The feasibility of using photographs to estimate historical heights of children

In studying the growth and development of children, records of body heights are often unavailable for certain periods of time. To evaluate the feasibility of using photographs to estimate body heights in the past, we obtained 23 photographs taken within two months of a documented measurement of height from a cohort of an on-going study. From each photograph, an object was selected as the reference, and the historical height of the child was estimated by proportional projection according to the height of the reference. Two independent sets of estimates were made, and both correlated well with measured heights (correlation coefficient = 0.99, p < 0.0001). The mean differences between the measured and estimated heights were 1.4 cm and 1.5 cm respectively, and both were not statistically significant (Wilcoxon signed rank test, p > 0.05). The average differences between the two sets of estimates was only 0.1 cm (p > 0.05 for Wilcoxon signed rank test), which indicates a high reproducibility. Photographs taken in the 'landscape' style tended to provide better estimates than those taken in the 'portrait' style. Without any high technology equipment, this simple approach can be readily applied to obtain satisfactory estimates for epidemiological studies.