Scottish children's physical activity levels: study analysis

Chapter 2 Methods

2.1 Sample

The Scottish Government and Scotcen Social Research have worked in partnership to provide the MRC/ CSO Social and Public Health Sciences Unit ( SPHSU), University of Glasgow with access to - subject to consent from families - the GUS Birth Cohort 1 ( BC1) sample. GUS is a longitudinal cohort study that began in 2005 with an aim of tracking the lives of children through childhood and beyond. BC1 is the first of two GUS birth cohorts to have been followed from age 10 months. Now aged 10 years, the cohort form the sample for this report ^[11] .

The data collection for this study took place between May 2015 and May 2016, following the eighth sweep of the GUS main interview stage ^[12] . Ethical approval was provided by the College of Social Sciences, University of Glasgow.

As part of the GUS sweep eight interviews, respondents were provided with brief information about a sub-study, Studying Physical Activity in Children's Environments across Scotland - SPACES ^[13] being conducted by SPHSU and asked if their contact details could be passed on to SPACES staff. From a possible 2,402 children who had participated in sweep 8 of GUS, 90% (n=2,162) consented to be contacted by SPHSU. This number represents 41% of the original sweep 1 families (n=5217).

Those who consented to be contacted were similar to the GUS sweep 8 sample, with very slight differences across certain socioeconomic demographics: the consenting sample contained slightly lower proportions of children from mothers with no educational qualification (4.6% compared to 5.1% in those consented to be contacted), and slightly higher proportions of children from families with household income between £38,000-£44,000 (8.6% compared to 8.1% who consented). No differences existed in Body Mass Index ( BMI).

2.1.1 Sample Characteristics

From the 2,162 GUS children who consented to be contacted, 1,096 children took part in the data collection, representing 46% of the overall GUS sweep 8 phase 1 sample. From these 1,096 participants, 859 children sent data back to SPHSU for processing, representing 40% of the possible recruitment sample. 774 children (mean age of 11.1 years old; representing 36% of possible recruitment sample and 71% of those who participated) provided at least 4 weekdays and 1 weekend day data, of which 417 (54%) were girls and 357 (46%) were boys. All weighting procedures and statistical analyses were based on these 774 participants (see Analysis section) ^[14] .

Unweighted demographic characteristics of those who returned valid data are summarised in Table 1. Means and standard deviations are presented for numerical variables and proportions for categorical variables.

The average height was 140.4cm (girls 140.0 cm, boys 140.9 cm). The average weight across the sample was 35.4kg (girls 35.4, boys 35.4). Just under half (47%) of the children in the sample lived in households where at least one adult had a degree level qualification, while 4% of the sample came from a household where the parents/carers had no qualifications or lower level qualifications. Around two fifths (42%) of the children lived in households where the annual income was over £50,000, and a small proportion (1.4%) lived in a household with an annual income of less than £10,000 per year. In relation to the child's location as defined by the Scottish Government Urban Rural Classification ^[15] , approximately 31% of the children lived in a Large Urban Area (a settlement of over 125,000 people) and around 10% were classified as living in a remote rural area (a settlement of less than 3,000 people and with a drive time of over 30 minutes to a settlement of 10,000 or more). In our sample, over one-third of the children lived in the least deprived areas and 8.3% in the most deprived areas.

Compared to the full sweep 8 phase 1 distribution (2,402 children), the sample contained slightly lower proportions of children classified as obese (13.0% compared to 17.3% in full sample), but greater proportions of children within the healthy weight (68.7% compared to 65.2%). Children whose parent/carer had no educational qualifications (1.8% compared to 5.1%), Lower level Standard Grades (2.2% compared to 4.4%), and Upper level Standard Grades (14.2% compared to 20.2 %) were underrepresented in the sample; children whose parent/carer held degree level qualifications (47.0% compared to 36.2%) were overrepresented. Compared to the sweep 8 phase 1 distribution, the sample within this report contained slightly lower proportions of children from households earning less than £9,999 (1.4% compared to 2.8%), £10,000-£19,999 (9.4% compared to 15.2%), and £20,000-£28,999 (10.6% compared to 13.2%), but contained higher proportions of children from households earning £38,000-£49,999 (16.8% compared to 14.7%) and households earning over £50,000 (42.0% compared to 33.1%).

