Risk Scoring is the subset of healthcare analytics in which organizations attempt to quantify the most complex but important measure for running their business – the health of each member under care. Although the healthcare industry’s central concern is improving member health, payers and providers alike have been slow to adopt Risk Scoring capabilities that provide a holistic depiction of member health. The most popular Risk Scoring models compute only one Risk Score for each member. These models measure members’ relative health but fail to provide critical insights relating to health improvement and illness prevention.
The Risk Scoring landscape is changing, however, with the emergence of big data technologies and advanced analytics. These innovations are challenging the status quo, rapidly fueling evolution of Risk Scoring capabilities towards providing specific and actionable insights. Unlike the calculation of one Risk Score for each member, the calculation of Risk Scores by disease state allows organizations to better understand and serve the medical needs of each member. Healthcare payers and providers are using machine learning predictive analytics to calculate disease-specific Risk Scores that identify members most likely to be admitted to the hospital. Early adopters of this Risk Scoring methodology have already seen results, decreasing hospital admissions and readmissions while improving member health and reducing medical expenses.
Gaining deeper insights into each member’s health to prevent avoidable hospitalizations is one of the top priorities of every healthcare organization. Risk Scoring by disease state is becoming a key competitive differentiator leading to healthier members and an improved bottom-line.
THE CURRENT STATE OF RISK SCORING
Many healthcare organizations currently utilize a Risk Scoring program as part of doing business. Health insurance companies, for example, calculate one Risk Score for each member to reserve for risk-adjusted payments to Medicare Advantage and Health Insurance Exchange plans. While the calculation of one Risk Score per member is effective for the distribution of risk-adjusted payments, it lacks insight into a member’s medical needs. The current methodology fails to provide a complete picture of members’ health and serves as more of a reactive than preventative tool.
For instance, a member with multiple well-managed disease states may have a higher Risk Score than a member with one, debilitating illness. The latter would be in much greater need of medical intervention and frequent follow-up with a Primary Care Physician (PCP) or Care Manager, but the former may have more resources dedicated to their care due to their high Risk Score. Additionally, when only one Risk Score is calculated for each member, a change in Risk Score fails to provide insight into the actions necessary to improve the member’s health. A worsening Risk Score could be due to the progression of a disease state the member already has, it could mean the member is developing a new illness, or it could be a combination of the two. In any case, the Risk Score fails to furnish relevant information necessary for truly understanding the member’s health and creating actionable insights enabling health improvement. Moreover, these Risk Scores are a quantification of the members’ demographic information and prior diagnoses, and as such suffer from being reactive and not predictive in nature.
RISK SCORING BY DISEASE STATE
To truly understand the health of their members and reduce medical expenses, healthcare organizations must be able to measure the root causes contributing to a change in the member’s overall health. Calculation of Risk Scores specific to a disease state allows organizations to pinpoint the illnesses impacting a member’s health and address the reason behind a worsening condition. By doing so, healthcare payers and providers gain specific and actionable insights into preventing avoidable hospitalizations, improving member health and reducing costs.
The tools enabling this sophistication in Risk Scoring have recently become available. Specifically, the use of “big data” tools and technologies – which allow organizations to analyze large quantities of enterprise data quickly and efficiently – is becoming increasingly prevalent in the healthcare industry. Big data technologies form a framework allowing organizations to leverage all relevant enterprise data when calculating a patient’s disease state-specific Risk Score. In a big data environment, data from Claims, Electronic Medical Records (EMRs), Lab Results, Medical Images, Care Management notes, member surveys, and demographic and psychographic data sources can be collectively analyzed to Risk Score patients and derive insights that were previously impossible to collect. The Unified Patient Record collates these disparate data points into a longitudinal view of the patient, facilitating analysis of and reporting on the patient data.i Advanced analytics techniques, such as those found in Aster Analytics, can be performed on the data to derive predictive insights.
Risk Scoring by disease state has several important advantages over the current system of calculating one Risk Score for each member. For starters, the information gleaned from Risk Scoring by disease state is specific enough to be actionable for both providers and payers. Monitoring the change in Risk Scores for the exact disease states that impact each member enables organizations to align the member with the appropriate resources and treatment plan to improve the member’s health and prevent subsequent hospital visits. Additionally, disease-specific Risk Scores can be calculated using predictive analytics instead of relying on the reactive Risk Scoring methodologies that are currently used. Machine learning is a discipline of data science in which algorithms learn and improve their accuracy as additional data is introduced to the system. Predictive machine learning algorithms can be leveraged in a big data environment to identify the members most likely to be hospitalized and determine the optimal message and channel of communication to connect these members with the appropriate care team, thereby helping to reduce avoidable hospital admissions and decrease medical expenses.
A WIN FOR ALL PARTIES INVOLVED
Better understanding patient health through disease-specific Risk Scoring achieves the Triple Aim of healthcare by improving the experience of care, improving population health and reducing costs. In doing so, it aligns the interests of the industry’s major stakeholders – healthcare providers, health insurance plans, and patients.
Healthcare providers are under major regulatory and financial pressure to reduce their 30-day readmission rates. The Hospital Readmissions Reduction Program initiated in 2013 by the Patient Protection and Affordability Care Act reduced Medicare payments to hospitals with high readmission rates for heart attacks, congestive heart failure (CHF), and pneumonia by 1%. High readmission rates for chronic obstructive pulmonary disease (COPD), total hip replacements, and total knee replacements now also trigger the fines, which have since increased to a 3% loss of payment for the offending hospital.ii It is now more critical than ever that providers are able to identify the patients most likely to seek readmission and target them for preventative care.
