Characteristics of different absenteeism
Results showed that the dynamic trend of absenteeism versus grade decreased first and then increased. DAR1 was the highest in both schools, and the lowest absenteeism was DAR4 in school A and DAR5 in school B (Table 1).
Table 1 Descriptive statistics of daily absence rates reported by FRSSS and school physicians.
Variable
|
School A (%)
|
School B (%)
|
Min
|
Max
|
Mean
|
SD
|
Min
|
Max
|
Mean
|
SD
|
DAR1
|
0.32
|
15.19
|
4.66
|
3.352
|
0.78
|
24.90
|
4.28
|
3.305
|
DAR2
|
1.04
|
16.19
|
3.76
|
2.933
|
0.00
|
14.49
|
3.21
|
2.561
|
DAR3
|
0.00
|
6.20
|
2.07
|
1.201
|
0.00
|
12.28
|
3.11
|
2.158
|
DAR4
|
0.27
|
5.22
|
1.84
|
0.862
|
0.00
|
11.93
|
2.98
|
1.989
|
DAR5
|
0.00
|
8.27
|
1.92
|
1.427
|
0.00
|
5.81
|
1.78
|
1.523
|
DAR6
|
0.00
|
6.40
|
1.81
|
1.023
|
0.00
|
7.74
|
2.32
|
1.434
|
DARL
|
1.14
|
15.74
|
4.19
|
2.793
|
1.08
|
16.42
|
3.78
|
2.485
|
DARH
|
0.63
|
4.92
|
1.91
|
0.815
|
1.02
|
5.68
|
2.56
|
1.050
|
DARX
|
1.07
|
6.44
|
2.71
|
1.151
|
0.00
|
8.14
|
3.01
|
1.426
|
DARY
|
0.00
|
4.39
|
0.77
|
0.734
|
0.09
|
6.04
|
1.84
|
1.224
|
DARZ
|
0.00
|
1.84
|
0.54
|
0.388
|
0.09
|
5.77
|
1.57
|
1.199
|
Among the six pair differences in absenteeism between school A and B, differences in grade three and four were the most prominent, and each exceeding one percentage point. The two schools were relatively close in DARX, separated by 0.3 percentage points (t=2.503, P=0.013), however, the gaps of DARY and DARZ between school A and B were larger, with the values of school B being more than twice that of school A.
According to the correlation matrix of absenteeism in six grades, factor analysis was performed by the principal component method, and the factor rotation method was maximal variance rotation. Two factors (i.e., grades 1-2 and 3-6) could be selected from this analysis in school A, which explained 62.2% of the total variation. Two factors (i.e., grades 1-2 and 3-6) could also be selected from this analysis in school B, which explained 64.4% of the total variation. Considering that the human immune system approaches maturation at about 8 years old [20], students at this age are in grade 3 in China. Therefore, absenteeism reported by FRSSS for the six grades can be divided into two levels: grades 1-2 and grades 3-6. The former was recorded as DARL, and the latter was recorded as DARH. Whether in school A (t=10.466, P<0.001) or B (t=5.731, P<0.001), DARL was significantly higher than DARH.
Correlation of different absenteeism
For school A, there were 9466 all-cause absences reported by FRSSS and 2609 all-cause absences reported by school physicians (of which 1811 were due to sickness). Accordingly, DARY and DARZ were 27.5% and 19.1% of DARX, respectively, while DARZ was 70.5% of DARY. For school B, FRSSS reported 6787 all-cause absences and school physicians reported 2652 all-cause absences (of which 2278 were due to sickness). Accordingly, DARY and DARZ accounted for 39.5% and 33.6% of DARX, respectively, while DARZ for 84.9% of DARY. Combined, DARY and DARZ were 32.6% and 25.2% of DARX, respectively, while DARZ was 77.3% of DARY.
Figure 1 Time-series of DARL, DARH, DARX, DARY and DARZ at two schools.
