Participants and settings
From May 2020 to November 2020, participants were recruited using a variety of community-based recruitment methods, such as attaching our questionnaire QR codes on posters or contacting community staff to help post the information on bulletin boards, in Fujian province, China. Eligibility criteria mirrored the Chinese Expert Consensus on the Screening and Management of Lung Cancer [20] and Lung Cancer Screening National Comprehensive Cancer Network Clinical Practice Guidelines in Oncology Recommendation [21] for individuals eligible for lung cancer screening and included individuals (1) aged 55 to 74 years and individuals who currently smoke and have a 30 pack-year tobacco smoking history or individuals who used to smoke and have quit within the past 15 years; or (2) ≥ age 40 years and currently smoking with a 20 pack-year tobacco smoking history with one of the following risk factors: a) history of environmental or high-risk occupational exposure (e.g., exposure to asbestos, beryllium, uranium, radon); b) pulmonary disease (e.g., chronic obstructive pulmonary disease, diffuse pulmonary fibrosis or previous history of tuberculosis); c) previous malignant tumor history; d) family history of lung cancer; or e) long-term second-hand smoking exposure.
Participants with previous lung surgery, metal implants or devices in the chest or back, obesity and chest thickness, or diagnosed with lung cancer were excluded. The sample size was determined based on a subject-to-item ratio of 5–10:1 [22] by assuming a non-response rate of 15%, thus the final sample size was 360 potential participants to which the recruitment materials and survey were mailed.
Design and procedures
After written permission was obtained from the original scale developer (Carter-Harris), we translated the LCSHB into the Chinese version (LCSHB-C) and then examined the psychometric properties of the LCSHB-C, which were found to be adherent to the COnsensus-based Standards for the selection of health status Measurement INstruments (COSMIN) checklist [23, 24]. We applied Brislin's translation model to the cross-cultural translation, which includes translation, back-translation, comparison, and linguistic adaption [25, 26], as showed in Fig. 1. When we compared the original and back-translated versions, we found four items in discrepancy and were re-translated and back-translated, including I7" I might put off having a lung scan because transportation would be a problem."(B7. I may postpone the lung scan because traffic will be a problem.); I14 "I might put off having a lung scan because I worry about feeling like a social outcast for smoking."(B14. I may postpone the lung scan because I am worried that I feel like a person abandoned by society because of smoking.); I15 "I might put off having a lung scan because I worry about being blamed for having smoked."(B15. I may postpone the lung scan because I worry about being blamed for smoking.); I28 "Compared to other people your same age who have never smoked, what would you say your risk of getting lung cancer is"(B28. Compared with other people who are the same age but not smoking, you think that your risk of lung cancer is).
In the stage of pilot testing, the interview used structured probes to uncover how participants interpreted items of the LCSHB-C to verify its comprehensibility and readability. Example probes included: “Tell me in your own words what this question is asking,” “How did you decide on your answer to this question?” and “What does [health beliefs about lung cancer screening] mean to you?” Interviews were audio recorded and transcribed verbatim. None of the participants reported confusion or incomprehension about any of the scale items. After this process, the LCSHB-C was ready for validation. In the survey, we adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement [27].
Ethical considerations
The research was approved by the institutional review board of Fujian Medical University (Grant number:FM2020097), and written informed consents were obtained from all participants. Finally, a total of 353 participants enrolled and completed the 15-min survey and received a $10 gift card at completion. Participant anonymity was preserved in all cases.
Data collection
Data were all collected through online surveys with QR codes on posters or bulletin boards by Wenjuanxi (the most popular online data collection platform in China, available at https://www.wjx.cn/). The study aim and data collection procedure were fully explained to the participants.
