4 Steps to Utilize A Hospital's Own Patient Data to Improve Self-Pay Collections

Healthcare spending has dropped to some of the lowest levels recorded, but out-of-pocket expenses for patients have soared. A recent report from the Health Care Cost Institute says that out-of-pocket expenses have jumped 4.8 percent to $768 per person per year. When considering older adults 55 to 64, out-of-pocket costs hit $1,265 per year.¹ High-deductible health insurance plans and co-payments have triggered the shift to greater self-pay (the amount owed by a patient), and consequently, these collections have become a larger percentage of a hospital's accounts receivable. According to research by the Healthcare Financial Management Association, more than one-third of hospitals surveyed saw their self-pay receivables increase by 10 percent or more year-over-year.²

Typically, hospitals rely heavily on third parties (collection agencies and credit bureaus) for credit data and outsourced account follow up. With tighter margins and regulatory pressures, hospitals need a more intelligent and cost-effective way to drive self-pay collections.

Technology innovation, like predictive analytics, can help hospitals leverage their own data to gather the right information for a more effective and efficient process for self-pay collections. By leveraging hospital's patient claims data, predictive analytic models can accurately estimate a patient's propensity-to-pay with accuracies of 90 percent or higher. These probability scores provide the analytic horsepower to increase yield, allowing hospitals to strategically align resources, focus on the right accounts at the right time, and ultimately improve the patient experience by proactively identifying alternative payment options for those in need.

Here are four steps to utilizing a hospital's patient data asset for more accurate and efficient self-pay collections.

1. Build and maintain a patient big data source. The first step to conducting any advanced data analytics is integrating and aggregating data sources from disparate hospital systems (financial, clinical and patient) into a data warehouse.

Specifically, data elements that link to the patient are collected including account, transaction, claim, diagnosis, procedure, charges, contract terms and reimbursement, as well as payment history. These pieces of information must be cleansed and validated, for example, removing duplicate records and verifying financial data is properly expressed. In addition, account-level transactions and interaction data need to be linked at the patient level to view historical care, charge, reimbursement and payment patterns — an essential step for accurate self-pay modeling.

2. Integrate all data elements relevant to the patient's payment history and ready for activation. The patient's payment history is the single most important factor in determining his or her propensity to pay the out-of-pocket portion of the hospital bill. All source data elements related to each patient must be linked together to generate a view of the patient currently and tracing back over their historic visits. At each of these points in time, the following crucial information must be extracted from the source data: patient visit detail, knowledge of the payer at these different points in time and insight into the reimbursement logic for those payers.

Once these elements are extracted and organized in the hospital data warehouse, it is ready to be activated with predictive analytics.

3. Apply predictive analytics to estimate the patient's propensity-to-pay score. The data elements that show historic self-pay amounts and the extent to which they've been paid are the raw material necessary to determine the factors that relate to a patient's propensity-to-pay to a bill. The model first cultivates a dataset containing historic examples of patients who have paid their portion of their healthcare expenses and those who have not. Predictive modeling algorithms then sift through the myriad of patient factors (e.g., the type of visit, diagnoses, procedures, financial breakdowns, etc.) to find those that indicate the propensity to pay.

Selecting self-pay predictive models that compute a propensity-to-pay score are needed to make the self-pay analytics actionable. For example, consider a model that, when given data about a patient, their current encounter and self-pay amount, provides a value between 0 and 10 where 10 means "very likely to pay" while 0 means "very unlikely to pay." This is precisely the information that fuels the decision-making process allowing hospitals to strategically focus resources, increase yield and improve patient experiences.

4. Leverage actionable insight from self-pay analytics to improve collections. Armed with propensity-to-pay scores, hospitals now have the high-value information to identify and implement strategies to improve collections.

The first benefit is to drive collections at point-of-service. By utilizing the predictive model and advanced workflow, the hospital can initiate collection efforts on a patient's open balance at time of registration or scheduling for an elective visit. Providing this intelligence across the patient access function at a hospital allows administrators to provide appropriate financial counseling.The second benefit aligns with those patients who are responsible for their entire bill. For these cases, self-pay analytics and workflow can proactively assist in identifying alternate insurance eligibility and payment options for those in need.

With the right technology innovation, like predictive analytics, hospitals can leverage their own existing information and resources to improve both their financial and clinical performance. Organizations can more strategically utilize internal and external resources for greater revenue yield, improved decision-making and more cost effective patient care.

Paul Bradley, PhD is chief data scientist at MethodCare, where he oversees research and development functions, including the development of new processes, technologies, and products. Dr. Bradley earned his PhD and MS degrees in computer science and a BS degree in mathematics from the University of Wisconsin.