eHealth and mHealth Advisor for the Measure Evaluation SIFSA (MEval-SIFSA) Project, Rob Allen, talks about laying the foundations to prepare Africa’s healthcare system for a big-data driven era.
MEval-SIFSA, a USAID-funded PEPFAR initiative, is a show partner for eHealthALIVE2016 and will be hosting a comprehensive M&E afternoon workshop to give individuals and organisations practical tools for assessing and making evidence-based decisions at critical points throughout a project’s lifecycle. Get your ticket now and be part of this transformational event.
Tell us about your background in healthcare and your involvement with SIFSA.
I originally came to South Africa 13 years ago on sabbatical from the UK to work with some of the organisations working on HIV and AIDS here, and have since continued to work at the intersection of health, technology and data. I now work for ICF International, an organisation which provides professional services and technology-based solutions to governments, commercial clients and healthcare organisations and is an implementation partner for the Measure Evaluation SIFSA project of South Africa (MEval-SIFSA).
Do you still see a huge difference between the use of mHealth and eHealth, or is it all one thing?
eHealth is an all-encompassing term, so anything that is digital in any sense is often defined as eHealth, but there are so many semantics that go into it; digital health is now the term often used over eHealth. mHealth is a subset of eHealth that focuses on the growth of mobile technologies (such as cell phones) and has opened up a number of other, new possibilities for eHealth. For instance, mHealth has more possibilities for patient engagement than traditional eHealth which has used fixed infrastructure. I would say that in the same way the term “eHealth” somewhat overran the term health informatics, mHealth has gone some way to change the way that we look at health informatics and eHealth as well.
What are the specific barriers that Africa needs to overcome to start leveraging big data for better health outcomes?
First, there must be a clear understanding of what big data is and what it means to different people. There’s a general perception that big data is far bigger, and more complex, than what we’re used to and able to process effectively. When it comes to healthcare in Africa, just about any reasonably sized electronic data set could be considered big data to the people dealing with it. Has Africa really ever had the capacity to even manage relatively small data sets and use them effectively? One of the biggest data challenges in African healthcare lies in getting access to the data we already have and using it.
On the MEval-SIFSA project, USAID has given us a unique opportunity to work with the Department of Health (DoH) in enhancing their capacity to analyse large quantities of data coming from a variety of routinely collected sources of data including, e.g., electronic patient records. It is common to talk about leapfrogging, or jumping over the steps that other countries have gone through. This may in some senses be appropriate, but we need to get the foundation right, such as: good governance, good operating procedures and good policies and then make sure they are actually implemented well, so there is a solid base around the issues that need to be addressed. The DoH in South Africa has done a good job of putting some of these fundamentals in place, as detailed in their eHealth strategy.
We also need to be aware of what we’re dealing with in big data, what we want to gain from it, and plan the roadmap around going ahead with it. Without that vision and full understanding of where we’re going to get to that data from, and how we are going to use it, it can end up being an expensive and relatively pointless diversion.
In addition, we need to start looking at how data actually works together, because another aspect of big data is that there are many different diverse data sets and information coming into a central place that need to be pooled together and synthesised. If we’re not structuring that framework right at the outset, as the DoH has done in their Normative Standards Framework for Health, then we’re never really going to be able to make sense of it or interpret it properly, and also there’s a good chance that we won’t be able to present it well or analyse it in any sensible way.
Talk to us about defining data sets and making sure we’re collecting the right information?
The healthcare industry has collected a lot of data, but now we need to decide on the most appropriate data sets and make sure we are collecting the right information. In South Africa’s public health sector, this is guided by the national indicator data set. Studies have shown that it’s often better to collect less data because you don’t want to make data collection overly burdensome. For example, having to collect 500 different data points can put too much pressure on the people who are charged to collect it, which can also compromise the quality of the data. It is better to collect key data that is relevant and accurate. With an HIV epidemic at the scale of South Africa’s, this effort leads to development of appropriate interventions that target the right people, in the right places, at the right time. This is why USAID felt that the SIFSA project has a role to play in strengthening the systems that manage this data. Every extra dataset adds to the level of complexity.
