BACKGROUND: The World Health Organization recommends confirmatory diagnosis by microscopy or malaria rapid diagnostic test (RDT) in patients with suspected malaria. In recent years, mobile medical applications (MMAs), which can interpret RDT test results have entered the market. To evaluate the performance of commercially available MMAs, an evaluation was conducted by comparing RDT results read by MMAs to RDT results read by the human eye.
METHODS: Five different MMAs were evaluated on six different RDT products using cultured Plasmodium falciparum blood samples at five dilutions ranging from 20 to 1000 parasites (p)/microlitre (µl) and malaria negative blood samples. The RDTs were performed in a controlled, laboratory setting by a trained operator who visually read the RDT results. A second trained operator then used the MMAs to read the RDT results. Sensitivity (Sn) and specificity (Sp) for the RDTs were calculated in a Bayesian framework using mixed models.
RESULTS: The RDT Sn of the P. falciparum (Pf) test line, when read by the trained human eye was significantly higher compared to when read by MMAs (74% vs. average 47%) at samples of 20 p/µl. In higher density samples, the Sn was comparable to the human eye (97%) for three MMAs. The RDT Sn of test lines that detect all Plasmodium species (Pan line), when read by the trained human eye was significantly higher compared to when read by MMAs (79% vs. average 56%) across all densities. The RDT Sp, when read by the human eye or MMAs was 99% for both the Pf and Pan test lines across all densities.
CONCLUSIONS: The study results show that in a laboratory setting, most MMAs produced similar results interpreting the Pf test line of RDTs at parasite densities typically found in patients that experience malaria symptoms (> 100 p/µl) compared to the human eye. At low parasite densities for the Pf line and across all parasite densities for the Pan line, MMAs were less accurate than the human eye. Future efforts should focus on improving the band/line detection at lower band intensities and evaluating additional MMA functionalities like the ability to identify and classify RDT errors or anomalies.