for children and youth core set using universal qualifiers to measure the relative levels of CC in dyadic communication among CAD who use AAC devices in their daily life.

Currently, there are four assessment tools for measuring CC that have been assessed in extant studies and in actual practice (please refer to Supplementary Materials for details), including the Communicative Competence Rating Scale (CCRS) [26], the Communicative Competence Scale (CCS) [27], the Communication Supports Inventory-Children and Youth (CSI-CY) [28,29], and the Dynamic AAC Goals Grid 2 (DAGG-2) [30]. However, considering the four domains and perceptions of CC (e.g., dyadic, co-constructed), each of the above-mentioned tools has limitations. First, the CCRS, CCS, and DAGG-2 do not reflect the perceptions of CC, which imply a dyadic, co-constructed characteristic. That is, these tools focus on children and/or adolescents only, and fail to consider dyadic CPs. Second, the CCRS, CCS, and CSI-CY were normed on 11-18-year-olds, 12-20-year-olds, and 5-20-year-olds, respectively, but not on 0-4-year-olds. Furthermore, the DAGG-2 does not specifically address CAD, but rather people of various ages. Third, the CCRS, CCS, CSI-CY, and DAGG-2 have been judged only by CPs and/or healthcare providers, without any self-assessment from the communicators themselves (i.e., CAD who use AAC devices). Fourth, the CCRS and CCS have been developed for specific research purposes (research applications) and may be unfamiliar to global, multidisciplinary practitioners and scholars. Moreover, definitions for each point of the Likert scale (i.e., strongly disagree; disagree; neither agree nor disagree; agree; and strongly agree) for each listed assessment item are lacking. Fifth, some domains of CC are not included in the CCS, which loosely addresses strategic and social competence but does not deal with linguistic and operational competence. Moreover, the four domains and the nature of CC are not reflected in these current assessment tools. In sum, existing tools for measuring CC cannot determine whether 0-18-year-old CAD would competently use the subsidized AAC devices in their daily communication life in Taiwan. Therefore, a universally accepted and standardized tool for determining the potential benefits of subsidized AAC devices for CAD in Taiwan is needed for social and educational services.

