Artificial Intelligence and Machine Learning
The latest tools in Artificial Intelligence and Machine Learning (AI/ML) are transforming today’s landscape. But even as these tools mature, enterprises are still struggling to take advantage of them. One of the biggest hurdles they face is how to strategically implement machine learning to achieve business goals and optimize current processes. International Data Corporation (IDC) forecasts revenues from AI/ML systems worldwide will almost double to $12.5 billion in 2018, and keep growing at a similar rate until they hit $46 billion in 2020.
Knowledgent is an AWS Advanced Partner that is focused on innovating in and through data to impact your business using Machine Learning, Artificial Intelligence, Robotics, and Natural Language Processing. Knowledgent and its team of more than 300 Informationists is a pure-play data intelligence company with a focus on Healthcare, Life Sciences, Financial Services, and Commercial industries.
Our Impact in AI/ML
Click on the left or right arrows to browse our use cases.
“Knowledgent and Celgene have partnered to develop ML/AI-based applications across the clinical operations value chain with the aim to make significant reduction in clinical trial length.”
Vice President of Global Clinical Operations, Celgene
Patient accrual remains a huge challenge in clinical research, despite heavy spending on recruitment. According to a 2011 Tufts Center for the Study of Drug Development (CSDD) report, two-thirds of sites don’t meet the enrollment requirements for a given trial. Failing to meet these goals causes delays in clinical trials that end up costing even more money. Consequently any time patient recruitment can be accelerated and be made more effective directly translates into saving for the sponsor company.
Knowledgent was recently engaged by a Life Sciences company to improved patient recruitment in clinical trials, including identification of potential candidates, disease dispersion, and risk states. The solution provided was driven from real-world data analysis using AI/ML techniques and EMR for data collection and processing.
Member Retention and Acquisition
“As a large regional plan serving one in eight New Yorkers, using information in new ways is essential to improving the quality of care our members receive, and the cost-effectiveness of that care. Generating analytic insights using an increasingly diverse and exponentially expanding set of data inputs is key to getting those hard-to-reach benefits that can make a difference to our business and to our members. We use a variety of AWS services to pursue AI/ML-based analytic outcomes affecting member retention and acquisition, risk-modeling and fraud detection. Knowledgent has been a critical partner. They brought both architectural vision and boots-on-the-ground leadership to this effort, which is a major strategic initiative for us.”
Vice President Clinical Analytics, Healthfirst
As user bases expanded, marketing departments have found it nearly impossible to personalize user communications. A stop gap measure has been the use of segmentation is a marketing tactic, where similar users are grouped together, and then the communication is tailored accordingly to the group’s traits.
With the advent of the latest techniques AI/ML, personalization can be implemented in a smoother fashion. Simply put, personalization helps organizations retain customers. A one on one experience is one of the major factors when it comes to driving higher user engagement and increasing conversion rates.
Knowledgent was engaged by an Insurance Provider to improve customer retention and to detect fraudulent claims. The solution provided was implemented a series of AWS services (including EC2, S3 and EBS) and AI/ML tools including Python and Jupyter notebooks.
The issue of unnecessary hospital readmissions is now front and center in the national conversation about the quality of health care. Thanks to Medicare’s readmissions reduction program, hospitals are working hard to bring their readmission rates down, and that’s good news – good news being drowned out by a chorus of complaints.
Avoidable readmissions are a strong indicator of a fragmented health care system that too often leaves discharged patients confused about how to care for themselves at home, and unable to follow instructions and get the necessary follow-up care. Readmissions are also a costly price to pay for a system that doesn’t have resources to spare; Medicare alone reports spending $17.8 billion a year on patients whose return trips to the hospital could have been avoided. For this reason, working to minimize hospital readmissions is of paramount importance to bend the health care cost curve.
Knowledgent designed and developed models to predict Member Readmissions for a major health insurance provider in North Carolina. The system focuses on the Medicare Advantage population. During implementation an existing AWS powered Cloudera environment was leveraged to munge and transform data sets. Subsequently the clean data was submitted to a machine learning model built in Python to predict members most likely to be readmitted within 30-days.
