Skip to main content

Publication, Part of

National Child Measurement Programme, England 2020/21 School Year

Official statistics, National statistics

National Statistics

Text added to Introduction section

Text has been added to the Introduction section to make information in the Methodology and Data Quality section and table A2 more accessible. This clarifies that many Local Authorities measured in additional schools and data on 299,000 pupils was submitted and used.

16 November 2021 18:00 PM

Methodology and Data Quality

Data collection and sampling

Following the disruption to the 2019/20 collection, NCMP did not officially open as usual at the start of the 2020/21 school year in September 2020. Local Authorities were not required to start collecting child measurement data. At the end of March 2021, following a period of school closure, Local Authorities were asked to use the remainder of the academic year to collect a representative sample of child measurement data to enable a national estimate of children’s weight status (including obesity prevalence) for 2020/21.

The sample of schools was stratified by deprivation and ethnicity, taking every 9th school, to yield 10% of children in the local area, after factoring in that there could be a higher than normal level of pupil absence. The aim was to have a representative sample of children measured in England in terms of deprivation and ethnicity mix.

Schools were ranked into deprivation quintiles and then within each quintile by the proportion of children measured in 2018/19 who were in the Asian ethnic group and then the proportion who were in the Black ethnic group. The starting point for the selection of every 9th school was randomised for each Local Authority. Where a school had been selected for measurements but had very small numbers of children in both Reception and Year 6, a school on the list either above or below was selected instead. This above or below selection was randomised.

Local Authorities were only required to measure in the sample schools selected but were advised that they could measure in additional schools and were provided with details on how to select these schools to maintain a representative sample. This involved selecting an alternative school directly above or below the selected schools, and ensuring that additional schools selected were spread evenly, in the ordered school list provided to them. If a sample school had permanently closed or refused to take part then a substitute school could be chosen by taking the school above or below at random, e.g. by tossing a coin.

If one of the sample schools on the list was a junior school but there was also an infant school on the same site that had not been selected, Local Authorities were advised to measure in both. Similarly, this would also apply if the infant school had been selected and the junior school had not.

The NCMP system contained the number of pupils at each school, based on Department for Education data on numbers of children in Reception and Year 6 in school year 2019/20. Normally Local Authorities could update these with the 2020/21 pupil figures, but this year were asked to do this only in sample schools.

Local Authorities were advised that measuring at least 75% of eligible children would produce robust estimates of obesity prevalence for their area provided the sample of schools was representative of all schools in their area.

Any measurements taken prior to the official NCMP start in March 2021 were still able to be entered onto the NCMP system and thus collected.

As per usual NCMP years, Local Authorities were only asked to collect data from mainstream state-maintained schools but collection of data from special schools (schools for pupils with special educational needs and pupil referral units) and independent schools is encouraged. Since the proportion of records from independent and special schools varies each year, this report excludes such records to ensure consistency over time. There are also concerns around how representative the participating independent and special schools would be.

The measurement of children's heights and weights, without shoes and coats and in normal, light, indoor clothing, was overseen by healthcare professionals and undertaken in school by trained staff. The Office for Health Improvement and Disparities (formerly Public Health England) provides guidance to Local Authorities on how to accurately measure height and weight.

More information can be found in the National Child Measurement Programme operational guidance.


Data validation

Full details about validation are provided in NHS Digital’s validation document and have been summarised below.

Local Authorities enter data into the NCMP system which validates each data item at the point of data entry. Invalid data items (e.g. incorrect ethnicity codes) and missing mandatory data items are rejected and unexpected data items (e.g. “extreme” heights) have warning flags added.

During the collection the NCMP system provides each Local Authority with real time data quality indicators, based on the data they have entered, for monitoring and to ensure the early resolution of any issues. At the end of the collection each Local Authority must confirm any data items with warning flags and sign off their data quality indicators. In cases where the data quality indicators breach the required thresholds (provided in Validation of National Child Measurement Programme Data linked to above), LAs are required to provide a breach reason. Breach reasons are not included in the 2020/21 report because Local Authorities were not actively followed up to investigate breaches following the suspension of the NCMP.

The below table shows the key data quality measures, at national level, since the first year of robust NCMP data was collected in 2006/07.

