Article Text
PDF
Abstract
Objective The aim of this systematic review was to summarise and update the existing literature on determinants of physical activity and sedentary behaviour in young people, considering the methodological quality of the studies.
Methods Prospective studies were identified from searches in PubMed and PsycINFO, from April 2004 through November 2010. The authors included studies investigating the association between determinants of overall physical (in)activity and/or sedentary behaviour in healthy children or adolescents. When a determinant was investigated for its association with physical (in)activity and sedentary behaviour assessed between ages of 4–12, or mean age ≤12, it was classified as ‘child determinant’. When a determinant was investigated for its association with physical activity and sedentary behaviour assessed between ages of 13–18 or mean age >12, it was classified as ‘adolescent determinant’. Included articles were scored on their methodological quality and a best-evidence synthesis was applied to summarise the results.
Results The authors identified 30 papers, of which seven were of high methodological quality. Intention was found as a determinant of children's physical activity. Determinants of adolescents' physical activity were age (ie, older children were more active), ethnicity (ie, being of African–American descent determined being less physically active) and planning. The authors found insufficient evidence for determinants of sedentary behaviour.
Conclusion Studies on determinants of physical activity and sedentary behaviour were in general of poor methodological quality. To develop long-term effective interventions that increase physical activity and decrease time spent in sedentary behaviours in young people, we need more high quality prospective evidence on the determinants of these behaviours.
Statistics from Altmetric.com
Introduction
Nowadays, ample evidence exists on the detrimental health effects of insufficient physical activity in youth.1,–,4 Although not as extensive as evidence on physical activity, evidence on the harmful effects of sedentary behaviours is accumulating as well.5 Physical (in)activity and sedentary behaviours ‘track’ from childhood into adulthood,6 7 meaning that children who are currently inactive and sedentary persist in such behaviours as they grow older. Numerous initiatives have been conducted to increase children's time spent in physical activity and to decrease their sedentary time.8 9 However, these programs have been shown to be only marginally effective, especially in the longer term.10 11
The ecological health approach12,–,14 supports the assumption that more attention should be paid to different types of correlates of behaviour (eg, psychological, environmental) during the development of interventions. As a consequence, recent reviews have tried to summarise correlates of physical (in)activity and sedentariness.15,–,20 However, most of the research included in these reviews had cross-sectional designs, and the quality of these studies was not assessed.
This review elaborates on the reviews of Sallis et al19 and Van der Horst et al,20 in which studies on biological, demographic, psychological, behavioural, social and physical correlates of physical activity, insufficient physical activity and sedentary behaviours were summarised. Based on the recommendations of Van der Horst et al20 we added keywords on sedentary behaviours (eg, television viewing, gaming) to the former literature search, and included studies with a longitudinal design only. Moreover, different from the review of Van der Horst et al,20 we also assessed the methodological quality of the included studies.
We aim to systematically review the determinants of physical (in)activity and sedentary behaviour in young people, taking into account the methodological quality of the studies.
Methods
Selection of the literature
To assure a good fit to the search performed by Van der Horst et al,20 we performed a computerised search in two electronic databases (PubMed and PsycINFO) from April 2004 through November 2010. The search strategy was based on the former search of Van der Horst et al20 and consisted of four elements: (a) age (eg, child, youth); (b) physical (in)activity (eg, exercise, inactivity); (c) sedentary behaviour (eg, TV viewing, gaming); and (d) possible determinants of physical (in)activity and sedentary behaviour (eg, demographic factors, social environment). Terms referring to these four elements were used as MESH-headings and title words in both databases (See supplementary material, e-appendix 1).
Inclusion criteria
To be included in the review, an article had to meet the following criteria:
Subjects were healthy children (age 4–12 years or mean age ≤12 years), or adolescents (age 13–18 years or mean age >12) when physical (in)activity and/or sedentary behaviour was measured
The dependent variable was a measure of overall physical (in)activity, or a measure of sedentary behaviour assessed before the age of 18 years
The determinants were tested for their association with the dependent variable
The study had a prospective design.
We only included full-text articles published after 2004 in the English language in peer reviewed journals. Studies reporting on the effects of an intervention were excluded. Additionally, studies that did not measure overall physical activity but rather used a certain type of physical activity as outcome measure (eg, sports participation, light physical activity) were excluded from the review. Studies using a single item instrument to assess physical activity were excluded as well. Because of the limited number of prospective studies on sedentary behaviour, in contrast to the studies on physical activity, all articles that used sedentary behaviour as the outcome were included.