Measures
LCSHB-C
The LCSHB consists of 35-items in total to evaluate an individuals' lung cancer screening health beliefs in the following four domains represented by the four subscales: (1) perceived risk of lung cancer, (2) perceived benefits of, (3) perceived barriers to, and (4) self-efficacy for lung cancer screening. All items in the perceived risk, perceived benefits, and perceived barriers subscales use a 4-point Likert scale (1 = strongly disagree to 4 = strongly agree) response option, and items in the self-efficacy subscale use the 4-point Likert scale of 1 = not at all confident to 4 = very confident. Items are all positively worded, and the total score is obtained by summing the scores of all items in each subscale. A higher total score on each individual subscale indicates higher perceived lung cancer screening health beliefs (perceived risk, perceived benefits, perceived barriers, and self-efficacy, respectively). The Cronbach’s alpha values ranged from 0.88 to 0.92 for the 4 subscales [17].
Lung cancer and screening knowledge questionnaire
This questionnaire was used to assess individuals' knowledge for lung cancer and screening, which was adapted from Carter-Harris' knowledge scale [18]. The questionnaire includes five questions with binary scoring (“correct” = 1, “wrong” = 0). The total score ranges from 0 to 5, and a higher total score indicates greater knowledge about lung cancer screening.
Socio-demographic questionnaire
We also collected participants' age, gender, marital status, educational level, monthly household income (yuan, RMB), residential location, religious belief, employment status, body mass index (BMI), health insurance, smoking status, family history of cancer, and frequency of lung cancer screening.
Data analysis
Data analyses were conducted using SPSS 23.0 (IBM, Chicago, IL, USA) and WINSTEPS 3.75.0 (Chicago, IL, USA) with a p-value < 0.05 was considered significant. Missing values were replaced using the multiple imputation calculation [28].
-
a)
Cross-cultural validity: The 4-point COSMIN checklist [29] was used to measure whether the description of the translation scale well reflected the items in the original scale [23, 24].
-
b)
Content validity: Content validity was evaluated from the translation validity index (TVI) adapted from the content validity index (CVI) described by Tang and Dixon [30]. A four-point scale was employed to rate the translational relevance of each item on the four subscales (1 = “totally different”to 4 = “equivalent”). The item TVI (I-TVI) was calculated by dividing the number of experts with a relevance rating of 3 or 4 by the total number of experts. And the mean value of TVI for each item was the TVI of the total scale (STVI).
-
c)
Structural validity: Confirmatory factor analysis (CFA) in CTT and Rasch analyses were performed in combination to assess the construct validity of the scale. In the CFA, the best fitting model of each subscale was tested using the maximum likelihood method. Absolute and relative indices [31, 32], including normed χ2 (χ2/df) between 1.0 and 3.0, Root Mean Square Error of Approximation (RMSEA; < 0.08), Comparative Fit Index (CFI), Tucker-Lewis Index (TLI), and Normed Fit Index (NFI) > 0.9, were employed to evaluate the model's goodness of fit. In Rasch analysis, the unidimensionality assumptions were first checked by the first contrast of the residual to ensure that it was not higher than 2 [19] and then the rating scale model (RSM) was used to assess person/item separation reliability, person/item separation index, category probability curves, test information functions and person-fit statistics [33, 34]. Infit and outfit mean squares, as well as difficulty (location) for individual items were involved in Pearson’ s fit statistics [35]. Items were tested for the differential item functioning (DIF) by gender (male and female).
-
d)
Construct validity: We estimated the convergent validity of the four subscales of the LCSHB-C using Pearson's correlations, with expected significant positive correlations with the lung cancer and screening knowledge total score.
-
e)
Known-group validity: Known-group validity was performed by determining whether the subscale scores of the LCSHB-C could discriminate among participants with different frequency of lung cancer screening participation behaviors.
-
f)
Internal consistency: We used Cronbach's alpha to assess the internal consistency reliability of the four subscales [32].
-
g)
Floor/ceiling effect: Floor effects were evaluated by examining the percentage of the respondents that achieved the lowest possible scores. Ceiling effects were evaluated by examining the percentage of respondents that reached the highest possible score.