The International Classification of Functioning, Disability, and Health (ICF)
The International Classification of Functioning, Disability and Health (ICF) and the International Classification of Functioning, Disability and Health for Children and Youth (ICF-CY), officially endorsed by the World Health Organization (WHO), use a universal, standardized language for clinical, public health, and research applications [31,32]. The ICF and ICF-CY document the functioning of individuals by considering the interactions between health conditions and contextual factors, and allow for consistent communication about health and health care across the world in various disciplines and sciences [31,32]. The ICF-CY documents bodily functions and structures, activity limitations, participation restrictions, and environmental factors emerging during infancy, childhood, and adolescence [31]. Despite limited knowledge in applying the ICF-CY among CAD who use AAC devices [33], several scholars [12,34] have suggested that the ICF-CY is very appropriate to be applied to CAD. The ICF-CY can be used by service providers, consumers, and others concerned with the health, education, and well-being of CAD [31], especially for describing communication and social participation [35]. Four components-Body Functions (b), Body Structures (s), Activities and Participation (d), and Environmental Factors (e)-are included in the ICF-CY. The physiological functions of body parts are encompassed by Body Functions (b), while anatomical body parts such as organs, limbs, and their components are subsumed under Body Structures (s) [31]. The interactions of CAD in their families, schools, and communities are focused on Activities and Participation (d) [36], while the physical, social, and attitudinal environments external to CAD are Environmental Factors (e) [31]. However, personal factors (e.g., personal attributes) are not classified due to their associations with extensive social and cultural variance [31].
The ICF-CY contains 1,645 alphanumeric codes (e.g., d132) [31,34,37] that are hierarchically organized at the first, second, third, and fourth levels of classification; they are mutually exclusive (i.e., not sharing the same attributes) [31,34]. There are 522 codes in Body Functions (b), 321 in Body Structures (s), 543 in Activities and Participation (d), and 259 in Environmental Factors (e) [31]. Five levels of universal qualifiers-(0) no impairment, difficulty or barrier to (4) complete impairment, difficulty or barrier-in the ICF-CY are used to document the severity or magnitude of the body's functions and structures, activity limitations, participation restrictions, and environmental factors using direct measurements, observation, first-hand interviews, and/or professional judgment [31].
To improve coverage of the transitions across the lifespan, and for easier maintenance and reduced redundancy of work, the World Health Organization [WHO] [38] announced the merger of the ICF-CY and the ICF, and that the codes in the ICF-CY will be added to the ICF. These codes in the ICF-CY were maintained as a specially derived classification and linearization of the ICF, but no further updates were made. Scholars, e.g., [39,40] concluded that CC can be measured through the ICF-CY, which offers a valid and reliable framework. First, potential assessment items can be converted from the universal alphanumeric codes (e.g., d132) [41], and assessment items can be standardized from the standardized language [37]. Second, the four domains of CC that determine the functioning of CAD interact better with the four components of the ICF-CY rather than with a traditional focus on single outcomes [28,42]. Researchers e.g., [24,28,36,41] have concluded that Activities and Participation (d) and Environmental Factors (e)-which contain codes related to communication, with a focus on the interactions of CAD with their families, schools, and communities-can better capture the performance of CC in an interpersonal dyad. Third, the perceptions of CC (e.g., dyadic, co-constructed) meet the frameworks of the ICF-CY [8,16]. Activities and Participation (d), a category focused on the interactions of CAD in their families, schools, and communities, and Environmental Factors (e), centered on the physical, social, and attitudinal environments external to CAD, consider the interactions (e.g., co-construction) of both CAD and others (e.g., CPs) in daily communication life not solely CAD. In addition, the five levels of universal qualifiers, which document activity limitations, participation restrictions, and environmental factors, are relative and dynamic (as opposed to absolute and static). The documentation-performed via direct measurement, observation, first-hand interviews, and/or professional judgment-derived from the ICF-CY is similar to the CC exhibited by CAD, CPs (e.g., family members), and healthcare providers (e.g., SLPs). Simeonsson, Björck-Åkessön and Lollar [36] further established that the CC of people in families, schools, and communities can be understood through the ICF-CY.
However, administering these codes limits the use of the ICF-CY in specific healthcare contexts and daily practices [37,43]. Simeonsson, et al. [44] and Pan, et al. [45] indicated that obtaining a certain minimum number of codes is necessary for numerous groups and settings. A core set is a selected set of codes that are relevant to most people with a certain health condition or in a particular healthcare context; the codes comprehensively portray people's level of function [46][47][48][49]. A core set is an evidence-based shortlist, including a brief core set (20-30 codes) and a comprehensive core set (70-100 codes) [37,50]. The core set allows global, multidisciplinary practitioners and scholars to jointly establish minimum standards for measuring functioning and health using clinical and multi-professional assessments while considering environmental factors [34,43,48]. Bernabeu, Laxe, Lopez, Stucki, Ward, Barnes, Kostanjsek, Reed, Tate, Whyte, Zasler and Cieza [34] and Pan, Hwang, Simeonsson, Lu and Liao [45] found that the core set can, over time, quantify the severity of impairments in Body Functions (b) and Body Structures (s), the limitations of Activities and Participation (d), and Environmental Factors (e) or barriers experienced by diverse kinds of people with communication disorders. The development of a core set can be used to assess, intervene, and gauge people's functioning for multiple disciplines [34,50,51].
Unfortunately, scant research has critically examined the ICF-CY core sets to measure CC among CAD aged 0 to 18 years who use AAC devices [52,53]. Hence, the current study aimed to develop an assessment tool by using the ICF framework to measure CC in dyadic communication among CAD who use AAC devices in Taiwan. This comprehensive core set focuses on (a) the four domains and nature of CC in AAC practice; (b) CAD aged 0 to 18 years who use AAC devices; (c) the availability of self-assessment, as well as the assessments of CPs and healthcare providers. Most importantly, this tool provides information regarding the follow-up benefits of subsidized AAC devices provided to CAD by the MOHW in Taiwan.

Methods
The present study was completed in 2019 and used two of the best-known consensus methods: the Delphi method and the nominal group technique (NGT), also known as the expert panel technique, to develop standards for the appropriateness of assessment in medical and health services research [54]. The Delphi method often pools and cultivates the options of a heterogeneous group to assess consensus about a given issue [54,55], and has been used to develop standards for the appropriateness of assessments in medical and health services research [54]. A correct response was not expected from the various rounds of the questionnaire. However, this technique avoids the dominance of a response from a single source, and shows summaries of distributed responses from the group [54]. NGT is structured as a face-to-face expert panel within the context of a focus group for developing solutions (e.g., categorizations) [56].

