JMIR Publications

ABSTRACT

Background:

Malaria impacts nearly 250 million individuals annually; specifically, Uganda has one of the highest burdens of malaria, with 13 million cases and nearly 20,000 deaths. Controlling the spread of malaria depends on vector surveillance, a system where mosquitos are collected and analyzed for mosquito vector species prevalence and density data in rural areas to plan control interventions accordingly. However, this relies on trained entomologists known as Vector Control Officers (VCOs) to speciate mosquitos via microscopy. This time-intensive process and a global shortage of entomologists cause significant time delays in reporting. VectorCam is a low-cost artificial intelligence-based tool that identifies a mosquito’s species, sex, and abdomen status with a picture, and can send these results electronically from surveillance sites to the decision makers, thereby de-skilling the identification process to Village Heath Teams (VHTs).

Objective:

This study evaluates the usability of the VectorCam system among Village Health Workers by assessing efficiency, effectiveness, and satisfaction.

Methods:

The VectorCam system has specialized imaging hardware and a phone application to speciate mosquitos. Two users are needed: (1) an Imager for imaging mosquitos using the application and (2) a Loader for loading and unloading mosquitos from the hardware. Critical success tasks for both roles were identified and VHTs were trained and certified by VCOs on the VectorCam system. In the first testing phase (Phase I), a VCO would assume the role of Imager or Loader, and a VCO would assume the other role. In Phase II, two VHTs are paired together, mimicking real use. Time to image each mosquito, errors of critical tasks, and System Usability Scores (SUS) were recorded per participant.

Results:

Fourteen men and seven women Village Health Team members, ages 20-70, 60% of whom had experience using smartphones were recruited. The average throughput for Phases I and II for the Imager were 70.0 and 56.1 seconds per mosquito respectively indicating a decrease in the length of time for imaging a tray of mosquitos. The Loader’s average throughput values for Phases I and II were 50.0 seconds and 55.7 seconds per mosquito respectively, indicating a slight increase in time. Regarding effectiveness, there was a 7.5% error rate for the Imager and a 13.75% error rate for the Loader in Phase I, and a 16.85% error rate from the Imager and an 11.27% error rate from the Loader in Phase II. The average SUS score of the system was 70.25, indicating an overall positive usability. A Kruskal-Wallis analysis demonstrated no significant difference in SUS scores between genders or smartphone experiences.

Conclusions:

VectorCam is a usable system for deskilling identification of mosquito specimens in the field in rural Uganda and was acceptable to VHTs. Upcoming design updates will address the voiced and noted concerns from users and observers.