Table 1 - Sociodemographic characteristics of the participants who returned valid data (unweighted)

Variable	Girls (n=417)	Boys (n=357)	All (n=774)
Height in cm ± SD	140.0 ± 6.8	140.9 ± 9.8	140.4 ± 8.3
Weight in kg ± SD	35.4 ± 8.1	35.4 ± 7.8	35.4 ± 7.9
Highest level of education of primary carer/parent
Missing *	3 (0.7%)	0 (0.0%)	3 (0.8%)
No qualifications	7 (1.7%)	7 (2.0%)	14 (1.8%)
Other	1 (0.3%)	2 (0.6%)	3 (0.4%)
Lower level Standard Grades and Vocational	9 (2.2%)	8 (2.2%)	17 (2.2%)
Upper level Standard Grades/Intermediate Vocational	65 (15.6%)	45 (12.6%)	110 (14.2%)
Higher grades /Upper level vocational	157 (37.6%)	106 (29.7%)	263 (34.0%)
Degree level academic and vocational	175 (42.0%)	189 (52.9%)	364 (47.0%)
SIMD quintile
1 (most deprived)	35 (8.4%)	29 (8.1%)	64 (8.3%)
2	55 (13.2%)	44 (12.3%)	99 (12.8%)
3	96 (23.0%)	73 (20.4%)	169 (21.8%)
4	110 (26.4%)	97 (27.2%)	207 (26.7%)
5 (least deprived)	121 (29.0%)	114 (31.9%)	235 (30.4%)
Urban/rural indicator
Large urban areas	124 (29.7%)	115 (32.2%)	239 (30.9%)
Other urban areas	125 (30.0%)	98 (27.5%)	223 (28.8%)
Accessible small towns	47 (11.3%)	33 (9.2%)	80 (10.3%)
Remote small towns	19 (4.6%)	7 (2.0%)	26 (3.4%)
Accessible rural	67 (16.1%)	64 (17.9%)	131 (16.9%)
Remote rural	35 (8.4%)	40 (11.2%)	75 (9.7%)
Household income category
Missing	22 (5.3%)	19 (5.3%)	41 (5.3%)
<£3,999-£9,999 pa	6 (1.4%)	5 (1.4%)	11 (1.4%)
£10,000-£19,999 pa	34 (8.2%)	39 (10.9%)	73 (9.4%)
£20,000-£28,999 pa	42 (10.1%)	40 (11.2%)	82 (10.6%)
£29,999-£37,999 pa	62 (14.9%)	50 (14.0%)	112 (14.5%)
£38,000-£49,999pa	80 (19.2%)	50 (14.0%)	130 (16.8%)
£50,000 pa	171 (41.0%)	154 (43.1)	325 (42.0%)

*Missing = 'Don't know', Refusal', or 'No Information' Percentage figures rounded to 1 decimal place

2.2 Recruitment to the Study

For each group of participants, study information, registration documents, and consent forms were issued by post using the main parent/carer as primary contact. The primary contact was asked to return registration documents and consent forms back to SPHSU before their son/daughter could be enrolled on the study. SPHSU staff phoned the primary contact one week following the postal date to check that participants had received the documentation. SPHSU staff were able to enrol participants over the phone if required (although completed consent forms were still required to be returned before any data could be used) and a start date for study equipment to be delivered to the home (or most appropriate location) was organised.

2.3 Data Collection

Participants were sent all necessary equipment through the post. Packaging and contents had been carefully prepared and piloted to ensure the contents would fit through most letterboxes. A pre-paid envelope was provided to return the study materials upon completion of the protocol.

2.4 Physical Activity Measurement

Physical activity was measured through two separate methods: objectively through an accelerometer activity monitor (the ActiGraph GT3X+), and through self-report, using the Physical Activity Questionnaire for Children ( PAQ-C), which had the terminology and language adapted for use in the UK.

2.4.1 ActiGraph GT3X+ accelerometer

The ActiGraph GT3X+ (ActiGraph, Pensacola, Florida) is an activity monitor that measures acceleration across three axes ^[16] . Small (4.6 x 3.3 x 1.5cm), lightweight (19g), and unobtrusive, the ActiGraph is worn at the hip by way of an elastic band. The acceleration signal is digitised and then processed to provide information regarding frequency, intensity, and duration of activity performed. Participants were asked to wear the device, during waking hours, for 8 consecutive days.

Participants were required to wear the devices for 10 hours on a week day, and 8 hours on a weekend day to be classified as having sufficient data ( i.e. wear time) to create a reliable estimate of daily physical activity. We assumed that a device was not being worn if there were 60 consecutive minutes of no acceleration recorded by the device and these periods were removed from any analyses. Following the International Physical Activity and the Environment Network ( IPEN), adolescent accelerometer data collection protocol ^[17] , children who provided at least five days including one weekend day were included in the analyses.

2.4.2 Outcomes measured

The activity monitors were used to extract three main outcomes:

1. A measure of total physical activity that integrates all movement that is recorded through the device. These are measured in a metric called 'counts'. To take into consideration the length of time which a device is worn, we standardise these 'counts' to 'counts per minute' ( CPM).

