• Big Data
  • Need a Big Data and AI overview; this’s it

    It seems that Big Data isn’t big after all. David Stephenson, in his book Big Data Demystified, published by Pearson, says “Big” significantly understates the volume and differences to conventional data. Understanding it needs to be in its context of AI and Machine Learning (ML). 

    He ranges over Big Data’s:

    UsefulnessEcosystemStrategyImplementationTechnology selectionTeam buildingGovernance and legal compliance.

    Case studies bring each of these into practical environments. While Stephenson’s keen on Big Data, his book’s not an exhortation to rush into initiatives. Instead, his “Keep in mind” boxes are valuable switches from his commentary that provide realistic insights for policy makers, strategists, executives, managers, practitioners, health workers and students.

    It’s clearly written and offers new, late and in between comers to Big Data many very valuable insights and case studies. Examples are his analyses of Big Data’ infrastructure requirements and its 3Vs, Volume, Velocity and Variety. His concept of a “data lake” draws a vivid perspective of Big Data’s difference to databases 

    He includes a salutary lesson. Many Big Data projects “Die on the launch pad because of inadequate preparation, internal resistance or poor programme management.” His case study was a $62m crash.

    As Africa’s health systems move towards more Big Data opportunities, Big Data Demystified will help to set scenarios that lie ahead. Investment in new skills is part of it.

  • Top ten algorithms that can help healthcare

    As algorithms become more prevalent in eHealth, it’s important to have a clear development path for their use. Two core principles are:

    No single algorithm works best for every problemA learning a target function (f) maps input variables (X) to an output variable (Y), so: Y = f(X), used for predictive modelling.

    An article by James Lee in Towards Data Science sets out ten top algorithms. They’re: 

    Linear regression, a long-standing techniques from some 200 years ago, but a good starting pointLogistic regression, suitable for binary classification problems and their two class valuesLinear discriminant analysis, where prediction rely on calculating a discriminate value for each class and making a prediction for the class with the largest valueClassification and regression trees represented by a binary treeNaive Bayes, a simple, powerful algorithm for predictive modelling using two types of probabilities, one of each class, the other the conditional probability for each class given each x valueK-Nearest Neighbours (KNN), a simple and effective algorithm, where predictions are derived from  new data points by searching  entire data sets for the K most similar instances, the neighbours, and summarizing output variables for those K instancesLearning Vector Quantisation (LVQ), a KNN relative, and an artificial neural network algorithm enabling choices of the number of instances to hang onto, learning precisely what the instances should look likeSupport Vector Machines (SPV) are possibly one of the most popular, using a hyperplane to separate points in input variables spaces by their class, either class 0 or class 1Bagging and Random Forest (BBR), another popular algorithm, called Bootstrap Aggregation or bagging, and can estimate quantities from data samplesBoosting and AdaBoost, an ensemble technique aiming to create strong classifiers from several weak classifiers by building a model from training data then creating a second model that attempts to correct the errors from the first model.

    Selecting algorithms in eHealth uses, four questions need answering, what’s:

    The size, quality, and nature of the dataThe available computational timeThe urgency of the taskThe data to be used for.

    The answers aren’t easy to find. Lee points out that experienced data scientist can’t tell which algorithm’s best before trying different ones. It seems that Africa’s eHealth needs time to ponder these before settling on a preferred short list.

  • What were the top ICT stories in 2017?

    Now 2017’s history, the significant ICT themes can be seen. A retrospective by Health IT Analytics found the top ten from its posts. They’re Big Data, Fast Healthcare Interoperability Resources ( FHIR) and machine learning are included. They’re:

    Top 10 Challenges of Big Data Analytics in HealthcareTop 4 Machine Learning Use Cases for Healthcare ProvidersWhat is the Role of Natural Language Processing in Healthcare?Judy Faulkner: Epic is Changing the Big Data, Interoperability GameHow Healthcare can Prep for Artificial Intelligence, Machine LearningExploring the Use of Blockchain for EHRs, Healthcare Big DataHow Big Data Analytics Companies Support Value-Based HealthcareBasics to Know About the Role of FHIR in InteroperabilityData Mining, Big Data Analytics in Healthcare: what’s the Difference?Turning Healthcare Big Data into Actionable Clinical Intelligence. 

