Applying Statistical Technique To Develop A Scale For Measuring The Sub-health Status On Chinese

The overall aim of the present study was to develop a sub-heath scale that is suitable for Chinese aged from 20 to 60 years old (the Chinese Sub-heath Scale, CSHS). This paper describes the development and validation of CSHS. After four phases of researches, we have proposed the CSHS-4, which includes 47 items, 13 facets (fatigue, digestion, sleep, indisposition, ache, dysfunction, immunity, constipation, depression and anxiety, memory, pressure, satisfaction, sex behavior) and 3 domains (physical, psychological and social accommodation) with the use of classical and contemporary statistical technique.The analytic results of the present field data (data2) show that the CSHS-3 is a reliable measuring instrument with good reliability and validity. The confirmatory factor analysis of CSHS-3 suggests that the 3 domain (physical, psychological and social accommodation) solution is appropriate and sex domain is recommended to be a new facet of social accommodation domain.The goal of the first phase study was to generate the initial SCHS questionnaire (CSHS-1). First of all, the research group worked to determine the concept and scope of sub-health and decompose it into four domains (physical, psychological, social accommodation and life style), 13 facets (fatigue, digestion, sleep, indisposition, immunity, aging, depression, anxiety, feeling of safety, learning and memory, social accommodation, life style, sex behavior). A draft item pool of SCHS questionnaire was developed based on the literature review and referring the wildly used measures for assessing health-related issues. These items were primarily screened by experts using Delphi technique and the items with more importance, high sensitivity, stronger independence, better representation, maneuverability and acceptability were reserved. After being pre-tested and modified by 60 middle-degree educated subjects, the initial questionnaire (CSHS-1) was determined, which covers 4 domains (physical, psychological, social accommodation and life style), 13 facets (fatigue, digestion, sleep, dysfunction, immunity, aging, depression, anxiety, feeling of safety, learning and memory, social accommodation, life style, sex behavior), and includes 74 items totally.The aim of the second phase research was to conduct pretest, select items and then generate the formal questionnaire of CSHS (CSHS-2). A pre-test was carried out in 5 centers nationwide, they were located in Shanghai, Nantong Hubei, Hunan and Yunnan. The instrument applied was the CSHS-1 questionnaire proposed in phaseⅠ. The participants were determined by group randomized sampling method. 544 valid questionnaires were selected. Items were tested and reduced by 6 methods, frequency analysis, dispersion trend analysis correlation analysis, exploratory factor analysis, cluster analysis and internal consistency analysis. As a result 46 items were reserved. Based on this result, the questionnaire was adjusted and 32 additional items were added to form the formal CSHS questionnaire (CSHS-2), which includes 4 domains: (physical, psychological, social accommodation and sex behavior), 15 facets (fatigue, digestion, sleep, dysfunction, immunity, indisposition, ache, constipation, depression, anxiety, feeling of safety, learning and memory, pressure, satisfaction and sex behavior), and 78 items.In the third phase research, a field research was conducted to collect data on the CSHS items for the purpose of item testing and reduction and proposing the CSHS-3. The CSHS-2 proposed in phaseⅡwas applied to 856 volunteers from 2 medical examination centers of Shanghai. 462 valid questionnaires were used as the dataset (data1) to test and reduce items after data processing. The statistical methods applied in item testing and item reduction were frequency analysis, item analysis, internal consistency analysis and exploratory factor analysis. Frequency analysis was used mainly for the purpose of checking ceiling and floor effect of item responds. The items with more than 20% ceiling or floor effect should be excluded. Item analysis was undertaken to examine the contributions of individual items within the scale. Item-total and item-facet score correlations were calculated, items of smaller coefficients and/or of larger correlation coefficients with other facets were excluded. Internal consistency of total, domains and facets was assessed using Cronbachαcoefficient. The item should be excluded if the corresponding facet's Cronbachαcoefficient greatly increases after deleting the item. Exploratory factor analysis was used for the purpose of selecting items according to their loading on each factor. The item should be excluded if the item has a factor loading of less than 0.35 or greater than 0.35 on two or two more factors. Through the analysis of the four methods mentioned above, the items should be excluded if they were recommended to be deleted by the results of two or more methods. Thus, 55 items are reserved while 23 items are deleted. The result of correlation analysis shows that the items of depression and anxiety facet are recommended to be merged as a new facet called depression and anxiety facet due to high correlation, as recommended by the result of exploratory factor analysis, the items of depression and anxiety facet have high loading on the same factor. Finally, the refined CSHS questionnaire (CSHS-3) is composed of 4 domains 4 domains(physical, psychological, social accommodation, sex behavior), 13 facets(fatigue, digestion, sleep, indisposition, ache, dysfunction, immunity, constipation, depression and anxiety, feeling of safety, memory, pressure, satisfaction, sex) and 55 items.A large scale field research was conducted in phaseⅣfor the purpose of assessing validity and reliability of CSHS-3 proposed in phaseⅢ. The participants were determined by group randomized sampling method. 