• No results found

RESEARCH STUDY QUESTIONNIARE

4.3. Scale Validation

For organizational reasons and since both the primary and secondary medication- related stressors and pharmacy support are measures included in the multivariable analysis of the dissertation, the factor analysis results are included here. The domains identified in these scale validation processes will be used in the methods and results chapters that will follow. This may appear to be presented out of order; however, it is important to remember that the scale development was a major portion of this dissertation. For a description of the

dissertation study questionnaire and the study population in which the scales were tested, refer to Chapters 5 and 6, respectively.

Data from the Survey of Caregiver Medication Concerns (SCMC) and pharmacy support portions of the dissertation study questionnaire (n=139) were analyzed using descriptive statistics (mean, SD, min, max) to confirm variability among responses. The responses were also examined to verify missing data. Mean imputation was performed for missing values in the medication stressor or pharmacy support items in an effort to preserve the number of observations used in the multiple variable modeling and since the amount of missing data was small with no more than 2.8% missing from any item. Any items that exhibited either a ceiling or floor effect were considered for elimination from instrument validation. Factor analyses (principal component with Promax rotation) was performed to identify the domain structure. Cronbach’s alpha was calculated to measure internal

4.3.1. Survey of Caregiver Medication Concerns Scale

Item-reduction Analysis: A descriptive analysis of the original 21-items related to medication stressors revealed that there were several items that were subject to either ceiling or floor effects where the responses were not well distributed. Descriptive statistics (mean, SD, range) are included in Table 4.2. The items examined for potential exclusion because of abnormal distribution were: (1) It is easier to keep track of my loved one’s medications when I use a pill organizer (mean=3.00, SD=1.27, range 0-4), (2) I have, on occasion, decided not to fill a prescription for my loved one because money was tight (mean=0.69, SD=0.97, range 0-4), (3) I have, on occasion, decided not to ask my loved one’s doctor for a prescription because money was tight (mean=0.63, SD=0.86, range 0-4), (4) There have been times when I have not had reliable transportation to the pharmacy to pick up my loved one’s medications (mean=0.59, SD=0.84, range 0-4), (5) There have been times when my own health has prevented me from giving my loved one medication (mean=0.92, SD=1.14, range 0-4), and (6) I have considered getting my loved one’s medications from outside the United States because they are less expensive (mean=0.98, SD=1.19, range 0-4). Two other items were identified for possible exclusion as well based on their ambiguity: (1) It is easy to manage my loved one’s medications and (2) It is hard to organize my loved one’s

medications. The responses were fairly well distributed; however, the questions were not as specific as some of the other items and their validity was questioned.

A correlation matrix was then used to determine correlations among these items with ceiling and floor effects to see if there were significant relationships between items that would warrant inclusion in the final instrument analysis. All of the items outlined above were correlated with other items in the instrument except for the question about using a pill

organizer to manage medications. Due to the specificity of this item and its relationship to being a physical task of organizing the medications, it was chosen for exclusion from the factor analysis.

In addition to some of the responses to items related to the affordability of

medications not being normally distributed, there was also a concern that the implementation of the Medicare Part D prescription drug benefit during the data collection phase might have created a bias in these questions. During development of the items, there was no prescription drug benefit under the Medicare program and both the focus groups and literature suggested that the cost of medications was a major concern of caregivers. Since several of the items related to affordability were subject to the ceiling and floor effect, we decided to also exclude all five items that asked caregivers about paying for medications or their ability to afford them. The following five items were eliminated from the factor analysis: (1) I have difficulty, on occasion, paying for my loved one’s medications, (2) I have, on occasion, decided not to fill a prescription for my loved one because money was tight, (3) I have, on occasion, decided not to ask my loved one’s doctor for a prescription because money was tight, (4) I have considered getting my loved one’s medications from outside the United States because they are less expensive, and (5) I worry that I will not be able to afford my loved one’s medications in the future. Once these items were excluded, along with the one previously mentioned, there were 15 items available for the factor analysis which will be discussed in the next section of this chapter. These items used in the factor analysis are in bold type in Table 4.2.

Table 4.2: Distribution Properties of Original SCMC Scale Items (n=139)

Question N Mean SD Min Max 1. It is hard for me to give my loved one medication more

than one time a day. 135 1.36 1.18 0 4 2. It is easy to manage my loved one’s medications. 136 2.40 1.18 0 4 3. It is hard to organize all of my loved one’s medications. 135 1.41 1.12 0 4

4. It is easier to keep track of my loved one’s medications

when I use a pill organizer. 136 3.00 1.27 0 4