The data analyses of individual food categories show that the FP 24-hR with digital photographs and a photo atlas was able to estimate the weights of food portion sizes adequately and gave results comparable to the actual consumed amounts recorded by the WFR. The modifications with digital photographs and a photo atlas added to the ability of the 24-h recall to minimize errors associated with the estimation of portion sizes, as well as the reduction of respondent burden. Therefore FP 24-hR represents a good alternative to the gold standard method (weighed food record) for estimating individual nutrient intakes, as it is demonstrated by the presented results.
Furthermore, recent studies show that the introduction of digital photographs taken by the subjects as a diet assessment method helps to estimate food intake and plate waste and this can reduce over and underestimates. This has been shown with children in cafeteria settings
[17, 20], with children at home
[24, 29], adults with intellectual disabilities living in the community
, obese patients in hospital
, as well as in college and university environments
[16, 18]. However, only a few studies have used digital photographs to estimate intake in free-living conditions
 and with general populations
. Besides, to our knowledge this may be the first study using photographs to assess nutrient intake in rural populations in low-income countries.
Studies comparing food estimates from digital photographs and weighed records have found that the use of digital photographs results in small differences in the amount of food, in the range of −9.1 g to 18.3 g
, and an underestimation of −6.6% for energy intake
, although these differences are very small it has been reported that these underestimations were significantly different from the values obtained by the weighed record method. In the present study the comparison of the food amount evaluated by FP 24-hR and WFR has shown mean differences from −14.4 g to 4.5 g between the different food categories, where the differences were not statistically significant, excepting for some food categories (rice, potatoes, eggs, vegetables).
In the analysis of nutrient intake energy was underestimated by −3.92% and the underestimations are in the range of −0.90% for vitamin C to −5.98% for total fat. These underestimations were significant excepting for certain nutrients (calcium, vitamin A, C, and β-carotenoids). The significant differences found even when FP 24-hR and WFR have identical mean, may be due to the variance within each group owing to the high variability among individual food consumption (i.e. portion size of rice consumption varies from 44 g to 400 g between subjects), which is subsequently reflected in significant differences in nutrient intake. The small underestimation of most of the food groups and all the nutrients may be due to some hidden foods in the photographs making it difficult to estimate portion sizes, and failure in memory of the respondents to identify all the hidden foods in the photographs.
Notwithstanding the significant differences, the FP 24-hR showed high correlation coefficients in estimating portion sizes, in the range of 0.75 (for egg) to 0.98 (for potatoes and cassava), comparable to those reported in previous studies (>0.74)
[12, 15, 19], where a photo atlas was used as a tool for quantifying portion size.
Moreover correlations between photographic food record and weighed dietary record, for energy intake, reported by previous authors were as high as: from 0.93 to 0.95
, from 0.44 to 0.48
 and 0.60
. The correlation coefficient for energy intake reported in this study is 0.99. A few studies have reported correlation coefficients for macro and micronutrients; for protein 0.83, 0.48, and 0.61; for carbohydrates 0.55, 0.52 and 0.68; for fat 0.82, 0.46 and 0.50 were reported respectively by
[24, 31, 32], the FP 24-hR found correlation of 0.99, 0.99 and 0.96 for protein, carbohydrates and fat respectively. Correlation for vitamins are reported in the range of 0.06 to 0.80
 and 0.30 to 0.86
, correlation for minerals from 0.34 to 0.57
, and from 0.21 to 0.74
, the present study reports correlations for vitamins and minerals in the range from 0.97 to 0.99.
The results found with the Bland Altman analysis showed that the majority of the measurements 95.2% for food categories and nutrient intake, were scattered along the mean difference line and close to the equality line (difference = 0). The plots thus show fairly good agreement between estimated and actual food consumed and indicate that the differences (including the outliers) were random and did not exhibit any systematic bias, being consistent over different levels of mean food amount. Results were similar to previous studies, which have reported that the bias between the use of digital photographs and weighed food records was consistent over different levels of energy intake, indicating that the two methods were comparable, and bias was very low
In the analysis of food categories, the geometric mean ratios are close to 1 (from 0.93 for rice to 1.09 for leafy vegetables), and limits of agreement are narrow for most of the food categories. The ratios of proportional agreement indicate that for about 95% of the cases the estimated amounts will be between the values of the ratio respect to the weighed amount, for example for bread the geometric mean is 0.98 with limits of agreement from 0.79 to 1.22; thus FP 24-hR when is compared with WFR gives values by between 0.79 to 1.22 times the weighed amount of bread. The limits of agreement are relatively broad for vegetables (0.65 to 1.43) and leafy vegetables (0.70 to 1.69); this may be because the dispersion of the values in these two food categories increases as the weight increases.
