Automated wearable cameras for improving recall of diet and time use in Uganda: a cross-sectional feasibility study

listen audio

Study Justification:
– Traditional recall approaches for collecting dietary intake and time use data are prone to recall bias.
– Automated wearable cameras have shown promise in collecting objective health behavior data and improving participants’ recall.
– This study aimed to evaluate the feasibility of using automated wearable cameras in rural Eastern Uganda to collect dietary and time use data.
Highlights:
– Mothers of young children wore automated wearable cameras for 2 non-consecutive days while continuing their usual activities.
– Participants’ dietary diversity and time use were assessed using an image-assisted recall.
– Most participants reported positive experiences with the automated wearable camera and image-assisted recall.
– None of the study withdrawals were definitively attributed to the camera.
– Challenges included device malfunction, insufficient lighting for image coding, labor-intensive processing and analysis, and difficulty interpreting images.
Recommendations:
– Future studies should use a structured data format to reduce image coding time.
– Data should be electronically coded in the field to eliminate ex post facto data entry.
– Computer-assisted personal interviews software should be used to ensure completion and reduce errors.
– In-depth formative work with key local stakeholders is needed to protect the ethical rights of study participants.
Key Role Players:
– Researchers from low-income countries
– Representatives from government and/or other institutional review boards
– Community representatives and local leaders
Cost Items for Planning Recommendations:
– Training for researchers and enumerators
– Equipment and devices (automated wearable cameras, tablets)
– Data processing and analysis software
– Community sensitization and engagement activities
– Ethical review and approval processes

The strength of evidence for this abstract is 7 out of 10.
The evidence in the abstract is moderately strong. The study had a relatively large sample size (n = 211) and participants reported positive experiences with the automated wearable camera and image-assisted recall method. However, there were some challenges identified, such as data loss due to device malfunction and difficulties in interpreting the images captured by the camera. To improve the strength of the evidence, the study could address these challenges by using a structured data format to reduce image coding time, electronically coding the data in the field to eliminate errors, and conducting in-depth formative work with local stakeholders to ensure ethical rights are protected.

Background: Traditional recall approaches of data collection for assessing dietary intake and time use are prone to recall bias. Studies in high- and middle-income countries show that automated wearable cameras are a promising method for collecting objective health behavior data and may improve study participants’ recall of foods consumed and daily activities performed. This study aimed to evaluate the feasibility of using automated wearable cameras in rural Eastern Ugandan to collect dietary and time use data. Methods: Mothers of young children (n = 211) wore an automated wearable camera on 2 non-consecutive days while continuing their usual activities. The day after wearing the camera, participants’ dietary diversity and time use was assessed using an image-assisted recall. Their experiences of the method were assessed via a questionnaire. Results: Most study participants reported their experiences with the automated wearable camera and image-assisted recall to be good (36%) or very good (56%) and would participate in a similar study in the future (97%). None of the eight study withdrawals could be definitively attributed to the camera. Fifteen percent of data was lost due to device malfunction, and twelve percent of the images were “uncodable” due to insufficient lighting. Processing and analyzing the images were labor-intensive, time-consuming, and prone to human error. Half (53%) of participants had difficulty interpreting the images captured by the camera. Conclusions: Using an automated wearable camera in rural Eastern Uganda was feasible, although improvements are needed to overcome the challenges common to rural, low-income country contexts and reduce the burdens posed on both participants and researchers. To improve the quality of data obtained, future automated wearable camera-based image assisted recall studies should use a structured data format to reduce image coding time; electronically code the data in the field, as an output of the image review process, to eliminate ex post facto data entry; and, ideally, use computer-assisted personal interviews software to ensure completion and reduce errors. In-depth formative work in partnership with key local stakeholders (e.g., researchers from low-income countries, representatives from government and/or other institutional review boards, and community representatives and local leaders) is also needed to identify practical approaches to ensuring that the ethical rights of automated wearable camera study participants in low-income countries are adequately protected.

