E. Cuando se brinda tratamiento autoadministrado (no observado por el personal de salud) los enfermos no cumplen estrictamente la toma de
4.1.5. Atención de enfermería al paciente con tuberculosis
Table 17 shows the high percentage agreement achieved in the inter-rater and intra-rater reliability studies for the coding of children's modes of communication.
Table 17 Mean inter-rater and intra-rater percentage agreement for mode categories used by children across groups
Mode category Inter-rater Intra-rater
Verbal 96.62 99.31 (N = 207) (N = 146) Vocal 94.79 94.44 (N = 96) (N = 72) Gesture 94.74 97.50 (N = 152) (N = 160) Physical 93.92 98.99 (N = 148) (N = 99) Verbal gesture 88.57 94.44 (N = 35) (N = 36) Vocal gesture 82.14 89.74 (N = 28) (N = 39) AAC * 100 ( N - 8 )
(See Appendix 7 for inter-rater and intra-rater percentage agreement scores for child modes within each subject group separately.)
5.2. Script Situation
To determine inter-rater reliability of coding communicative functions in the script procedure the second independent rater, who had also coded data for the reliability study on the Conversation Situation, coded twenty randomly selected Script Situations, representing fifty percent of the data (ten fi'om Group One and ten from Group Two). For intra-rater agreement the author recoded twenty separately randomly selected Script Situations (ten each from Group One and Group Two), at least two months after their original coding. As functions on the Script form are noted as present or absent, percentage agreement (number of agreements divided by number of agreements plus disagreements, where disagreement includes act coded as absent) was calculated. Inter-rater and intra rater agreement for each category is given in Table 18.
Table 18 Percentage inter-rater and intra-rater agreement for functions elicited in the Script Situation
Function Inter-rater Intra-rater
Request for joint attention 80 87.5
Request for information 82.5 90
Request for object/action 85 90
Request for clarification 96.67 98.33
Provision of information present 95 94.63 absent 95 95 Provision of clarification 95 95 Confirmation/denial 93.75 81.25 Expression of self 95 90 Mean 90.95 90.86
5.3. Summary and conclusion
The kappa scores obtained across groups indicate high levels of inter-rater and intra-rater agreement for all three levels of analysis for the Conversation Situation, all falling into the 'substantial' or 'almost perfect' ranges proposed by Landis and Koch (1977). However, the inter-rater percentage agreement calculations for individual categories show that although overall agreement was high, there was considerable variation in agreement for individual categories within each level of analysis between the two raters. Some low agreement scores (less than 70%, Barlow & Hersen, 1984) were achieved for categories which appeared infrequently in the Conversation Data (less than 50 times in the reliability data), mostly falling into the act level of analysis. The low agreement scores were most often found in the child act data where there were fewer acts coded, due to the high level of No Response and No Response Requested moves. However, patterns of disagreement were noted. For the most part it was observed that the first rater, the author who had conducted the sessions with each of the children, coded the children's acts as intelligible (for example, expression of self, request for attention) and the second rater coded them as unintelligible. This difierence in coding is possibly due to the second rater not being as familiar with the children and having diflBculty recognising their often idiosyncratic communicative signals. The first rater's knowledge of the children and their mothers may also explain the higher levels of intra-rater agreement across most categories. Similar diflSculties in raters coding the communication of physically disabled children whom they do not know have been observed in previous studies (Jolleff et al, 1992; Winyard, 1993). It is noteworthy that higher levels of inter-rater reliability were achieved for data from children in Group 2 than Group 1, indicating that the intentional communication of non-speaking physically disabled children is more difficult for observers to interpret than that of verbal physically
disabled children, supporting similar findings by Yoder (1987). It may have been possible to increase the level of inter-rater agreement in the present study if more detailed information about each child had been given to the second rater before coding began
In the Script Situation higher levels of inter-rater agreement were achieved for each of the acts for which low percentage agreement was obtained in the Conversation Situation. This is to be expected, as in the Script Situation acts were elicited as responses and coded as present or absent. Such acts should be easier to code than communication signals which appear in conversational discourse, whose code assignment is dependent upon previous and following messages produced by both speakers, and which are coded in terms of both their move and act category.
The predominantly high levels of agreement observed for each level of analysis were judged acceptable for analysis to proceed with some revisions to the coding system. For
the Conversation Situation it was considered necessary to:
- collapse child repetitions and revisions as provisions of clarification into one group,
- combine maternal confirmations and denials with acknowledgements,
- retain the categories request for joint attention and expression of self for children in spite of the low level of inter-rater agreement for Group 1, due to the frequent coding of the act as unintelligible by the second rater and high intra-rater agreement for the two categories,
A decision was taken not to undertake further inter-rater reliability checks on child act data due to the infrequent production of the act categories on which low agreement was achieved in conversation and the higher agreement achieved for these categories in the Script Situation.