Obamacare: Predicting Outcomes
With the October 1 opening of the Obamacare Health Insurance Exchanges swiftly approaching, we’re about to experience a major shift in the way our country’s healthcare system works. For better or worse, there are definite changes that are already starting to take place to accommodate the new healthcare plans. Besides shifting the way health insurance is distributed, the Affordable Care Act (ACA) is also introducing incentives to improve outcomes and patient safety. These incentives are based on the quality of visits rather than the quantity, and introduce penalties for care that isn’t as valuable as it could be. These changes constitute new and expanded uses for predictive modeling within various fields related to healthcare. Here’s how predictive modeling will be useful during the transition into Obamacare and beyond in four use cases:
30 Day Patient Readmission
With new ACA guidelines, Medicare and Medicaid are penalizing hospitals that have a higher than expected 30-day readmission rate by withholding reimbursements. As of last year, Medicare was spending about $17.5 billion annually in hospital bills related to readmitted patients. Their penalties (which were actually implemented last October) are levied by withholding one percent or more of every payment for a patient stay which adds up to millions annually. The money that comes from the penalties is being held in a bonus pool so hospitals that do better than average or show great improvement can be rewarded. By building models to predict readmits, hospitals can get a handle on which patients are likely to be readmitted to a hospital within 30 days of discharge. This will allow them to intervene in cases that are likely to be readmitted to make sure each patient is getting the best care possible and won’t be readmitted – a win-win for both the hospital and the patient.
Predicting Patient Health
What if providers were able to identify the populations most at-risk for commonly managed diseases? With the introduction of the HITECH act, providers are required to use and maintain electronic health records (EHRs) with a standardized set of data points, making this analysis possible. Researchers at Columbia University and MIT have attempted to target the populations who are at-risk for commonly managed diseases by using EHR data on gender, ethnicity, prescriptions and other medical history to predict an individual’s health over time. Because these things weren’t necessarily being tracked or reported previously, the amount of data available for analysis is growing exponentially, which means we can start analyzing new areas of overall health. Now, when providers encounter a patient with a high risk trajectory, they can do their best to prevent infection and administer targeted interventions more efficiently. By targeting these patients more specifically, they can improve patient outcomes and provide more quality care with limited stress on hospital resources, as well as create a more positive population health outcome.
By accurately forecasting patient inflow, hospitals can schedule nurses to meet the expected patient demand. It’s especially important to have a handle on the scheduling of nurses in order to distribute resources in the most efficient way possible. If we know that a hospital is historically very busy on Saturday mornings, but not so busy on Monday evenings, that information can be leveraged into a predictive model to tell that hospital what the expected patient inflow will be for a given time frame. Accurately predicting inflow allows the hospital to adjust nursing shifts accordingly which reduces cost; taking potential overtime into account, the savings could add up to millions of dollars annually. Additionally, this increases the overall efficiency of a hospital, and because the workload is more balanced for nurses, can increase patient satisfaction.
Health insurance fraud is a major problem, estimated to total between $48 and $98 billion for Medicare and Medicaid alone. It is expected to compound as more people are enrolling in the insurance exchanges. Fraud models are being implemented to freeze payments to providers with suspicious Medicaid claims before they are paid out. These models assign a risk score to each claim, and by targeting the high risk claims, investigators can find fraudulent claims prior to paying them, and build a case to send to an attorney general or an auditor for investigation before paying. Before the ACA, many systems detected these trends through a post-payment analysis after they occurred. Targeting fraudulent claims earlier in the process prevents providers from taking advantage of the system and frees up resources for people who have legitimate claims.
As our healthcare program shifts towards outcome-based compensation, predictive modeling will become more important in all aspects, for hospitals, staff, and insurers. In addition to the changes above, the increased availability of EHRs means that the data available for building such predictive models will increase exponentially into big data territory. With the additional data from EHRs, we can increase model accuracy, resulting in more specific targets to drive efficiency in various arenas within healthcare. This new data means we’re at the tip of the iceberg as far as predictive modeling in healthcare goes; the more data we have available, the more specific and diverse models we can start to build to affect outcomes. With billions of dollars at stake, it is more important than ever to make accurate forecasts to drive more data-driven decisions, and predictive modeling is one technique that will help for a more efficient distribution of resources throughout the healthcare system.
Harness the power of many.
Create and share reports and datasets across the enterprise, and put analytical power in the hands of everyone. Veera creates a truly data-driven culture. Try it for yourself today.