Literature review

Health examination surveys such as the HSE and the National Health and Nutrition Examination Survey (NHANES) in the United States (USA) enable the study of reporting bias when they collect both self-reported and measured data on height and weight from the same participants. Comparisons of such data have shown that adults tend to overestimate their height and underestimate their weight, which leads on average to an underestimation of BMI and obesity (Connor Gorber et al., 2007; Maukonen, Männistö and Tolonen, 2018).

Available evidence consistently shows that men overestimate height to a greater extent than women, whereas women underestimate weight more than men (Connor Gorber et al., 2008). For example, analysis of the HSE 2011 data showed that men overestimated height more (by 1.7cm and 1.0cm among men and women, respectively), and women underestimated weight more (by 1.4kg and 2.4kg among men and women, respectively), each working out to an average 1kg/m² difference in mean BMI (0.9kg/m² men, 1.2kg/m² women) (Sutton, 2012).

Misreporting also increases with age (Kuczmarski, Kuczmarski and Najjar, 2001) and with higher measured BMI (Connor Gorber et al., 2007). The association between age and misreporting may include a tendency for overestimation of height among older adults who report their height as measured earlier in life, prior to becoming shorter due to changes in bone and muscle, whereas underestimation of weight is typically greater among younger age groups, especially among women, due to a greater social desirability for thinness (Stommel and Schoenborn, 2009). The higher accuracy of self-reported body weight in smokers than among non-smokers found in a Swedish study may reflect the higher levels of body weight awareness among smokers (Nyholm et al., 2007).

Variations in the magnitude of misreporting may be partly explained by differences in perceived norms and desirable body weight across different educational levels, cultures and societies. The recent review by Freigang et al. (2020) found that participants in lower socioeconomic groups were less likely to classify their BMI status accurately compared with those in higher socioeconomic groups.

With regards to ethnicity, in an analysis of NHANES 2007-2008 data, adults in the non-Hispanic White group were more likely than those in the non-Hispanic Black and the Hispanic groups to under-report BMI (a dichotomous variable: BMI derived from self-reported height and weight being lower than BMI based on measured height and weight), a finding which the authors suggested could be due to a greater ideal for thinness among non-Hispanic White adults (Wen and Kowaleski-Jones, 2021).

Changes over time in the gap between self-reported and measured height and weight would affect the accuracy of attempts to monitor obesity levels using BMI derived from self-reported data only, as well as decrease the accuracy of prediction equations (due to becoming less stable over time).

The few studies which have examined whether the differences between self-reported and measured height and weight (and BMI derived from these) have changed over time have produced mixed findings.

Four studies found no change over time in the magnitude of the differences between self-reported and measured mean height, weight and BMI, (Connor Gorber and Tremblay, 2010; Bibiloni et al., 2017; Shields et al, 2011) including a high-quality study using NHANES data from 1999 to 2008 (Hattori and Sturm, 2013). The authors of that study concluded that the observed increases in mean BMI derived from self-reported height and weight were likely to reflect actual increases in mean BMI rather than change over time in the accuracy of self-reported height and weight (Hattori and Sturm, 2013).  

However, concentrating solely on the differences between self-reported and measured mean height, weight and BMI may mask change over time in the classification accuracy of BMI derived from self-reported height and weight.

In the USA from 1999-2000 to 2015-2016, the over-reporting of mean height increased over time for both sexes, while under-reporting of mean weight increased among obese men but not among other men and not among women. These changes over time did not result in any changes in the difference between self-reported and measured obesity prevalence (Flegal et al, 2019). In a different study in the USA, the difference between self-reported and measured obesity prevalence stayed relatively constant from 1976–1980 to 2003–2004 (Connor Gorber and Tremblay, 2010). No statistical tests for change over time in misreporting were carried out. Other studies found the changes over time in the reporting accuracy of obesity to be not statistically significant (Bibiloni et al., 2017; Hattori and Sturm, 2013).

Two studies found significant increases in misreporting over time. One study conducted in Ireland, based on data from the Survey of Lifestyle Attitudes and Nutrition (SLAN), found an increase over time in the proportion of participants under-reporting their height and weight, leading to an incorrect BMI classification (Shiely et al, 2010). This study also showed a decline in the sensitivity estimates for the obese category from 80% in 1998 to 64% and 53% in 2002 and 2007, respectively (Shiely et al, 2010). ‘Sensitivity’ refers to the proportion of adults classified as obese according to BMI derived from measured height and weight who were correctly classified as obese based on BMI derived from self-reported height and weight. No statistical tests for the change over time in misreporting were carried out. However, the comparisons were performed on a small subsample of data from two of the three periods (1988, 2002 and 2007), bringing into question the comparability of the data.

In Canada, based on data from the Canadian Heart Health Survey (CHHS) and the Canadian Community Health Survey (CCHS), the gap between self-reported and measured obesity prevalence estimates was found to increase between 1986-1992 and 2005, doubling from 4% to 8% (Connor Gorber and Tremblay, 2010). However, measured data was collected only from a subsample of participants in 2005: furthermore, the elapsed time between self-reports and measurements differed between surveys (within two weeks versus on the same day).

Based on NHANES data over a twenty-year period (1988-1994 to 2005-2008), BMI classifications based on self-reported height and weight improved in accuracy among overweight and obese adults, whereas the accuracy remained the same for those who were underweight or normal weight (Stommel M and Osier N, 2013). Overall, this study observed a slight increase in the proportion of adults who were correctly classified as obese, from 60% in 1988-1994 to 69% in 2005-2008. No statistical test for change over time in misreporting was carried out.

A similar finding of decreasing misreporting over time was found in Australia between 1995 and 2008. However, findings in 1995 were based on a random subsample who were measured, and the elapsed time between self-reports and direct measurement differed between surveys (three weeks versus on the same day) (Hayes, Clarke and Lung, 2011). Such improved accuracy over time in reporting of height and weight may be explained to some extent by an upward shift in mean BMI, resulting in an increased acceptance of obesity status (Stommel M and Osier N, 2013) or due to a greater knowledge of one’s own height and weight in response to growing awareness of obesity-related health problems (Hayes, Clarke and Lung, 2011).  

Overall, there were mixed findings regarding change over time in the differences between self-reported and measured height and weight (and BMI derived from these), which is likely due to differences in the geographical or temporal periods assessed. These may reflect changes in attitudes towards obesity and growing awareness of obesity-related health problems which might influence frequency of body weight monitoring.

Furthermore, there were methodological challenges in some of the studies due to not using like-for-like methodology for the data they were comparing, for example the timing of the questions and measurements, or sampling differing between survey years. The higher quality studies conducted in the USA found no differences over time in reporting accuracy (Hattori and Sturm, 2013; Connor Gorber and Tremblay, 2010), or found improved reporting accuracy (Stommel M and Osier N, 2013). Those studies which found improved reporting accuracy quantified the changes over a longer time-period, over ten or twenty years. We found no studies that examined change over time in the differences between self-reported and measured height and weight using data from England.