Posted by: Jenny Laurello
ACO, Big data, Disaster recovery
Big data is a hot health care topic, particularly when it affects decision-support systems, accountable care organizations (ACOs), best practices for addressing chronic conditions and identifying symptoms and disease-vector tests. Have you done the heavy lifting necessary for dealing with these issues in your organization?
If you answered no, and want to know where to start, ask yourself:
- Are you prepared to address master data management and data governance using semantics, as different departmental systems feed into an analytics engine that generates various meanings (i.e., semantics) for the same fields?
- Do you have the resources to create a data dictionary that defines the values, types, and formations of every data element across all of your departmental systems?
- Do you have the data scrubbing and enhancing capabilities to address the fact that many fields driving your indexes won’t be filled in, and sparse or unscrubbed data won’t allow you to build an efficient database?
Next, consider the importance of “big pipes,” because big data demands them. You’ll need them to architect a system that can:
- Handle bulk data loads
- Move data from source systems (e.g., electronic medical record (EMR) systems) to your departments’ and your partners’ systems
- Securely back up data
- Maintain a disaster-recovery plan
- Start, stop, and re-index all data on a regular basis
The following questions might help you think through what’s needed:
- Who will have access to your big data as it arrives in the database, and before and after it moves through these big pipes?
- Will you de-identify data by removing personally identifiable information from it before or after it enters your analytics engine?
- Will you be able to re-identify the data in the event that you need to find and validate a patient with a particular set of circumstances, conditions or symptoms? (Remember, you’re analyzing a mammoth amount of data. Statistical anomalies will arise, and you’ll need to be able to identify the source data, the original system that provided the data, and even the original patient on demand.)
When it comes to health care data, how do you align the present and the past? This is an item that is rarely discussed. How do you enable a clinician to use the aggregated, statistically aligned, scrubbed information about symptoms and best practices and apply that past information to a current patient?
These questions involve both data governance and identification of the axis of interest. You must ask:
- Are you matching patients based on demographics (e.g., body mass index, gender, age brackets, blood pressure ranges, blood sugar ranges)?
- What did you record in your earlier data reports?
- How are you going to match that data, especially if yours is an old institution and your original records are organized around one data value, and your latest system is organized around a different data value?
Actually applying the recommendations to achieve best practices requires a detailed, consistent, and managed approach, similar to how you match the past to the present. But when will we insert that past information? Are we going to insert it before the clinician sees the patient (i.e. provide recommendations at the beginning of the encounter)? Are we going to apply recommendations as the clinician is formulating a diagnosis or after the diagnosis has been made?
The post-diagnosis application is what we often see when ACOs manage certain client conditions like diabetes. ACOs know certain things can be done to provide the best outcome for a patient once they’ve been diagnosed. The data is provided to a care manager who ensures the patient gets a glaucoma test, has their circulation checked, and gets any cuts or wounds evaluated.
Applying recommendations made while a clinician is forming a diagnosis must be inserted into the patient encounter in real time, which means the clinician’s preference for working with the recommendation and seeing the data live must be accommodated.
If you hadn’t realized it before, you may now appreciate the issues that go along with all that hefty data we’re collecting, analyzing, and using to help inform our health care decisions, plans for patient encounters and more. You can see why the issues call for strong fundamentals in IT infrastructure data governance, robust pipes that can move the data back and forth while recognizing HIPAA requirements, the capacity to use past data effectively in a present situation, and the ability to align and integrate big data engines with EMR systems.
Don’t be afraid of the heavy lifting. It may not be easy, but it’s definitely worthwhile, like many endeavors in life. Instead, become an Olympic infrastructure weightlifter for your organization. Be the architect who will be remembered for laying the big pipes that revolutionized the way the organization handles big data. Most importantly, strive to apply that big data to improve the current and future decisions you’ll soon be making for your patients. They’re depending on you.