“Laura’s heart attack didn’t come with a warning” were the words of the television commercial voice-over as a woman opened a note that warned her about a heart attack in two days. The announcer went on to explain how the “Laura’s” doctor now had her on a plan to avoid another heart attack.
While the television ad cleverly pointed out how difficult it is to predict a health issue, the reality is that healthcare does have a tool that can improve diagnoses, treatments, and yes, even the likelihood of developing a disease or chronic condition.
Predictive analytics is employed in many industries to evaluate risk for a loan applicant, project length of life for a life insurance customer, or forecast potential sales in a specific region. Predictive analytics in healthcare is less effectively used, often limited to predicting outcomes or likelihood of readmission for populations versus personalized care for individuals.
Before the widespread adoption of electronic health records (EHR), the healthcare industry did not have the breadth and depth of data available for meaningful analytics. Although the advent of big data has changed the amount of information available to healthcare organizations, there are challenges to using clinical data, financial data, and health history information included in EHR data. However, these challenges can be overcome.
Healthcare data within a healthcare system is often found is disparate, siloed systems with limited integration and difficult access across business divisions such as clinical, financial, and administrative. The separation of information and the inability for a clinician to access multiple types of data to develop a comprehensive look at how staffing levels, processes, and costs relate to treatment efficacy and outcomes, limits the value of predictive modeling.
Even when a healthcare system is committed to integrating systems to enable interoperability, the data needed for healthcare predictive analytics is often buried in a myriad of structured and unstructured forms and formats that require specialized expertise to find.
While there are many talented individuals working in healthcare information technology (IT), the pressures on an in-house staff to manage and oversee day-to-day IT activities leaves little time to tackle implementation of a major technology overhaul — and today’s healthcare budgets leave little room to add IT staff or make major capital investments to replace legacy systems.
A cost-effective solution for a healthcare system is the use of a platform that runs “under” the myriad of applications and systems currently in use. By relying on one technology to connect, aggregate, integrate and harmonize data so multiple users can access it easily to establish patient registries, clinical data repositories, or ACO enablement. Serving as a big data repository, the platform can provide on-demand, self-service access to clean, quality data to which predictive analytics can be applied.
Relying on a third-party to manage the technology – ensuring up-to-date standards are in place and highly trained personnel are on hand to maintain the platform – frees in-house IT staff to focus on activities that provide the best return on investment for the organization.
The financial challenges faced by healthcare systems are not just operational in terms of staffing resources and capital funds for technology, but they also impact clinical practice.
Value-based health care initiatives incentivize healthcare providers and hospitals to ensure that the right treatment is provided at the right time and at the right cost. Clinicians who are uncertain about a diagnosis might over-order diagnostic procedures to gather as much information as possible. If the information in a patient’s medical or family history is not readily available, the provider might undertreat the patient, unaware that what appear to be symptoms of one condition are actually precursors of a hereditary predisposition such as Alzheimer's.
Predictive models that use all of the patient’s healthcare data to evaluate potential diagnoses and outcomes can streamline physicians’ decision-making — focusing on patient-specific diagnoses rather than applying the same risk to all patients, which ensures that high-cost therapies are provided to high-risk individuals.
For example, less than 0.05 percent of newborns have an infection confirmed by a blood culture, but 11 percent receive antibiotics. Kaiser Permanente of Northern California used predictive analytics to reduce overuse of antibiotics for newborns by developing a predictive algorithm using the mother’s clinical data and the baby’s condition at birth to guide OB/GYNS decision regarding the need for antibiotics.
The ability to access healthcare data across multiple sites and systems allows for more robust analytics systems that can ensure patients receive the best care and that providers are reimbursed appropriately. This not only produces healthy patients but also healthy hospital and health systems.
Just because there is a wealth of data, don’t assume that all predictive analytics in healthcare are created equal. Healthcare decisions are complex, requiring different information for each patient and each clinical decisions. For this reason, predictive models must be specific to the clinical decision the physician must make.
Algorithms that predict the risk of readmissions, for example, must take into account a variety of factors including the initial reason for hospitalization or emergency room visit, the capability to care for one's self, the need for home care, and the type of follow-up care required.
Parkland Health and Hospital System in Dallas, Texas, reduced readmissions for patients with heart failure by 26 percent with an EHR-based algorithm that identifies patients who are high risk for readmission. Once a patient is determined to be high risk, the clinician has a menu of evidence-based interventions that include education, telephone support, outpatient follow-up, and a physician appointment. By focusing the algorithm on one population (patients with heart failure) and tying it to clinical interventions proven to work with this population, predictive analytics in healthcare were proven successful.
Increasing Sources of Data
Healthcare data will continue to grow and will begin coming from sources outside the healthcare provider’s walls. Studies predict that in 2017, mobile health apps will be downloaded by 50 percent of smartphone users.
The use of mobile devices will not only be used by the general public to record health and fitness information, track diet and exercise, or record blood pressure, blood sugar levels and heart rate. Physicians are increasingly using mobile health technology to engage patients in their own care. By encouraging the capture and sharing of health data, physicians and patients collaborate on monitoring care and evaluating results, leading to greater compliance with physician-ordered medication therapies and better outcomes.
As the volume and type of data collected by EHR and other health provider systems grow, it is critical to be sure the organization has the infrastructure necessary to aggregate, integrate and harmonize data effectively. Ensuring the ability to share information across different locations, providers, and systems increases the viability of healthcare analytics and predictive algorithms that can inform clinical decisions that benefit patient care and health systems.
To answer the question posed in the headline: It is a fact that predictive analytics in healthcare have the potential to transform the healthcare industry. It is fiction that implementation of the technology needed to leverage big data for healthcare analytics is financially impossible. Effective use of predictive modeling and healthcare data to improve patient care is the future of health care.
What are your future predictions?