9 centers nationwide involved in the research, they were located in Shanghai, Xian, Yangzhou, Kunming, Luzhou, Shijiazhuang, Zhengzhou, Henan and Wuhu. The CSHS questionnaire proposed in phaseⅢwas applied to 10000 participants.The centers were requested to consider the balance of gender, age and occupation of volunteers when they were recruited. The presented 876 questionnaires were only from 4 sites (Shanghai, Xian, Henan and Hebei) because the field research is on going now. Internal consistency, test-retest reliability and split-half reliability were used to check the reliability of CSHS-3. Construct validity, criteria-related validity and discriminant validity were used to examine the validity of CSHS-3. Confirmatory factor analysis was conducted with LISREL8.53 to test if the scale construction was consistent to that of hypothesized model.The result of internal consistency analysis showed a better internal consistency of CSHS-3. The Cronbachαcoefficient for total score was 0.93. Except sex behavior domain (includes two items); Cronbachαcoefficient for other three domain score ranged from 0.83 to 0.93 and Cronbachαcoefficient for 10 facet score (except constipation, memory and sex behavior facet that includes two items) ranged from 0.67 to 0.84. The result of test-retest analysis shows good test-retest reliability. Correlation coefficient for the total score was 0.91, for 4 domains ranged from 0.76 (sex behavior) to 0.93 (physical), for 13 facets ranged from 0.63 (fragile) to 0.87 (constipation). The result of split-half reliability analysis indicates a high correlation between halves. The split-half correlation coefficient for CSHS-3 was 0.95, for physical domain (0.92), psychological domain (0.88) and for social accommodation domain (0.83). The result of construct validity analysis showed that most domains and facets had better construct validity except sex behavior domain. The Spearman correlation between domain score and total score ranged from 0.32 (sex behavior) to 0.73 (physical domain). The Spearman correlation between facet scores and total score ranged from 0.32 (sex behavior) to 0.77 (depression and anxiety). Except that the sex behavior domain had single facet, The Spearman correlation between facet score and domain score ranged from 0.56 (constipation) to 0.97 (Pressure). The result of criteria-related validity analysis showed a weak criteria-related validity. The Spearman correlation between domain/facet score and subjective health score ranged between 0.10 (sex behavior) to 0.27 (physical domain). The result of Discriminant validity analysis showed that the 55 items had strong Discriminant validity. Each of the 55 items had statistically significant difference among healthy, sub-healthy and patient groups, and also between the upper 30% and the lower 30% total score groups. Three domains and the whole CSHS-3 were tested using confirmatory factor analysis (CFA). The result of physical domain showed that three items F602, F903 and F302 were suggested to be excluded because of abnormal parameters and lower factor loading on its facet. After excluding these four items F302,F602,F903 and F901 (only 1 item left in the facet),the CFA was recomputed on the remaining items. The result showed better goodness of fit index:NNFI =0.955,RMSEA =0.0641,CFI =0.961,the parameters estimated were reasonable and statistically significant(t>2). The result of psychological domain showed better goodness of fit index: NNFI =0.968,RMSEA =0.0832,CFI =0.978;The factor loading of items ranged from 0.490 to 0.834, and were statistically significance(t>2);The determinant coefficients in testing formula were between 0.314(F1201) and 0.860(F1202).The result of social accommodation showed better goodness of fit index after modifying the model: NNFI =0.966,RMSEA =0.0682,CFI =0.973 ; The parameters estimated were reasonable and statistically significant(t>2). The results of all items in CSHS showed that 4 items F1402, F18, F2002 and F2001 were recommended to be deleted because of negative covariance (F1402, F18, F2002) and only one item left (F2001). The result of CFA on the remaining items showed better goodness of fit index: NNFI =0.915,RMSEA =0.0866,CFI =0.915; The parameters estimated were reasonable and statistically significant(t > 2). The CFA was conducted at domain level to test the goodness of fit index of 3 latent variable (physical, psychological, social accommodation) model and 4 talent variable (physical, psychological, social accommodation, sex behavior) model. The result supported 3 latent variable model and suggested sex behavior domain to be a new facet of social accommodation domain.Finally, the refined CSHS-4 includes 3 domains (physical, psychological, social accommodation), 13 facets.