December 2, 2015

expert reaction to screen time and physical activity during early adulthood linked to midlife cognition

Levels of physical activity and patterns of television viewing and their effect on a person’s cognition is the subject of a paper published in the JAMA Psychiatry, with the authors reporting that high levels of inactivity and low levels of television viewing were associated with poorer cognitive function.

Dr Andrew Przybylski, Experimental Psychologist and Research Fellow at the Oxford Internet Institute, University of Oxford, said:

“The research casts a wide empirical net with the goal of identifying the relationships – if any – between television viewing in early adulthood and cognitive functioning measured 25 years later. The primary strengths of this study include a relatively large sample of 3247 adults (64% of those who enrolled in the study), and geographically diverse (recruited from four separate American cities). As a whole it appears the study design and data collection were competently conducted. That said there are three shortcomings that should provide serious food for thought before the general public, policymakers, and health practitioners consider the work relevant to their work or everyday behaviour.

“First, these data rely entirely on a potentially problematic self-reported measure for television time. Participants were asked how much time they spend watching TV each day, presumably this used an open ended response (i.e. ranging from 0 to 24 hours). Though the paper is not clear on this, the authors go on to define “low to moderate” as fewer than 3 hours a day and more than this as “high”. This is an error that introduces a statistical artefact that exaggerates the small differences between those who might watch for 2.99 vs. 3.01 hours a day and distorts (minimises) the large differences between those who might watch 3.01 and 10.00 hours a day.

“Second, these data do not account for baseline cognitive performance at the start of the study. This means that the data are essentially correlational, while there are covariates introduced and tested (e.g. socioeconomic status, gender), it is not possible to draw any inferences about an effect. This is mainly because nearly 1 in 3 participants did not complete the study. Without a baseline, it is simpler, and potentially more accurate, to assume that participants dropping out may explain the findings reported and a single time point, correlational design might have produced more useful findings.

“Finally, there are two interrelated issues surrounding the nature of the data analysis that merit serious attention. First, because the number of observations are quite large most NHST (null hypothesis significance testing) models used will show a statistically significant result even if any link between measurements is miniscule. Second, there are at least three separate statistical tests performed on cognitive outcomes and these do not correct for multiple statistical comparisons. Taken together, it is not clear how large the net effect of television watching on these grouped outcomes is without the appropriate multivariate tests. Without knowing the covariance or proportion reduction in error statistic between the measures and what the authors had as apriori estimate of what constitutes a practically (vs. statistically) significant relationship it is difficult to judge if the research supported or disconfirmed the researchers’ expectations.

“Taken together, the work should provoke continued conversation about the nature of different forms of interactive media and underline the value of open science methodology including open datasets, pre-registered analysis plans, and robust and open peer review process. Until these innovations are introduced into this research literature we will be left scratching our heads at studies like this.”

‘Effect of Early Adult Patterns of Physical Activity and Television Viewing on Midlife Cognitive Function’ by Hoang et al. published in JAMA Psychiatry on Wednesday 2^nd December. 

All our previous output on this subject can be seen at this weblink: http://www.sciencemediacentre.org/?s=%22screen%20time%22&cat=

Declared interests

Dr Andrew Przybylski declares that he has no relevant interests.