The values were not calculated for all data quality measures in 2006/07.

Table A1 : LA data quality report for NCMP 2006/07 to 2020/21                         
Year 2006/071 2007/08 2008/09 2009/10 2010/11 2011/12 2012/13 2013/14 2014/15 2015/16 2016/17 2017/18 2018/19 2019/20 2020/21
Percentage of records with heights rounded to whole numbers 30% 21% 18% 19% 17% 17% 17% 17% 18% 16% 16% 17% 17% 17%
Percentage of records with weights rounded to whole numbers 23% 15% 12% 12% 10% 10% 10% 10% 10% 10% 10% 10% 10% 10%
Percentage of records with missing home/child postcodes 3% 2% 1% 0% 1% 0% 0% 0% 0% 0% 0% 0% 0% 0%
Percentage of records with missing ethnicity codes2 33% 23% 17% 17% 15% 14% 9% 10% 9% 7% 7% 6% 9% 6%
Percentage of records with missing NHS numbers3 39% 33% 30% 19% 17% 15% 18% 12%
  1. The values were not calculated for all data quality measures in 2006/07
  2. Missing codes excludes 'Not stated' which is considered a valid code for ethnicity
  3. This data quality measure only became available from 2013/14 when the new NCMP IT system was launched

Table A within the accompanying Data Quality tables shows the same data quality measures at submitting Local Authority level for 2020/21.

 

The proportion of missing data items has improved over time although it still remains quite high for NHS number and ethnicity. In 2020/21, 26 LAs had over 25% of records with missing NHS numbers and 12 LAs had more than 25% of records without an ethnicity code.

 

By chance, 10% of height and weight measurements would be expected to be whole numbers. However, there is some evidence of LAs rounding heights to whole numbers as nationally 17% of heights were whole numbers in 2020/21. This was over 30% for three LAs.


 

After the collection has closed NHS Digital carries out further data validation which includes:

  • Querying breach reasons that do not fully explain the reasons for the data quality issues, however due to the unique circumstances of the 2020/21 school year, COVID-19 was assumed to be the general reason for any otherwise unexplained breaches.
  • Comparing each Local Authority’s dataset with their previous year’s dataset and querying unexpected changes.
  • Looking for clusters of unexpected data items to identify data quality issues affecting particular schools.

Data collected

In a ‘normal’ recent year like 2018/19 there would have been around 600,000 Reception and 600,000 Year 6 children measured. The table below contains the number of measurements in 2020/21 and compares these with an estimate of the total number of children in state maintained primary schools who could have been eligible to be measured.

Table A2: Number of measurements collected in 2020/21 NCMP year
    Reception Year 6 Total
Number of records with measurements entered in sample schools            59,123         60,123          119,246
Number of records with measurements entered         156,879       142,597          299,476
Number of children in state maintained primary schools in year group1         617,357       623,352        1,240,709
Estimated percentage of children measured by the sample   10% 10% 10%
Estimated percentage of children measured    25% 23% 24%
  1. Number of children is headcount figures as at January 2021 taken from year group and gender tables published by the Department for Education 

 

The collection achieved much more than the desired 10% of children, with the majority of LAs submitting data from both sample and non-sample schools, as per the table below.

Table A3: Distribution of Local Authority submissions for 2020/21
  Number of Local Authorities
  Reception Year 6
No data submitted 8 8
Data from sample schools only 44 47
Data from non-sample schools only 0 0
Data from both sample and non-sample schools 97 94
Total 149 149
     
  Number of Schools
  Reception Year 6
Total schools in Local Authorities  16,382 16,382
Total schools in which children were measured 4,455 3,711
     
Schools selected for sample 1,897 1,897
% of total schools selected for sample 12% 12%
Sample schools in which children were measured  1,550 1,512
Sample schools in which no measurements taken 347 385
% of sample schools with no measurements taken 18% 20%
     
Extra non-sample schools in which children were measured 2,905 2,199
% of total schools in which children were measured 27% 23%

 

Within the data collected:

  • Six Local Authorities did not submit any data and two submitted only Reception and two only Year 6 data
  • There are some sample schools for which no data was collected. For Reception, 18% of sample schools were missing and, for Year 6, 20% of sample schools were missing.
  • Approximately twice as many schools participated than required for the sample.
  • Measurements were submitted for 27% of schools for Reception children and for 23% of schools for Year 6.
  • Some Local Authorities measured in higher proportions of their schools than others.
  • In some schools, fewer pupils were measured than expected given their pupil numbers.
  • Most data related to measurements taken between March and July 2021. However, some Local Authorities took measurements prior to this and were able to submit them to the collection.

