The scale was developed in two phases: a) item generation and b) psychometric evaluation. The ‘Strengthening the Reporting of OBservational studies in Epidemiology’ (STROBE) guidelines for the reporting of observational studies were followed (see Supplementary file 1).
Phase 1: Item generation
1) Literature review
In this stage, we searched English and Chinese databases, including PubMed, Web of Science, Embase, Google Scholar, and the Cumulative Index of Nursing and Allied Health Literature (CINAHL), and the Chinese databases of China National Knowledge Infrastructure (CNKI), Wan Fang and VIP, with the keywords ‘tuberculosis’, ‘pulmonary tuberculosis’, ‘supportive care’, ‘healthcare’, ‘practical’, ‘spiritual’, ‘social’, ‘psychological’, ‘informational’, ‘emotional’, ‘physical’, ‘support’ and ‘need’. This review was undertaken to identify existing knowledge regarding the supportive care needs of patients with pulmonary tuberculosis. Items were identified and categorized into one of the domain outlines in the SCNF.
2) Expert consultation
We organized a two-round expert consultation to examine the content validity of the SCN-PTB draft. To ensure heterogeneity, 15 experts were recruited from general hospitals, TB prevention and control centres, and universities in Xi’an (n=9), Beijing (n=1), Sichuan (n=1), Shanghai (n=1), Dalian (n=2) and Hunan (n=1). They were asked to rate the feasibility and relevance of each item on the draft scale from 1 (irrelevant) to 5 (highly relevant) [28]. The content validity index (CVI) for an item is the proportion of experts who rated it as 4 or 5. The CVI was calculated for each item and scale. The scale level content validity index (S-CVI) of the last round of the expert consultation was used to evaluate the content validity of the SCN-PTB. The S-CVI should be larger than 0.8, suggesting that the content validity of the scale is good. The experts were also required to evaluate each item’s accuracy and clarity and then provide their specific suggestions regarding the item. Items with a mean score > 3.5, a coefficient of variation (CV) < 0.25, and an item level content validity index (I-CVI) > 0.8 were retained [29].
3) Pilot study
The content-validate items were designed as a self-administered scale with a 5-point Likert-type response format for each item. Each item asked patients to consider their level of need for help with the item by choosing one of the following response options: 1=no need-not applicable; 2=no need-already satisfied; 3=low need; 4=moderate need; or 5=high need. No items needed to be reversed scored, and higher scores reflected a higher level of need.
We conducted the pretest of the scale on 50 patients to evaluate its clarity, understandability, and feasibility. Patients were recruited from Xi’an Chest Hospital using convenience sampling methods based on the following criteria: a) age ≥18 years, b) confirmed diagnosed of TB, c) able to read and understand Mandarin and d) willing to participate in this study. Patients who had cognitive deficits, a history of mental illness, or any other severe physical problems or serious organ injuries were excluded.
Phase 2: Psychometric evaluation
Sample and setting
A convenience sampling strategy was used to recruit PTB patients from four institutions (Shaanxi Province Tuberculosis Hospital, Xi’an Chest Hospital, Huashan Hospital, and the designated TB hospitals of the Baqiao District) in Shaanxi Province between 10 September 2019 to 20 January 2020. The inclusion criteria were: a) having a confirmed diagnosis of PTB; b) being 18 years of age or older; c) being conscious and able to answer questions, and d) willing to participate in this study.
The sample size was determined based on the number of items in the developed scale and the sample size requirements of factor analysis. In factor analysis, 5 participants per item are the minimum recommended sample size [30]. Since the initial number of items was 25, 125 participants were required. The exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) samples should be two independent samples, with a size of at least 100 participants and 200 participants, respectively [31, 32], hence the minimum sample size required for factor analysis is 300. Considering the possibility of an invalid questionnaire, there is a need to increase the sample size by 20%, and thus the final sample size is n ≥ 360. In consideration of the above, the sample size of this study should be no less than 360. Applying the inclusion and exclusion criteria, a sample of 518 participants was recruited.
