Smartphone photography as a possible method of post-validation trachoma surveillance in resource-limited settings

Abstract

Background

Validation of trachoma elimination requires monitoring after discontinuation of trachoma program activities, though such evaluations are not commonly done.

Methods

Conjunctival examinations and smartphone photography were performed on a random sample of pre-school children from 15 villages in a region of Burkina Faso thought to have eliminated trachoma.

Results

No clinically active trachoma was detected by in-field or photographic evaluation. Smartphone images demonstrated high agreement with field grading (>99% concordance).

Conclusions

Trachoma appears to have been eliminated from this area of Burkina Faso. Smartphone cameras may be a useful aid for monitoring in resource-limited settings.

Introduction

The World Health Organization (WHO) has targeted trachoma for elimination as a public health problem by 2020. Countries that have eliminated trachoma may apply to WHO for validation of elimination.¹ Validation is multifactorial but relies to a great extent on trachoma surveillance through in-person conjunctival examinations. Post-validation surveillance is required, although current guidelines do not provide specific recommendations. Post-validation surveys based on conjunctival examinations present special challenges since the conjunctival examiners may not have seen a case of trachoma in many years, likely increasing misclassification error. Conjunctival photography could offer a simple solution, but has not been widely tested.²

Methods

This cross-sectional, population-based study took place in 15 villages in the Nouna Health and Demographic Surveillance System (HDSS) site in Burkina Faso, July 2017.³ Burkina Faso is thought to no longer have endemic inflammatory trachoma, although areas of prevalent trichiasis may still persist.⁴ Mass azithromycin distributions were distributed in at least some parts of the country from 2007–2016. The most recent trachoma surveys in Nouna district were in 2007, at which time the prevalence of TF among 1–9 year-old children was 3.6%.⁵ For the present study a simple random sample of 15 children aged 6–59 months per village was selected from each of the 15 villages based on the most recent HDSS census (March 2017). This study population consisted of a subset of children selected for another study and thus does not match the WHO recommendation for trachoma surveillance; the sample size would allow estimation of the prevalence of trachoma with an absolute precision of ±2.6% assuming an expected prevalence of 1% and an alpha of 0.05.^6,7 After obtaining verbal informed consent from the guardian, trained clinical nurses everted the right upper eyelid of each child and assessed the conjunctiva with a pair of 2.5× loupes for the presence of TF and trachomatous inflammation—intense (TI) according to the WHO simplified grading system.⁸ Field graders had completed a training administered by an ophthalmologist (JDK) and had subsequently achieved a Cohen’s κ≥0.7 relative to an expert trachoma grading panel on a standard set of 100 photographs (photographs from a different study; the set included 54 cases of TF). Three photographs of the superior everted tarsal conjunctiva were taken from each participant’s right eye using the native 8 megapixel camera from a Nexus 5 smartphone (LG, Seoul, Korea) coupled to a custom-made Corneal CellScope attachment.⁹ Each photograph was subsequently compressed (Adobe Photoshop CC 2018; bicubic sharper resample setting) and transferred to an online file storage system (Box, Redwood City, CA). Masked graders assessed photographs for the presence of TF and TI from each eye’s triad of photographs. Photograph quality was graded by assessing image clarity and coverage of the tarsal plate. Two certified photo-graders graded all images; graders had achieved a κ≥0.7 on the same set of 100 standard photographs mentioned above. Discrepancies were discussed and adjudicated by the two graders. Ethical approval was obtained from the UCSF Committee on Human Research and the Comité Institutionnel d’Éthique du Centre de Recherche en Santé de Nouna.

Results

Conjunctival examination and photography were performed in 226 children from the 15 communities but data were lost for 26 children due to a lost smartphone. The mean age of the 200 children in the study population was 37 months (95%CI 34–38 months) and 95 were female (47%; 95%CI 38–56%). No cases of TF or TI were identified from the in-field examination (trachoma prevalence 0%, one-sided 97.5%CI 0–1.8%).

All 600 photographs were successfully compressed (total size 156MB). Compressed images were uploaded from Nouna using two different mobile devices and two different mobile service providers, with total upload times ranging from 28–49 minutes for one network and 97–108 minutes for the other. Image quality was satisfactory: 127 participants (64%, 95%CI 56–70%) had photographs deemed to be excellent quality and 72 (36%, 95%CI 29–43%) of moderate quality. The one ungradable set of photographs was too blurry to be graded with confidence. Randomly selected examples of excellent- and moderate-quality photographs are shown in Figure 1. Neither TF nor TI was identified in any of the 200 participant photographs (prevalence 0%, one-sided 97.5%CI 0–1.8%). The two graders had perfect agreement for each grade. Assuming the field grade as a reference standard, both specificity and negative predictive value were 100% (one-sided 97.5%CI 98–100%) for the 200 sets of gradable photographs. Results were identical for compressed and uncompressed images.

Figure 1.

Representation of photograph quality definitions, selected at random. The top row represents photographs judged to be of excellent quality and the bottom row represents photographs judged to be of moderate quality.

Discussion

This study demonstrates that smartphone photography may have a role for post-validation trachoma surveillance. Workers mastered the photography skills much more quickly than they did the trachoma grading (e.g., 30-minute training for photography; 3-hour training for trachoma grading), and were enthusiastic about incorporating digital technology into their work. An online repository allowed for assessment by trained graders from anywhere in the world with an internet connection. Image compression did not demonstrably affect image quality, increasing the chances that a tele-ophthalmology system such as this would be feasible in resource-limited settings with poor internet connectivity. Images could easily be audited, which may be helpful for validating elimination. Moreover, if such a practice were expanded, the resulting library of images and assigned grades could be used to develop automated algorithms for classifying trachoma. A limitation of this study was the lack of trachoma in the study population, making it difficult to assess the sensitivity of smartphone photography for trachoma surveillance. Further studies in areas with different endemicities of trachoma would be helpful to explore the diagnostic accuracy of smartphone photography for validating trachoma elimination.

Authors’ contributions:  

• Dr. Oldenburg had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

• Study concept and design: Oldenburg

• Acquisition, analysis, or interpretation of data: Snyder, Sié, Tapsoba, Dah, Ouermi, Arzika, Oldenburg, Keenan

• Critical revision of the manuscript for important intellectual content: Snyder, Sié, Tapsoba, Dah, Ouermi, Keenan, Oldenburg

• Statistical analysis: Snyder, Keenan

• Obtained funding: Oldenburg

• Administrative, technical, or material support: Sié

• Study supervision: Sié

Acknowledgements: We thank Genevieve LaCon (International Trachoma Initiative) for providing trachoma prevalence information for Burkina Faso.

Funding: This study was supported by That Man May See, Research to Prevent Blindness, the Calvin Wilson Scholarship for Future Leaders in Global Health Scholarship, and by the Doris Duke Charitable Foundation through a grant supporting the Doris Duke International Clinical Research Fellows Program at the University of California San Francisco School of Medicine. Blake M. Snyder is a Doris Duke International Clinical Research Fellow.

Competing interests: None declared.

Ethical Approval: Ethical approval was obtained from the UCSF Committee on Human Research and the Burkina Faso Comité national d'éthique pour la recherche en santé.