Introduction

This concept describes how the Manitoba Centre for Health Policy (MCHP) measures obesity using the body mass index (BMI) . It discusses data sources used and how BMI has been defined for adults and children in various MCHP reports.

Body Mass Index: A Measure of Obesity

Body Mass Index (BMI) is a statistical measure used to classify and compare individuals according to their height and weight. It is a measure of health risk that is correlated with body fat based on height and weight and can be applied to both adults and children. BMI is calculated as weight (in kilograms) divided by height (in metres) squared and, typically ranges from 15 to 45. BMI is a derived variable calculated from either self-reported or measured height and weight.

BMI = Weight (kg) / (Height (m) x Height (m))

MCHP researchers use the calculated body mass index value and a classification guideline to measure and identify obesity.

Data Sources

Unfortunately, there are no existing variables or measures available in the data housed in the Manitoba Population Research Data Repository (Repository) at MCHP that can specifically identify obesity in the population. There are ICD-9-CM and ICD-10-CA diagnosis codes for obesity but these diagnoses are rarely used. As the diagnosis is specified at the fourth-level digit in ICD-9-CM, individuals must be hospitalized to receive this diagnosis, thus it is a very poor measure. Using diagnosis codes to measure obesity would severely underestimate its prevalence.

For the BMI calculations, survey data was used to collect height and weight information, as administrative databases do not contain this type of information. The following surveys were used by multiple MCHP studies for the calculation of BMI:


• Canadian Community Health Survey (CCHS)
  • Lix et al. (2005); Finlayson et al. (2007); Brownell et al. (2008); Fransoo et al. (2009); Martens et al. (2010); Fransoo et al. (2011); Chartier et al. (2012)
• National Population Health Survey (NPHS)
  • Brownell et al. (2001); Fransoo et al. (2011)
• Manitoba Heart Health Survey (MHHS)
  • Fransoo et al. (2011)

Note 1: Height and weight were self-reported in all versions of these surveys except for in the MHHS and CCHS 2.2, where it was measured for some survey respondents.

Note 2: Martens et al. (2015) used all three surveys to calculate BMI measures.

Assessment of Weight in Adults

The following MCHP reports have used BMI values and different classification guidelines to assess weight in adults (over 18 years of age).

1. Lix et al. (2005)

Lix et al. (2005) used uncorrected self-reported data from the CCHS 1.1 (2000-2001). BMI was calculated for adults aged 20-64 years, excluding pregnant women. Weight (CCHS variable name = HWTADSW) was classified using the following BMI cut off points:



Classification	BMI (kg/m 2 )
Underweight	<20.0
Acceptable Weight	20.0-24.9
Some Excess Weight	25.0-27.0
Overweight	>27.0

2. Finalyson et al. (2007)

The same BMI classification system was used in Finalyson et al. (2007) as in Lix et al. (2005) for measuring BMI from the CCHS. Finalyson et al. (2007) also excluded individuals who were less than 0.914 metres, or greater than 2.108 metres tall.

3. Fransoo et al. (2009)

Uncorrected self-reported heights and weights from CCHS data were used to calculate BMIs for respondents that were 18 years of age or older by Fransoo et al. (2009) . Data from CCHS 1.1 (2000-2001), CCHS 2.1 (2003), and CCHS 3.1 (2005) were included. The advantages of combining many surveys are that estimates of obesity prevalence can be calculated for smaller geographic regions such as RHA Districts, it provides information on a large number of people, and trends in prevalence can be examined over time.

The CCHS categorized BMI into the following categories based on the World Health Organization (WHO) classifications:



Classification	BMI (kg/m 2 )
Underweight	<18.5
Normal	18.5-24.99
Overweight	25.00-29.99
Obese	>=30
Obese - Class I	30.00-34.99
Obese - Class II	35.00-39.99
Obese - Class III	>=40

Note: The CCHS also classifies survey respondents as BMI "not applicable", which would apply to survey respondents who were pregnant or respondents under 18 years of age, or "not stated", which would apply to survey respondents that did not answer one of the questions required to calculate the derived BMI variable. Respondents with missing values for height or weight, respondents for which the calculation is not applicable (i.e., age less than 18, pregnant women), and respondents who refused to answer either question were excluded from analyses.