This study is presented per the Strengthen the Reporting of Observational Studies in Epidemiology (STROBE) protocol [58]. This study was nested within a cross-sectional study of women with a child aged between 12 and 23 months inclusive (n = 211), to examine the impact of a labor-saving technology on women’s time for childcare, food preparation and dietary practices. It was conducted between January and February 2018 in Bugiri and Kamuli Districts, Eastern Region, Uganda. It validated the use of three methods of collecting data on dietary practices and women’s activities, which were the automated wearable camera-based image-assisted recall, interactive voice response collected via a mobile telephone, and 24-h recalls, using direct observation as the reference method. Only results related to the automated wearable camera-based image-assisted recall are reported here. In our study, maternal and child dietary diversity and women’s time allocation were assessed via an image-assisted recall using photos captured the previous day with an automated wearable camera. The methods are described in detail elsewhere [3, 59]. However, in brief, for each respondent, dietary intake and time allocation data were prospectively collected using photographs automatically taken every 30 s by a wearable camera attached to the participant. The next day, using the photos captured by the automated wearable camera during the previous day, an enumerator first independently coded the images for foods / beverages consumed by the mother and child and activities performed by the mother. Then, the enumerator administered an image-assisted recall to the participant. On the day before data collection began, a structured socio-demographic questionnaire was administered, and anthropometric measurements were made. On the final day of data collection, a structured questionnaire was administered to assess participants’ perceptions of the automated wearable camera-based image-assisted recall method. Each participant wore the automated wearable camera for two non-consecutive days and completed two image-assisted recalls, which meant enumerators met participants on a total of four days to collect two days of data on dietary practices and women’s activities. Ethical approval was obtained from the Uganda National Council for Science and Technology (UNCST) (A24ES), the London School of Hygiene & Tropical Medicine Observational Research Ethics Committee (Project ID: 1420), and the University of Greenwich Faculty of Engineering and Science Ethics Committee (Project ID: B0501). Community sensitization was done to ensure the study participants and other community members understood the study objectives and data collection methods. It included a review of key aspects of informed consent, demonstration of the devices used in the study (i.e., automated wearable camera, mobile phone, and GPS tracker), and a detailed description of the methods that would be used in the study, and time was allowed for questions. Following community sensitization, written informed consent (signature or thumb print) was obtained from all respondents who participated in the study. Twelve mother–child dyads were randomly selected from 22 purposefully selected villages in two districts of Eastern Region Uganda, as described elsewhere [3]. Mother–child dyads were excluded if the child was less than 12 months or greater than 23 months of age, was not yet eating solid foods on a regular basis, or was a multiple-birth child; the mother was unable to communicate in Lusoga, Luganda or English; either the mother or child had a severe disability; the mother was not the biological mother of the child; the mother was a co-wife with a selected mother; or either the mother or child was not available for the duration of the study. Participants were given a bar of soap, one kilogram of sugar, a half-liter of cooking oil and a t-shirt on the final day of the study. Also on the final day, they were given a photo of their family taken by a supervisor using a polaroid camera. The enumerators administered two structured questionnaires to the respondent. The first questionnaire collected information on household socio-demographics and assets, and factors related to women’s empowerment. The second questionnaire, which was administered on the final day of data collection, collected information on household mobile phone access and ownership, and participants’ perceptions of their experiences with the automated wearable camera-based image-assisted recall and other data capture methods assessed in this study that are not reported here (i.e., direct observation, 24-h recall, and mobile phone-based interactive voice response). Specifically, participants were asked to rate the automated wearable camera-based image-assisted recall method using a 4-point Likert scale (very bad, bad, good, or very good). Participants were also asked to select their favorite and least favorite method among the four data capture methods assessed, and whether they would be willing to participate in an automated wearable camera-based image-assisted recall study again. Although not specifically requested, any comments provided by the participants in answering these questions were translated and transcribed by the enumerators. A brief “innovative methods’ questionnaire was also completed at the end of each data collection day to assess participants’ experiences wearing the automated wearable camera, including any technical issues or reactions from members of their households or communities. Each participant was also asked to reconfirm her consent to use the images captured by the automated wearable camera. No data on automated wearable camera acceptability among other members of the household or community were collected. As described elsewhere [3], a small, lightweight, automated wearable camera (iON SnapCam Lite, dimensions 42 × 42 × 13 mm) was attached to a t-shirt worn by the respondent at approximately 06:00 and removed at approximately 21:00. Participants were instructed to wear the automated wearable camera while continuing their usual activities, and to cover or remove the camera as needed for privacy. A