Tuesday 16th October, 2012
4:25pm to 5:05pm
Healthcare industry generates 100’s of TBs of data in any given month and is the most regulated and documented. There is also the complex challenge of many silos of data (with different data formats) from a multiplicity of sources – Clinical data, Prescription data, Patient profiles, Treatment data, and more. Therefore the volume and variety of health related data are Big Data challenges.
Much of these data contain valuable insights into patient behavior, prescription trends, drug response patterns, clinical data findings, disease characteristics, insurance claims etc. If these are successfully mined, they can lead to successful more accurate treatment recommendations and will lower the cost to serve by observing macro trends while keeping the ability to personalize the treatment. In the healthcare industry, in particular, providing personalized patient care is partly about understanding what happened (history); the other part of increasing treatment effectiveness lies in knowing what’s going to happen – in other words, understanding patients’ responsiveness to different treatment options are crucial. Both of these have strong correlations with user demographics, doctor’s criteria for prescriptions and other characteristics with the ultimate goal to provide the best possible care at the lowest cost.
Hospitals, healthcare providers and insurance companies are looking for the following big data analytics solutions:
- Operational data analytics - Claims data analytics: insurance providers want to understand patient behavior and match and compare patients in their network to predict their actions - Clinical data analytics: The objective here is for various entities in the healthcare system to use this data to evaluate and understand which drug is useful for which patient and why.
Traditionally the industry has used standard statistical and Bayesian approaches for data modeling and predictive analytics. But these approaches fall short.
Given the Big Data business context, in order to use past data and predict future events, there needs to be sufficient technology that is not merely extrapolating events. Powerful machine learning algorithms in association with combinatorial and graph algorithms are imperative to make accurate predictions about patient responsiveness to different potential treatment options, based on cross-dimensional correlations of diverse sources of data, time-event correlations and recognizing patient response patterns that would not be identified or understood using standard statistical, AI or Bayesian techniques. Handling the scale and the complex relationships within the data requires purpose-built algorithms that are specially designed to handle the needs of personalized patient care & treatment.
In this session, we will talk about methods that will extract insights from large volumes of data from various sources and diverse types, machine learning and combinatorial algorithms that can effectively break-down and correlate key data, and predictive analytics techniques that can result in personalized recommendations.
Experienced high tech marketer with specific interests in Big Data, analytics, BI, social media, cloud-based delivery models and data management. bio from Twitter
Sign in to add slides, notes or videos to this session