One of the great advantages with eHealth and mHealth is more robust information because data collection can be automated, or at least made easier. If we start to automate collection of specific data, maybe based on location or social interaction, the data can be accurately, and instantly, generated and that’s a big help. We’ll immediately have data that is more accurate, relevant, and timely than data that’s captured manually or retrospectively, which is where errors largely occur. This was a key finding in a study that the Gauteng DoH conducted with MEval-SIFSA support to establish a baseline for data quality in their province.
Automating data collection is compelling but then the question becomes, how do you automate this process? Some data can be collected through a larger investment in patient information systems, which are designed in collaboration with healthcare workers and drive the capture of information they want to use effectively at the bedside or on the ground. It’s also very important to determine data sets from the patient perspective. This is often lacking and results in a lack of patient focus.
Should we be factoring in new technologies such as Natural Language Processing and machine learning to help us improve the quality and the use of our data?
Definitely, but it needs to be done in a way that follows proper protocols and procedures. The landscape around the adoption of technology is filled with projects that have failed, often because we haven’t gotten the foundations right first. For instance, South Africa needs to find a solution to the current connectivity challenges at health facilities. If you start to get the connectivity issues sorted out, then you can start to implement electronic systems that can be connected and provide more relevant patient-based data. After this you can start to have national or large scale systems that could work. The computing power is certainly there, and certainly available for any country if it wishes to invest in it. Economies of scale can bring down costs related to big data systems such as server space and analytics. There is room for the private sector to make a contribution here.
To put effective systems, including machine learning, in place, we need good governance and standards so that data collected from different places is interoperable and can be passed between different systems. If we’ve got different systems doing their own thing, it starts getting very complicated very quickly and diminishes the gains. What South Africa has done so far to address this is set up the Health Normative Standard Framework for eHealth and define a master patient index which will be built up by the Health Patient Registration System that is being rolled out.
Once we have built the foundations and solved the capacity, connectivity and interoperability issues then we start to build up data that is far beyond what we actually envisage now and what we are using now. Then we start talking big data and then there is a whole set of other questions around how we effectively use it and what’s the cost-benefit ratio, because at the end of the day we are working within resource-constrained environments.
So, what are the actual benefits? That might be one of the largest challenges that we have to face, because currently we can’t quantify what the overall benefits of doing this would be and that poses a huge challenge to make the investment case for it. You can say that we’re going to spend X billion rands on a particular set of projects based around using big data effectively and implementing eHealth systems, but unless you can actually say that this will point to something like an improvement in life expectancy then it’s very hard to build a case for investing such a huge amount of money. It’s all got to be done with cost in mind, trying to maximise the value and decrease the risk. We also need to make sure the systems are built with sustainability in mind from the beginning. We certainly don’t want to get tied into contracts with businesses that will continue to raise costs well above inflation and make it difficult to sustain the system going forward.
What sort of impact do you predict harnessing Big Data will have on human capacity in the sector?
There needs to be a concerted effort to build up capacity within the health sector along with the roll out of systems. Specific initiatives like the eSkills institute can make a significant contribution towards skills development but more needs to done. Academic institutions have a huge role to play in developing a cohort of people who have sufficient skills in health informatics, health analysis, health economics and data science, etc. To get systems running effectively we certainly need more skills than are currently available. This is one area in which I believe that a strong, professional, but independent, body for people who work in the space could make an impact. There’s a conversation around the IT side, monitoring and evaluation, and data science. However, within the eHealth and health informatics space there isn’t a professional body or means of accreditation that could drive people towards skill sets that will be useful in the future. The contribution being made by MEval-SIFSA and some NGOs is valiant but needs to be aligned to much greater effort.
Beyond the skilled professionals there’s also a lot of IT skills capacity that needs to be built around individual healthcare workers on the ground. If we’re going to implement larger eHealth systems and start generating data sets from digital systems, we’re going to need to train all the people who are involved at all levels to ensure that they have the required skills to work with these systems.