For more on predictive analytics for healthcare, attend Predictive Analytics World for Healthcare New York, Oct 29-Nov 2, 2017.
With widespread usage of electronic medical records, big data has truly arrived in healthcare. In addition to the volume of data, there is considerable variety as data is extracted from clinical services, health insurance claims, administrative functions, and patients themselves. There is even exploration into acquiring social and behavioral data from patients, such as illicit drug use, credit scores, criminal records, and other socioeconomic indicators. With so much data, it can be hard to know where to begin.
Right now, most healthcare institutions compile data into reports so they can comply with quality reporting regulations. This is time consuming work, and it can dictate how you use data. Before you know it, the goal becomes managing the data in these reports without gaining any meaningful insights and the purpose is lost. You can do so much more with data, and it starts by deciding how you will prioritize the use of the data you have. Here are five questions you should be asking your predictive analytics professionals about population health to start gaining control over your data so that you can genuinely improve healthcare quality and reduce costs.
1. Which patients are at the greatest risk of further health decline?
One of the most powerful insights you can get from data is risk stratification. By using predictive analytics to identify which patients are the most likely to get sick in the near future, you can start to plan how you will mitigate their decline before it happens. According to the 2012 Medicare Chartbook [1] on chronic conditions among Medicare beneficiaries, Medicare patients with 0 or 1 chronic conditions cost an average of $2,025 per year, whereas those with 6 or more chronic conditions cost an average of $32,658 per year. Moreover, 14% (1 in 7) of Medicare beneficiaries have 6 or more chronic conditions. If you can avert even a fraction of the excess costs in these highest-risk patients by employing any of a number of risk-reducing interventions, the savings could be staggering. Moreover, if a patient is so high-risk that treatment may seem futile, that patient can be steered toward hospice and palliative care that is cheaper and far more compassionate. Can you see the possibilities here?
2. Which patients are most likely to be readmitted in the first 30 days after hospitalization?
With hospitals being graded on their 30-day readmission rate, it is imperative to know who is likely to come back to the hospital. With predictive analytics, it is possible to identify which patients are at highest risk of readmission so that steps can be taken before hospital discharge to reduce this risk. In 2010, Medicare patients with 4 or more chronic conditions accounted for 90% of all hospital readmissions. With the cost of hospital care far exceeding that of most outpatient interventions, reducing readmission rate by even a small amount can save considerable money, as well as prevent patients from succumbing to hospital-acquired conditions that are expensive to treat and unlikely to be reimbursed. Can you see the possibilities here?
3. How many ER visits will our patients likely make in the next 12 months?
Nearly any healthcare professional can tell you that the emergency room is usually not the best place to get care for non-emergencies. Yet, this is precisely where many patients get treatment for conditions when they can’t get into the doctor’s office. With predictive analytics, you can estimate the number of ER visits individual patients will make in the future. In 2010, 27% of Medicare beneficiaries with 6 or more chronic conditions had three or more ER visits during the year, whereas only 2% of those with 0 or 1 chronic conditions visited the ER three or more times. With the cost of an average ER visit at over $1,200 [2], not to mention the downstream effects on ER wait times and patient flow from ER overcrowding, preventing inappropriate ER visits can save a lot of money. If these patients arrive by ambulance, that is at least another $289-$481 [3] per patient in savings from averting an ambulance ride, on average. Can you see the possibilities here?
4. How much will it cost to care for the patients in our network in the coming year?
Healthcare institutions are continuously shifting the contracts they have with insurance providers. These decisions are based, in part, on the financial risk of having to care for the patients covered by a given plan. With predictive analytics, you can more accurately estimate the future cost of patients in a network to determine what the likely financial burden will be on the institution. In 2010, Medicare beneficiaries with 6 or more chronic conditions accounted for 46% of all Medicare spending, whereas those with 0 or 1 chronic condition accounted for only 7% of total spending. By knowing the true cost of higher-risk patients, you will be better guided in entering into contracts with insurance providers and can bargain for more appropriate reimbursement. Can you see the possibilities here?
5. For which disease entities can we be providing better value?
There is a tendency to over-treat patients as a precaution from further harm. However, overtreatment is more expensive, may not necessarily help the patient, and may even harm them. Consider the use of antibiotics with conditions like pneumonia, urinary tract infections, MRSA skin infections, and gastroenteritis. By using predictive analytics, physicians can receive a risk score that indicates the true risk of infection when the condition is suspected, but not confirmed, to guide proper decision making about antibiotics at the point of care. In a University of Maryland study [4], implementation of an antibiotic stewardship program at one institution saved $17 million over seven years. With each course of inpatient antibiotics costing hundreds-to-thousands of dollars to administer, avoiding treatment can produce significant savings, not to mention the reduced risk of C. diff colitis, antibiotic resistance, and other side effects. This method can be reasonably applied to almost any clinical condition with a little ingenuity. Can you see the possibilities here?
Don’t forget that big data can also help with risk-reducing interventions once you know who to target. Providers can send messages to patients reminding them to take medicines, check blood sugars, or log symptoms. Moreover, smartphones and smartwatches can record patient data that can be sent to providers for quick and convenient review. If deterioration can be detected sooner, a costly ER visit or hospitalization can be averted. Keep in mind that many of these procedures, like patient messaging, can be automated, so you don’t necessarily have to hire a whole new team of providers to act on the insights that predictive analytics can provide with population health management.
Predictive analytics professionals love to analyze and solve problems, and they have high levels of intellectual curiosity. If you ask your predictive analytics team these five questions and give them the time, resources, and data to answer them, you might be pleasantly surprised with the answers they provide.
Can you see the possibilities here?
References:
[1] Centers for Medicare and Medicaid Services. Chronic Conditions among Medicare Beneficiaries, Chartbook, 2012 Edition. Baltimore, MD. 2012.
Dr. Karnani is a subject matter expert in predictive analytics in healthcare, as well as a physician with a medical foundation in academic medicine. He provides a strong skill set in Analytical and Statistical Decision Making, Project Development & Management, and Education & Training. His focus is working with healthcare data to gain actionable insights in the areas of population health management, value-based care, and patient experience.
Dr. Karnani can be reached via email at rajilkarnani [at] yahoo [dot] com or on LinkedIn here.
 |
 |
 |