Selection process and data extraction
Two reviewers (LU and JN) independently screened a respective half of the yielded articles based on title and abstract. When an article seemed to fulfil all inclusion criteria, or if the title and abstract did not provide sufficient information, the full text article was obtained and checked. When one reviewer (LU or JN) could not decide on the inclusion of an article, the second reviewer was consulted. If the two reviewers (LU and JN) could not agree on the in- or exclusion of an article, a third author (MC) was asked for advice.
Two reviewers (LU and JN) extracted the following data from the identified studies:
Subject characteristics
Follow-up time
Measure of physical activity or sedentary behaviour
Type of determinants
Statistical analyses
Statistical methods
Results.
Because varieties of statistical techniques (eg, regression, random coefficient models, latent growth modelling (LGM)) were used to evaluate the longitudinal relationships, and most studies reported the adjustments for other potential correlates investigated, only results from the final fully adjusted model are presented. Furthermore, this review only reports on direct relationships, excluding moderator and/or mediator effects.
Categorisation of variables
Determinants of physical (in)activity and sedentary behaviour were categorised into five groups:
Demographic and biological determinants (eg, body mass index (BMI), parent education)
Psychological determinants (eg, attitude, intention)
Behavioural determinants (eg, past physical activity, screen time viewing)
Social determinants (eg, parental physical activity, social support)
Environmental determinants (eg, weather conditions).
This categorisation was also used in previous studies19 20 and is based on the ecological approach of determinants of health behaviour. Conceptually comparable determinants were combined (eg, self-efficacy and perceived behavioural control). Determinants were grouped by age; when a determinant was investigated on its association with physical (in)activity and sedentary behaviour assessed between ages 4 and 12, or mean age ≤12, it was classified as ‘child determinant’. When a determinant was investigated on its association with physical activity and sedentary behaviour assessed between age 13 and 18 or mean age >12, it was classified as ‘adolescent determinant’. Subsequently, determinants were grouped by behaviour (physical (in)activity or sedentary behaviour). We categorised determinants as showing a significant positive (+), significant negative (−) or no significant (0) relationship with physical (in)activity or sedentary behaviour.
Some articles reported results separately by sex or age groups (eg, age 13–14 and 15–16).21 Other studies used different cohorts in their analyses and presented separate results.22 Results for different time spans (eg, associations between predictors measured at T1 and the outcome measure at T2, and associations between predictors measured at T2 and the outcome measure at T3)23 were reported as well. Therefore, we decided to present the number of samples instead of studies that were associated with physical (in)activity and/or sedentary behaviour. This approach was consistent with earlier reviews on correlates of physical activity in children.19 20
Methodological quality assessment
Two reviewers (LU and JN) independently scored the methodological quality of the included articles using a 10-item criteria list, adapted from Ariens et al.24 The criteria list (table 1) assessed the methodological quality in four dimensions1 study attrition and follow-up duration (three items);2 assessment of determinants (two items);3 assessment of physical activity and/or sedentary behaviour (two items) and4 data analysis (three items). Criteria had a positive,1 negative (0) or don't know (?) answer format. If the study provided information on a quality item and met the criterion, we gave a positive score. If the study provided information on a quality item but did not meet the criterion, we gave a negative score. In case of no or insufficient information, we scored the quality item with a question mark. When an article referred to another study containing relevant information for scoring the quality items, the study of interest was retrieved. If the additional study did not provide the requested information, a question mark was given.
Quality criteria and overall positive scores on each item
Because several articles investigated multiple determinants (eg, biological, psychological and social), criterion D and E were scored on a scale ranging 0–1. For example, if a study assessed four different types of determinants of which one was measured with a reliable tool, a score of 0.25 was given.
We assessed the methodological quality score of each study by calculating the percentage of items that were scored positively, relative to the number of applicable quality items. If the quality score was ≥70% the study was considered to be of high methodological quality. A score <70% was considered as low quality.
Level of evidence
To synthesise the methodological quality of the studies, we applied a best evidence synthesis,25 26 consisting of the following three levels:
▶ Strong evidence: provided by generally consistent findings in multiple (≥2) high quality studies;
▶ Moderate evidence: provided by generally consistent findings in one high quality study and one or more low quality studies, or in multiple (≥2) low quality studies;
▶ Insufficient evidence: only one study available or inconsistent findings in multiple (≥2) studies.
We considered results to be consistent when at least 75% of the studies showed results in the same direction, which was defined according to significance (p<0.05). If there were two or more studies of high methodological quality, we disregarded the studies of low methodological quality in the evidence synthesis.
To test the stability of our conclusions based on the methodological quality assessment, we performed a sensitivity test after the final evidence synthesis, taking into account the high quality studies twice and low quality studies once.
Results were again considered to be consistent when at least 75% of the studies showed results in the same direction.
