Predictive analytics is an area of data science that is getting a lot of attention in health care.
Predictive analytics offers a tantalizing solution to problems plaguing resource-restrained hospitals. Namely, if providers can predict which patients will be readmitted within 30 days, or who will acquire an infection in the hospital, they can apply scarce resources to those high-risk patients and change the predicted outcome. This has the potential to improve quality outcomes and lower costs.
Predictive analytics is made possible with the widespread adoption over the past five years of electronic health records. Providers can mine the EHRs for previous information on patients to predict future trends and outcomes.
Patient care is not the only area where predictive analytics is proving useful. For example, one of integrated health giant Kaiser Permanente’s state groups is using predictive analytics to reduce no-show rates for medical appointments. Kaiser Permanente Colorado launched a predictive analytics tool to identify patients who are at high risk of missing appointments. Kaiser then developed an interactive voice response and text messaging service to target those patients identified by the algorithm as being at high-risk of missing an appointment.
The results indicated that the validated predictive model was accurate in identifying patients at high risk of missing appointments, and the intervention reduced no-shows among those patients. Medical Care, the journal of the American Public Health Association, published the results.
With the proliferation of predictive analytics in health care, journalists should be prepared to ask hard questions about how this technology is being applied to patient care. Here are some ideas:
Data: What are the data sources for making predictions, and how reliable are those data sources? Does the health system gather data from various sources, or from one source? Is the system conducting data validation to ensure the accuracy of coded data, for instance?
Interoperability: Does the provider pull data from other hospitals or medical groups in the region? Is the outside data integrated into the patient’s electronic medical record? That’s because data from other systems on patient health history can provide a broader picture and improve the accuracy of predictions.
Social Determinants of Health: Does the system gather information about a patient’s housing status, food security and other social determinants of health? Again, the more data sources, the better the predictions.
Workflow: How are front-line caregivers learning about the predictions? How are predictions being integrated into their workflows so caregivers can respond to them?
Response Resources: How are resources allocated to respond to predictions? For instance, on avoidable readmissions predictions, does the hospital have ample care coordinators and outpatient services to intervene when a patient is at high risk for readmission?
Patient Satisfaction: Is the provider or hospital tracking patient satisfaction with the predictive model and resulting interventions? What are the results?
Predictive analytics makes a great story. Reporters can go beyond the hype to probe whether these predictions will help providers do their jobs better and ultimately improve health outcomes for patients.
Further Reading
- RockHealth report on predictive analytics
- Integrating predictive analytics into routine patient care, Harvard Business Review
- Predictive analytics roundtable, U.S. News and World Report