Participants
A diverse combination of 14 Mandarin-speaking practitioners who are AAC experts, including five SLPs, five special education teachers, and four scholars signed informed consent forms and completed the entire research procedure [54]; the recommended minimum of seven participants for the Delphi method was met [57,58]. The participants were familiar with the framework and coding system of the ICF-CY and spoke Mandarin as their first language. The SLPs and special education teachers have worked for at least five years in clinical practice and special education, respectively, and have at least one year of experience providing AAC services in Taiwan (e.g., assessment and/or intervention) to Mandarin-speaking CAD. The scholars had to have earned their master's or doctoral degree in communication disorders or special education, and conducted AAC-related research on Mandarin-speaking CAD with complex communication needs. Table 1 outlines the participants' demographic information.

Procedures and Data Analysis
Second-level (i.e., 3-digit) ICF-CY codes are recommended for surveys and the evaluation of clinical outcomes [31]. The current study aimed to develop a core set using the full range of second-level codes [59]. The first step entailed using the Delphi method to identify codes relevant to the four domains of CC, and the second step required using the NGT to categorize the selected codes into one of the four domains. The first step was performed using the Delphi method to reach a consensus among the participants regarding the selected codes [60,61]; none of the participants were aware of the other participants [62]. The second-level codes were not pre-chosen to avoid bias [40]. A three-round Delphi method was conducted through an online, close-ended questionnaire (i.e., Google Docs) provided in Mandarin [60,61]. The questionnaire was sent out by e-mail to all participants and contained: (a) the questionnaire's purpose; (b) instructions; (c) a detailed timeline; and (d) demographic information (e.g., e-mail address, gender, age, role/profession, years of work experience, and years of providing AAC services); (e) definitions of the four domains of CC from the work of Light [10] and Light, Beukelman and Reichle [11]; and (f ) the second-level codes [62]. The participants had four weeks to respond, and reminders were sent out approximately one week before the deadline.
The first round of the questionnaire asked the participants to review the definitions of the four domains of CC. Next, they were asked to consider whether each second-level ICF code was relevant to any domain of CC. To turn the codes into a core set, the participants were provided with the following directions as suggested by Rowland, Fried-Oken, Lollar, Phelps, Simeonsson and Granlund [28]: According to the definitions of the four domains of [CC] proposed by Light [10] and Light, Beukelman and Reichle [11], please rate the potential relatedness of each [ICF-CY] code to each domain of CC on a 5-point Likert scale, including (1) the item is not relevant; (2) the item is relevant; (3) the item is relevant and [we should be concerned about it]; (4) uncertain (considered relevant); and (5) uncertain (considered not relevant).
After the first round, the relatedness of each code was analyzed for selection and elimination. Responses of (2) the item is relevant; (3) the item is relevant and [we should be concerned about it]; and (4) uncertain (considered relevant) were grouped under the "relevant group," while responses to (1) the item is not relevant, and (5) uncertain (considered not relevant) were placed in the "irrelevant group." Descriptive statistics were used to examine the frequency with which the participants rated the potential relatedness of each ICF-CY code. Codes that received less than 40% of all participant ratings in the "irrelevant group" were automatically removed, while codes that received at least 75% of all participant ratings in the "relevant group" were automatically included [49,63]. Codes that received 40 to 74% relevance from all participants were deemed ambiguous and listed in an upcoming round of the questionnaire [49].
The questionnaire of the second round was similar to the one carried out in the first round but included: (a) a summary list of all ambiguous codes and (b) percentages of the participants who had considered the ambiguous codes relevant to CC [49,62]. The participants were asked to consider whether each of the second-level ICF-CY codes was relevant to any domain of CC. Similar descriptive statistics were conducted.
The third round of the Delphi method was performed in the same way as the second round; the participants were also asked to consider whether each of the second-level ICF-CY codes was relevant to any aspect of CC. The Delphi method was discontinued when all codes were automatically included (i.e., codes rated as the "relevant group" by at least 75% of all participants) or removed (i.e., codes rated as the "irrelevant group" by less than 40% of all participants). Conversely, it continued to be utilized when the codes received 40 to 74% relevance from all participants [49]. Three rounds of the Delphi method were carried out for the current study.
In the second step, each selected code from the Delphi method identified in the first step was placed under one of the four domains of CC. The NGT (involving a face-to-face expert panel), performed in Mandarin, was used to reach a consensus on the categorization of the codes. The written definitions of the four domains of CC proposed by Light [10] and Light, Beukelman and Reichle [11] were provided for discussion. Next, each participant independently categorized each item. Two rounds of the NGT were carried out, similar to the Delphi method, and the percentages of the categorizations of each code into the five groups were tallied. Each participant was provided with the following instructions: According to the definitions of the four domains of CC proposed by Light [10] and Light, Beukelman and Reichle [11], please categorize each item into one of the following five groups, including (1) linguistic competence; (2) operational competence; (3) social competence; (4) strategic competence; and (5) none of these.