Fransoo et al. (2009) grouped the BMI values into three categories:



Classification	BMI (kg/m 2 )
Underweight/Normal	<25.0
Overweight	25.0-29.0
Obese	>30

4. Martens et al. (2010)

Martens et al. (2010) also used uncorrected self-reported heights and weights from CCHS data to calculate BMIs for respondents that were 18 years of age or older. The same methodology was used as in Fransoo et al. (2009), however, CCHS 2.2 was used in addition to CCHS 1.1, CCHS 2.1, and CCHS 3.1 to calculate BMI from self-reported and measured heights and weights. BMI values were grouped into the following categories, as per the WHO classification above:



Classification	BMI (kg/m 2 )
Underweight	<18.5
Normal	18.5-24.9
Overweight	25.0-29.9
Obese	>30

5. Fransoo et al. (2011)

Fransoo et al. (2011) classified BMI values in adults (18 years of age or older) using the World Health Organization (WHO) groupings. Various groupings and exclusions were used, depending on the analysis. Consult the full report for details.

BMI Corrections

As self-reported data for height and weight are often inaccurate, Fransoo et al. (2011) derived formulae based on data and methodology from Statistics Canada to correct for this:

Males: Corrected BMI = Self-reported BMI * 1.0531 - 0.4082
Females: Corrected BMI = Self-reported BMI * 1.0505 + 0.0849

These formulae were derived from simple linear regression models that predicted measured BMI values based on self-reported BMI, separately for males and females. Fransoo et al. (2011) utilized national unlinked survey data from the CCHS 3.1, subsample 2, which included both self-reported and measured height and weight. Fransoo et al. (2011) found that the resulting correction formulae increased BMI by 4% and 5% for males and females, respectively.

Note: The corrected data reported in Fransoo et al. (2011) cannot be directly compared to non-corrected BMI values in other research.

Fransoo et al. (2010) calculated BMI using corrected, self-reported values of height and weight from the following surveys:


• CCHS 1.1 (2000-2001), CCHS 1.2 (2002), CCHS 2.1 (2003), CCHS 2.2 (2004), CCHS 3.1 (2005), CCHS 2007, and CCHS 2008
  
• NPHS (1996-1997)
  • Only the cross-sectional survey was included.
• HHS (1989-1990)

6. Chartier et al. (2012)

Chartier et al. (2012) used the same classification system as Fransoo et al. (2009) :



Classification	BMI (kg/m 2 )
Underweight/Normal	<25.0
Overweight	25.0-29.0
Obese	>30


The CCHS 1.1 (2000-2001), CCHS 1.2 (2002), CCHS 2.1 (2003), CCHS 2.2 (2004), CCHS 3.1 (2005), CCHS 2007, and CCHS 2008 were the data sources used by Chartier et al. (2012) . The results were age- and sex- adjusted.

7. Martens et al. (2015)

In Martens et al. (2015) the weighted crude prevalence of normal/underweight, overweight and obese persons were calculated for survey respondents aged 12 and older as the percentage of respondents with BMI values less than 25, 25 to less than 30, and 30 and over, respectively. BMI values were based on either measured or self-reported values of height and weight. In the case of self-reported values of height and/or weight, BMI was corrected using the formulae below, specific for men and women. Due to a different sampling frame for the Manitoba Heart Health Survey (MHHS), such as the inclusion of First Nations living on reserve, respondents of the MHHS had a different set of correction formulae. These corrections are not applicable to children, and so were only applied to adults age 18 and older. Women pregnant at the time of survey and respondents who answered "don't know" and those with missing or invalid data were excluded from the prevalence calculation.

