HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Extract Full Text (PDF) Submit a response Purchase Article View Shopping Cart Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Chen, J. J. Search for Related Content PubMed PubMed Citation Articles by Chen, J. J.

John J. Chen, PhD

John J. Chen is with the St Louis University School of Public Health, Saint Louis, Mo.

Correspondence: Requests for reprints should be sent to John J. Chen, PhD, St Louis University School of Public Health, 3545 Lafayette Ave, Salus Center 300, Saint Louis, MO 63104 (e-mail: chenjj{at}slu.edu).

Chiu et al. reported the dramatic impact of the Taiwan motorcycle helmet use law on the mortality and morbidity of head injuries 1 year before and after the law's implementation.¹ The authors appear, however, in several places to be confused about the interpretation of statistical test results and inappropriately use these results to support their conclusion.

In Table 2,^1(p765) they compare Glasgow Coma Scale [GCS] scores for the severity of motorcycle-related head injuries before and after implementation of the Taiwan helmet law. They report a 34.3% reduction in injury frequency for the severe injury group, a 35.5% reduction for the moderate injury group, and a 32.3% reduction for the mild injury group; they also report that a highly significant likelihood ratio ² (P < .001) was associated with the results. But the likelihood ratio test performed here could not be used to support the claim that the number of severe head injuries decreased. It tested the proportional difference of various severity levels as a fraction of the total number of head injuries before and after the law was passed. More precisely, the test evaluated whether the proportions for the 3 severity levels before implementation of the law (484/5260 = 9.2% for the severe group, 521/5260 = 9.9% for the moderate group, and 4255/5260 = 80.9% for the mild group) were similar to those after implementation (318/3535 = 9% for the severe group, 336/3535 = 9.5% for the moderate group, and 2881/3535 = 81.5% for the mild group).

Reanalyzing the GCS data, I could not find any of the differences (P = .77) that Chiu et al. mentioned in their Results section (P < .001). This is not surprising, given the very similar proportions for the 3 severity levels before and after implementation of the law that are shown in Table 2. Even though the overall number of injuries decreased by 33%, from 5260 to 3535, a quite different picture was shown by the likelihood ratio test. There is no statistical difference in the proportions of head injuries of various GCS severity levels during the 2 periods.

Second, regarding the Glasgow Outcome Scale (GOS) scores results also reported in Table 2,¹ although a significant likelihood ratio ² (P < .001) was found for the changes in severe outcomes, it only indicates that the proportions of some of the outcomes, as a fraction of the total number of cases, were not statistically the same before and after the helmet law. Even though the proportions presented in the table appear quite similar, this significance is not surprising given the large sample size of the study. But the significant test result does not necessarily suggest that "outcomes were better (lower [GOS] score) after implementation of the law," as the authors claimed.¹ In order to show the decrease of the GOS scores, the authors should have applied an ordered categorical data analysis approach. The application of a mean score test using integer scores before and after the law showed no statistical significance (P = .12)—that is, the average GOS scores are quite similar.

Third, reanalyzing Chiu and colleagues' Table 3,¹ I found that all significant likelihood ratio tests for associated injuries, except for "chest," resulted in an increased odds of associated injuries, such as "total associated injuries" (P < .001), "cervical spine" (P < .001; this contradicts the authors' conclusion of nonsignificance for "cervical spine"), and "upper extremity" (P < .001). These results suggest a significant increase in the proportion of various associated injuries. Again, the authors inappropriately interpreted the significant ² results and misused the statistical test results to support their conclusion of a reduction of associated injuries.

In order to arrive at meaningful and convincing conclusions, it is crucial that appropriate statistical analysis methodologies and proper interpretation of the results be applied. Without a proper denominator for the total number of motorcycle-related injuries (or accidents) before and after implementation of the law, it is very difficult to convince the reader of the law's effectiveness on the basis of a percentage change in head injuries alone. To determine the true impact of the helmet use law on the mortality and morbidity of motorcycle head injuries in Taiwan will require further study.

Acknowledgments

The author would like to thank Drs. B. Levin and G. Evans for their careful comments.

References

1. Chiu W-T, Kuo C-Y, Hung C-C, Chen M. The effect of the Taiwan motorcycle helmet use law on head injuries. Am J Public Health. 2000;90:793–796.[Abstract/Free Full Text]

This Article Extract Full Text (PDF) Submit a response Purchase Article View Shopping Cart Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Chen, J. J. Search for Related Content PubMed PubMed Citation Articles by Chen, J. J.