Results
Three rounds of the Delphi method were performed through the online closed-ended questionnaire to reach a consensus on the related codes from the participants, and the NGT was employed to categorize each selected code into one of the four domains of CC. The second-level codes listed in the ICF-CY were initially presented. After the first round of the Delphi method, 117 codes-48 (41%) from Body Functions (b), 6 (5%) from Body Structures (s), 51 (44%) from Activities and Participation (d), and 12 (10%) from Environmental Factors (e)-were included. Five codes receiving 40 to 74% relevance from all participants were considered ambiguous and listed in the second round of the questionnaire. After the second round of the Delphi method was carried out, 114 codes-47 (41%) from Body Functions (b), 6 (5%) from Body Structures (s), 50 (44%) from Activities and Participation (d), and 11 (10%) from Environmental Factors (e)-were included. Two codes receiving 40 to 74% relevance from all participants were considered ambiguous and listed in the second round of the questionnaire. After the third and final round of the Delphi method was performed, 112 codes-47 (42%) from Body Functions (b), 6 (5%) from Body Structures (s), 50 (45%) from Activities and Participation (d), and 9 (8%) from Environmental Factors (e)-were included. Appendix A contains the 112 s-level codes in the core set, with their components and chapters listed in the ICF-CY.
These 112 codes were fully categorized into the four domains of CC through the NGT, as reported in Appendix A. None of them were categorized under (5) none of these. Twenty-eight codes were further classified under linguistic competence, 38 under operational competence, 32 under social competence, and 20 under strategic competence. Among these, "b125 dispositions and intra-personal functions", "b140 attention functions", "b164 higher-level cognitive functions", "d155 acquiring skills", and "d230 carrying out daily routine" were placed under social and strategic competence. The code "b130 energy and drive functions" was classified under operational and social competence. The code "b163 basic cognitive functions" was placed under linguistic and strategic competence.

Discussion
The ICF-CY component Activities and Participation (d) contained the largest number of codes, followed by Body Functions (b), Environmental Factors (e), and Body Structures (s), respectively. There are several possible explanations for these outcomes. First, the components of Body Functions (b), Activities and Participation (d), and Environmental Factors (e) are the areas that SLPs and special education teachers mainly work on during clinical practice and special education, respectively, compared to Body Structures (s) [28]. Second, several codes in Activities and Participation (d) and Environmental Factors (e) link communication well, and soundly describe CC [24,36,41]. The codes in Environmental Factors (e) refer to the supports and barriers of Environmental Factors (e), such as socio-relational skills in social competence. Third, barriers to participation, as considered in the participation model in AAC practice, can be comprehensively reflected in CC [17,52]. The codes in Body Functions (b), Activities and Participation (d), and Environmental Factors (e) play a more critical role than those of Body Structures (s) in this ICF-CY core set developed for measuring CC in dyadic communication with CAD who use AAC devices.
The codes in the core set were fully categorized into the four domains of CC (i.e., linguistic, operational, social, and strategic). Operational competence (38 codes) encompasses the largest number of codes, followed by social competence (33 codes) and linguistic competence (27 codes). Strategic competence (21 codes) contains the smallest number of codes. This is a logical outcome because the four domains of CC are interrelated [8] and the first three competencies develop simultaneously, while strategic competence emerges later [15]. Further, the small number of codes in strategic competence might be explained by the fact that adaptive skills are learned and used later than other competencies to minimize and compensate for restrictions in linguistic, operational, and/or social competence [14,64]. Different CPs (e.g., immediate and extended family members) use diverse adaptive skills to communicate with CAD who use AAC to co-construct their CC [11,16,23].
Several codes were simultaneously placed under two domains. This finding is corroborated by Blischak, Loncke and Waller [15], who confirmed that more than one CC domain might be concurrently involved in communication. First, the codes b125, b140, b164, d155, and d230 were classified under social and strategic competence. In social competence, b125 illustrates personal communicative confidence in initiating, maintaining, developing, and terminating communication to attain social closeness with diverse CPs. The appropriate adaptive skills selected by CAD according to varied CPs and environments (strategic competence) are also reflected in b125. Social competence in making decisions on how to initiate, maintain, and terminate communication, judge communication breakdowns, and then meta-cognitively plan to select appropriate adaptive skills in strategic competence is exemplified in b140 and b164. The learned skills in sequentially initiating, maintaining, and terminating communication (social competence), and planning adaptive skills that match communication breakdowns according to varied environments and CPs (strategic competence) are exemplified in d155 and d230.
Second, b130 was categorized into operational and social competence; b130 reflects mental functions of physiological (e.g., cognition) and psychological (e.g., motivation) mechanisms that enable people to satisfy their individual needs and goals [31]. Motivation (e.g., trusting AAC systems and devices) and cognition (e.g., understanding the required motor planning steps)-which are needed for operational competence-are reflected in b130. Social communication skills to meet communication goals in social competence are contained in b130 as well. Third, b163 was placed under both linguistic and strategic competence and reveals the ability to comprehend and express speech and AAC symbols, as well as to organize and apply them in communication.