The analysis of nutrient intake showed that the mean differences between FP 24-hR and WFR were low and the limits of agreement acceptable, for example the average energy intake estimated by the FP 24-hR was 5854 KJ, the mean difference when it was compared to WFR was −3.92% and the limits of agreement were from an under-estimate of −11.5% to an over-estimate of 3.5%, most of the nutrients showed similar narrow limits of agreement. The widest limits of agreement resulted for the intake of β-carotenoids which presented a small mean difference −1.27%, but the wide limits of agreement from an under-estimate of −25.9% to an over-estimate of 25.4%, similar for vitamin A. In spite of this, the limits are in an acceptable range to guarantee that the FP 24-hR can be used in place of the WFR for all the nutrients presented.
The small differences, high correlations and good agreement of the FP 24-hR with the WFR, may be because the food patterns in the study area are simple and less diversified than in urban populations where the food availability is wider and includes more processed food ready-to-eat, which might be more complicated to evaluate, in addition the use of digital photos and a 24-h recall questionnaire carried out by an interviewer make possible for the respondents to describe the hidden foods in the photographs or describe poor quality photographs, thus obtaining the most complete data possible. At the same time the volunteers were motivated with the FP 24-hR which involves the use of a simple but interesting and new device like a digital camera, because rural populations in developing countries are not so familiar with digital cameras. Another important factor that could enhance compliance with the method is that it is simple and fast, demands less than 2 minutes to take two pictures of each meal, which implies a maximum investment of 10 minutes per day to take digital photos of food consumption.
Very limited data are available about food and nutrient intake in rural areas in Bolivia. In this study we found an apparently low daily energy intake: mean 5.9 MJ, from 3.6 to 9.8 MJ in women 35 ± 8.6 years old. However, similar low energy intake for women in rural areas in South America has been reported previously, using different methods for measuring food consumption. In a study conducted in Calchaqui - Argentina, a 24-h recall and a semi-quantitative food-frequency questionnaire were applied and energy intake was estimated to be 6.6 MJ
 in women 43 ± 15.2 years. Furthermore, in Ura Ayllu, Peru, low energy intake such as 5.3 to 7.5 MJ was reported by the weighed food record method in women 31 ± 6.3 years
. Also in a study using multiple pass 24-h recall in a Mexican population the energy intake was 5.9 MJ in women 32 ± 0.3 years
The common food pattern in the currently studied population is based mainly on carbohydrates like: tubers (potatoes, cassava) and cereals (rice, bread, pasta); accompanied by small portions of protein from eggs or meat (mainly beef and chicken); oil or tallow as sources of fat, and a few vegetables and fruits. The composition of macronutrients as a percentage of total energy reflects the food pattern, in total carbohydrates 72 E%, protein 13 E%, and total fat 15 E%. The macronutrients consumption of the study group is within the dietary recommendation from the World Health Organization (Total carbohydrates 55–75 E%, protein 10–15 E%, and total fat 15–30%)
. However, the carbohydrates intake is nearly in the upper limit and the fat intake is nearly in the lower limit.
Despite the lower energy intake, 56% of the women had normal BMI (22.80 ± 1.64), 26% and 11% respectively were overweight or obese, and only 7% were underweight. These results are comparable to those found in rural areas with low energy intake such as in an study in Calchaqui- Argentina, which reported 39% of the women presenting normal weight
A possible limitation in this study might be the undiversified food patterns of the population under study; the photo atlas was designed and developed in accordance of the specific food patterns in the area, as the method is aimed to be used in further studies of dietary assessment in the same area, another limitation is the relatively small number of the volunteers.
On the other hand the strengths of the study are: it was performed under the normal living conditions without disruption of the eating behavior, therefore the food consumed was representative of their habitual diet, and the inclusion of a digital camera which is a simple but interesting device for rural populations in developing countries may enhance the compliance with the method, and it may be used equally by both genders.