Risk Scoring by disease state has shown impressive results among early adopters that seek to reduce readmissions by leveraging their stores of patient data. Machine learning predictive analytics are being used to identify patients that are admitted to a hospital with the highest probability of readmission due to a specific disease state, enabling providers to focus follow-up efforts and resources – including Case Management, nursing attention, and specialist visits – on the most vulnerable individuals.
For example, Texas Health Harris Methodist Hospital Hurst-Euless-Bedford has been able to reduce their 30-day readmission rate for heart failure from 23% to 12% by performing predictive analytics on their EMR data and assigning their members a Risk Score for CHF readmission.iii Likewise, the Carolinas Healthcare System leveraged predictive analytics to calculate a COPD readmission Risk Score; since doing so, they have reduced their COPD readmission rate from 21% to 14%.iv By looking at their member’s health by disease condition and not in the aggregate, these organizations were able to target resources to their most vulnerable patients, improving patient health while significantly reducing readmissions.
Health Insurance Plans
Health insurance plans, including all commercial and government payers, also have a significant incentive to implement a sophisticated Risk Scoring program. When member health improves, members tend to experience fewer adverse events (e.g. fewer hospitalizations). In turn, this leads to lower medical expenses for which the insurance company is responsible.
The Risk Scores that insurance companies calculate for risk-adjustment purposes can be used as a barometer to identify the sickest patients, but they don’t provide specific or actionable insights into improving the member’s health. Health insurance companies have enormous stores of data on their members including demographic and psychographic membership data, medical data from claims, EMRs, labs, medical images, and RX prescriptions, along with customer complaint data from call centers and various other valuable sources of data that can be used in the evaluation of member health. By leveraging these stores of data, health insurance companies can use big data processing and advanced analytics to uncover insights into improving member health and reducing medical expenses that the current system of Risk Scoring cannot.
Disease-specific Risk Scoring is emerging as an important capability for health insurers seeking to control medical expenses by preventing avoidable hospitalizations and the development of costly disease conditions. An early-adopting health insurer has shown the model’s incredible potential by achieving a dramatic reduction in hospital admissions. Independence Blue Cross implemented a CHF Risk Scoring model using predictive analytics that, coupled with proactive preventative care, has reduced their expected CHF admission rate by over 40%. After success with this application of predictive analytics, Independence is now also creating a disease-specific Risk Scoring model to identify the members most likely to develop diabetes – only 11.8% of pre-diabetics are aware they have the condition – and engage them with a medical professional in hopes of illness prevention and medical expense reduction.v, vi
There are many others ways that Risk Scoring by disease state can enable health insurance companies to better understand member health. The calculation of disease-specific Risk Scores enables insurance companies to identify undiagnosed disease states and notify the members’ PCP. Newly-diagnosed disease states will yield additional risk-adjusted revenue for the insurance company, while ensuring the member gets the care necessary leads to improved member health and a decrease in the payer’s medical expenses. Additionally, the Risk Scores can be used to monitor the change in a disease state over time. This enables the payer to identify members with worsening symptoms of a previously-identified disease state and immediately notify the member’s Care Manager or PCP. These care providers can work with the patient to understand why the condition is worsening and improve the specificity of the treatment plan to address the member’s needs.
The sickest patients and members under care benefit the most from disease-specific Risk Scoring. As hospitals and insurance companies attempt to reduce avoidable and costly hospital admissions through disease-specific Risk Scoring coupled with preventative interventions, the health of the highest-risk patients will improve while the cost they pay for healthcare will decrease.
These members will be identified in the predictive Risk Scoring models as having their health worsening or as one of the most likely candidates to return to the hospital. As a result, they will be engaged with medical professionals who are experts at dealing with the patient’s most problematic disease conditions. These experts will update the member’s treatment plan to better address the root cause of their medical needs and keep the member engaged in their health plan through routine follow up visits, which over time will lead to improved member health, fewer readmissions, and lower out of pocket healthcare costs.
With expenses rising and margins shrinking across the healthcare landscape, it’s critical that payers and providers alike leverage their member demographic and medical data to gain predictive and informative insights into the health conditions faced by each member.
The emergence of big data technologies enables first movers to gain a competitive advantage by coordinating the delivery of the right care to the right patient at the right time, a huge win for their members and the company’s bottom line. Early-adopting hospitals are already seeing major reductions in readmission rates due to predictive analytic models of disease-specific Risk Scoring. Health insurance companies are also using these technologies to reduce expected admissions and to identify those members most likely to develop a disease state.
All healthcare organizations are under pressure to increase the quality of care while reducing the cost, and innovations in Risk Scoring technologies will play a large role in this mission. A healthcare organization’s ability to quantify member health by disease state and identify risky members through predictive analytics will quickly become the most important tool for improving member health, decreasing medical expenses, and expanding margins.
i Shuster, Mitchell. “Accelerating Patient-Centric Analysis with Unified Patient Records.” Knowledgent Perspectives. June 24, 2015. Accessed July 20, 2015.
ii “Readmissions Reduction Program.” CMS.gov. August 4, 2014. Accessed July 20, 2015.
iii Conn, Joseph. “Using Big Data to Target Preventable Readmissions.” Modern Healthcare. August 2, 2014. Accessed July 20, 2015.
iv Vecchione, Anthony. “Predictive Analytics Lowers Readmissions.” Healthcare IT News. September 15, 2014. Accessed July 20, 2015.
v Bookman, Todd. “Insurer Uses Patients’ Personal Data to Predict Who Will Get Sick.” Kaiser Health News. June 8, 2015. Accessed July 20, 2015.
vi Gopalan, Anjali, Ilona S. Lorincz, Christopher Wirtalla, Steven C. Marcus, and Judith A. Long. “Awareness of Prediabetes and Engagement in Diabetes Risk-Reducing Behaviors.” American Journal of Preventative Medicine, 2015. Accessed July 20, 2015.