Study Design and Setting
A cross-sectional survey was conducted using a convenience sample (18) of nursing students from two Public Nursing and Midwifery Training Colleges attached to Teaching hospitals. These colleges were selected to represent the northern and southern sectors of Ghana. The northern sector is primarily Muslim dominated and share ethnic, and cultural traditions which are slightly different from the ethnic and cultural traditions of the Christian dominated southern sectors of Ghana. Second- and third-year nursing students were invited to participate in the study because they had more clinical experience than first-year students which was relevant for answering the questionnaires.
Using the sample size calculation based on our main study (17), 170 nursing students were invited to participate. The 170 nursing students met the Sahin rule of sample size estimate of at least 150 to determine item parameters accurately in unidimensional dichotomous scales (19).
Data were collected using survey questionnaires between December 2019 - March 2020. An information session by the principal investigator (DAB) was conducted in the second- and third-year classrooms. An envelope that contained the questionnaires, information sheet and consent form, and invitation to participate in the second phase of our study and a token of appreciation was given to participants who volunteered. The questionnaire had four sections. Sections A focused on the sociodemographic data, Section B included the KOP-Q, Section C included the KOAP scale and Section D was the Self-efficacy to Care for Older Adults’ (GES-COA) scale. Sections A, B and C were the focus of this paper. Completed questionnaires were either dropped in a sealed box that was placed at the reception desk at the school or handed to the principal investigator in a sealed envelope onsite.
The Knowledge about Older Patients Quiz (KOP-Q) (4, 11), is a unidimensional scale that consists of 30 true or false items to measure nursing students’ knowledge in the care of older adults. The KOP-Q examines nurses/nursing students’ knowledge on the normal ageing process, depression and delirium in older adults, geriatric syndromes, nutrition and fluid imbalance, polypharmacy, referrals, and family caregivers’ support. A correct response scored is +1, and an incorrect response scores zero. The scores are summed for a total score. Face validity of 0.91 and construct validity of 0.70 have been reported (11, 20) in the Netherlands and USA.
The Kogan’s Attitude Towards Old People Scale (KAOP) (9) consists of 32- item statements. The items of the KAOP scale focused on issues such as the housing arrangements of older adults, the environment and personal appearance, interpersonal relationships with other age groups and dependence of older adults. Item responses are rated on a 5-point Likert scale, ranging from 1 strongly disagree, to 5 strongly agree. Half of the statements are negative statements and the other half are positive statements regarding attitudes toward older individuals (21). Negative statements were reverse coded, so that there is one total positive score summed which ranges between 32-160. Using the neutral response score of 3, the maximum neutral response 96 was determined as the cut-off point. Higher scores above 96 indicated positive attitudes and scores lower than 96 indicated negative attitudes. In addition to the two scales, socio-demographic data of the participants were collected to describe the sample. These included age, gender, years of school (second /third year), living with/having lived with an older adult, religion, and region (southern /northern sector).
All analysis were undertake using SPSS (IBM version 26)
Analysis of KOP-Q
Item difficulty and discrimination indices were calculated as follows. First, the participant knowledge scores were summed, and the total knowledge scores calculated. Next, data were categorized in ranks from the lowest to the highest scores. Participants were categorized into quartiles; the lowest 25%, the middle 50% and the highest 25%. The lowest 25% was calculated using the total number of correct answers in the quartile and divided by the number of participants in that quarter (42 participants). The highest 25 % was calculated using the total number of correct answers in the highest 25% divided by the number of participants in the quarter (42 participants). Participants within the lowest and highest, 25% were used for the discrimination indices (22, 23).
Item difficulty is the percentage of participants who answered an item correctly. Item difficulty was calculated using the formula:
The score ranges between 0-100 % (24). The lower the difficulty level, the harder the items, suggesting fewer students chose the correct response. The higher the difficulty level, the easier the items, indicating majority of the students chose the correct response (22, 23).
Optimal item difficulty levels are somewhat higher than halfway between guessing and students actually knowing the correct response to an item (22). For dichotomous (true/false) response choice questions, optimal difficulty ranges between 30-85%. Items less than 30% are considered too difficult and items more than 85% are too easy and are recommended to be removed or revised (25).
However, for the selection and examination of final items for the psychometric property analysis of the scale, discrimination indices (DI) was considered (26). The discrimination index (DI) was calculated using the formula:
Various factors such as number of items, unidimentionality of the scale and item difficulty impact the DI threshold of items (27). A discrimination index threshold of ³ 0.2 is recommended to determine which items to retain or remove (27, 28). Using our content judgement some items with a DI of 0.19 were approximating 0.2 and included for analysis (27). The DI calculations were repeated after items were removed one after the other which enhanced the DI for the items retained. The Kuder Richardson (KR20) reliability for the retained items was calculated. KR20 is the dichotomous equivalent of the Cronbach alpha coefficient and appropriate for use with survey instruments that measures dichotomous items such as true/false or yes/no responses (29). A KR20 of 0.70 and above is considered acceptable (30).
Analysis for the KAOP scale
To evaluate and improve the internal consistency of KAOP, we calculated the corrected item-to-total correlations and Cronbach alpha if an item was deleted. Corrected item-to-total indicates how much each item correlates with the overall scale (31). Negative and very low corrected item-to-total correlation suggest that items do not correlate adequately with the overall scale and should be removed (32). Negative and very low corrected-item-to-total correlations were deleted one after the other; the deletions increased the overall Cronbach alpha of the scale (31, 33).
The extent and pattern of missing scale item values were examined using descriptive statistics, 95.8% of the sample provided complete data at the item level. With such a small proportion of missing data, complete case analysis was employed. The complete case analysis based on the amount of missing data did not impact the statistical efficiency due to the minimum required sample estimates (19).