bespoke plastic clip using a safety pin was designed to keep the device firmly attached at the neckline of the t-shirt and well-positioned to minimize interference with clothing (Fig. 2). The wearable camera automatically recorded a picture every 30-s, storing all photos (approximately 1,500) on a micro-SD memory card and with the image number (e.g., 4) as the filename (e.g., SNAP0004.JPG). Examples of the photos obtained by device are provided in Supplementary Figure 1. The AWC affixed via a bespoke clip to the neckline of participants’ clothing The automated wearable camera was turned on at the beginning of the day and turned off at the end of the day by the enumerator. The t-shirt was provided by the study and worn by the participant over her clothing, so that if the participant needed to remove the camera, she would remove the entire t-shirt rather than handling the device. For administrative purposes, at the start of each data collection day, the enumerator took a single picture of the participant, her child and a placard displaying her study ID using the designated function of the device. After attaching the automated wearable camera to the project-provided t-shirt, the enumerator reminded the participant of key points covered during sensitization, i.e., that at any time during the study she could remove or cover the device or request all images to be deleted; that the device was splash proof but could not withstand immersion in water; and to do exactly the activities she would have normally done. Upon picking up the automated wearable camera at the end of the data collection day, the enumerator completed the innovative methods questionnaire. In addition, the first author (ALSB) kept records of inoperable devices, and members of the data collection team monitored issues (e.g., negative rumors about the automated wearable cameras) that may have affected study participation or compliance. Upon collection of the devices used by the study participants, ALSB saved a copy of the images recorded on each automated wearable camera memory card to a local drive, and assigned two participants’ memory cards (i.e., data for two participants) to an enumerator who had not been engaged in direct observation of the participant the previous day. The following day, the enumerator inserted the assigned memory card for the first participant into a tablet (16 GB Samsung with a 10″ screen, using Simple Gallery software for image display) to review the photos captured by the automated wearable camera. Using the image-assisted recall form, the enumerator annotated the foods she thought were consumed and the activities she thought were undertaken by the participant, and their corresponding image numbers, based on what she could see in the photos. Based on her interpretation, the enumerator demarcated the series of foods and activities for review later that day with the respondent. After completing the annotation for the first assigned participant, the enumerator completed the same steps for the second assigned participant. Upon meeting with the participant, the enumerator oriented the mother to the photos captured by the automated wearable camera by viewing, on the tablet with the inserted memory card, five pre-selected images: a picture of the mother herself, a picture of her child, a picture of her home, a picture of her garden, and a picture where her own hand is visible while performing a task (e.g., while preparing or cooking food, digging, or using a mobile phone). The enumerator rated the participant’s ability to recognize the content of these five photos on a three-point scale: recognized, recognized with help, or failed to recognize. The enumerator then administered the image-assisted recall. During this interview, the enumerator used “verbal probing” [60, 61] to elicit from the participant additional relevant information about the activities performed, for example to elaborate on what she was doing, who she was with, where she was going and why, etc. The enumerator revised her original annotations of foods / beverages consumed and activities undertaken by the participant, as needed, based on the participant’s feedback. The image-assisted recall protocol was adapted from one described by Kelly et al. (2015) [14]. The protocol followed ethical guidelines for automated wearable camera research to ensure privacy of the participants was maintained [51]. All protocols were pilot tested and refined prior to the start of the study. Enumerator training for all devices, protocols, and instruments took place over one week (December 18–22, 2017). The training comprised classroom training, role-play practice, and an assessment with individualized feedback. Training also included two days of field practice. In this study, feasibility was assessed using administrative data (non-compliance and withdrawal; camera malfunction; image quality; researcher time allocated to data coding and analysis); participants’ self-reported ratings of their experiences with the automated wearable camera and other methods used in the study; enumerators’ ratings of participants’ ability to interpret the images captured by the wearable camera; non-technical (e.g., fear of health or spiritual harm caused by the automated wearable camera) and/or technical (e.g., depleted battery) issues regarding the wearable camera reported by study participants or members of the data collection team; and requests by participant to delete wearable camera captured images. Participants’ self-reported experiences with the automated wearable camera and method ratings were double entered via EpiData. Administrative and demographic data and participant image-assisted recall orientation ratings were entered via Excel. Information about the data processing and analysis of demographic, dietary diversity and time-use data has been previously published [3, 62]. Because the data were not normally distributed, the Mann–Whitney U test and Fisher’s Exact test were used to compare method ratings for participating households and households lost to the study. Data were analyzed using Stata/SE version 17. P-values less than 0.05 were considered significant for all tests.