Prevalence rates for BMI found in this report are available in the following tables:


• Table 4.3: Basic Descriptive Statistics of the Estimated-Population-Based Sample* by Ever Smoked and Never Smoked Categories
  
• Table 4.7: Basic Descriptive Statistics of the Estimated-Population-Based Sample* by Smoking Status Categories

BMI Corrections

NPHS and CCHS:

Males: Corrected BMI = Self-reported BMI x 1.0531 - 0.4082
Females: Corrected BMI = Self-reported BMI x 1.0506 + 0.0850

MHHS:

Males: Corrected BMI = Self-reported BMI x 0.9764 + 0.8411
Females: Corrected BMI = Self-reported BMI x 1.0005 + 0.2095

Assessment of Weight in Youth

The following MCHP reports have used BMI values and different classification methods to assess weight in youth (under 19 years of age).

Brownell et al. (2001)

Brownell et al. (2001) used NPHS data from 1994/95 and 1996/97 surveys to assess weight in children. In the report, Assessing the Health of Children in Manitoba: A Population-Based Study (Brownell et al., 2001) , BMI was calculated for children (0-19 years of age) using the age- and sex-specific percentiles based on the National Health and Nutrition Examination Survey (NHANES) to determine overweight (85th percentile) and obesity (95th percentile) in children. Please see Table G.13: 85th Percentile of BMI from NHANES I, Ages 6 to 19 Years for cut off points for classifying a child as being overweight.

Note: These percentiles have been criticized as being arbitrary and their use internationally has been questioned given they are based solely on American children ( Cole et al., 2000 ). The International Obesity Task Force recommended converting adult cut-offs using age- and sex-specific centiles from large nationally representative surveys from six countries to form child-specific BMI cut-offs for overweight and obese (Cole et al., 2000) .

Brownell et al. (2008)

In the Child Health Atlas Update (Brownell et al. 2008) , self-reported and measured data from the CCHS (CCHS 1.1 (2000-2001), CCHS 1.2 (2002), CCHS 2.1 (2003), CCHS 2.2 (2004), CCHS 3.1 (2005)) were used to assess BMI in children, aged 12-19 years. Cut-offs developed by the International Obesity Task Force (based on the widely accepted classification of adult BMI values of 25 kg/m2 and 30 kg/m2 as cut off points for overweight and obesity, respectively) were used.

See Table 4 from Cole et al., 2000 for the overweight and obese cut-off measures for children age 2 - 18 years of age, for both sexes. A link to the PubMed abstract is available from the Cole reference below, and the abstract provides a link to the free, full-text article on PubMed Central.

Fransoo et al. (2011)

Fransoo et al. (2011) assessed obesity in youth (aged 12-18 years) using the same BMI cut off points developed by the International Obesity Task Force as in Brownell et al. (2008) .

Self-reported and measured height and weight were used from the:


• CCHS 1.1 (2000-2001), CCHS 1.2 (2002), CCHS 2.1 (2003), CCHS 2.2 (2004), CCHS 3.1 (2005), CCHS 2007, and CCHS 2008
  
• NPHS (1996-1997)
  
  • Only the cross-sectional survey was included.

More information on overweight and obesity in Manitoba children can be found in A Report on the Weight Status of Manitoba Children by Manitoba Health and Healthy Living (2007) . The same definitions of overweight and obese children from the International Obesity Task Force (Cole et al., 2000) are used by Manitoba Health as in Brownell et al. (2008) and Fransoo et al. (2011) .

Cautions/Limitations

• BMI is not the most sensitive indicator of obesity (compared to percent body fat), however it is a realistic measurement for use in epidemiology studies (Fransoo et al., 2011).
• BMI does not directly assess body composition (i.e. fat and lean muscle mass).
• The WHO BMI classification may not be appropriate for pregnant women, older adults or individuals of certain ethnic backgrounds (i.e. Aboriginal or Asian) (Fransoo et al., 2011).
• As the self-reported BMIs were corrected by Fransoo et al. (2011), this data cannot be directly compared to other obesity research that used "uncorrected" self-reported values.