Formative Evaluation of Pictorial Methods for Portion Size Estimation in Malawi

Courtney Schnefke1, Chrissie Thakwalakwa Kantukule2, Mary Muth1, John Phuka2, Jennifer Coates3, Beatrice Rogers3, Brooke Colaiezzi3, Winnie Bell3, Valerie Flax1 1RTI International, Research Triangle Park, NC, 2Centre for Social Research, University of Malawi, Zomba, Malawi, 3Tufts University Friedman School of Nutrition Science and Policy, Boston, MA

■ Women 18 to 45 years with either 4 or fewer years of education or 5 or more years were selected to participate.

■ Women of reproductive age were chosen as the target population because they are typically the respondents in both large and small studies assessing dietary intake.

■ The sample was purposively selected so that half of the women were from an urban area in Blantyre and half were from a rural area of Chikwawa district.

■ We prepared food and water images in collaboration with a Malawian photographer and a U.S.-based photographer and graphic designer.

■ We used a stepwise approach with six separate groups of participants to assess the visual variables of interest.

■ Ninety-six women served themselves portions of five foods (groundnuts, bananas, nsima, bean relish, vegetable relish) and water, which were individually weighed before and after consumption.

■ We administered a meal recall using a digital or printed portion size estimation aid and a structured interview 30 minutes after the meal.

■ Across participants and foods, we calculated the percentage of portion sizes selected within 20% of the actual gram weight of food or water consumed.

■ Participant preference and ease-of-use perceptions favored photos, 3 and 5 portion sizes, a 45-degree photo angle, and simultaneous presentation (Figure 1).

Figure 1. Participant Perceptions of Ease of Use for Visual Variables

Note: n represents the number of responses across all foods for each variable from participants who consumed that food and did not have data missing.

■ Participants expressed it was difficult to relate the shapes to the actual food they had eaten, to know what the in-between portion sizes were, and to remember the different portion size options in addition to what they ate when seeing portion size options sequentially.

● Shapes versus photos: “Now the shapes were difficult because I was not able to see the actual food, I was just seeing the size.”

– Participant with higher education in urban area ● Number of portion sizes:

“It was difficult (to select a portion size) because the other plates were not visible.” – Participant with higher education in rural area

● Simultaneous versus sequential“Mmm! It is difficult to remember because you have to see the pictures in different pages and keep it in mind then choose the one representing your food.”

– Participant with higher education in urban area

■ Some participants also thought remembering how much they had eaten was difficult because they do not typically think about how much they are eating.

“We just eat these things. We don’t know the measurements.” – Participant with lower education in rural area

■ When making portion size selections, some participants had to be reminded to select the portion size that represented only what they ate, not the entire amount of food they served themselves. Some participants talked themselves through accounting for their leftovers.

“I drunk like that (portion size) because I left some.” – Participant with higher education in rural area

■ About the same number of participants reported it was and was not difficult for them to only think about what they ate when selecting the correct portion size using the portion size estimation aid.

“It was a bit difficult to remember because I had to first of all think about the amount of leftover food before choosing the right picture representing the amount of food which I ate.”

– Participant with higher education in urban area

“It was not difficult, I know what I ate.”

– Participant with lower education in rural area

■ Overall, approximately half to three-quarters of participants found the portion size options for each food to be an appropriate amount (Figure 2).

Figure 2. Participant Perceptions of Portion Size Appropriateness

Note: n represents the number of participants who consumed that food and did not have data missing.

■ Overall, shapes with five portion sizes were slightly more accurate than photos with three portion sizes (Table 1). A 45° versus 90° photo angle and simultaneous versus sequential presentation of images were more accurate (Table 1).

Table 1. Percentage of Participant Portion Size Selections within 20% of Actual Gram Weight Consumed

Food

Pictorial Depiction and Number of Portion Sizes Photo Angle Presentation Sequence

Shapes Photos45˚ 90˚ Simultaneous Sequential3 3+ 5 3 3+ 5

Groundnuts 16.7 23.3 33.3 20.0 16.7 20.0 40.6 40.6 50.0 40.6Bananas 54.8 45.2 51.6 46.7 40.6 37.5 53.1 56.3Nsima 31.3 28.1 25.0 25.0 21.9 25.0 40.6 37.5 50.0 50.0Bean relish 25.0 28.1 40.6 40.6 37.5 31.3 50.0 37.5 41.4 34.5Vegetable relish 12.9 3.2 6.5 9.7 3.2 6.5 3.1 3.1 9.4 9.4Water 24.1 24.1 41.4 24.1 24.1 24.1 53.1 43.8 43.8 37.5Total 27.6% 25.4% 29.4% 28.6% 24.9% 21.6% 37.0% 33.3% 41.3% 31.2%

Note: Shaded cells indicate we did not have any responses for those cells because we did not have images of bananas with five portion sizes.

■ When asked how images could be improved, most participants said that no changes were needed. Those who did have suggestions said that shapes should be made into real pictures, the in-between portion sizes should become visible, and the photo angle should be from the side.

“Change [shapes] to actual pictures. Those shapes do not make sense. It is just a circle ... Actual pictures should remain the same.”

– Participant with lower education in rural area

“The hidden portions in the in-between should be made available.”

– Participant with higher education in urban area

“(The 90° pictures) should look like the 45° pictures to make them easy to see.”

– Participant with higher education in urban area.

■ Participants reported the foods in the photos looked like the foods when participants prepared and ate them at home. Many reported the plates were different, such that participants use metal or plastic plates instead of what they interpreted as clay or plastic plates in the photos or they use bowls.

“No, [the plates] are not the same [as the ones we use at home]. These look like melamine while at home we use plastic and metal plates.”

– Participant with higher education in urban area

■ This study investigated the preference for, ease-of-use perceptions of, and accuracy of visual variables in portion size estimation aids for dietary recall in Malawi.

■ Visual variables assessed included pictorial depiction, number of portion size options, photo angle, and sequence of image presentation.

■ The results from this study informed the development of portion size estimation aids used in a validation study.

1. Objectives

2. Study Population

3. Methods

4. Results

■ The process of recalling foods eaten on the previous day, especially when considering leftover food, may be difficult for participants.

■ Portion size estimation aid visual variables can be optimized to improve participants’ experiences during dietary recalls and enhance the recalls’ accuracy.

■ Results of this formative evaluation and the associated validation study can inform the development of portion size estimation aids for dietary intake data in low-resource settings.

5. Conclusions

Acknowledgments ■ RTI International provided

research funding for this study.

■ The authors thank Cathleen Prata, Zachary Gersten, and Rosalind Gibson of the International Dietary Data Expansion (INDDEX) Project for their collaboration.

■ The authors would like to thank consultants Denview Magalasi and Austrida Gondwe for their research assistance.

■ The authors have no conflicts of interest to declare.

More InformationPresenting author: Courtney H. Schnefke, MPH, RD 919.541.6314 cschnefke@rti.org

RTI International 3040 E. Cornwallis Road Research Triangle Park, NC 27709

www.rti.org RTI International is a registered trademark and a trade name of Research Triangle Institute. The RTI logo is a registered trademark of Research Triangle

Institute.