    It’s a valuable checklist for Africa’s health informatics and ICT professionals for there personal development plans. eHealth leaders can use it too to ensure their eHealth strategies either include initiatives for the top ten, or lay down the investigative and business case processes for future plans. 

  • Managing high risk populations’ health needs better information

    Successful population health management need health organisations to learn and know how to manage risks, outcomes, utilisation and well-being of high and increasing risk communities. Components Necessary for Managing High-Risk Population, a report from Cerner, available from EHR Intelligence, sets out ways that organisations can use information to manage people’s care as part of health risks cohorts and identify opportunities to reduce avoidable costs.

    The report says about 5% of the population are high risk. Another 20% are grouped as rising risk. Globally, these health risks are increasing. In a report, on global health risks, WHO says “Health risks are in transition: populations are ageing owing to successes against infectious diseases; at the same time, patterns of physical activity and food, alcohol and tobacco consumption are changing. Low- and middle-income countries now face a double burden of increasing chronic, non-communicable conditions, as well as the communicable diseases that traditionally affect the poor. Understanding the role of these risk factors is important for developing clear and effective strategies for improving global health.”

    Cerner’s report focuses on care management requirements and patients. The principles apply to health promotion and illness prevention too. Selecting the right people for care management plans is essential to improve their health and enable healthcare to cost outcomes. Cerner proposes three components:

    Risk stratification strategiesHealth IT needs for managing high-risk populationsChoosing the right care management approach.

    These are supported by six eHealth requirements.

    Longitudinal healthcare recordsChronic condition and wellness registries for patient cohortsCare management and co-ordination systemsLongitudinal plansData analytics and modellingReferral management system. 

    Linked to information on social determinants of health, some of this approach can support health interventions in high risk communities in low and middle income countries. It could include local data and predictive analytics to identify changes communities’ behaviour, and needs and demands for healthcare and related services such as education and social care.

  • Is machine learning the new buzzword for healthcare?

    By now, it’s old news that big data will transform healthcare. Electronic health records and health information systems have arrived, data flows, and there’s a lot of it. However, all this data, is only useful when it has been analyzed, interpreted and acted on. So will it be algorithms that perform this analysis that will really transform healthcare?

    Access to lab results via a mobile app, has helped clinicians diagnose and treat patients faster. Imagine how much more useful these results would be if they also showed the patient’s risk for cardiovascular disease or renal failure, based on the last several years of the patient’s lab reports. This is where machine learning might help physicians to make better decisions at point of patient care.

    Machine learning is an area of artificial intelligence (AI) that is starting to attract interest in healthcare. It is a set of algorithms that help a system to automatically learn and predict outcomes, after continuous exposure to variable datasets. The value of machine learning in healthcare is its ability to process huge amounts of clinical information, beyond that of human capability, and then reliably convert analysis of that data into clinical insights. This will help physicians plan better and ultimately lead to better outcomes, lower costs of care and increased patient satisfaction.

    Machine learning is already making headlines in healthcare. Google has developed a machine learning algorithm to help identify cancerous tumors on mammograms. Stanford is using a machine learning algorithm to identify skin cancer. A JAMA article, last year, reported the results of a deep machine-learning algorithm that was able to diagnose diabetic retinopathy in retinal images. Others, like Philips, are transforming TB screening, using machine learning algorithms that can offer an objective opinion to improve efficiency, reliability, and accuracy.

    Machine learning puts a new arrow in the quiver of clinical decision-making.

  • EU’s BigData@Heart aims to improve heart disease treatments

    A report on Cardiovascular Disease (CVD) published by Springer says CVD prevalence in sub-Saharan Africa’s increasing. Limited access to prevention and continuing care are seen as constraints to improvements. The EU’s BigData@Heart project may contribute to the development of treatments for heart disease patients. It has lessons for Africa’s health systems.

