This study describes feasibility and user experience through private healthcare providers in Kisumu of a novel digital approach to malaria diagnosis that directs conditional payments for malaria treatment: ConnDx. We demonstrate significant potential for increasing efficiencies of malaria service delivery in the Kenyan private healthcare sector concerning better diagnosis, reducing over-prescription, selecting correct 1st and 2nd line drug combinations and reducing malaria transaction costs, while at the same time generating valuable real-time data on malaria prevalence and incidence that can be fed into the DHIS-2, that captures routine health service data.
ConnDx proved through this pilot its potency to monitoring malaria epidemiology in semi-real time and generate important data for malaria management. Considerable variation was revealed between providers, with malaria positivity rates ranging from 7.4% (provider B) to 30.2% (provider D). This led to verifiable assumptions such as Provider B being a referral hospital and therefore less likely to serve primary malaria cases, while provider D, being located near wet rice fields, serving known hotspots for malaria. During months with more rainfall, there were significantly more malaria tests done at the providers than months with less rainfall (p-value <0.05, chi-square test). However, no relation was found between months with more rainfall and positive malaria test results. Providers located in or near low-income settlements (A and E) appeared to have higher malaria positivity rates. A chi-square test showed a relation between poorer patients and positive malaria tests (p-value <0.05, chi-square test). The quantitative analyses showed a relatively low participation rate of children: only 16.5% of reported patients were aged <5 years, with a positivity rate of 17.7%. Through our qualitative interviews, it was learnt that more children were tested for malaria, but clients experienced challenges subscribing their children to M-TIBA as dependents and sometimes reported them incorrectly as adult primary members. This was noticed later during the campaign and corrected but could have contributed to general underreporting of pediatric malaria cases. All in all, it was demonstrated that ConnDx can facilitate in semi-real time important healthcare provider-differences in malaria case management. Such information, when collected at a larger-scale level could help policymakers and health system managers to target their efforts for (human) malaria capacity building.
Secondly, this study demonstrated the overall potency of ConnDx to monitor provider prescription behaviours and identify practices that are significantly aberrant from the Kenyan National Guidelines. Important overall over-prescription was recorded of antimalarials (28.0%), varying between providers from 4.6% (provider D) to 63.3% (provider B). There are multiple reasons for over-prescription, ranging from monetary considerations of private providers to patient expectation and pressure to receive drugs, avoidance of clinicians to take the risk of a false negative diagnosis and subsequent fatality, etc. [14]. Our qualitative interviews showed patient pressure was mentioned as a reason by most of the interviewees when addressing this issue. Furthermore, ConnDx revealed an unexpected and erroneously high level of prescription of 2nd line antimalarials (overall 28.0%). Provider A revealed 2nd line prescription levels of overall 53.7%, at times going up to even 100%. This is remarkable, as 2nd line antimalarials are generally used for severe cases of malaria, which represent on average <2% [37], or in (rare) cases of suboptimal parasitological response with 1st line antimalarials (resistance). When probed with this observation, provider A reported a prolonged stock-out of 1st line antimalarials and therefore switching to 2nd line. Over-prescription of 2nd line antimalarials was more often found with more affluent and uninsured participants. This could indicate providers are aware of the socioeconomic status of their clients and they incorporate this into their prescriptions. Moreover, it appeared there was a very dichotomous, almost exclusive usage of either branded (provider C and D) or generic (provider A, B, E) antimalarials. One possibility could be that providers serving more affluent customers prefer procurement of branded versus generic antimalarials. Conversely, more affluent customers might request for branded instead of generic antimalarials. Often, generic medicines are considered to be of poor quality and treated with more suspicion than branded medicines [38, 39].
Third, this study indicates that ConnDx can increase efficiency in malaria service delivery by decreasing costs in several ways. Over-prescription of antimalarials can be monitored, aberrations identified, and actions can be undertaken to address those. A 2013 study conducted in four providers in western Kenya, noted that presumptive malaria treatment can lead to misdiagnosis rates as high as 53% in public facilities. [13]. ConnDx can play an important role to reduce such a figure in facilities. Further cost reductions can potentially be realized by ConnDx, such as decreasing paperwork in health providers; such automated systems saving time, manpower and being more accurate in reporting cases [40]. In addition, ConnDx implies less dependency on expensive and maintenance-dependent microscopy, and less electricity will be required to perform diagnostics tests. Moreover, due to its user-friendliness ConnDx, will provide more opportunities for lower trained lab-staff to perform such tests, saving personnel costs.
Finally, and most importantly, ConnDx can facilitate a much more targeted bottom-up payment for malaria services to providers and clients, creating unprecedented transparency as compared to current top-down payment systems [manuscript in preparation]. During this pilot, patients also benefitted from the reduction in costs as they received a free RDT test and treatment. However, when these would be obtained in a private facility outside this pilot, patients would have to pay on average 150 KES ($1.38 at the time) for an RDT and 150 – 500 KES ($1.38 – $4.59 at the time) for malaria treatment when tested positive, depending on whether they require first line or 2nd treatment. In public facilities, the average costs to patients per malaria visit were found to be 112 KES ($1.03) in 2016, including registration fees, diagnosis, and treatment [41]. However, in public facilities, there is a gap in the availability of both testing and treatment of malaria [32].