Data collection
With the assistance of the healthcare providers of the target investigation site, questionnaires were distributed to the patients by two investigators. Potential participants were given a cover letter informing them about the purpose of the study, its voluntary nature, and anonymity concerning participation, along with instructions for completing the questionnaires. Those subjects who agreed to participate were asked to sign an informed consent form and filled out the questionnaires independently, and then the investigator collected them immediately.
Data analysis
Data were analysed using SPSS 25.0 (SPSS Inc., Chicago, IL, USA) for all statistical analyses except for CFA, where we used AMOS software. Continuous variables are presented in mean and standard deviation (SD). Categorical variables are presented as numbers and frequency. The psychometric properties of the SCN-PTB were tested by validity and reliability. Generally, the α level was 0.05.
Item analysis
Item analysis aims to determine whether each item is correlated with the total score. An item was eliminated if it met one or more of the following criteria: a) the mean of the item was extreme or its variance was zero; b) the critical ratio value of an item was found to be insignificant; or c) the item-total correlation coefficient was < 0.30 or > 0.70.
Validity analysis
Content validity
The content validity index (CVI) calculated in the final round of the Delphi survey was used to evaluate the content validity of the scale.
Construct validity
We used EFA and CFA to examine the construct validity of the scale. The total sample was split into two subsamples using the SPSS random-assignment function. The first split half was used for EFA and the second was used for CFA. The cases included in each subsample (n=259) satisfied the requirement for the sample size for EFA and CFA, and their equivalence on demographic characteristics was examined through Mann-Whitney U tests (for continuous variables) and Chi-Square tests (for categorical data).
Exploratory factor analysis
Before conducting EFA, Bartlett’s test of sphericity and the Kaiser-Meyer-Olkin (KMO) test were used to check for the factorability of the data [33]. EFA using principal-components analysis with varimax rotation was performed to explore the underlying factor structure. The number of factors was determined by the eigenvalues and the scree plot. Factors with eigenvalue > 1 were extracted, and the result was considered good when at least 60% of the variance was explained by the identified factors [34]. According to the scree plot, the number of factors is indicated by the point at which the line indicating the slope begins to flatten [35]. Meanwhile, we assessed the best EFA solution based on multiple criteria as primary factor loadings > 0.40 [36], cross-loadings, the threshold for item communality (h2) was > 0.40 [37], the interpretability of the factors structure and the theoretical sense of the factors [38].
Confirmatory factor analysis
CFA was performed to test whether the data fit the hypothesized measurement model, which was extracted by EFA. The maximum likelihood estimation method was used for CFA. The fit of the CFA model was assessed using the following fit indices: the normed X2 (X2/df < 3), the root-mean-square error of approximation (RMSEA < 0.08), goodness-of-fit index (GFI > 0.90) and adjusted goodness of fit index (AGFI > 0.90) [39]. Additionally, the average variance extracted (AVE), construct reliability (CR) and the correlation coefficients between factors were calculated to validate the discriminant validity and convergent validity of the sub-factors of the scale. The AVE > 0.70, CR > 0.50 indicated good convergent validity, and the square root of AVE larger than the correlation coefficient between factors, indicated a good discriminant validity [40, 41].
Reliability analysis
Internal consistency reliability was tested using the Cronbach’s alpha coefficients for the overall scale and each domain. The split-half reliability was used to divide the scale items of SCN-PTB into odd-even parts according to the number, and the correlations of the score between the two parts were computed. The Pearson relation coefficient was calculated between the scores of the 50 patients who completed the SCN-PTB twice at a two-week interval to determine the test-retest reliability. A statistically acceptable reliability coefficient of the total scale should be > 0.70, and a statistically acceptable reliability coefficient of a domain should be > 0.60.
Ethical considerations
This study was approved by the Ethics Committee of Xi’an Jiaotong University (No. 2020-1244), and written informed consent was obtained from all participants prior to filling out the survey. The investigation process adhered to the principles of confidentiality, with the questionnaires completed anonymously, and the research data were used only in this research.