Predictive Big Data Analytics for Personalized Patient Care and Response Prediction in Cystic Fibrosis ?

Here’s are the excerpts of a talk published by Medscape on  personalized patient care and prediction of responses in Cystic Fibrosis.Direct link to the talk gives a registration page.Hence copied and pasted here verbatim.An audio version of the talk is embedded here.

Dr. Boyle: Scott, at this meeting we have heard a lot of exciting things about cystic fibrosis transmembrane conductance regulator (CFTR) modulation, and some of the other new therapies that are coming along. Tell me some of the things you have learned about here at the conference that you believe will affect the way you are going to take care of your patients in the future.

Dr. Donaldson: This meeting struck me as being not only about randomized controlled trials, but also about raising new ideas to explore how we take care of patients. One especially interesting area was that we really need to personalize our approach to patient care of individuals with cystic fibrosis (CF).

Dr. Boyle: Absolutely.

Dr. Donaldson: For instance,  ” clinicians are looking into ways of predicting how patients will do in the future, especially those on the younger end of the age spectrum. One approach is to use high-resolution CT scanning. The group from Wisconsin has shown that having a CT scan early in life predicts what lung function will be several years down the road. That kind of information would be very useful in identifying patients who might have a worse outcome so that we can be more aggressive with them and, hopefully, provide a better outcome. “

Dr. Boyle: Is that something you imagine using in your own patients?

Dr. Donaldson: As an adult pulmonologist, I think it probably applies more to pediatrics, but it is a very attractive approach to use that information early in life, and right now we do not have many other good outcome measures.

” Another common theme of this meeting is using existing databases. Doing so will enable us to look at patients early in life and predict how they are going to do down the road. Jessica Pittman and the pediatric group at the University of North Carolina looked at a large group of patients, and they are beginning to pick out those factors that predict a worse outcome. ” Some of the findings were not so surprising: poor nutritional status, more symptoms, almost any typical CF pathogen. Perhaps one surprise is that nonmucoid Pseudomonas was not a predictor, whereas mucoidPseudomonas was. It warmed my heart to learn that one therapy, hypertonic saline, seemed to predict a better outcome. Perhaps it is an early hint, at least from this study, that maybe we can be doing good things by providing hypertonic saline early in life.

Dr. Boyle: And some of those risk factors are things we can address early on.

Dr. Donaldson: That is really the point. That is right.

Dr. Boyle: As clinicians, we spend so much of our time focused on exacerbations. Was there anything at the meeting that struck you about that?

Dr. Donaldson:” More and more data are accumulating, such as that from the Toronto group, showing that frequent exacerbations have a profound impact on outcomes.”  As a community, we are thinking more about how best to take care of patients having exacerbations. There are still some basic questions that we really have not even started to answer, but at this meeting we are learning that different durations of therapy might be better than others. One comparison showed that 2 weeks appear to be better than 1 week; that’s perhaps not surprising, but it is good to have the information in hand. It still leaves a lot of questions unanswered. Would 3 weeks be better than 2? Are 10 days okay? The more important question remains: How can we personalize it to an individual patient?

Dr. Boyle: What do you usually do in your practice in terms of length of therapy?

Dr. Donaldson: We do try to personalize it. I like to get a lot of lung function measurements. My treatment goal is to have lung function return to baseline. Not every patient achieves that, so sometimes you wind up treating for a very long time hoping that you will get there [with patients]. But that’s certainly one of my strategies.

Dr. Boyle: What other factors might you consider in personalizing treatment?

Dr. Donaldson: One interesting poster and talk at the meeting was the use of C-reactive protein (CRP) as an inflammatory biomarker.[4] Reduction in CRP during exacerbation correlates with time to the next exacerbation. That’s another piece of information that would help us figure out how long we need to treat our patients.

From big data analytics point of view,this discussion between two eminent healthcare practitioners proves that by leveraging the analytic tools at our disposal ,we can certainly make a huge impact on personalized patient care and response prediction.

Our Team @ BigDataCognition will be attending ‘Hack For CF” on October 5th.We hope to make an impact.

Quantified Self ,Personalized Medicine & Big Data

The Quantified Self is a movement to incorporate technology into data acquisition on aspects of a person’s daily life in terms of inputs (e.g. food consumed, quality of surrounding air), states (e.g. mood, arousal, blood oxygen levels), and performance (mental and physical). Such self-monitoring and self-sensing, which combines wearable sensors (EEG, ECG, video, etc.) and wearable computing, is also known as lifelogging. The primary methodology of self-quantification is data collection, followed by visualization, cross-referencing, and discovery of correlations. If we add genetic data to all the social and personal medical sensors data then the dream of data driven personalized medicine in healthcare is not too far.

Personalized medicine or PM is a medical model that proposes the customization of healthcare – with medical decisions, practices, and/or products being tailored to the individual patient. Traditional clinical diagnosis and management focuses on the individual patient’s clinical signs and symptoms, medical and family history, data from laboratory and imaging evaluation to diagnose and treat illnesses. This is often a reactive approach to treatment, i.e., treatment/medication starts after the signs and symptoms appear. A proactive approach using big data technologies and application of predictive analytics can help us take effective preventive measures.

Life and health tracking information from various applications can be difficult to integrate and find meaning. But innovative and readily available open source technologies have delivered efficient tools in the hands of data scientists. Now we can combine, visualize, and analyze the cacophony data to identify correlations between specific input factors and treatments (independent variables) and psychological or physiological outcomes (dependent variables).

Today, creative data scientists can arrive at amazing conclusions about how various factors correlate with one another and affect a person’s psychology or physiology. We can combine this insight and apply predictive algorithms to build models, tools to predict how a person would respond to a device, situation, treatment or medicine. Open source innovators behind path breaking  frameworks such as Hadoop, Hive, MapReduce, Yarn, KNIME, WEKA and Apache Mahout have built a set of components that can help us achieve the goals of personalized medicine.

Learning all these technologies, building the analytical tool from scratch and then running data analysis is a set of daunting tasks. The team at Big Data Cognition  is helping  data scientists face this challenge. We have integrated all these technologies together  to build a single platform, Saarus. Using Saarus,data scientist can focus on creative use of data without wasting precious time and effort on the technology plumbing.

Saarus, as is the case with all it’s component technologies, remains open source and is designed to make life easier for data scientists. It helps data scientists harness the capabilities of all these technologies and provide a cost-effective graphical analytic work bench. Big Data Cognition is committed to work on integration of all promising statistical algorithms and technologies to help healthcare data scientists achieve the goal of personalized medicine and preventive care.