Results
The search in PubMed and PsycINFO yielded 9537 articles (see figure 1). After excluding duplicates, non-English articles and articles that were only published as an abstract, we screened the title and abstract of 7781 papers. In total, 371 full papers were obtained. Also, references of relevant reviews were manually searched, resulting in an additional 16 full papers. Eventually, 30 articles were identified as relevant for the review. None of these articles was included in the review of Van der Horst et al.20 Table 2 provides an overview of the characteristics of the included studies.
Flowchart of papers included in the present review.
Characteristics of studies on determinants of physical activity and sedentary behaviour sorted by methodological quality score
General findings
Nine21 27,–,34 of the 30 articles were carried out in the US, six23 35,–,39 in Canada, five40,–,44 in Great Britain, three45,–,47 in Australia and one in each of France,48 The Netherlands,49 Estonia,50 Sweden51 and New Zealand,52 respectively. Two papers reported results of multiple countries, one paper from Poland and China,22 and one from Great Britain, Hungary, Greece, Estonia and Singapore.53 Ten studies23 27 28 35 36 40 45 46 48 52 investigated determinants of physical activity and/or sedentary behaviour in children; 1822 29,–,34 37,–,39 41,–,44 49,–,51 53 focused on determinants of physical activity and/or sedentary behaviour in adolescents; and two21 47 studies examined both age groups. Twenty-two articles21,–,23 28,–,35 37,–,41 44 48,–,51 53 examined only determinants of physical activity and five articles27 36 42 43 45 only determinants of sedentary behaviour. Three articles46 47 52 examined determinants of both physical activity and sedentary behaviour. There were no studies that focused on determinants of physical inactivity or insufficient physical activity.
Methodological quality assessment
Scoring the quality of the 30 studies led to a disagreement between the reviewers of approximately 20%. All disagreements were discussed and resolved. When required, a third author (MC) was asked for advice. The methodological quality score ranged from 20%28 44 49 53 to 80%.47 52 Seven of the 30 included papers had a quality score ≥70%35 37 38 40 45 47 52 and were thus considered as being of high methodological quality. Only four articles (13%) reported a participation rate at baseline ≥80%.36 38 40 51 As has been previously mentioned, several articles investigated more than one determinant. To assess the proportion of reliable and valid measurements, we calculated the number of times the determinant groups were measured with a reliable and valid tool relatively to the total number of determinant groups that were measured in all studies. In 29% of the cases, the use of a reliable tool (item D) was reported. Use of a valid tool (item E) for assessing determinants was reported in 28% of the cases. Seven35 40 41 46 47 51 52 and five40 46 47 51 52 articles, respectively, used a reliable (item F) or valid (item G) tool for measuring physical activity or sedentary behaviour. Table 1 presents the quality criteria and the percentage of studies scoring positive per item. The full methodological quality scoring of the included studies is presented in table 3.
Methodological quality rating of the included studies sorted by quality score
Determinants of physical activity
Based on two low quality studies, we found moderate evidence for a significant positive relationship between intention and physical activity in children (table 4).
Summary of prospective studies on determinants of physical activity in children and adolescents
One high quality and one low quality study showed a significant positive relationship between age and adolescent physical activity (the older the participants, the higher their physical activity level), providing moderate evidence. We also concluded moderate evidence for a significant negative relationship between ethnicity (being of African–American race) and physical activity based on two low quality studies. Based on multiple low quality studies we concluded moderate evidence for a significant positive relationship between planning and past physical activity, respectively, and physical activity in adolescents (table 4).
Determinants of sedentary behaviour
For all determinants of sedentary behaviour we concluded insufficient evidence for both children and adolescents (table 5). The two most frequently studied types of sedentary behaviour in children were TV viewing and gaming. In adolescents, research mainly focused on sedentary behaviour in general. Biological/demographic determinants of sedentary behaviour were most frequently investigated in multiple samples.
Summary of prospective studies on determinants of sedentary behaviour in children and adolescents
The sensitivity test, taking into account the high quality studies twice and low quality studies once, did not change our conclusions. However, additional strong evidence was found for a positive association between being male and adolescent physical activity and BMI and child sedentary behaviour.
A total of 57 assessed determinants were investigated in one sample only: 35 regarding child physical activity, 18 regarding adolescent physical activity, and four regarding child sedentary behaviour27 28 32 35,–,37 40 45,–,48 (table 6). Based on the principles of evidence synthesis we could only conclude insufficient evidence for those determinants. For the sake of readability they were excluded from tables 4 and 5. Results from two high quality studies35 37 and one low quality study41 are not reported.
Determinants that were reported on in only one study sample (insufficient evidence)
Discussion
This review aimed to summarise and update the existing literature on determinants of physical activity and sedentary behaviour in young people, considering the methodological quality of the included studies.
