The Penn Foundation, a regional Pennsylvania Behavioral Health provider, struggles to meet demand for inpatient beds designated to treat substance abuse. Furthermore, this facility, like others within the industry, has the goal of minimizing the number of patients who Left Without Completing Treatment (LWCT). Aside from the negative financial implications associated with patients leaving prematurely, the Penn Foundation wanted to optimize clinical resources by admitting patients with a higher likelihood of completing treatment or at least identifying patients at higher risk of not completing inpatient treatment and thus treating them accordingly. An attempt to identify the causes behind this program abandonment led the facility to FoundationDx.
A healthcare analytics company, FoundationDx leverages machine learning for business performance and quality optimization in underserved and complex data environments. Many common analytical tools are only effective when the number of outcome factors is limited and easily identifiable. FoundationDx created a process to leverage machine learning on a small scale so that all factors associated with outcomes of interest are automatically identified and presented according to their level of contribution to the outcome. Factors associated with poor outcomes (in this case LWCT) can be used as part of an admission screening checklist in advance of more automated ways to integrate risk into the admission screening process. FoundationDx, for example, identified the fact that patients admitted on Wednesdays were at greater risk of LWCT. This insight led the staff to realize that patients typically feeling better after 72 hours would leave the facility when staffing was at its lowest on the weekend. Staff was alerted so that mid-week admissions could be better monitored on weekends. Other factors such as employment status, marital status, living arrangement, age, gender, and the like were also used in the learning process.
Learning that occurred over a four month period showed an opportunity to reduce the LWCT rate by 30 percent in a subsequent four month period by identifying program candidates that fit into a lower LWCT risk category.
Cases as such are indicative of FoundationDx’s commitment to enhancing business performance and quality optimization by delving deep into complex data environments and presenting useful information that is not easily identified. As healthcare providers focus on offering quality service and experience, FoundationDx closes the quality feedback loop by identifying opportunities for tactical change within a larger quality process. This is made possible through frequent learning on periodic datasets accompanied by quality measurements. The company also identifies factors leading to undesirable HCAHPS (patient satisfaction) survey results. Example findings such as specific service line, unit, discharge time and disposition have been associated with poorer quality scores. The company aims to enhance various quality ratings of its clients while at the same time providing a better experience for patients. “We optimize ongoing outcomes by embedding our technology in the space between a best practice process and ongoing measurable results by focusing on automated tactical adjustment.” states Michael Alterman, co-founder of FoundationDx.
JR DeFeo, co-founder of FoundationDx, adds, “We have created a niche in the industry by providing valuable insight into what drives quality measure outcomes using readily available and relatively small datasets.” The company’s strategy-based approach to solving its customers’ quality problems involves analyzing periodic data associated with a process to distinguish potentially problematic process factors. As tactical change is brought into the continuous learning process, client outcomes improve over time.
The company can take on smaller creative projects overlooked by others and test the opportunity for machine learning to positively impact outcomes. While FoundationDx is focused on the quality and patient experience initiatives in acute care, it also sees opportunities in non-acute care settings like the behavioral health space, which have had limited exposure to leading-edge technology. The company recognizes important aspects of healthcare that go beyond the scope of basic medicine and uses small data sets to provide more value in value-based care.