 |
|
 |
|
 |
 |
|
 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
To the Editor.—
The report by Christakis et al1 suggests that early-age television viewing is associated with attention deficits later in childhood. It has been argued that television promotes inactivity, obesity, promiscuity, and possibly aggression in adults, and thus the impact of television viewing by children is certainly a concern. The message resonates in a society seemingly obsessed with public health villains. Although television eventually may earn a place among contemporary threats to our well-being, Christakis et al do not provide a convincing case for this conclusion.
The use of 1.2 standard deviations above the mean on a hyperactivity scale as the threshold for attention disorders is problematic. The authors defend the arbitrary threshold by noting that it corresponds to previous reports of attention-deficit/hyperactivity disorder (ADHD) prevalence in age-matched community samples. However, this misapplication of statistical data seriously compromises the very foundation of the authors’ conclusions.
In the first place, the choice of 1.2 standard deviations above the mean as the threshold for attention disorders is specious and begs the question by assuming the conclusion before advancing the argument. According to the American Diabetes Association, 6% of the US population has diabetes, but this does not imply that everyone with blood glucose >94th percentile has diabetes, and measurement of blood glucose is a far more reliable test than subjective responses to behavioral surveys. Second, the 10% incidence of ADHD cited by the authors is not entirely consistent with recent estimates. Previous reports state that ADHD affects anywhere from 4% to 12%2 or from 3% to 10% of children in the United States.3 Finally, the authors admit that the surveys used in this study are not equivalent to a clinical diagnosis of ADHD, yet they are comfortable with choosing a cutoff that fits the bona fide incidence of ADHD.
We also suspect that the authors selectively indict television viewing because it fits their hypothesis, overlooking other possibilities apparent in their results. Inspection of the data in their Table 2 reveals that maternal self-esteem is more highly correlated with the outcome measure of attention deficit than the hours of television watched per day, a result never mentioned in the discussion.
Proving a causal relationship between television viewing and behavioral traits has always been difficult. The authors of this study acknowledge that content is quite likely a critical determinant in television-viewing influence, but their conclusion does not take this important factor into account. Children raised in an environment that promotes attention-seeking behavior may have greater exposure to television because of factors that limit more wholesome activities. On the other hand, it makes just as much (or more) sense to hypothesize that poorly behaved children are exposed to more television by parental choice.
Thus, particular care must be taken when interpreting statistical arguments built on such precarious footing. When reviewers and readers evaluate conclusions based on statistical analysis of mined data, would be wise to consider the likelihood that the statistics are being used, in the words of Andrew Lang, "... as a drunken man uses lampposts—for support rather than illumination."
REFERENCES
Frederick J. Zimmerman, PhD
Department of Health Services
Child Health Institute
University of Washington
Seattle, WA 98115
In Reply.—
We appreciate the opportunity to reply to Drs Bertholf and Goodison.
Their first concern is with our cutoff for attentional problems. The use of standard deviations from the mean to define an "abnormal" level is in fact a common approach. Although it is by no means diagnostic of a problem (and we never make a claim that it is), it suggests that the reported value is far from the mean and at minimum would represent worrisome attentional problems. Although one could set this threshold either higher or lower, we chose that point in part because it is consistent with the recent population-based estimates of attention-deficit/hyperactivity disorder (ADHD), a clinical diagnosis that, by definition, represents abnormal ability to attend. In one multicenter study, 9% of children presenting for nonacute primary care visits had an attentional disorder,1 whereas the percentage of elementary school children in a general population sample who meet Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition criteria for ADHD was estimated recently at 11.4%.2 Furthermore, we believe that a cutoff of the 10th percentile has face validity. The readers are free to decide on their own if the associations we found between our primary exposure (television) and being at or below that percentile for attention are clinically meaningful.
They also complain about the problem of measurement error because of the subjectivity of parental surveys, but they are incorrect in their assessment of what the likely effects of measurement error would be on our analysis. Generally, random measurement error will reduce the ability to detect an effect in data by biasing toward the null,3 so our significant finding despite the existence of measurement error suggests that, if anything, the true effect may be stronger than what we were able to identify.
As to the question of the relative contribution of the different covariates, we believe that their concern is misplaced. Following standard scientific methods, we conducted this study to explicitly test an a priori specific hypothesis that has existed in the literature for a long time, in support of which there has been sound, if only circumstantial, evidence to date. The other variables in our model were control variables that might plausibly confound our primary relationship of interest but about which we did not in fact have hypotheses articulated ex ante. Although we did not discuss them in the article, their directionality is in fact plausible, and they themselves may warrant additional investigation. Finally, the fact that they are controlled for in our multivariate model suggests that television viewing is associated with attentional problems even when these other independent predictors are adjusted for, which is a strength of our analysis.
For what it may be worth, the letter writers are also incorrect in their assessment of the relative effects of early television exposure and mother’s self-esteem. It is true that the odds ratio presented in our table is greater for self-esteem than for hours of television watched. However, the variables are not scaled the same: an inspection of Table 1 in the article will reveal that the standard deviation of mother’s self-esteem is 1.0, whereas the standard deviation of television hours watched per day is 2.9. Adjusting for the difference in the scaling of the variables, the magnitude of the effects of a 1-standard-deviation change in mother’s self-esteem and television hours watched per day are similar.
Bertholf and Goodison rightly argue that our analysis cannot be seen as the definitive word on the relationship between television and attentional problems. We acknowledged the limitations of the data in our article and couched our conclusions with the appropriate caveats. Nevertheless, we believe that our findings are important and innovative and have moved many people to respond, most positively and some defensively. As with all new findings, however, only comprehensive subsequent research effort by many teams of researchers will reveal this article to be either seminal or a red herring. We invite the letter writers to contribute constructively to this important effort.
REFERENCES
This article has been cited by other articles:
 |
|
 |
|
  T. Stevens and M. Mulsow Viewing Television Before Age 3 Is Not the Same as Viewing Television at Age 5: In Reply Pediatrics, July 1, 2006; 118(1): 435 - 436. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
||
|
||
Home | My Pediatrics | Journal Information | Current Issue | Past Issues | Subscriptions & Services | Contact Us Click here for faster international access © 2004 American Academy of Pediatrics. All rights reserved. |
||
|
||