 


Assessing whether a representative sample was collected

This section outlines the data quality work carried out to assess whether the submitted data is representative of the pupil population measured by NCMP in England in terms of sex, deprivation and ethnicity mix.

Comparisons have been made between all data submitted in 2020/21 (the dataset) and all data submitted in previous years. The data for only the sample schools (sample) has also been compared to all data submitted in previous years. The 2019/20 collection was not complete due to COVID-19 national school closures so data for 2018/19 has also been used for comparison as the last full ‘normal’ year.

 

Sex

Within the dataset and the sample, females accounted for 51% of measurements and males 49%. This can be observed in Reception and Year 6. These are the same proportions as in previous years suggesting the 2020/21 dataset and sample are representative for sex.

 

Deprivation

Overweight and obesity prevalence is associated with Index of Multiple Deprivation (IMD), as reported in previous NCMP publications. Therefore, analyses have been undertaken to assess whether the dataset and sample are representative by IMD decile.

The proportion of children measured in each IMD decile was compared between 2020/21 and previous years and the following information is included in the outputs:

  • Number of children measured in 2018/19, 2019/20 and 2020/21 in each IMD decile
  • Percentage change in the numbers measured in each IMD decile between years
  • Proportion of children in each IMD decile in 2018/19, 2019/20 and 2020/21
  • Percentage point change in proportion of children in each IMD decile between years

Table B of the accompanying data quality tables includes all the data submitted in 2020/21 (dataset) compared to 2018/19 and 2019/20. Table C includes data for only the sample schools compared to all 2018/19 and 2019/20 data

 

Reception

The proportion of children measured in each IMD decile in the 2020/21 dataset has changed compared to previous years. Generally, there are larger proportions of children within the more deprived areas (over representation) and lower proportions within the least deprived areas (underrepresentation) compared to a “normal” NCMP year.

The 2020/21 sample distribution is closer to previous years than the 2020/21 dataset. A comparison of the 2020/21 distribution by IMD decile from that of 2018/19 is illustrated in the chart below, which shows the change in percentage points for each IMD Decile.

 

 


 

In the sample, the proportion of children from the most deprived decile is 0.9 percentage points higher and the proportion from the least deprived decile is 0.6 percentage points lower than 2018/19, with little change for other deciles. The differences for the dataset were larger, with the proportions of children from the two most deprived deciles being 2.1 and 0.8 percentage points higher and the proportions from the two least deprived deciles 1.3 and 1.8 percentage points lower respectively.

 

Year 6

For year 6, there were also differences in the proportion of children in each IMD decile in 2020/21 compared to earlier years.


 

For the dataset, children from the most deprived deciles were overrepresented and those from least deprived were underrepresented. For the sample, the differences were smaller but there is still some over representation of children from the most deprived decile and a lower proportion from the least deprived decile.

Changes in proportions within IMD deciles are similar for both males and females.

The changes in the distribution across IMD deciles indicate that the dataset is not closely representative in terms of deprivation. The sample distribution is a closer match to previous years, though the differences mean that the sample is still not fully representative of the population.

 

Ethnicity

Overweight and obesity prevalence varies between different ethnic categories, as reported in previous NCMP publications. Therefore, data quality checks have been undertaken to assess whether the dataset and sample is representative by ethnic category.

The proportion of children measured in each ethnic group was compared between 2020/21 and previous years and the following information is included in the outputs:

  • Number of children measured in 2018/19, 2019/20 and 2020/21 in each ethnic group
  • Percentage change in the numbers measured in each ethnic group between years
  • Proportion of children in each ethnic group in 2018/19, 2019/20 and 2020/21
  • Percentage point change in proportion of children in each ethnic group between years

Table D of the data quality tables includes all the data submitted (dataset) in 2020/21 compared to 2018/19 and 2019/20. Table F includes data for only the sample schools compared to all 2018/19 and 2019/20 data.

