In many countries including South Africa, patients may receive an HIV test at one clinic, and look for follow-up treatment at another. This is particularly common in high HIV risk groups like migrant workers and truck drivers. The patients routinely have to be subjected to repeat testing, wasting resources as well as time to begin ARV treatment. The core problem is lack of consistent identity across clinics. Governments, organizations and clinics have been reluctant to put centralized biometric database systems in place due to privacy, security and compliance concerns. We have developed a federated biometric identity system which enables one organization to accept another organization’s biometrics for the purposes of biometric matching, which removes the requirement of centralization of biometric data and removes single point of failure, security-breach, or misuse, which are often perceived or real barriers to mass deployment of biometric solutions especially in politically volatile countries.
Approximately 300 HIV or TB patients in South Africa were tracked between 2 federated biometric stations in different parts of a clinic using a desktop server version of the federated biometric identity architecture, and using only multimodal biometric matching (iris + fingerprints) with no dependence on ID number, and with approximately 305,000 biometric cross comparisons, resulting in 0 patient tracking errors. Results from a survey questionnaire showed very high acceptance of biometric usage by HIV and TB patients when described as a means to improve efficiency of retrieving their medical information.
We are now beginning to support federated biometric capability using off-the-shelf mobile devices. Low-cost and highly-portable devices will enable very rapid and scalable deployments in clinics, and allow doctors to positively ID patients in villages without any permanent medical facilities.
DISEASE RISK ASSESSMENT
In 2013 UNAIDS established clear targets, which if achieved, will begin to end the HIV/AIDS epidemic.
By 2020: 90% of all people living with HIV will know their HIV status --> 90% of all people with diagnosed HIV infection will receive sustained antiretroviral therapy-->90% of all people receiving antiretroviral therapy will have viral suppression.
Achieving these targets means 73% of all people living with HIV worldwide will have their disease virally suppressed - a two to three-fold increase over the current global status.
Our projects in South Africa with partners like WITS RHI have proven the effectiveness of subjective assessment (responses to a set of specific questions) as a substitute for RDT strips to identify individuals at high risk for HIV, and have been tested on over 1000 patients. This is significant since the continual supply of RDT strips at scale is expensive and does not reach remote areas easily, while mobile-phone based assessment scales much more easily. Also, WHO guidelines dictate that a second test (supervised RDT test or more advanced tests) should be performed upon a positive initial RDT test, so the initial test, be it by machine-guided risk assessment or RDT, is in any case always only an initial triage step in knowing HIV status.
Disease risk assessment using a mobile application also engages the patient with the healthcare system, and has demonstrated a higher follow-through rate than any education, self assessment, or classroom based program to drive care-seeking behaviors.
We are in the process of exporting a WHO-proven, mobile-phone based machine-guided risk assessment approach for maternal-fetal health and transferring the capability to Africa.
The key to managing high-risk pregnancy is early identification. Ordinarily requiring little in the way of procedures or medications, an early identification of elevated risk pregnancy can trigger more frequent follow-ups, counseling and lifestyle changes, all leading to improved outcomes for mother and baby. However, scaling a capability proven in one region does not necessarily result in success in another region. Language and cultures are different, risk factors are different, training processes are different, and the clinical workflows are different. We are working on modularizing core machine-guided risk-assessment capabilities and infrastructure so that a local NGO or field partner can customize aspects, for example local risk factors, without having to build the baseline infrastructure that delivers mobile-phone-based service in the field at scale.
RDT TEST STRIPS
In partnership with Audere we have developed automated RDT reading for flu test strips which includes image acquisition, detection and analysis. The system runs on an Android-based phone with an optional back-end service.
The algorithms developed within this project can identify and analyze test strips at various scales, rotations, lighting conditions and can aid in situations where less trained professionals cannot accurately interpret the results. It also enables the capture of a digital record of an RDT reading, whether anonymized or personalized, which enables real-time surveillance at scale, for example. Most RDT readings are lost to paper record logs or disconnected digital systems.
Our image enhancement technique is able to read lines on test strips that may be too light or faint for a healthcare worker to accurately interpret.
Original image of test strip
Image enhanced to interpret the blue control line
Image enhanced to interpret the red line indicating a positive flu test
ACCELERATING INTEROPERABILITY OF CAPABILITIES
E.G. HEART RATE DETECTION
Traditional medical diagnostic devices are beginning to be replaced by more flexible, off the shelf, mobile devices. For example, to the right is a heart rate detection tool (developed by Conrad Tucker at CMU) which uses video from a mobile phone to detect the human pulse. Typically such capabilities remain isolated and are not useful by themselves. We are developing FHIR-based standards for such emerging capabilities that make it easier for developers to integrate them and other capabilities into total solutions that are then capable of scale.
HIGH-VALUE DATA-CENTRIC SERVICES FOR UNDERSERVED COMMUNITIES
A key first step in enabling high-value data-centric healthcare services for underserved communities is the specification and adoption of data standards that form the basis of the services. Work on OpenHIE standards has advanced the vision and brought people, processes and technologies together. More recent work on FHIR standards is rapidly gaining acceptance in the developed world, and has the capability to scale even in underserved communities with limited mobile infrastructure, bandwidth and full-time connectivity.
We are developing reusable software Application Programming Interfaces (APIs) to complement the existing and emerging standards, and enable integrators worldwide to easily build and deploy innovative and highly scalable healthcare systems for their local populations without having to develop specialized component capabilities themselves.
We have focused our initial work on developing a first set of proposed FHIR standards for projects primarily in Africa that target underserved communities at scale, and that are led by organizations within our partner network. The result is a set of FHIR-compatible standards together with FHIR-based implementation guides for an initial set of diagnostic measures, such as heart rate and SPO2 levels, that we believe serves as a blueprint for the process and outcome of developing FHIR-based standards for more complex use-cases within the larger partner community working on data-centric projects at scale.