    It’s a large-scale, five-year, €19 million project. Its aim’s to use data and advanced analytics to develop a translational research platform of phenotypical resolution to improve patient outcomes and reduce societal burdens of atrial fibrillation (AF), heart failure (HF) and acute coronary syndrome (ACS). Data sources include real-world evidence, best-practices in drug development and personalised medicines.

    Four outputs are:

    ·         New universal, computable, definitions of diseases and outcomes relevant for patients, clinicians, industry and regulators

    ·         Informatics platforms linking, visualising and harmonising data sources, completeness and structures

    ·         Data science techniques to develop new definitions of disease, identify new phenotypes, and construct personalised predictive models

    ·         Guidelines that allow for cross-border use of big data sources acknowledging ethical and legal constraints and data security.

    It can have considerable value for Africa. The continents health systems and cardiologists could move their CVD services ahead in the wake of BigData@Heart’s progress. 

  • Africa’s eHealth isn’t far behind South America and Mexico

    Africa’s eHealth has a different profile to the average for South America and Mexico, but it’s not trailing significantly. Comparing findings from the WHO and GEO’s Global eHealth Survey 2015 shows similar simple coverage rates of just over a third. A global average’s near 50%.

    Perhaps the most important difference’s the emphasis on Big Data. South America and Mexico are at about 12%. Africa’s at about 2%. The survey didn’t ask for data about Artificial Intelligence (AI) or coverage of cyber-security. It doesn’t provide data about the quality, sophistication or maturity of the coverage. It’s not easy data to assemble, quantify or analyse. 

    Suffice to say, it highlights the need for Africa’s health systems to catch up on AI. A first step may be using their existing data more extensively. A second may be supporting public health specialists and clinical teams in local AI initiatives. The comparison seems to indicate no need for a sudden rush. A steady, imaginative plod along the AI road should be sufficient.

  • AI and Big Data will transform healthcare

    There’s a substantial progression of digital capacity in which data is produced and stored. In 2013, the amount of available digital data encompassed 4.4 zettabytes and it is estimated to reach 44 zettabytes, which is 44 trillion gigabytes, or ten times more by 2020. There is no denying that the world of Big Data’s enormous. Healthcare’s challenge is how to make best use of it.

     Artificial intelligence (AI) is a vital and increasing part of it. It’s quickly becoming a necessity for healthcare. AI’s already widely used in everyday life. It’s in cars, Google searches, Amazon suggestions and other devices, but is yet to extend towards large-scale, routine AI for sophisticated activities such as healthcare and its clinical decisions.

    AI services such as Apple’s Siri, Microsoft’s Cortana, Google’s OK Google, and Amazon’s Echo can extract questions from speech using natural-language processing and perform limited set of useful things, such as looking for restaurants, provide  directions, find an open slot for a meeting, or run a simple web search. In addition, a 19-year-old British programmer launched a bot last September which is successfully helping people to appeal their parking tickets. It’s an “AI lawyer” that sorts received parking tickets. In both London and New York, the bot has a success rate of 64%, which translate to 160,000 of 250,000 parking tickets successfully appealed.

    The same efficiency is essential in healthcare. An article in Medical Futurist reports that in years to come, AI in healthcare and medicine may organise patient pathways and treatment plans, and provide doctors with all information they need to make better informed decision. Several companies have a stake in AI in healthcare. They include Dell, Hewlett-Packard, Apple, Hitachi Data Systems, Luminoso, Alchemy API, Digital Reasoning, Highspot, Lumiata, Sentient Technologies, Enterra, IPSoft and Next IT. Their ultimate goal is to transform medicine and healthcare in way that ensures that it’s widely available to the average, mainstream users and not only the richest medial institutions or to a handful of experts.