Within the dataset, the proportion of children measured in each ethnic group in 2020/21 has changed compared to previous years. A higher proportion of children in the White ethnic group was measured in Reception, along with a lower proportion of “not stated”. In Year 6, a higher proportion of children in the Asian ethnic group and lower proportions in the White and “not stated” groups were measured.

Within the sample, the distribution of ethnic group has also changed compared to previous years, though in a different way. In Reception, lower proportions of children in the White and higher proportions in the Mixed and Asian ethnic groups were measured. In Year 6, lower proportions of children in the White and Asian ethnic groups were measured and a higher proportion “not stated”.

The distribution across ethnic group does not appear to have changed as much as deprivation, and some of these changes are likely due in part to the fluctuations in the use of “not stated”. Tables E and G show the distribution across ethnic groups when children with no ethnicity recorded have been excluded. For Reception children within the dataset, the distribution now shows a much smaller change from previous years, suggesting that improved data quality has been a factor in the change. However, the distribution across ethnic groups still exhibits a change from previous years among Year 6 children. Within the sample data, the distribution for Year 6 children is now a closer fit to previous years but a difference still exists among Reception children. 

Changes in proportions within ethnic groups are similar for both males and females.

These investigations showed that the dataset is not fully representative, by IMD and ethnic group, of the pupil population measured by NCMP. The sample has smaller differences but has fewer children measured.  


 

Proportion of children measured within schools

Within the NCMP data collection system, a figure for the number of children in Reception and in Year 6 is held. The proportion of children within each school for whom measurements were submitted was calculated to investigate whether low measurement rates within schools could be affecting the distribution of the dataset and/or sample.

Table A4: Measurement rates of schools from 2018/19 to 2020/21
  Number of schools
  2018/19   2019/20   2020/21
Proportion of children measured  count %   count %   count %
<=25% 2,685 14   5,451 28   96 1
>25% to <=50% 82 0   1,518 8   211 3
>50% to <=75% 315 2   1,761 9   789 10
>75% to <=100% 16,778 84   10,824 55   6,547 86
Total 19,860     19,554     7,643  

The majority of schools in the 2021/21 dataset measured over 75% of eligible children, with only 4% of schools measuring less than or equal to 50%. The proportion of schools with high measurement rates is similar to 2018/19 levels and there is a smaller proportion of schools within the 2020/21 dataset with low measurement rates compared to previous years.  This suggests that the impact from school classes being sent home due to positive COVID-19 tests had a minimal impact on measurements.

Schools appeared to be similarly distributed across IMD deciles in each proportion banding. Attempts were made to refine the sample by excluding data from the schools with less than 75% participation rate, however the distribution obtained was still not closely representative in terms of deprivation.

 
 

Examining missing data

A number of Local Authorities did not make a submission to the 2020/21 data set. Some Local Authorities submitted only one year group.
 
Table A5: Local Authorities with missing data in 2020/21
Local Authority (LA) LA code Region School Year Missing
Manchester City Council 306 North West Year 6
Barnsley Metropolitan Borough Council 204 Yorkshire and the Humber Both
Lincolnshire County Council 503 East Midlands Both
Hertfordshire County Council 606 East of England Both
Luton Borough Council 611 Year 6
Bedford Borough Council 625 Both
Central Bedfordshire Council 626 Both
Milton Keynes Council 613 South East Both
West Sussex County Council 807 Reception
Brighton And Hove City Council 816 Reception

 

For the above Local Authorities (non-submitting LAs), data from the most recent complete NCMP collection in 2018/19 was examined. Together these non-submitting LAs accounted for 7% of children measured in NCMP in 2018/19. The proportions of children measured across each IMD decile for the non-submitting LAs was compared to the data from the remaining LAs in 2018/19. See Table H of the data quality tables.

Within 2018/19 data the non-submitting LAs had higher proportions of children measured within the least deprived deciles and lower proportions in the most deprived deciles, this being most apparent in Reception.