Real World Evidence
Real-world data (RWD) and real-world evidence (RWE) are increasingly playing a larger role in making health care decisions. The FDA uses RWD and RWE to keep track of post-market safety and to make regulatory decisions. The health care community is using this information to facilitate coverage decisions and to come up with guidelines and to develop decision support tools for use in clinical practice. Medical device makers are using RWD and RWE to support clinical trial designs and studies to develop innovative, new treatment approaches. The 21st Century Cures Act, passed in 2016, puts a greater focus on using these types of data to further support regulatory decisions.
Over the course of the past few decades, computers have allowed us to gather and store vast quantities of health-related data. This data has the potential to allow us to design and conduct better clinical trials. Problems that previously seemed insurmountable are now possible. Furthermore, with the development of sophisticated, new analytical capabilities and tools, we can better process this data and to use the analysis results for better and faster development and approval of medical products.
Knowledgent designed, developed and deployed an advanced Real-World Evidence (RWE) analytics platform for a major pharmaceutical company. The platform is powered by an Amazon Web Services infrastructure and leveraged a series of technologies including S3, Redshift, Lambda, Informatica EIC and Tableau. Knowledgent enabled ingestion of Healthcare Claims, Electronic Medical Records, Lab Information and Clinical Trials data, mapped those data sets to the OMOP industry standard, and enabled analytics and visualizations for Patient Journey, Treatment Pathways, Drug Utilization and Switching Analysis. This platform will enable the firm to perform advanced patient analytics, at scale, to discover, develop and deliver life enhancing medications.
It is well recognized that the clinical response to drug administration varies widely between individuals and that most of this variability is pharmacokinetic in origin. In general, variability arises because of inter-individual differences in rates of drug absorption, drug distribution and elimination, either by metabolism or excretion. Variability of drug response is also a consequence of a variety of drug interactions which may influence pharmacokinetic parameters.
For this reason, pharma companies are keenly interested in personalized medicine and managing different reactions from patients to different drug protocols.
In this engagement, Knowledgent analyzed biomarker data to assess and predict drug treatment outcomes to address uncertainty about which candidates held the most promise. Six different classes of machine learning models were applied including CARTs, Random Forests, Linear Models, and Support Vector Machines. As part of the delivery a sophisticated user interface was provided to facilitate visualization of the insights.
Next Best Action
“Next Best Action” is a customer-centric marketing paradigm that considers the different actions that can be taken for a specific customer and decides on the ‘best’ one. The Next Best Action (an offer, proposition, service, etc.) is determined by the customer’s interests and needs on the one hand, and the marketing organization’s business objectives, policies on the other. This is in sharp contrast to traditional marketing approaches that first create a proposition for a product or service and then attempt to find interested and eligible prospects for that proposition. This practice, direct marketing, typically automated in the form of a campaign management tool, is often product-centric, and usually always marketing-centric.
Knowledgent has spent many hours at their labs to accelerate the deployment of these techniques.
For example, with one of our health insurance clients, we demonstrated how Next Best Action can enable run a highly targeted marketing, increased ROI on marketing spend, and created a more meaningful dialogue with the members. Machine learning models were built in SAS using SAS Enterprise Miner and marketing orchestration was enabled through client’s proprietary marketing orchestration and workflow management application.
In medicine, compliance (also adherence, capacitance) describes the degree to which a patient correctly follows medical advice. Most commonly, it refers to medication or drug compliance, but it can also apply to other situations such as medical device use, self-care, self-directed exercises, or therapy sessions. Both the patient and the health-care provider affect compliance, and a positive physician-patient relationship is the most important factor in improving compliance, although the high cost of prescription medication also plays a major role.
Worldwide, non-compliance is a major obstacle to the effective delivery of health care. Estimates from the World Health Organization (2003) indicate that only about 50% of patients with chronic diseases living in developed countries follow treatment recommendations. In particular, low rates of adherence to therapies for asthma, diabetes, and hypertension are thought to contribute substantially to the human and economic burden of those conditions.
During an engagement with one of the largest health insurance providers we identified members with low adherence to their treatment plan and create value-based partnership with care-providers to improve outcomes. Ensemble machine learning techniques were used to predict patient’s level of adherence. Supervised Machine Learning models were combined with unsupervised models to predict patient-level adherence down to the month. Machine learning models were custom-built for the client. In addition a dashboard to run reports was provided to visualize insights.