    AI’s used in several areas in healthcare. It includes mining medical records, designing of medical plans and medication management. Googles AI research branch in cooperation with the Moorfields Eye Hospital NHS Foundation Trust in London has launched a Google Deepmind Health project. It mines medical records to help to provide better and faster health services.

    This year, a British subscription, online medical consultation and health service launched Babylon. It offers medical AI consultation based on personal medical history and common medical knowledge. Users report the symptoms of their illness to the app, which checks them against a database of diseases using speech recognition. After taking into account the patients’ histories and circumstances, Babylon offers appropriate courses of action. The app also reminds patients to take their medication and finds out how they’re feeling.

    Molly was the first virtual nurse developed by Sense.ly, a medical start-up. It aims to help people to monitor their condition and treatment. Nurses use machine learning to support patients with chronic conditions in-between doctor’s visits. Customised monitoring and follow-up care is part of its service too, with a strong focus on chronic diseases.

    The AiCure app maintained by The National Institutes of Health monitors patients’ medication compliance. It uses a smartphone’s webcam and AI to confirm patients’ medication ingestion and helps them to manage their conditions.

    AI still has a long way to go. It will be exciting to see how Africa’s health systems adopt it in their eHealth strategies and use it to transform health and healthcare.

  • eHealth, Google and others are revolutionising healthcare in emerging markets

    Access to basic healthcare information is a challenge in many parts of the world. It’s especially demanding in developing countries. Google’s latest move in India may help increase access to healthcare information for millions of people.

    An article in The Market Mogul says Google identified this gap in Indian and has added health information to its Knowledge Graph. It’s a sematic search base that Google uses to supplement organic search results with summarised information.

    So, the next time someone in India uses Google to search common health conditions, it’ll show information cards illustrated with images. This information will include typical symptoms, details on how common the condition is, whether it’s critical, if it’s contagious and which age groups it affects. Google said that it’ll provide a condensed version if users have limited Internet connections. This goes some way to deal with India’s slow and intermittent mobile connections.

    An article in TechCrunch says Google’s initiative is in response to start-ups dedicated to democratising India’s healthcare. Lybrate, an online, web-based healthcare service’s an example. It aims to increase access to doctors and quality healthcare information. It’s app service allows users to ask doctors questions online, search surgeries nearby and make and manage appointments.

    Other developing countries benefit from mHealth initiatives too. Successful start-ups include Docway, Beep Health and Dr Vem! in Brazil. These use apps and the web to connect patients and doctors. Doctors have to be registered with the app, and  set their own consultation rates. Users can also browse doctor’s resumes before deciding to book appointments. Most users are parents looking for paediatricians. The next big group of users are elderly people with limited mobility.

    Are these online initiatives coming to Africa on a big scale? A more appropriate question may be when will they be available?

  • Big Data is not big in Africa’s eHealth - unpacking the 3rd Global Survey on eHealth

    As a relatively new part of eHealth, Big Data has a negligible effect on Africa’s health systems and eHealth programmes. Big Data insights are in Chapter 8 of the WHO and Global Observatory for eHealth (GOe) publication eHealth Report of the third global survey on eHealth Global diffusion of eHealth: Making universal health coverage achievable. WHO Global Survey 2015 provides the data source for the report.

    It hasn’t taken off globally yet. Fewer than a fifth of countries say they have a national policy or strategy for regulating Big Data in health and healthcare. In Africa, it’s about 2%. This Big Data deficit isn’t much of a cause for concern. As the eHNA posts about WHO’s report show, Africa’s health systems have many other eHealth priorities. One that wasn’t included in WHO’s survey is stepping up cyber-security. Acfee’s report from its African eHealth Forum (AeF) our priorities include  cyber-security and others, such as Interoperability (IOp), cloud computing, eHealth governance, regulation and capacity building are well ahead of Big Data.

    WHO found that a lack of integration, privacy and security are major barriers to Big Data adoption. It’s constructive that Africa’s health systems are focusing on these as part of their expanded eHealth initiatives. Acfee’s activities in 2017 will support them.