The BMI category prevalence rates for the non-submitting LAs were compared to those for the remaining LAs. See Table I. For the non-submitting LAs:

  • In Reception, healthy weight children accounted for 1.9 percentage points higher, and obese children 1.3 percentage points lower, than the remaining LAs. However, the BMI category prevalence rates for the remaining LAs were still very similar to the overall England rates using all LAs. This is due to the small proportion of data from non-submitting LAs.
  • In Year 6, prevalence did not differ much from that of the remaining LAs and is similar to all LA prevalence.

The above suggests that the missing data may be contributing slightly to the over representation in most deprived deciles and underrepresentation in least deprived deciles observed within the dataset and sample. But this is unlikely to have an effect on the BMI prevalence rates at national level.

 

School cohort analyses, 2018/19, 2019/20 and 2020/21

The school cohort analyses are designed to assess the comparability of the dataset and sample submitted during 2020/21 with the data from those same schools submitted in 2018/19 and 2019/20. These analyses assume that the children attending the same schools in these years are likely to share similar characteristics. This allows a comparison of change over time where assumed IMD and ethnicity are similar.

Proportions of children from the dataset school cohort and the sample school cohort measured across each IMD decile can be found in Tables J and K. The sample school cohort has a similar distribution to the same schools in earlier years with the largest change being 0.6 percentage points in both Reception and Year 6. For the dataset school cohort, the largest change is 0.6 percentage points for Reception children and 1.3 percentage points for Year 6, among the most deprived. These differences are smaller than when comparing to all data from previous years.    

Proportions of children from the dataset school cohort and the sample school cohort measured across each ethnic group can be found in Tables L and M. Whilst there is some change in distribution across ethnic groups compared to previous years this is relatively small and, in most cases, smaller than the change when comparing to all previous data (in Tables E and G).

As the distributions are similar across years for the school cohorts, this suggest that performing cohort analysis on both the sample and dataset may help to give an indication of change in weight prevalence. The results for England are shown in the below table.

Table A6: Estimates of obesity prevalence from school cohort analysis, 2020/21 
      Percentage points 
England Prevalence of obese in 2020/21 (%)   Change from 2019/20 Change from 2018/19
Reception        
Sample
school cohort
14.4   +4.5 +4.7
Dataset
school cohort
14.7   +4.6 +4.7
         
Year 6        
Sample
school cohort
25.5   +4.2 +5.0
Dataset
school cohort
26.2   +4.6 +5.5


Tables N and O show BMI category prevalence data for the sample school cohort broken down by deprivation and ethnicity.

  • Across all IMD deciles in Reception and Year 6 the obesity prevalence has increased and the proportion of healthy weight children has decreased. This is more apparent in the most deprived areas, with some percentage point changes over 7, and changes of around 3 to 4 in least deprived areas.
  • Across most ethnic groups in Reception and Year 6 the obesity prevalence has increased, and the proportion of healthy weight children has decreased.

Tables P and Q show weight prevalence data for the dataset school cohort broken down by deprivation and ethnicity.

  • Across all deciles in Reception and Year 6 the obesity prevalence has increased, and the proportion of healthy weight children has decreased, although slightly less than in the sample. This is more apparent in the most deprived areas, with some percentage point changes over 6, with changes of around 3 in least deprived areas.
  • Across most ethnic groups in Reception and Year 6 the obesity prevalence has increased, and the proportion of healthy weight children has decreased. 

By using cohort analysis where the same set of schools are compared over time, there is evidence of an increase in obesity prevalence of over 4 percentage points. Increases are seen across all IMD deciles and most ethnic groups. There is variation in the proportions in each BMI category across IMD decile and somewhat larger variation across ethnic groups.

 

Weighting data

It is desirable to use as much of the collected child measurement data that we can, and the dataset has measurements for 299,476 pupils, compared to 119,246 from the sample schools. Weighting the dataset to adjust the distribution to better fit the profile of the pupils by IMD and ethnicity would make use of all the data collected and be an appropriate method for producing national estimates. 

Weighting was carried out on all 2020/21 data submitted to make it more comparable to a more normal NCMP year in terms of demography and geography.

Upper tier LA of school and IMD 2019 quintile based on the child’s postcode and grouped ethnicity were investigated as potential weighting variables because they have an association with obesity prevalence.

