Participants
The study adopted convenient sampling method, recruiting participants among first- to fifth-year medical students in China between May 1 and May 30, 2019. The study took place at Fenyang College of Shanxi Medical University and Tongji Medical College, Huazhong University of Science and Technology, located in Fenyang Shanxi province and Wuhan Hubei province with approximately 6,400 and 4,100 undergraduate students, respectively. Inclusion criteria were following: a full-time medical student in school, informed consent, voluntary to participate in this investigation. Exclusion criteria were: students returning to school after working first or providing incomplete responses in the questionnaire. A sample size of 1000 for factor analysis is recommended as ideal [23], we need 1000 for exploratory factor analysis and 1000 for confirmatory factor analysis. So, the initial sample included a total of 2,034 students, of which 1,974 students completed the whole set of questionnaire, the effective response rate of questionnaire was 97.05%.
Instrument
Demographic Information
Demographic information including age, gender, grade and department.
The Fraboni Scale of Ageism
FSA was developed by Fraboni, including 29 items [19]. The items were responded by a likert scale, ranging from 1 (strongly disagree) to 4 (strongly agree). It's worth noting that item numbers 8, 14, 21, 22, 23, 24 are positive statements and scores should be reversed. Total scores range 29 to 116, the higher the score, the greater the ageism. An exploratory factor analysis found FSA measure three levels of prejudices: antilocution, avoidance, and discrimination [19]. The questionnaire took approximately 5 minutes to complete.
Procedures
This study proposed to evaluate validity and reliability of the Chinese version of FSA in medical students. We conducted the study in two steps: translated and validated scale. All translation processes was conducted based on modified Brislin's translation model [24]. In the forward-translation, we invited two bilingual translators who are Chinese and had studied in the United States ever to translate the FSA to Chinese independently. In back-translation, we invited two professional bilingual translators who is blind to FSA to translate back to English (from Chinese to English). Then they had a meeting to look back to the back-translations, find differences in meaning to achieve the most refined culturally equivalent meaning. Until the members were all agreement on the culturally same meaning in the two version of the FSA, the meeting wasn’t ended. Then we selected 20 medical students for the pre-survey by the FSA (Chinese version) and collected their opinions to make the scale items clear and easy to understand. Before data collection, we invited six experts (2 geriatric nurse clinical specialist, 2 health care providers who work at nursing home, 2 associate professors whose research interest is elderly abuse) to evaluate the content validity of the FSA (Chinese version) based on relevance as 1 (high relevant), 2 (quite relevant), 3 (somewhat relevant), and 4 (not relevant).
Data collection
The study was approved by the Institutional Review Board of Tongji Medical College, Huazhong University of Science and Technology. Trained investigators recruited eligible subjects at Fenyang College of Shanxi Medical University and Tongji Medical College, Huazhong University of Science and Technology. For students from Tongji Medical College, Huazhong University of Science and Technology, investigators obtained written inform consent and collected information from all the participants who filled in the questionnaires by themselves. The data collection procedure for students from Fenyang College of Shanxi Medical University was as follows: First, one author sent an email to the investigator who is a nursing faculty member in this university. The email described the purpose, content and eligibility criteria of this study, and attached a link to the online questionnaire. Second, the email was sent to the undergraduate medical students by the investigator, then the students finished the survey online. In order to prevent repetition, we designed it so the same mobile phone number could only fill in the questionnaire once and the questionnaire could only be submitted when it was completely filled out. In addition, it is recommended the minimum sample size for test-retest reliability is 15 subjects [25]. So, 30 subjects were selected to fill in the same questionnaires 4 weeks following the first time for filling out the questionnaire for assessment of the test–retest reliability. Before the survey, students were informed that participation was voluntary, anonymous and confidential. If medical students agreed to participate this study, they signed a consent form and then completed the questionnaires.
Statistical analysis
The Statistical Package for Social Sciences, version 21.0 (SPSS IBM Corp), was performed statistical analyses except confirmatory factor analysis (CFA). The CFA was carried out using SPSS Amos, version 21.0. Content validity was evaluated by the content validity index (CVI). The reliability was assessed by Cronbach's a coefficients and acceptable level was set at 0.70 [26]. As for test-retest reliability, it was determined by computing the intraclass correlation coefficient (ICC), the minimal acceptable value was set at 0.60 [27]. The exploratory factor analysis (EFA) and CFA were used to perform construct validity. First, the data is divided into two parts, 967 samples for EFA and 1007 samples for CFA. When the value of Kaiser-Meyer-Olkin (KMO) is > 0.60 and Bartlett's test of sphericity is significant, the samples are appropriate for factor analysis. The number of factors was determined by eigenvalues > 1 and scree plot. Factor loading > 0.30 were considered appropriate. The model fitness was performed by CFA. Factor loading reach a significant level the chi square degree of freedom ratio (CMIN/DF) < 3, the goodness-of-fit index (GFI) > 0.90, the incremental fit index (IFI) > 0.90, the compare fitting indices (CFI) > 0.90, and the root-mean-square error of approximation (RMSEA) < 0.08 indicated the model fit the hypothesized model well [28].