Optimizing a Patient’s Discharge with Cognitive Machines
Optimizing a patient’s length of stay (LOS) directly improves health outcomes by reducing the risk of a hospital acquired condition and improving the overall patient experience. Moreover, by addressing LOS outliers—those patients who have longer than expected stays in the hospital—we lower the cost of care, improve resource allocation, and help to drive to improved financial outcomes.
LOS outlier patients in particular tend to use higher levels of resources across staffing, clinical testing, and pharmaceuticals than other hospital patients. In a recent study, LOS outlier patients accounted for only 5.4% of the study sample but accounted for almost 15% of the total hospital costs and 25% of the total inpatient days. But the challenge with LOS optimization overall is that it has multiple dependencies and complex interrelationships that flow across the hospital setting.
In a recent study, LOS outlier patients accounted for only 5.4% of the study sample but accounted for almost 15% of the total hospital costs and 25% of the total inpatient days.
The Cognitive Clinical Success Machine is uniquely capable of identifying outlier patients across inpatient episodes and delivering the recommended actions that will optimize LOS while reducing the likelihood of avoidable complications. The machine does this using the underlying Eigenspace architecture.
This approach enables the development of clinically relevant Eigen Spheres, which are used to find at-risk patients and extend to pin point the relevant clinical and socioeconomic factors driving the likelihood of an extended stay. A patient’s risk propensity and clinically relevant Eigen features are synthesized by the machine and translated into the specific patient-level actions that will help move what would have been a potential outlier to an optimized LOS. The machine does this day one of an inpatient stay regardless of the Diagnosis Related Group (DRG) and is able to derive the clinical conditions for the stay even if a working DRG is not available.