Population Level Health Management and Predictive Analytics
There has been much discussion of population health management coupled with predictive analytics recently in the health care field. Why? Most who are discussing these topics see it as a means of improving the health of patients while reducing the costs of doing so. Providing better care at lower costs is becoming necessary as payers are beginning to pay for quality results as they move away from fee-for-service.
What is population health and how does predictive analytics fit in? Let me begin by defining population health and illustrative predictive analytics. In statistics, population reiterates the complete set of objects of interest to the investigation. For instance, it could be the temperature range of adolescents with ratios. It could be the individuals in a rural town who are prediabetic. These two are of interest in healthcare. Population also applies to any other field of research. It could be the income level of adults in a county or the ethnic groups living in a village.
Typically, population health management refers to managing the health outcomes of individuals by looking at the collective group. For instance, at the clinical practice level, population health management would refer to effectively caring for all the patients of the practice. Most practices segregate the patients by diagnosis when using population health management tools, such as patients with hypertension. Practices typically focus on patients with high costs for care so that more effective case management can be provided to them. Better case management of a population typically leads to more satisfied patients and lower costs.
Population health from the perspective of a county health department (as illustrated in last month's newsletter) refers to all the residents of a county. Most services of a health department are not provided to individuals. Rather, the health of residents of a county is improved by managing the environment in which they live. For instance, health departments track the incidence of flu in a county in order to alert providers and hospitals so that they are ready to provide the levels of care needed.
You should be able to see that the population which health is being managed depends upon who is providing the service. Physician practices' population is all the patients of the practice. For county health departments it is all residents of a county. For the CDC it is all residents of the United States.
Once the population is identified, the data to be collected is identified. In a clinical setting, a quality or data team is most likely the body that determinates what data should be collected. Once data is collected, trends in care can be identified. For instance, a practice may find that the majority of the patients who are identified as being hypertensive are managing their condition well. The quality team decides that more can be done to improve the consequences for those who do not have their blood pressure under control. Using the factors from the data that it has collected the team applications a statistical approach called predictive analytics to see if can find any factors that may be in common among those whose blood pressure is not well managed. For instance, they may find that these patients lack the money to buy their medication consistently and that they have trouble getting transportation to the clinic that provides their care service. Once these factors are identified, a case manager at the clinic can work to overcome these barriers.
I will finish this overview of population health management and predictive analytics with two examples of providers using the approach correctly. In August 2013 the Medical Group Management Association presented a webinar featuring the speakers Benjamin Cox, the director of Finance and Planning for Integrated Primary Care Organization at Oregon Health Sciences University, an organization with 10 primary care clinicians and 61 doctors, and Dr. Scott Fields, the Vice Chair of Family Medicine at the same organization. The title of the webinar was "Improving Your Practice with Meaningful Clinical Data". Two of the objectives of the webinar were to define the skill set of their Quality Data Team, including who the members were, and describing the process of building a set of quality indicators.
The clinics were already collecting a large variety of data to report to various groups. For instance, they were reporting data for "meaningful use" and to commercial payers as well as employee groups. They decided to take this data and more and organize it into scorecards that would be useful to individual physicians and to practice managers at each clinic. Some of the data collected was patient satisfaction data, hospital readmission data, and obesity data. Scorecards for doctors were designed to meet the needs and requests of the individual physicians as well as for the practice as a whole. For instance, a physician could ask to have a scorecard developed for him that identified individual patients who diabetes indicators showed that the patient was outside of the control limits for his diabetes. Knowing this, a physician could devote more time to improving the quality of life of the patient.
Scorecards for the clinic indicated how well the doctors at the site were managing patients with chronic conditions as a whole. With predictive analytics the staff of the clinic could identify which processes and actions helped improve the health of the patients. Providing more active case management may have been demonstrated to be effective for those with multiple chronic conditions.
Mr. Cox and Dr. Fields also stated that the quality data team members were skilled at understanding access, structuring data in meaningful ways, at presenting data to clinicians effectively and in extracting data from a variety of sources. The core objectives of the data team were to balance the competitiveness agendas of providing quality care, making sure that operations were efficient and that patient satisfaction was high.
A second example of population health management focuses on preventing cardiovascular disease in a rural county in Maine-Franklin County. Over a 40-year period, starting in the late 1960's, a volunteer nonprofit group and a clinical group worked together to improve the cardiovascular health of the residents of the county. As the project advanced, a hospital joined in the efforts.
At the beginning of the prevention efforts, the cardiovascular health of this poor county was below the state average. As volunteers and clinical groups became more active in improving the health of its residents, various cardiovascular measures improved significantly and actually were better in some respects than more affluent counties in the state that had better access to quality health services. The improvements were driven by volunteers who went out into the community to get those identified as being at risk of developing cardiovascular problems involved in smoking cessation classes, in increasing their physical activity and in improving their diets. This led to lower blood pressure, lower cholesterol rates and improving endurance.
The results and details of this 40-year effort in Franklin County has been published in the Journal of the American Medical Association in January 2015. The article is "Community-wide CVD prevention programs linked with improved health outcomes".
As you can see, a population level approach to healthcare provides effective results. A clinic can improve the outcomes of its patients with chronic diseases while balancing costs through improved efficiency by focusing on data at the population level. A community can improve the lives of its residents by taking a population level approach to preventive care. Population level approaches to healthcare are varied and can be very successful if population level theory is correctly implemented. Better results can be obtained pairing it with predictive analytics.