Examine the Data Quality in Texas Youth Fitness Study

Background/Purpose: Texas Senate Bill 530 passed in 2007 as a mandate for all public schools to increase physical activity and conduct fitness testing with FITNESSGRAM. More than 2.5 million students were tested in fitness and their scores were entered and uploaded to Texas Education Agency (TEA) since 2008. Only grade level Health Fitness Zone (HFZ) achievement was obtained through TEA and it is important to explore the strategies to handle large scale group level fitness data.

Method: Fitness date was collected by PE teachers inTexas through 6,913 out of a total of 8,526 public schools in 2011. A total of 2,922,851 students have been reported being tested by PE teachers. Grade level student enrollment in 2011 were also obtained through TEA and being merged with fitness data. Percentage of the students being tested were calculated and histograms of the distribution were examined. Two screening methods were applied. One is based on minimal number of students tested per grade is 10 for boys and girls, respectively. The other is at least 50% students per grade must be tested and no more than 120% students reported have been included. Linear contrast was used to examine the difference in aerobic capacity HFZ achievement by using different screening methods.    

Analysis/Results: The percentage of the students being tested were reported from 1% to 18000% (e.g. indicating some schools reported only 1% students were tested and some schools reported 180 times students being tested than the total number of students enrolled). The total number of students being tested per grade were reported from 6 to 1600. The linear contrast results showed that aerobic capacity HFZ with two screening method were significantly (P<0.05) different from the results in the unscreened data but the difference in the estimated HFZ was small (diff_{method1-no filter}=0.49% and diff _{method2-no filter}=-0.29%).  

Conclusions: Although the difference between non-screening data and screening data was trivial, it is important to rule out some unreasonable entered data to improve the quality of large scale statewide fitness data.