The majority of the articles focused on biological/demographic, psychological, behavioural and social determinants of physical activity. However, due to the large variety of determinants that were investigated, we concluded sufficient evidence for only few determinants of physical activity. With regard to sedentary behaviour, prospective research was scarce, especially in adolescents. Again, because most determinants were studied only once, we concluded insufficient evidence for determinants of sedentary behaviour. Overall, few studies focused on environmental determinants of physical activity and sedentary behaviour. The review of Ferreira et al,18 in which environmental determinants of physical activity were established, confirms that little prospective research on this relationship has been conducted.
We found moderate evidence for intention as a determinant of children's physical activity, and age (ie, older children were more active), ethnicity (ie, being of African–American descent determined being less physically active) and planning as determinant of adolescent physical activity. Van der Horst et al20 and Sallis et al19 did not include articles that assessed the association between intention and child physical activity, or planning and adolescent physical activity. In addition Van der Horst et al20 found insufficient evidence for age as a determinant of adolescent physical activity, while the review of Sallis et al,19 reported a negative (ie, younger children were more active) association between age and physical activity in adolescents. Van der Horst et al20 found insufficient evidence for ethnicity as a determinant of physical activity, although Sallis et al19 reported a positive association between being of Euro-American race and physical activity in adolescents. Overall, the majority of the findings of the former reviews were not similar to our results. Moreover, few determinants that were included in previous two reviews were also investigated in the studies included in the current review. In addition, we only included prospective studies, further complicating comparison of results between these three reviews.
Methodological quality
The overall quality of the included studies was poor to moderate, with only seven out of 30 studies scoring ≥70%. The strengths of the included studies were the follow-up period, the number of cases relative to the number of variables (ie, the majority of the studies had a large sample size), and the inclusion of confounders in statistical analyses. The most common shortcomings in the methodological quality of the included studies were the lack of valid and reliable measures of both determinants and outcome measures (ie, physical activity and sedentary behaviour). Few studies40 46 47 51 52 provided information on both the validity and reliability of their measures. The provided information on reliability and validity was often not accessible through the provided references, or was not relevant (eg, the validation study was performed in middle-aged men). For the majority of studies, we had to check other articles on the reliability and validity of the measurement instrument54 than was referred to in the initial paper.
Due to our strict definitions of reliable and valid measurement instruments, few studies received a positive score on the reliability/validity of determinants. We encourage the development and use of reliable and valid measures of both determinants as well as outcome measures, an area in which there is still a lot to gain.
For future research we recommend:
To focus on determinants of child and adolescent sedentary behaviour
To focus on environmental determinants of physical activity in both children and adolescents
To use reliable and valid measures of both determinants as well as the actual behaviours
To conduct prospective studies of high methodological quality.
Strengths and limitations
The major strength of the review is the quality assessment of the included studies. Also, because it was limited to prospective studies, we were able to report on longitudinal relationships and not only on cross-sectional associations. We also extended our search with specific keywords on sedentary behaviour which resulted in an overview of determinants of sedentary behaviour.
However, a few limitations should be mentioned. First, because we used two databases for our computerised search, we may have missed potentially relevant articles that were only available in other databases. Nevertheless, because there is considerable overlap between electronic databases and we also manually searched for potential relevant articles, we are confident that the use of additional databases would have only yielded few new articles. Second, we excluded all non-English studies and this may have biased our results. Third, because of the large number of studies using physical activity as their outcome measure, we included only studies using overall physical activity. Hence, possible differences in associations with various specific sub-behaviours (eg, moderate physical activity, sports participation) were not established. Fourth, although two reviewers rated the articles independently, the scoring may have been biased by subjectivity. However, we performed a sensitivity test in order to test the stability of our conclusions, which remained similar (except for additional strong evidence for a positive association between gender (being male) and adolescent physical activity, and BMI and child sedentary behaviour). Finally, as we did not contact authors to gain more information on insufficiently described quality items, we may have scored a quality item with a question mark, although additional information could have been available elsewhere. However, experience has taught us that if information is available, authors will report on it in their article or refer to another article containing the relevant information.
Conclusion
Few high methodological quality studies on determinants of physical activity and sedentary behaviour have been conducted. Studies performed in this area are heterogeneous regarding the determinants and measures of physical activity and sedentary behaviour. To develop long-term effective interventions that increase physical activity and decrease time spent in sedentary behaviours in young people, we need more high-quality evidence on the determinants of these behaviours.
Acknowledgments
The authors would like to thank René HJ Otten for his contribution during the literature search.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Web Only Data - This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
-
Competing interests None.
-
Provenance and peer review Not commissioned; externally peer reviewed.
Linked Articles
- Electronic pages