This paper builds upon the experiences of an implemented ConnDx approach in Kenya (15). We use retrospective quantitative data to demonstrate how an ConnDx approach can contribute to channel financial health funds.
2.1 Study setting
The ConnDx field experiment was conducted in Kisumu, Western Kenya, which is one of the counties with the highest malaria prevalence of the country. Although recent figures point towards a prevalence decrease from 27% in 2015 to 19% in 2021, these figures remain substantial and could worsen again with ongoing climate-change (16,17). Malaria misdiagnosis in Kisumu was previously determined very high, at 47% (18). Currently, Kisumu is rolling out a social health insurance program (Marwa) with most of its population (~1M) subscribed with CarePay M-TIBA through their mobile phones with high penetration rate (19).
2.2 Study period and study sample
The experiment ran from October 2017 to December 2018 across five health facilities, where ConnDx was presented as a ‘Malaria Test and Treat Campaign’ to improve access to quality malaria diagnosis and treatment. The technical evaluation of this experiment is reported elsewhere (15). The selection of providers was determined by factors such as high patient throughput recorded in DHIS-2, proximity to water in their geographic location and the willingness of the management to participate. The included providers varied from smaller facilities with specific operating hours to large 24/7 hospital facilities. Patients with malaria symptoms were informed of the campaign through posters in the waiting rooms and verbally during consultations with clinicians. The experiment operated alongside existing systems.
2.3 The ConnDx process for malaria in Kenya
M-PESA – the mobile money transfer system in Kenya - laid the foundations for M-TIBA: a healthcare exchange platform that collects real-time data on healthcare transactions creating transparency and de-siloes information from providers, patients, and payers (www.mtiba.com). For patients, M-TIBA serves as a mobile health wallet. In countries where most pay for care out-of-pocket, M-TIBA provides options for bundling different payments sources into pre-payment, create risk pools and linkage to insurance that can have big impact on health seeking behavior. We have applied the ConnDx process for malaria (15), but ConnDx could be applied to any medical condition that can be digitally diagnosed and for which vertical funding exists. Figure 1 visualizes the ideal ConnDx process for malaria. In our first pilots many of these steps were put into practice, as described below.
The process starts when a patient presents with fever at a health facility and the clinician suspects malaria. The patient can be invited to enroll into M-TIBA using his/her personal phone, which enables digital collection of clinical information and transactions into a cloud database. Subsequently, the clinician refers the patient to the lab for a malaria rapid diagnostic test (RDT). The lab technician performs the RDT and can interpret the result in real-time and inform the clinician to take pertinent action. In addition, the lab technician inserts the RDT into a network-connected reader that makes a digital photograph in a standardized manner and uploads the data into a dedicated cloud database (we used a reader from FioNet; Figure 1). Alternatively, the lab technician uses a smartphone App to photograph the RDT result and upload accordingly, while artificial intelligence software interprets the picture. The reader/App should feature acceptable sensitivity and specificity, be able to operate without electricity for a considerable time (20), allow for off-line data collection, and can be electronically maintained from a distance and be validated in LMIC settings (21–23). In the next step, the RDT cloud database exports its information to a payment platform using customized application programming interphases (APIs). An anonymized unique identifier-code is used (e.g. M-TIBA transaction code) that enables the combination of data from both platforms according to GDPR regulations. The combined data allow for downstream conditional payment algorithms: funds to the providers for performing RDTs (on the condition of correct performance) and funds for the patients for 1st or 2nd line malaria treatment (on the condition of being tested malaria positive and according to the applicable National Guidelines). Simultaneously, important data are collected and fed back into the system, like numbers, timing and geographical location of positive malaria cases, timing of provision of medical services, patient behavior in terms of selecting providers, provider behavior in terms of prescription of anti-malarials (generic/branded, 1st line, 2nd line).
2.4 Data analyses
Identification of hotspots and vulnerable groups
To assess the opportunities of ConnDx to contribute to improved intervention targeting we wished to assess its ability to identify hotspots and vulnerable population groups. We used the timing and geographical location of positive malaria tests to identify hotspots. For the identification of vulnerable groups, we estimated the socio-economic status (SES) by using three indicators (access to electricity, toilet type and education level of household head). The selection of these SES indicators was based on its association with the computed wealth index in the MICS dataset (24). Principal component analysis was applied to derive three wealth categories. Further details are described elsewhere (15).
Identification of patient healthcare seeking behavior
The ConnDx field experiment generated insights in patient’s travel distances to clinics because it registered geographic coordinates of participating clinics as well as the community units where patients live. Accordingly, we estimated which facilities were chosen by the patients and what the average distances traveled from their home was. The actual location of homes was not known; however, the participants indicated to which Community Unit they belonged (aggregate group of ~5,000 citizens). The geo-position of their homes was assumed to be in the geographical middle of the pertinent Community Unit they are living in. In addition, we also reported the time and day of the diagnostic performance which we analyzed to identify peak hours.
Measuring health staff compliance to guidelines
The ConnDx field experiment provided data on prescription behavior by providers. Prescription of anti-malarial drugs is recorded in the Kenyan National Malaria guidelines and should follow strict procedures regarding body weight of the patient and choice of 1st line versus 2nd line medicines, usually in the 95%/5% range (24). Discrepancy from these guidelines was quantified.
Estimation cost model
We created a cost model to generate insights in the standard care and ConnDx care approach. For this model we used the data insights regarding unnecessary over-prescription of antimalarial. Additionally, we made some reasonable assumptions in the model based on the findings of the ConnDx pilot. We assumed ConnDx would support lowering the usage of microscopy while increasing the RDT usage (ratio change of 50/50 to 20/80) because of RDT’s user-friendliness, shorter time-to-result, less dependence on qualified lab staff, electricity, equipment and reagents. Costs of uploading data into the digital RDT reader and eventually into a digital payment platform such as M-TIBA are both set at a reasonable 10% of diagnostics costs. We assumed that with active promotion of ConnDx the annual malaria population testing rate in Kisumu would (slightly) increase from the current 40% to 50%. For current annual antimalarials usage rate the figure of 46% was used (based on pediatric data (24)). Under ConnDx ideally only those who test positive will receive treatment, thus usage decreasing to the currently reported decreased malaria prevalence of 19%. During our field experiment we observed in participating private facilities the ratio 1st/2nd line antimalarial drugs to be 75%/25% respectively. We assumed that with implementation of ConnDx this ratio could change to 95% and 5% due to improved adherence to treatment National Guidelines. We also assumed that sensitivity and specificity of microscopy and RDT are comparable in LMIC field situations (25), although microscopy remains gold standard.