Study Design
An anonymous survey was conducted in a cross-sectional design. Study reporting was done per the STROBE guideline (18).
Setting
The study was conducted within February 2021 in Munich. Participants were German- and Turkish-speaking patients of Turkish-speaking family physicians. Munich is a city with 1.472 million citizens (19), of which around 40 thousand are of Turkish origin (17).
Participants
A list of the 612 family physicians working in Munich was obtained from the The Bavarian Association of Statutory Health Insurance Physicians “Kassenärztliche Vereinigung Bayerns (KVB),” which was reviewed for Turkish names. Additionally, an internet search and snowball inquiry by contacting the Turkish-speaking family physicians was made, which returned a list of 10 such physicians. All targeted family doctors were visited by the principal investigator in their offices and inquired about their willingness to support the study. Six family physicians agreed to distribute the study questionnaires to their patients. Inclusion criteria were being aged 18 and above, having sufficient knowledge of the German or Turkish language, and volunteering to participate. Participants who did not disclose their demographic information or had more than 50% missing items in the questionnaire were excluded from the analysis. A flow diagram of physician and patient participation is given in Figure 1.
Variables
The primary outcome measures of the study were COVID-19 knowledge scores and intention to get vaccinated for COVID-19. Other study variables were related to attitudes and behaviors regarding COVID-19 and demographic information.
Citizenship/nationality, native language, and place of birth were considered as additional variables for subgroups to take the complex construct of a “migration background” into account. Participants with either citizenship/nationality, country of origin, native language, or place of birth being non-German were categorized as “Having a migratory background.”
Construction of the Questionnaire
The study questionnaire was developed by the researchers after reviewing the literature for common knowledge and guideline recommendations related to COVID-19. Two team meetings were conducted to refine the questionnaire items. A pilot test on five Turkish and five German-speaking participants was conducted to further modify the questionnaire. The final questionnaire consisted of a demographic information section (6 questions) and three subdomains, including items on knowledge, attitudes, and behaviors concerning COVID-19.
The knowledge domain contained 25 items, which were scored as true or false. Hence, the COVID-19 knowledge scores were calculated by summing up the correct answers, providing a score of minimum 0 and a maximum of 25.
The attitude and behavior domains included both 7 items arranged in a 5-point Likert scale (1=disagree/never, thru 5=agree/very frequent). Items number 2, 3, 5, and 6 in the attitude domain and items 4, 5, 6, and 7 in the behavior domain were reversely coded. Attitude and behavior scores were calculated by adding the scores of each item and dividing by 7 (the total number of items), which revealed the minimum and maximum possible scores of 1 and 5. Greater scores indicated more sensitive attitudes and behaviors.
The two-page paper questionnaire was made available in German and Turkish.
Data sources/ Measurement
For recruitment, the practice personnel was asked to consecutively approach patients speaking German or Turkish and invite them to participate in the questionnaire survey.
All interested individuals were given the patient information letter and the study questionnaire. If the individual either spoke Turkish or German (or one language much better than the other), he/she received the material in the appropriate language. If the individual spoke both languages, he/she was asked for his/her preference. Questionnaires were self-administered in a silent place and collected after completion by the practice personnel or the primary author.
Data Quality and Bias
Each participant was asked to fill in the questionnaire alone without interference by others. After digitalizing the data, error checking and debugging were done to eliminate questionnaires with missing data or conflicting information. An effort was given to minimize selection bias by recruiting concomitant patients. However, although the principal investigator spent time in the practices reassuring data collection according to the protocol, this was not always possible due to the local conditions and the rules and regulations discouraging prolonged stay of patients in practice and accommodating multiple patients at a time.
Sample Size
A sample size of n=400 in the cross-sectional study was determined to be sufficient for consistent estimation of the coefficients of a multiple logistic regression model, including the factor variable ‘group’ and the demographic variables for adjustment. Further, the sample size is efficient for subgroup analyses, e.g., by vaccination hesitancy, migratory background, etc. (20).
Statistical Methods
The data were entered into the IBM SPSS Statistics spreadsheets (IBM Corp, Armonk, NY). The data distribution was described within and across the study groups by frequencies, percentages, means, and standard deviations (SD), as appropriate. Cronbach’s alpha was computed to assess the reliability of the items belonging to the knowledge, attitude, and behavior domains. Corresponding hypothesis testing of univariable group differences was performed by Chi-squared tests, Fisher’s exact tests (or Fisher-Freeman-Halton test), and independent samples t-tests or Mann-Whitney-U tests. Multiple binary logistic regression models were fit to the data to adjust the effect estimation of group differences by potential predefined confounders, such as demographics. Significant variables affecting vaccination intention in the univariate analyses were included in the model. Hypothesis testing was performed at exploratory two-sided 5% significance levels.