  • Data from the last three complete NCMP years of 2016/17 to 2018/19 was used as the target population the weighting methodology was trying to achieve.
  • Data from three years was used to account for any atypical high or low cell counts from the combination of the weighting values which may occur if only one year of data was used.
  • 2019/20 data was not used as it was not a full NCMP collection year due to schools closing on 20 March 2020.
  • Where a child’s postcode was missing, the IMD quintile was derived from the postcode of the child’s school. This affected 0.1% of records.

The following table uses a simple example to demonstrate how weights were calculated using three variables.

Table A7: Example of weight calculation using all three variables
LA IMD Quintile Ethnicity 2016/17 – 2018/19 2020/21 Weight
A 1 X 300 100 3
A 1 Y 250 125 2
A 2 X 100 80 1.25
A 2 Y 400 200 2
A 3 X 150 50 3
A 3 Y 80 50 1.6
A 4 X 90 45 2
A 4 Y 200 150 1.33
A 5 X 90 60 1.5
A 5 Y 125 75 1.67
B 1 X 50 30 1.6
B 1 Y 10 5 2
etc          

To address potential issues arising from small numbers or missing measurements in some table cells, different weighting methodologies were explored. These used fewer variables to decrease the likelihood of a combination of values having no or few measurements.

As an additional check, some sensitivity analysis was carried out for the weighting methodologies which used LA because these had more cells and could have more cells with very few records in them. These checks excluded child records with weights greater than 4 and compared the results to using all records.

 

Weighted estimates of the change in obesity prevalence compared to 2019/20 and 2018/19

The below table shows the obesity prevalence obtained using the different weighting methods and also the data with no weighting applied.

Table A8: Obesity prevalence obtained using weighting methods and non weighting
    Obesity Prevalence (%)  
  Variables used for weighting Year R Year 6
A No weighting using dataset 14.7 26.2
No weighting using sample 14.4 25.5
B LA, IMD quintile,
ethnicity (5 groups)1
14.8 25.2
C As B but weights trimmed 14.4 25.5
D LA, IMD quintile 14.6 25.5
E As D but weights trimmed 14.5 25.5
F IMD decile and ethnicity
(17 groups)
14.3 25.4
G IMD decile 14.3 25.4
  1. The five groups used were White, Black, Asian, Other and Not Stated/Unknown

 

Table R within the accompanying data quality tables provides a comparison to previous years.

All methods used for calculating obesity prevalence point towards an increase for both Reception and Year 6 of approximately 4.5 percentage points compared to the previous year. This suggests the large increase compared to 2019/20 is real and is not materially affected by any inaccuracies in the weighting methodology.

Using three variables: IMD quintile and ethnic group and controlling for the size of the child population in each Local Authority is optimal. The number of 2020/21 measurements in some of the cells in the table was very small leading to those measurements having a large weight. This might result in just a few measurements having a large impact on the overall results and therefore trimming was used – dropping child records where the weight would be 4 or greater. This resulted in dropping just 2% of child records using Method C (see table below).

Table A9: Summary of weighting method C using LA, IMD quintile and ethnicity (5 groups)
  Reception Year 6 Total
All submitted child records   156,879       142,597       299,476
Unweighted counts after trimming records where weight would be >=4       153,409        140,144        293,553
Weighted bases (sum of weights)       129,586         124,230         253,816
Child records dropped due to trimming rule           3,470            2,453 5,923
% records dropped 2% 2% 2%

Summary

A number of exploratory analyses carried out on all the data collected in 2020/21 and only data collected from sample schools showed that the data was not closely representative in terms of deprivation and ethnicity.

School cohort analyses provided estimates of obesity prevalence of over 4 percentage points compared to 2019/20 and 2018/19, with increases over all IMD deciles and ethnic groups.

Various weighting methods were applied to all data collected (to retain the most measurements) and all methods gave a similar obesity prevalence and large increases from 2019/20 and 2018/19. The weighting method chosen (method C) corrected for ethnicity and deprivation and used Local Authority. Trimming was applied to prevent a small number of measurements with large weights having a large impact on the overall results.  


Last edited: 16 November 2021 5:42 pm