2. Estimates of time spent in different physical activity intensities as categorised using count thresholds:

• Time spent sedentary, and in MVPA is presented as an absolute measure ( e.g. hours or minutes).
• There is a continuing debate within the literature regarding the appropriate 'cut points' at which to classify children's physical activity data into sedentary, light, moderate, and vigorous. Cut points use the CPM outcome to classify activity into the varying intensity levels. The Evenson (2008) count threshold was used within this analysis as it has been shown to accurately represent physical activity across all levels (sedentary, light, moderate, and vigorous; see Table 2) in children between 7 and 15 years old (Trost, 2007; Loprinzi, Moore, & Pfeiffer, 2011). As a reference point, the MCS conducted their own calibration study to ascertain the appropriate cut points to be used within their study population (7 years old). 'Counts' of less than 100 CPM were classified as sedentary time and counts of more than 2296 CPM were classified as MVPA. As can be seen below, these are similar to those reported by Evenson.

Table 2- Classification of different physical activity intensities (and sedentary) using the Evenson (2008) cut point thresholds

Intensity Classification	Counts per minute ( CPM)
Sedentary	≤100
Light	101-2295
Moderate	2296-4011
Vigorous	≥4012

3. Measuring the proportion of children who meet the CMO physical activity guidelines (Table 3)

• Two approaches were taken to measure the proportion of children who met the current CMO physical activity guidelines - specifically a minimum of 60 minutes of MVPA per day: i) Children had to meet, or surpass, 60 minutes of MVPA on each day that they wore their accelerometer - termed the 'Threshold' method; ii) Children's average MVPA per day, across all days that they wore their accelerometer, had to be greater than or equal to 60 minutes - termed the 'Averaging' method.

Table 3 - Description of the two approaches used to measure the proportion of children who meet the CMO physical activity guidelines

Prevalence estimate	Description
Threshold Estimate	Must meet at least 60 minutes of at least moderate intensity activity on each valid day to be identified as meeting the guidelines.
Average Estimate	MVPA will be averaged across valid days and participants with a mean time ≥60mins/day marked as meeting the guidelines.

2.4.3 Physical Activity Questionnaire - Children ( PAQ-C)

There are number of things to consider when choosing a questionnaire to assess population level physical activity in children, not least the age of the child (Biddle, et al., 2011). Trost (2007) has suggested that this type of method may not be appropriate for children under 10 years old as recall and cognition issues may affect the validity and reliability of the data. However, the PAQ-C has been designed to address some of these concerns, particularly the issue of cognitive capabilities (Crocker, Bailey, Faulkner, Kowalski, & McGrath, 1997). This is largely due to the greater weight being placed on measuring 'general' physical activity levels, rather than trying to extract reliable data on complex constructs such as 'intensity'. Designed in the USA by Peter Crocker and colleagues, the PAQ-C is a 7-day recall questionnaire that asks a number of questions to assess physical activities engaged in over the previous 7 days. It encompasses questions that assess levels of physical activity during school intervals, lunch, P.E classes, after-school, evenings, and weekend and results in a summary score ranging from one to five (see Appendix A for questionnaire).

PAQ-C scoring:

• The questionnaire provides a summary physical activity score derived from nine items, each scored on a 5-point scale.
• Each of the nine items will have a score between 1 and 5 and the overall PAQ-C activity summary score is the mean of all 9 items.
• 1 denotes low physical activity
• 5 denotes high physical activity

2.5 Comparison between ActiGraph & PAQ-C

The ActiGraph accelerometer, for the purposes of this report, was considered to be a 'criterion' method of analysis. That is to say it was considered to be the measurement approach that most accurately captures levels of physical activity - to which we wanted to compare the questionnaire. An important question being asked by this report was whether self-reported approaches can demonstrate similar - ideally the same - results and patterns as an objectively measured approach ( i.e. an accelerometer). Specifically, the PAQ-C and accelerometer outputs were compared to assess their ability to extract similar physical activity levels, and patterns of activity between gender, and quintiles of SIMD. As the questionnaire and accelerometer do not produce the same type of outputs, we ranked the scores from the accelerometer (average time spent in MVPA) and questionnaire to see if a relationship existed between both. This was tested by the Spearman's rank order correlation test.

2.6 Analysis

All analyses allowed for the stratification and clustered survey design of GUS, and the data was weighted using cross-sectional weights developed and supplied by Scotcen ^[18] , to compensate for potential response bias in the sample. General Linear Modelling ( GLM, IBM SPSS 21) was used to investigate statistical differences between boys and girls, and sociodemographic characteristics ( SIMD), across the outcome measures ( CPM, MVPA, sedentary time, and PAQ-C scores). These analyses were adjusted for season of measurement, device wear time, and number of valid days worn.