Abstract

The high prevalence of anemia in resource-constrained settings calls for easy-to-use, inexpensive screening tools. The Sanguina Smartphone App, an innovative tool for non-invasive hemoglobin estimation via color-sensitive, algorithm-based analysis of fingernail bed images, was validated in the United States. This study evaluates the performance of the App in a population with different socio-economic, ethnic, demographic and cultural composition in rural Bihar, India. For 272 mainly adult patients of a private health centre, hemoglobin measurement with the App is compared with the gold standard laboratory blood analysis. For a second sample of 179 children attending pre-schools, hemoglobin measurement with the App is compared to the results of the HemoCue Hb 301, a point-of-care device using a small blood sample, serving as the reference standard for field-based settings. The App reaches ±4.43 g/dl accuracy and 0.38 g/dl bias of comparator values in the clinic-based sample, and ±3.54 g/dl and 1.30 g/dl, respectively in the pre-school sample. After retraining the algorithm with the collected data, the validity of the upgraded version is retested showing an improved performance (accuracy of ±2.25 g/dl, bias of 0.25 g/dl), corresponding to the results of the original validation study from the United States. The initial version of the App does not achieve the accuracy needed for diagnosis or screening. After retraining the algorithm, it achieves an accuracy sufficient for screening. The improved version with the potential for further adaptions is a promising easy-to-use, inexpensive screening tool for anemia in resource-constrained point-of-care settings. Hemoglobin is a protein in the blood that is required to move oxygen around the body. Anemia is a condition that occurs when levels of hemoglobin are too low. Anemia can lead to serious problems with the heart and lungs. This study assesses the accuracy of using a Smartphone App to measure hemoglobin levels by taking photos of a person’s nails in rural Bihar, India. Results from the App are compared with those obtained by measuring hemoglobin in blood samples. Initially, measurements of hemoglobin obtained using the App were inaccurate, however when the App was modified using data from some of the people in India, the App was more accurate. This study shows that the App can and should be adapted for use in different populations and can enable anemia to be diagnosed outside of hospitals and other healthcare settings. Haggenmüller et al. evaluate the performance of a smartphone app for non-invasive hemoglobin estimation that was developed in the USA in rural India. Performance improved when the app was retrained on the data collected in India.