Limitations
Although the ICF-CY core set was decided upon through two steps (the Delphi method and the NGT), there are still some limitations. First, the participant sample (i.e., 14) might not be representative of the larger population of practitioners who are AAC experts. In addition, the consensus of the recruited participants might differ from that of other observers, including CAD who use AAC devices, parents, or healthcare providers (e.g., the MOHW). Powell [65] stressed that the Delphi method is represented by the qualities-not the number-of participants. Linstone [57] further stated that a suitable minimum sample size is seven. A small number of participants in qualitative research typically produces a large amount of information [66]. Second, using the definitions of the four domains of CC proposed by Light [10] and Light, Arnold and Clark [13] in the Delphi method may have excluded some other appropriate definitions. However, several scholars [8,16,17,67] have argued that the definitions proposed by Light [10] and Light, Arnold and Clark [13] have dominated AAC research and practice for many years. Third, the ICF-CY core set was solely decided upon using the Delphi method, and the selected codes were placed into one of the four domains of CC through the NGT. Also, this tool was not validated, so the results should be interpreted with caution.

Implications
The assessment tool developed in this study for measuring CC by using the ICF-CY codes and guidelines for coding, acknowledged by the World Health Organization [WHO] [31,32], has several implications for clinical practice, despite the merger of the ICF and ICF-CY. First, the follow-up benefits of using the AAC devices subsidized by the MOHW among CAD and their CPs in daily communication life can now be assessed through the ICF and ICF-CY universal qualifiers (e.g., (4) complete impairment, difficulty or barrier), and, consequently, these subsidized AAC devices might not be abandoned. Second, the limitations and restrictions of the four domains of CC in dyadic communication can also be identified through the aforementioned universal qualifiers. Required eligibility for receiving and transitioning to social and educational services can be gauged [28,41]. Third, CC in dyadic communication among CAD who use AAC devices, and their peers in educational services, can be measured by practitioners across disciplines (e.g., special education teachers) while considering environmental factors in [48]. Consequently, the limitations, restrictions, and barriers of CC can be highlighted and overcome [68][69][70].

Conclusions
The comprehensive codes of the developed ICF-CY core set corresponded more frequently to Body Functions (b) and Activities and Participation (d) compared to the other two components, and were most frequently placed under operational, social, and linguistic competence. The follow-up benefits of the AAC devices subsidized by the MOHW can be measured through this ICF-CY core set regarding CC in dyadic communication among CAD who use AAC devices in daily life. In addition, the five levels of universal qualifiers (e.g., (4) complete impairment, difficulty or barrier) that establish the relative levels of CC across diverse communication contexts (e.g., discussing food) and dyads (e.g., peers, family members) can be used in educational services for support and transitions. This core set can be rated by the primary communicators (e.g., CAD who use AAC devices), their CPs (e.g., family members who communicate with them), and healthcare providers (e.g., SLPs and special education teachers). However, further studies are needed to validate this ICF-CY core set in social services (i.e., AAC devices subsidized by the MOHW) and educational services (e.g., special education).

Supplementary Materials:
The following supporting information can be downloaded at: https://www. mdpi.com/article/10.3390/bs12110467/s1. References [71][72][73] are cited in the Supplementary Materials. Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data presented in this study are available in Table 1