For school A, DARX and DARY (r=0.398, P<0.001), DARX and DARZ (r=0.225, P<0.001), and DARY and DARZ (r=0.843, P<0.001) were all significantly correlated. For school B, DARX and DARY (r=0.486, P<0.001), DARX and DARZ (r=0.483, P<0.001), and DARY and DARZ (r=0.933, P<0.001) were also significantly correlated. Between school A and B, the correlations of two DARLs (r=0.148, P=0.115), DARXs (r=0.079, P=0.296) and DARZs (r=0.103, P=0.197) were not significant, while the correlations of two DARHs (r=0.464, P<0.001) and DARYs (r=0.321, P<0.001) were significant. Data showed a clear gap between school A and B in the trend of five indicators (DARL, DARH, DARX, DARY, and DARZ) over time (Figure 1); however, the time series of these five indicators were relatively consistent in the same school.
For school A, it could be observed that there was an obvious divergence between the time-series of indicators reported by FRSSS (DARL, DARH and DARX) and those reported by school physicians (DARY and DARZ) during May 2022. For school B, a similar divergence took place during November 2021. According to the information provided by school physicians, we investigated that there was a severe influenza outbreak in school B during November 2021 and a serious pertussis epidemic in school A during May 2022. After excluding the data from these two periods, DARX's correlation with DARY and DARZ significantly increased: two coefficients were 0.672 (P<0.001) and 0.443 (P<0.001) for school A, and 0.910 (P<0.001) and 0.914 (P<0.001) for school B.
Comparison of surveillance effectiveness among indicators
For school A, the effective surveillance time was 178 days, of which 49 days (39 days for grades 1-2 and 27 days for grades 3-6) showed infectious disease outbreaks. According to the control charts of DARX, DARY and DARZ (Figure 2), the days to meet the warning standard (i.e., three standard deviations above the average) were 18 days, 13 days, and 10 days, respectively. Meanwhile, the time when the above three indicators sent out the warning signal simultaneously was only four days.
Figure 2 The control charts of DARX, DARY and DARZ for school A.
For school B, the effective surveillance time was 159 days, of which 44 days (34 days for grades 1-2 and 28 days for grades 3-6) showed outbreaks. According to the charts of DARX, DARY and DARZ (Figure 3), the days to meet the warning standard were 22 days, 16 days and 17 days, respectively. Meanwhile, the time when the above three indicators sent out the warning signal simultaneously was 15 days. There was only one day (December 14, 2021) when DARX, DARY and DARZ from both schools issued the warning signal simultaneously. After splitting DARX into DARL and DARH, the time for these two indicators to meet the warning standard was 20 and 14 days at school A, 19 and 15 days at school B, respectively. Either at school A (14-17, December, 2021) or B (November 15, December 13, 14, 17, 2021), the time when DARL and DARH issued outbreak signal simultaneously was four days (Figure 4).
Figure 3 The control charts of DARX, DARY and DARZ for school B.
Figure 4 The control charts of DARL, DARH for school A and B.
According to the detailed information reported by school physicians, all dates of infectious disease outbreaks in school A and B were identified from September 1, 2021 to June 24, 2022. Then, based on the warning dates shown in Figures 2-4, the surveillance accuracy of DARL, DARH, DARX, DARY and DARZ was calculated for the detecting of infectious disease outbreak in the two schools. Whether in school A or B, the data showed that indicators reported by FRSSS (DARX) outperformed indicators reported by school physicians (DARY and DARZ) in the accuracy of infectious disease surveillance, meanwhile the separation of DARX into DARL and DARH resulted in a significant increase in the specificity and Youden index (Table 2).
Table 2 Infectious disease surveillance effectiveness of different absenteeism indicators.
School
|
Surveillance indexes
|
Accuracy of infectious disease surveillance
|
sensitivity (%)
|
specificity (%)
|
Youden index (%)
|
A
|
DARL
|
95.0
|
88.0
|
83.0
|
DARH
|
92.9
|
92.1
|
85.0
|
DARX
|
100
|
80.6
|
80.6
|
DARY
|
100
|
78.2
|
78.2
|
DARZ
|
100
|
76.8
|
76.8
|
DARL or DARH
|
90.9
|
87.8
|
78.7
|
B
|
DAXL
|
100
|
89.3
|
89.3
|
DARH
|
100
|
91.0
|
91.0
|
DARX
|
100
|
83.9
|
83.9
|
DARY
|
100
|
80.4
|
80.4
|
DARZ
|
100
|
81.0
|
81.0
|
DARL or DARH
|
100
|
88.5
|
88.5
|