The development and evaluation of a sub-health self-rating scale for university students in China

The items for the draft questionnaire were identified according to the WHO’s definition of health and Chinese social culture. In addition, we conducted a literature review and solicited expert opinion (comprising two health managers, four physicians, a psychologist and two clinical epidemiologists). The final questionnaire included 83 items.

The second stage was to evaluate the scale. A total of 6000 questionnaires were handed out in a medical university; 5792 questionnaires were handed back (96.53% response rate). Stratified sampling is used to sample at each grade. There is no difference in the age and gender of the sample. The 193 students who did not complete the questionnaires were excluded. Ultimately, 5599 questionnaires were analyzed. Cronbach’s α of the draft questionnaire was 0.949. Cronbach’s α values of the physiological, psychological and social subs-cale were 0.906, 0.881 and 0.865, respectively. Construct validity used exploratory factor analysis (Varimax). According to the value of eigenvalues above 1, we extracted 16 factors (include positive feeling factor, passive feeling factor, pain factor, menstruation factor, sleep factor, capability& self-respect factor, fatigue factor, social relationship factor, urine factor, eyes factor, constipation factor, diarrhea factor, anxious factor, irritable factor, forced factor, hair factor). The cumulative proportion was 52.29%. The Kaiser-Meyer-Olkin (KMO) statistic was 0.960. The criterion validity showed that the scale was highly related to the WHOQOL-BREF (r > 0.60), which was statistically significant (P < 0.001). The correlation coefficients between WHOQOL-BREF and the scale were 0.812. There was a notable difference in mean scores between the health and sub-health state in the dimensions (t > 35.00, P < 0.000). A total of 63 items were eventually extracted from the draft SSS.

The third stage was to determine the final items. After the draft questionnaire was evaluated, focus groups were again used to discuss whether the items were relevant, clear, unambiguous and written in a language that would be understandable to potential respondents. According to the results of factor analysis, we increased 1 factor (skin factor), put together 2 factors (constipation factor, diarrhea factor) and removed 6 factors (menstruation factor, eyes factor, anxious factor, irritable factor, forced factor, hair factor). From a professional perspective, we deleted nine items, maintained 54 items, and added four items for skin factors. The scale ultimately consisted of 58 items, which had satisfactory sensibility, representation and internal consistency.

The raw score is the sum of the item score in the total scale or the sub-scale. The scores of 16 of the symptoms are inversely transformed before adding up. The conversion formula is as follows:

The linear T score formula is as follows: T = 50 + 10 (X + ⁻X)/S, where X is the raw score, ⁻X is the total mean score, S is the population standard deviation.

A total of 6205 university students from grades 1 to 5 were recruited to participate in the survey. The exclusion criteria for participants included having organic disease diagnosed in a clinical laboratory examination. The 176 students, whose medical reports were abnormal, were excluded. Ultimately, 6029 students (2606 male and 3423 female), aged 15 to 29 years (mean age 20.88 years, SD = 1.5) were analyzed. The acceptance rate and finish rate were 89.03 and 99.67%, respectively.

Internal consistency analysis shows that Cronbach’s α of the total scale was 0.942. The dimensions of physiological, psychological and social had a high reliability: 0.915, 0.856 and 0.850, respectively. Cronbach’s α of each factor was between 0.7 and 0.88. The split-half reliability of the total scale was 0.938. The dimensions of physiological, psychological and social were 0.933, 0.890 and 0.881, respectively. The split-half reliability of each factor was between 0.73 and 0.90.

The KMO measure of sampling adequacy was 0.947 and the Bartlett test of sphericity was statistically significant (χ² = 7778.7; P = 0.000). The method of construct validity was exploratory factor analysis (Varimax). Confirmatory factor analysis (CFA) showed a reasonable fit of the data in the factor structure: χ² = 222,264.171, RMSEA = 0.0536, GFI = 0.867, AGFI = 0.853, CFI = 0.965. There were 12 factors for which the eigenvalue was above 1. The cumulative proportion was 60.53%. Based on scree plot and related theories, 10 factors were extracted. The cumulative proportion was 56.63%. The correlation coefficient between the total scale and sub-scale was high. The dimensions of physiological, psychological and social were 0.929, 0.803, and 0.774, respectively. The correlation coefficient between each factor and its own domain was 0.52–0.89 (Table 2). The correlation coefficient between each item and factor was 0.51–0.88 (Table 3).

Our previous studies show that the total prevalence of sub-health was between 57.2 and 65.1% [10‐13, 17‐22]; the prevalence of physiological sub-health was between 42.5 and 55.1%; the prevalence of psychological sub-health was between 32.5 and 35.8%; the prevalence of social sub-health was between 32.5 and 40.7%. The percentage method was used to determine the cut-off point for sub-health in SSS: the reference line of the total scale is 60%, was 50% for the physiological field, was 35% for the psychological field, and was 35% for the social field. The raw score, transformed score and T score of the total scale were 224, 72, and 52, respectively. The raw score, transformed score and T score of the physiological sub-scale were 128, 72 and 50; of the psychological sub-scale, they were 48, 60 and 46; of the social sub-scale, they were 26, 56 and 45, respectively. The lower the score, the worse the health condition (Table 4).

There are five expert evaluated 1000 questionnaires which were randomly selected from the valid 6205 cases. The expert evaluation of sub-health status was performed according to the clinical guidelines for Sub-health published by the China Association of Chinese Medicine [23]. The 5 expert evaluated 200 scales respectively. If the expert cannot sure, the scale will be discussed together. If they are still uncertain, they will interview the student for confirmation. The expert evaluation of 1000 questionnaires showed that the total prevalence of sub-health was 64.49%; physiological sub-health was 47.78%; psychological sub-health was 37.15%; and social sub-health was 36.84%. The fit between expert evaluation method and the scale method was 0.758. The fit in the physiological, psychological and social sub-scale was 0.815, 0.797 and 0.787, respectively.

We had presented the ROC curve of the 1000 questionnaires to detect the cut-off point of the scale (Figs. 1, 2, 3, 4). The area under the curve in the physiological, psychological, social sub-scale and total scale was 0.987, 0.985, 0.990, 0.958, respectively (Table 5). The SSS was significantly positive to evaluate the sub-health state (P = 0.000). Table 6 shows the Youden’s index in each dimension. The maximum value of Youden’s index in the physiological, psychological, social sub-scale and total scale was 0.90, 0.94, 0.94, 0.81 (Table 6). The optimum cut-off point in the physiological, psychological, social sub-scale and total scale was 126.5, 48.5, 25.5 and 221.5. Ultimately, according to the standard above, 2150 were healthy and 3879 were ‘sub-healthy’ in 6029 students.

First of all, this scale is only a self-assessment scale, suitable for the student population. Only a few schools have been investigated. The next step is to expand the sample size to investigate more schools and improve the scale. Second, the current There is no objective measurement standard for sub-health. It is difficult to accurately measure the sub-health status only through the self-rating scale. Some people, who is in psychological and social mental illnesses, may be mistaken for sub-health. As a result, we must use a variety of methods and tools to measure sub-health status. For example, combined with current medical history, past history, physical examination results, plus sub-health self-rating scale, anxiety self-rating scale, depression self-rating scale, WHO neurosis screening table, etc. for comprehensive evaluation, and finally obtain more accurate results.