Imputation

In the 2017-18 National Health Survey (NHS), voluntary measurements of height, weight and waist circumference were collected from respondents aged 2 years and over, whilst voluntary blood pressure measurements were also collected from adult respondents (aged 18 years and over). These measurements provide information on overweight and obesity (using Body Mass Index (BMI)), risk of developing chronic disease, and high blood pressure amongst the Australian population.

Physical measurement variables have a relatively high rate of non-response, compared to other variables, due to respondent sensitivities and the voluntary nature of these questions. Non-response rates for physical measurements were higher in 2017-18 than in the 2014-15 NHS, for example, the non-response rate for BMI for adults in 2017-18 was 33.8% compared with 26.8% in 2014-15. Despite this, the information is still considered reliable for use.

Non-response rates for the physical measurement data items are shown in the table below.

		Non-response rates
Age group (years)	Total persons in sample no.	Body Mass Index(a) %	Waist circumference %	Blood pressure %
Children
2–4	841	44.4	48.2	. .
5–7	804	44.8	47.0	. .
8–11	1 004	41.5	43.0	. .
12–15	1 025	43.5	45.2	. .
16–17	614	46.9	47.4	. .
Total 2–17 years	4 288	43.9	45.9	. .
Adults
18–24	1 174	31.8	33.5	31.1
25–34	2 576	32.7	34.2	30.0
35–44	2 844	32.9	34.7	31.2
45–54	2 800	36.3	37.3	34.3
55–64	2 789	34.4	36.2	31.7
65–74	2 456	32.6	34.6	30.1
75 years and over	1 731	34.7	36.9	32.9
Total 18 years and over	16 370	33.8	35.4	31.6
. . not applicable

In addition to the voluntary measured items, respondents in the 2017-18 NHS were also asked to self report their height and weight measurements. Self reported measurements were not collected in the 2014-15 NHS. The majority of measured BMI non-respondents (76% of adults and 61% of children) provided self-reported height and weight measurements. This provides valuable information about the height and weight that can be used in assisting in the imputation for those with missing values. A future article will contain detailed analysis of the comparison between self report and measured height and weight.

In both the 2014-15 NHS and the 2017-18 NHS, missing values were imputed using the 'hot decking' imputation method. In this method, a record with a missing response (the 'recipient') receives the response of another similar record (the 'donor'). A number of characteristics with which to match recipients to donors were used. For adults they were:

• age group
• sex
• part of state (capital city and balance of state)
• self perceived body mass (underweight, acceptable, or overweight)
• level of exercise (sedentary, low, moderate or high)
• whether or not has high cholesterol (as a long-term health condition)
• self reported BMI category (calculated from self reported height and weight)

For example, a female recipient aged 35-39 years who lives in a capital city, has a self reported BMI category of overweight (calculated using self reported height and weight), has a self perceived body mass of healthy, has high cholesterol and lives a sedentary lifestyle will match to a donor record who has the same profile (female, 35-39, self-reports as overweight, etc).

For BMI, around 86% of imputed records with self-reported BMI used all seven variables to match to a donor record. The remaining 14% could not be matched using all seven variables and were therefore matched using fewer variables. For example, around 7% of imputed records with self-reported BMI were matched to donors by age group, sex, self-reported BMI, self perceived body mass, level of exercise and cholesterol.

For children 2-14 years, age group, sex, self reported BMI and part of state were used as imputation variables, while for 15-17 year olds, level of exercise and self perceived body mass (only if a person answered for themselves) were also used as imputation variables, due to the other variables not being collected for children aged 2-17 years.

For analysis purposes, the 2017-18 NHS data was processed using both the 2017-18 imputation method and the 2014-15 imputation method. The key difference between these two imputation methods was the addition of self-reported BMI category with which to match imputation recipients to donors. The table below shows that the results are comparable with a sufficiently small impact. This indicates that time series changes between 2014-15 NHS and 2017-18 NHS are unlikely to be due to a change in the imputation method.

---

	Measured only				Measured and Imputed
					Using 2017-18 imputation method				Using 2014-15 imputation method
Body Mass Index category	2-17 years (no.)	2-17 years (%)	18 years and over (no.)	18 years and over (%)	2-17 years (no.)	2-17 years (%)	18 years and over (no.)	18 years and over (%)	2-17 years (no.)	2-17 years (%)	18 years and over (no.)	18 years and over (%)
Underweight	174	7.2%	123	1.1%	325	7.6%	189	1.2%	317	7.4%	194	1.2%
Normal	1604	66.7%	3370	31.1%	2846	66.4%	4968	30.3%	2827	66.1%	5052	30.9%
Overweight	424	17.6%	3903	36.0%	758	17.7%	5837	35.7%	765	17.9%	5853	35.8%
Obese	202	8.4%	3446	31.8%	359	8.4%	5376	32.8%	368	8.6%	5271	32.2%
Total	2404	100.0%	10842	100.0%	4288	100.0%	16370	100.0%	4277	100.0%	16370	100.0%
Total overweight/obese	626	26.0%	7349	67.8%	1117	26.0%	11213	68.5%	1133	26.5%	11124	68.0%
Whether measured
Measured	2404	56.1%	10842	66.2%	..	..	..	..	..	..	..	..
Not measured	1884	43.9%	5528	33.8%	..	..	..	..	..	..	..	..
Total	4288	100.0%	16370	100.0%	..	..	..	..	..	..	..	..

---

.. not applicable
(a) Using National Health Survey 2017-18 sample counts

Estimates of physical measurement data (BMI, waist circumference and blood pressure) from NHS 2017-18 are of suitable quality and are directly comparable to 2014-15. For comparisons to earlier years, the ABS recommends using proportion comparisons only as imputation was not used on the physical measurement data prior to 2014-15 NHS.