The individuals in this test group were then asked to discuss what they thought each item meant to ensure that it was being read as intended. Examples of items deleted based on the cognitive interviews include “Many people can take care of their own asthma with the help of a doctor (T/F)” and “Most children with asthma come from poor homes (T/F).”
Based on information from the expert review and cognitive interviews, three new items were constructed, 26 items were retained without changes, 24 were modified, and nine items were deleted. This resulted in a new pool of 52 questions.
Step 3: Item Performance Measurement
This new pool of items became the survey for quantitative item performance measurement. Two samples were used for this phase of testing. First, data were obtained from the community-drawn convenience sample described above (n = 83). In addition to participating in the cognitive interviews, these individuals were also asked to complete self-administered surveys. A larger sample was also obtained via a random-digit dialing telephone survey of the Chicago area, using residential telephone numbers purchased from a commercial vendor (SDR Sampling Services; Atlanta, GA). For this telephone survey, a total of 222 respondents aged > 18 years were interviewed. canadian health & care mall
The purpose of this step was to ensure that responses to the items in the final survey did not manifest floor and ceiling effects, and would manifest sufficient variation to allow differentiation across segments of the population. For these reasons, items to which the two samples provided uniform responses were candidates for exclusion. After examining the distribution of SDs for the responses, a numerical decision rule for exclusion was adopted. Likert scale items with SDs < 0.95 were deleted. For binary response items (true/false or yes/no), “don’t know” responses were assigned a numerical value for calculation of a mean and SD. Items with an SD < 0.4 were deleted. Frequency distributions were also used to eliminate certain items based on the floor/ceiling of the responses. In general, items with responses > 80% at the highest or lowest values were deleted. In addition to deleting items based on uniform responses, several items were excluded because of extreme variability in the responses, suggesting unclear wording of the questions. Examples of the 22 items deleted as a result of this quantitative item performance measurement step are seen in Figure 2.
Figure 2. Examples of item reduction based on quantitative item performance measurement (Steps IIIA and IIIB) during development of the Chicago Community Asthma Survey (CCAS-32).