Description
Introduction-Objectives: Kenyan National Malaria Guidelines mandate parasitological testing for all suspected malaria cases through microscopy or malaria rapid diagnostic tests (mRDTs). Testing before treating suspected malaria cases saves lives since there are other causes of fever in this region.
However, adherence to this guideline remains difficult since most people with symptoms of malaria seek care first at private providers, where no tests are available or offered. This haphazard usage of malaria parasitological testing risks overprescription, increased population drug resistance and poor individual case management. The private health sector remains a black box for evidence of service delivery and quality, as epidemiological data is not reported by private providers to African health systems. With Connected Diagnostics (ConnDx) we aim to increase diagnosis of malaria, reduce irrational use of antimalarial drugs, improve the quality of care and improve quality and timeliness of data reporting; thus, complementing public sector efforts to reverse the recent decline in control of malaria in Kenya.
The objective of this study is to analyze data collected by a malaria digital diagnostic platform (ConnDx), assisted by Artificial Intelligence (AI) to increase malaria surveillance across public and private sectors in (semi-) real time.
Methods: The ConnDx for malaria project started in April 2023, using mRDTs, namely CareStart and Abbott Bioline, in parallel with ongoing routine malaria care across 27 healthcare facilities in Kisumu County, Kenya. Clinicians were requested to perform mRDT for suspected malaria cases and treat when the test result was positive. The mRDT results were digitalized through mobile phone photography using an App and uploaded to a cloud for AI interpretation. The AI-interpreted results were displayed on digital dashboards, accessible to clinicians and decision makers. AI mRDT interpretation accuracy was evaluated by calculating the sensitivity and specificity, using human mRDT interpretation as the gold standard. Clinicians were incentivized for the correct upload of the mRDT image via automatic payment of data bundles.
Results: 3,237 mRDTs were reported from the 27 participating healthcare facilities from April 2023-September 2023. Of the 3,237 RDT results submitted-due to a shift in procurement of the type of mRDTs to Paracheck-only 23.8% of the mRDT results (n=770) were eligible for interpretation by AI. The prevalence of malaria in Kisumu based on human interpreted mRDT results was 25.1% (805/3,209). The AI mRDT classification showed substantial agreement with clinician’s interpretation with a sensitivity and specificity of 86.6% (95% confidence interval [CI]: 79.6‒92.1) and 98.0% (95%CI: 96.6‒98.9) respectively (Cohen’s kappa 0.85, 95%CI: 0.78‒0.92). Overall, the False Negativity Rate was 13.4% while False Positivity Rate was 2%.
Conclusion: ConnDx successfully reported malaria testing results using AI in semi-real time to the local health surveillance system from both private and public sectors, making dashboards with results available for policy makers, and validating results uploaded by the clinicians. Additional qualitative research on user experience, usability and acceptability of this digitized malaria testing will be conducted between February and March 2024 as well as updated data analysis of the AI interpretation of the new Paracheck mRDT. Preliminary results will be reported.
| Contact Geneva Health Forum | I just want to receive information about the GHF 2024 conference / Je souhaite simplement recevoir des informations sur la conférence GHF 2024 |
|---|