Overall user experiences of ConnDx from the perspectives of providers were positive. For all providers, the key challenge was reservations of their staff to adopt the use of RDTs instead of microscopy. In general, microscopy was still seen as the golden standard for malaria testing in Kenya, despite National Guidelines indicating equality of RDT and microscopy as diagnostic procedure [42]. The fact that microscopy can identify different malaria species and quantify the severity of infection is counterargued by the fact that 1st line treatment for uncomplicated malaria is identical for Eastern African malaria species (see below), independent of their load.
Indeed, ConnDx is dependent on the use of RDTs instead of microscopy for malaria diagnostic testing. Apart from National Guidelines, also the international literature reports sensitivity and specificity of malaria RDTs equal to microscopy [43, 44]. RDTs are also recommended by WHO [3]. Several studies demonstrated impaired sensitivity of microscopy in actual field situations in Africa as compared to perfectly controlled laboratory circumstances, with regular refresher training being required [33, 34]. RDTs have the added advantage that, in contrast to microscopy, these can easily be externally quality controlled by visual inspection of independent third parties. This opportunity is further enhanced through the ConnDx feature of making digital photographs of every test result, stored in secure cloud-based databases that can be accessed anywhere in the world. An additional advantage of RDT is that results can be digitalized, which accelerates data collection to (semi)real-timeliness, allows for telemedicine-based quality control and improves quality and completeness of data collection (versus paper-based malaria files being entered into national DHIS-2 systems on a several-time-per-year basis). These options are all much more problematic when performing microscopy. Moreover, in the sub-Saharan African reality, RDTs can readily detect Plasmodiumfalciparum, which causes the highest malaria morbidity and mortality and represents 99.7% of cases [35]. RDTs indeed are less available that specifically detect P. vivax, but this species is virtually absent in the region. Finally, indeed RDTs can provide a false-positive result with patients who had recent malaria episodes. This can be addressed by building a feature into the ConnDx algorithm that patients should be asked whether (s)he experienced malaria episodes in the past 1-2 months and if so, microscopy should be prioritized.
The above-outlined challenges suggest diagnostics tests for febrile diseases, such as RDTs, should be embedded in a digital infrastructure of logistics and human decision support to raise to the next level of effectiveness and cost reductions. In the future, ConnDx could be deployed for bacterial infections if these can be diagnosed by RDTs (e.g. for C Reactive Protein), leading to better-informed antibiotic prescriptions, which is important in fighting antimicrobial resistance [45].
This study has several strengths and weaknesses. Strength is the important innovation of ConnDx providing reliable, geo-tagged and semi-real-time insights in malaria diagnostic and therapeutic services by private sector providers in a semi-rural setting in Kenya. Kisumu county hosts 94 private providers, which deliver approximately half of all primary healthcare services to its population (the government supplying care through 148 additional providers). With the majority (1 million) of Kisumu citizens currently connected to M-TIBA [46], ConnDx could in principle rapidly be scaled to all private providers and supplement the governments’ DHIS-2 database with valuable real-time private sector information. Another strength is the pioneering nature of this study that was supported by the local health authorities to run in parallel to existing malaria services. This allowed for rapid collection of important data, leading to actionable information for policymakers who demonstrated strong involvement.
In terms of weaknesses, this study was observational and not a formal clinical trial. Therefore, there are no statistically validated results on (improved) diagnostic performance and (improved) clinical outcomes for malaria. Moreover, in this study, there were no special provisions taken for febrile patients who tested negative for malaria and pertinent consequences concerning changes in clinical decision making. For example, it was not studied what the effect was of reduced malaria prescription on provider prescription of alternative drugs for fever (in particular, antibiotics) and what were the clinical consequences of such decisions. Furthermore, as the ConnDx process was not yet fully digitalized, several steps were still performed manually, such as linkage of cloud-databases and payout mechanisms. Therefore, the study did not allow for direct and automated feedback-loops with any of the participating stakeholders (patients, providers, payers, policymakers). For this reason, progress observed concerning quality improvement or increased cost-efficiency during this pilot was modest and likely due to the realization of providers that they were being remotely observed by the ConnDx intervention. There were also external factors that influenced the ConnDx pilot, such as civil unrest due to national elections, which hindered the uptake of participants due to security issues. Several strikes of medical staff put constraints on general malaria service provision. Moreover, M-TIBA is using Safaricom as the mobile operator (with a market share of 70% of Kenya), which became political in Kisumu where most of the population is from another tribal background than the Safaricom ownership, resulting in temporary boycotts of usage of this platform. Finally, it should be kept in mind that ConnDx was implemented in parallel to existing malaria services covered by the NHIF and the MoH. Thus, health providers could in principle benefit by participating in two parallel financing mechanisms, which could potentially create perverse incentives. This would not be the situation when ConnDx is fully integrated into NHIF or any other UHC prepayment mechanism and made a compulsory condition for payouts.