The innovation described in the study is the use of automated wearable cameras to improve recall of diet and time use in rural Eastern Uganda. The cameras automatically take photos every 30 seconds, providing objective data on participants’ dietary intake and activities. This method aims to reduce recall bias and improve the accuracy of data collection.

The study found that the use of automated wearable cameras was feasible, with most participants reporting positive experiences and expressing willingness to participate in similar studies in the future. However, there were challenges such as device malfunction, insufficient lighting leading to “uncodable” images, and labor-intensive data processing and analysis.

To improve the quality of data obtained, the study recommends several innovations:

1. Use a structured data format: Future studies should use a structured data format to reduce the time required for image coding.

2. Electronically code data in the field: Data should be electronically coded in the field as an output of the image review process, eliminating the need for ex post facto data entry.

3. Utilize computer-assisted personal interviews software: The use of computer-assisted personal interviews software can ensure completion of data collection and reduce errors.

4. Conduct in-depth formative work: Collaboration with key local stakeholders is needed to identify practical approaches for protecting the ethical rights of study participants in low-income countries.

By implementing these innovations, the quality and efficiency of data collection using automated wearable cameras can be improved, ultimately enhancing access to maternal health information.
AI Innovations Description
The recommendation to improve access to maternal health based on the study is the development and implementation of automated wearable cameras for collecting objective health behavior data in rural areas. This innovation can help improve the recall of diet and time use, which are important factors in assessing maternal health.

The study conducted in rural Eastern Uganda evaluated the feasibility of using automated wearable cameras to collect dietary and time use data from mothers of young children. The participants wore the cameras on two non-consecutive days while going about their usual activities. The images captured by the cameras were then used for an image-assisted recall to assess dietary diversity and time use.

The study found that the majority of participants had positive experiences with the automated wearable cameras and image-assisted recall method. They reported that the method was good or very good, and expressed willingness to participate in similar studies in the future. However, there were challenges encountered, such as device malfunction, insufficient lighting for some images, labor-intensive processing and analysis of images, and difficulty interpreting the captured images.

To improve the quality of data obtained and overcome these challenges, the study recommends several strategies. First, using a structured data format can reduce image coding time. Second, electronically coding the data in the field can eliminate errors associated with ex post facto data entry. Third, utilizing computer-assisted personal interviews software can ensure completion and reduce errors. These strategies can streamline the data collection and analysis process, making it more efficient and accurate.

Additionally, the study highlights the importance of conducting in-depth formative work in partnership with key local stakeholders to ensure the ethical rights of study participants are protected. This involves collaborating with researchers from low-income countries, representatives from government and institutional review boards, and community representatives and local leaders. By involving these stakeholders, practical approaches can be identified to address the unique challenges faced in low-income country contexts.

In conclusion, the recommendation based on the study is to develop and implement automated wearable cameras as an innovative method for collecting objective health behavior data in rural areas. This can improve access to maternal health by providing more accurate and reliable information on dietary intake and time use, which are crucial factors in assessing maternal health.
AI Innovations Methodology
Based on the provided information, the study evaluated the feasibility of using automated wearable cameras in rural Eastern Uganda to collect dietary and time use data from mothers of young children. The study aimed to improve recall of diet and time use, which are prone to recall bias when collected through traditional methods. The participants wore an automated wearable camera for two non-consecutive days while going about their usual activities. The images captured by the camera were then used in an image-assisted recall to assess dietary diversity and time use.

To simulate the impact of recommendations on improving access to maternal health, a methodology could be developed as follows:

1. Identify the recommendations: Based on the study findings and analysis, identify the specific recommendations that can improve access to maternal health. For example, one recommendation could be to use a structured data format to reduce image coding time.

2. Define the indicators: Determine the indicators that will be used to measure the impact of the recommendations on improving access to maternal health. These indicators could include metrics such as the number of women accessing maternal health services, the quality of care received, and the reduction in maternal mortality rates.

3. Collect baseline data: Gather baseline data on the current state of access to maternal health in the target population. This could involve conducting surveys, interviews, or analyzing existing data sources.

4. Develop a simulation model: Create a simulation model that incorporates the baseline data, the identified recommendations, and the defined indicators. The model should simulate the impact of implementing the recommendations on the selected indicators.

5. Run the simulation: Use the simulation model to run different scenarios that represent the implementation of the recommendations. This could involve adjusting variables such as the adoption rate of the recommendations, the availability of resources, and the level of community engagement.

6. Analyze the results: Evaluate the outcomes of the simulation by comparing the indicators before and after implementing the recommendations. Assess the impact of the recommendations on improving access to maternal health and identify any potential challenges or limitations.

7. Refine and iterate: Based on the analysis of the simulation results, refine the recommendations and the simulation model if necessary. Repeat the simulation process to further optimize the impact of the recommendations on improving access to maternal health.

By following this methodology, policymakers and stakeholders can gain insights into the potential impact of specific recommendations on improving access to maternal health. This can inform decision-making and resource allocation to effectively address the challenges and barriers faced in maternal health care.

Partagez ceci :
Facebook
Twitter
LinkedIn
WhatsApp
Email