In an article posted earlier this year on this blog I argued that hospitals have traditionally done a sub-par job of leveraging what has now been dubbed “big data.” Effectively mining and managing the ever rising oceans of data presents both a major challenge – and a significant opportunity – for hospitals.
By doing a better of job connecting the dots of their big data assets, hospital management teams can start to develop the crucial insights that enable them to make the right and timely decisions that are vital to success today. And, better, timelier decisions lead to improved results and a higher level of quality patient care.
That’s the good news. The less than positive story is that hospitals are still way behind in using the mountains of data that are being generated within their institutions every day. Nowhere is this more apparent than in the advanced data management practice of predictive modeling.
At its most basic, predictive modeling is the process by which data models are created and used to try to predict the probability of an outcome. The exciting promise of predictive modeling is that it literally gives hospitals the ability to see into (and predict) the future. Given the massive changes and continuing uncertainty that are buffeting all sectors of the healthcare industry (and especially healthcare providers), having a clearer future view represents an important strategic advantage for any hospital leader.
So, why aren’t more hospitals using predictive modeling to support their decision making and strategic planning activities? There is no question that the hospital industry stands to gain a lot from using these advanced analytics methodologies to gain a better understanding of future trends and needs. Indeed, most other major industries are already active users of predictive modeling, and have been for a number of years now. In the healthcare sector, this includes payors, who have been using predictive modeling to map out where the industry is going so they can make the right moves regarding marketing, risk management and pricing.
Hospitals Lack Consistent Data ‘Plumbing’
One of the principal reasons why hospitals have been so slow to leverage predictive modeling goes back to the data “plumbing” problem that I mentioned in my earlier blog post. Hospital data “management” really is all over the map, with something like 90 percent of those data not even kept by hospitals (according to the McKinsey Global Institute big data study). The technology “plumbing” that formats and stores these datasets can be strikingly different both within and between hospitals, creating huge data integration problems and disconnects.
Add to all of that the fact that hospitals are adopting proprietary data management systems from competing technology vendors, and I think the problem becomes much clearer. Hospitals just don’t have the consistent data types and formats that are essential for creating models that help predict outcomes within the hospital itself, and in relation to changing market conditions.
A growing array of institutions has already attempted a limited version of predictive modeling. This approach has been ineffective because it is based on a fatal flaw: Using past trends to predict future ones. It is like the legal proviso that accompanies every financial investment you make: “Past performance is not a predictor of future results.” Given the market disruption from healthcare reform, we all know that the past will not be a good predictor of the future for hospitals.
True predictive modeling is not just about making decisions based on what happened before. It is predicated on creating new models that are built upon a wide variety of data inputs, which are then “crunched” and analyzed with custom algorithms. That data-driven, forward-looking methodology is why predictive modeling really does work in helping management to see where the future can and probably will go.
Start With the Low Hanging Fruit
All of this does not necessarily foretell a bleak future for predictive modeling in hospitals. Actually, there are more and more positive signs to point to in our industry. As I write this, there are enterprising hospital leaders who are working with their technology partners to overcome the aforementioned structural and technological impediments. In doing so, they are making progress in figuring out how to identify, develop and format the datasets that are needed to drive effective predictive modeling initiatives.
Some of the “lower hanging fruit” opportunities that are especially ripe for predictive modeling include the following key questions that confront hospital leaders today:
·Where is the next infectious disease hot-spot going to occur in our hospital?
·How do we expect utilization to change with the Medicaid expansion that is part of the Affordable Care Act?
·How can we predict patient demand and minimize our use of contract nursing?
As an example of this last question, a medium-sized hospital system in Florida modeled its patient demand and thus was better able to predict where and when they would need various nursing roles. This predictive modeling approach has allowed the hospital to save millions of dollars in higher expense contracting nursing utilization.
With the historic risk shifts now taking place in the healthcare provider industry, hospitals increasingly will need the predictive capabilities that they clearly don’t possess today. This is not a matter of just using an “Excel” spreadsheet to crunch some numbers. The predictive modeling capabilities that I have been describing here represent a quantum leap over the analytical and planning practices of a great many hospitals. That said, today’s drastically changed healthcare landscape demands an equally large leap by hospitals – in how planning gets done, decisions are made, and changes implemented.
We now face an emerging “digital divide” in the healthcare provider market. In this unfortunate scenario, 10 percent of the larger, more sophisticated hospitals and health systems will have the resources and capabilities to use sophisticated data analytics – like predictive modeling – to guide their critical long range planning. The other 90 percent of hospitals will be severely disadvantage because they lack the resources to deploy advanced data management programs. This growing divide is indeed troubling for the health system, and society overall.
It does not have to be that way. There are resources out there that can help hospitals of all sizes and types to leverage state-of-the-art predictive modeling to strengthen strategic decision making for today and tomorrow.
Ultimately, the ability to use sophisticated and powerful predictive modeling capabilities will be a true strategic advantage for hospitals. And that clearly benefits the hospitals’ patients and other key stakeholders, as well as the communities they serve.
Russ Richmond is the CEO of Objective Health, part of the global McKinsey Healthcare practice, which serves hundreds of public- and private-sector organizations worldwide. He is passionate about the use of data to manage health and to improve healthcare performance. Russ served over 50 hospitals as a McKinsey consultant, across the USA, Europe, and Asia.