Now that BigDip is over, it’s time to share some of the key
insights shared during the conference. Some of the top thought leaders in the
industry shared their insights with the attendees. If you attended, please add
your takeaways in the comments below. If you were not able to attend, enjoy
this recap and feel free to participate in the discussion as well.
* * * * * * * * * * * * * * *
- Review current publications
concerning your products or drugs in development to understand how comparative
groups are being assigned today.
- List any gaps that may exist in
current research that could be addressed via observational data research
provided the right data elements are collected.
- Consider the value of closing
these gaps via a registry environment only if the right data are not collected
at the right times in the right patients for the right products.
-
Billy Franks - Astellas
* * * * * * * * * * * * * * *
Complex Business Problems Drive Analytics Innovation.
Enable Disruptive Information by following these three principles:
- Shared: The biggest
difference in practical analytics may not be in how we analyze or interpret
results but rather how that information is shared and consumed by all of
us.
- Enabled:
Data/analytic portals and interactive analytic tools available to business
leaders, analysts and scientists alike.
- Transparent: Information isn’t
sequestered in static reports/memos and data isn’t hidden in disjointed data
warehouses but rather is open and accessible to everyone.
-
Christian Nimsch - SAS
* * * * * * * * * * * * * * *
- Get Parallel - The difference
between a lot of data and big data is parallel processing. Via Netezza, sharded
SQL, Hadoop, Apache Spark, etc. If you are not yet in an environment where you
can do parallel processing, you will soon run out of capacity.
- Get to the Decision - If you can
apply your analytics where you make the most decisions, you will generate the
most value.
- Greg Hayworth, Humana
* * * * * * * * * * * * * * *
- In healthcare analytics a focus on
causal conclusion is key.
- Maximize the use of the
longitudinal data structure healthcare records provide.
- Three keys that will unlock the
potential of healthcare database networks: size, depth, and speed.
- Sebastian Schneeweiss – Harvard Medical School
* * * * * * * * * * * * * * *
- Evaluate how quickly you are
getting answers to critical questions.
- Look at the bottlenecks in data
processing and ask how well your system can scale if data volumes go up by a
factor of ten or 100.
- Honestly assess what the total
cost of your system and its upkeep are, and ask if it has become unwieldy.
-
Brian Bradbury - Amgen
* * * * ** * * * * * * * * * * * * * *
- Think of patient data as the new
resource that will power drug discovery in the next decade. Are your systems
ready?
- Everyone should at least have
pilots on how to manage large patient data. Genomic datasets are expected to
grow beyond 100,000 patients, but current systems are not ready to handle even
a few thousand patients.
- Push the edge on data analytics
and facilitate the combination of knowledge across disciplines (chemistry,
biology, genetics, statistics etc.).
-
Francesca Milletti - Roche
* * * * * * * * * * * * * * *
- No development of any RWE BIG DATA
platform is complete without the ability to provide an end- to -end view of the
patient journey.
- Think about, conceptualize and
address “confounding by indication” & “ selection bias” every time real
world BIG Data platforms are being developed and analyzed.
- BIG Data stacking in R&D spans
different methodologies that need to be coordinated simultaneously for any
reliable analytic exercise.
-
Usman Iqbal - AstraZeneca
* * * * * * * * * * * * * * *
- Decide beforehand what are the
questions that you plan to ask of the data. This will determine how you should
store your data as the performance of the storage schema is impacted by the
users queries.
- Context is crucial - in many cases
a single sensor may not suffice. Use multiple sensors to complete the picture
- Data is dirty and outliers may
skew your results. Ensure that you have a set of rules in place that will
enable you to differentiate between "good" and "bad" data.
-
Yadid Ayzenberg - MIT
* * * * * * * * * * * * * * *
- To assure meaningful patient
centered outcomes, plan for patient input through all stages of measures
development, including concept elicitation, determination of item language and
relevance, and testing.
- Consider differences in patient
preferences in the design and logistics of clinical trials based on condition,
severity, demographics, and other patient level variables.
- The potential utility of wearables
is promising but requires a carefully considered engagement experience to
enhance collection of meaningful data.
-
Emil Chiauzzi – Patients Like Me
Posts
