Design and implementation of a predictive risk model to improve oncology patient management
| Autor: | Whitney Church, Ashley Roper, Garrett Young, Tiffany Hoy |
|---|---|
| Rok vydání: | 2018 |
| Předmět: | |
| Zdroj: | Journal of Clinical Oncology. 36:154-154 |
| ISSN: | 1527-7755 0732-183X |
| Popis: | 154 Background: Methods to identify cancer patients that are likely to seek inpatient care for conditions that can be treated in the outpatient setting are not well defined. In light of the rise in value-based care programs, there is increased pressure on oncology providers to decrease costs of care while continuing to achieve high-quality outcomes. One way this can be achieved is by decreasing avoidable, high-cost utilization through population management. We sought to investigate the feasibility of creating a model to predict which patients were more likely to utilize emergency department (ED) and inpatient hospital (IP) services as well as the impact such a model may have on patient management. Methods: For a sample of 1093 patients, historical information was gathered across four domains: Cancer-specific Attributes, Comorbidity Burden, Social Factors, and Utilization History. This data set was then used to design a predictive model which assigned a numerical “acuity” score to each patient. Patients were ranked by numerical score, and their subsequent ED and IP utilization was measured. The developed model was then tested on an independent set of 591 patients to assess broader validity. After development, the predictive model was built into a tool for use in patient management by a team of nurse care managers. Results: Of the 100 highest-ranking patients in the training sample, 29% had at least one future hospitalization and 33% had at least one future ED visit, compared to 13% and 19% in the entire sample, respectively. In the validation sample, 11% of the 100 highest-ranking patients had at least one future hospitalization and 20% had at least one future ED visit, compared to 5% and 13% in the entire sample, respectively. Before model deployment, an average of 39 follow-up touchpoints were completed for high-risk patients per month. In the 6 months after deployment, the average rose to 171 touchpoints per month, an increase of 335%. Conclusions: Models can be developed to proactively identify cancer patients at higher risk of utilizing ED and inpatient hospital services. Implementation of such a model can also increase targeted outreach and communication with patients at greater risk of utilizing high-cost services. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|