The field of clinical analytics enables cooperation between patients, providers, and payers. This is sometimes easier said than done. The end goal for healthcare organizations is to improve outcomes and reduce costs. With so many factors in play regarding clinical analytics, Health Management Technology was curious about current issues surrounding the topic. We asked a roundtable of relevant solutions providers to expand on how they achieved a higher quality of care and patient safety, the challenges they are facing in the industry, and what their predictions for the future of clinical analytics are as well as their newest advancements and technologies. Here’s what they had to say.
Achieving higher quality care and patient safety
Medicare Advantage plans that are categorized as Special Needs Plans (SNPs) face especially significant challenges in working to improve the healthcare services provided to Medicare/Medicaid (dual eligible) beneficiaries. Due to a lack of effective analytics and clinical support tools, one such SNP demonstrated average compliance rates of less than 35% for the three 2014 Care for Older Adults HEDIS measures across its multiple contracts.
Upon initiating a partnership with Inovalon and implementing Inovalon’s Quality Spectrum solution, the SNP initiated targeted supplemental member encounters and the collection of auditor-certified supplemental data to support gap closure. The following year, this client achieved average compliance rates of more than 60% across all three Care for Older Adults HEDIS measures—functional assessment, pain screening, and medication review. Through two years of continued partnership, leveraging Inovalon’s advanced data analytics, point-of-care decision support through Inovalon’s proprietary HIPAA-compliant platform, ePASS, and a multipronged targeted intervention strategy for identifying, addressing, and closing multiple gaps in care, the client achieved an average rating increase of 1-Star across multiple contracts employing the Quality Spectrum solution. The increase from a 3- to 4-Star Rating within two years earned this plan a Quality Bonus payment resulting in increased revenue of more than $40 million. Additionally, Inovalon’s targeted communications include personalized patient education as part of the intervention plan and execution to ensure care continuity and improved health outcomes.
LexisNexis Health Care uses clinical analytics to help healthcare organizations find individuals at high risk using socioeconomic data (data about social determinants of health) in combination with predictive modeling by stratifying patients within classes of prevalent conditions. Similarly, we help to identify high-risk individuals using socioeconomic data in the absence of medical or claims data. This socioeconomic data includes data relevant to individuals’ social, economic, and environmental factors. For example, we worked with a member engagement specialist firm that supports payer clients. The firm needed a comprehensive picture to determine population health risk and the capability to precisely target high-risk members with appropriate engagement incentives. With so many new consumers lacking traditional data, like claims and medical records, the firm chose to test and measure the capability of non-medical, socioeconomic data to fill in major gaps in member health profiles and to more accurately predict health risks.
The firm validated the accuracy of the LexisNexis Socioeconomic Health Score, which combines hundreds of clinically validated data attributes that correlate to health outcomes into a key risk prediction score. Findings indicate an individual’s health risk over the next 12 months based on cost, for overall risk and risk among seven conditions including diabetes, cancer, cardiovascular conditions, end stage renal disease, musculoskeletal conditions, gastrointestinal conditions, and respiratory conditions. Members with the top 10% of Socioeconomic Health Scores did have higher health risk than average. Prevalence of chronic conditions was also higher. As a result, the firm has implemented a new strategy that blends the Socioeconomic Health Score with other unique data sets to create more advanced modeling to power its member engagement product.
The Vizient Clinical Data Base/Resource Manager (CDB) is our proprietary analytics platform for care performance improvement. The data base is populated by hundreds of health systems and community hospitals nationwide, including nearly all academic medical centers in the United States. This rich data base helps our members understand their performance, compared to peers, across a number of comparative benchmarks such as demographics, mortality, length of stay (LOS), complication rates, readmission rates, diagnoses, procedures, resource utilization, and other metrics.
Using this data, Vizient ranks participating members each year on quality, safety, financial, and operational metrics. These rankings are then published for members as a motivator for improvement. We have seen, on average, overall improvement in all metrics each year, but some individual hospitals improve more rapidly than others. For example, one academic medical center improved in their overall ranking from 73rd to 6th in three years by substantially improving in mortality, LOS, costs, readmission, and safety metrics.
The first step in this hospital’s journey toward improvement was structural. The hospital CEO and chief quality officer—drawing from data on from other Vizient member hospitals—promoted hospital rankings throughout their organization and educated staff on the availability of CDB’s analytics tools. The hospital was also restructured into service lines, and an analytics/quality resource was attached to each service line to provide monthly dashboards that tracked performance and improvement opportunities. Dashboards were further customized by creating custom compare groups for each of the hospital’s service line, comparing them against other hospitals (by name). This transparency was further reinforced by presenting the dashboards to hospital leadership at monthly board meetings.
When performance improvement opportunities were identified, the hospital’s analysts worked with Vizient to understand the specific metrics in question, talk to other members that had improved against those metrics, and gain access best-practice resources.
Using the same hospital as an example, when they analyzed their relatively high, observed mortality in the area of cardiology, they found they also had a number of other metrics that contributed to this high mortality, including higher LOS, higher complications, lower discharges to hospice and SNF, lower use of palliative care, and higher sepsis mortality. In response, the hospital implemented hospice beds in their main hospital, revamped their palliative care program, and developed an early sepsis predictive software program. In addition, they realized that reducing LOS would get them improvements in cost, utilization complications and mortality. The longer patients stayed in the hospital the higher the complication and mortality rate.
After analyzing LOS in the CDB, talking with experts at Vizient, and developing relationships with other AMC’s that do well in LOS, the hospital decided to completely change their case management department. They hired a full-time medical director, a nurse practitioner, a social worker, and two RNs. This group proactively sought out “non-connected” patients (those without a primary care physician) and those most likely to be readmitted. Each of those patients received a care plan and follow up.
Challenges
One of the biggest challenges with clinical analytics is that the clinical quality measures should not exist in a silo. While clinical quality measures are important, it is also essential to understand a broader 360-degree view of each patient encounter, physician, service line, and risk contract analysis. This includes integrating clinical measures and EMR data with cost data, patient satisfaction data, supply chain data, and claims data to understand the full picture of the trade-offs between quality, cost, revenue, and risk. The challenge is that these data structures are quite different and are often sourced from different vendor IT systems. In particular, healthcare providers struggle with allocating costs in a meaningful way and they struggle to handle adjudicated claims data for PMPM cost analysis.
A great example is bundled payment analysis. Clinical quality measures like readmissions are certainly part of the equation for understanding the success of bundles, but so are profit margins across the episode of care and CDI coding analysis to ensure maximum reimbursement. These measures, and the data required to support them, must work together to provide a complete view of a bundled payment or another type of risk contract.
Elsevier and its sister company LexisNexis Risk Solutions have extensive experience in the insurance and payor markets aggregating and making sense of big data. In the healthcare market, we have solutions that must find, filter, and present the most relevant reference information to physicians, nurses, and pharmacists from hundreds of thousands of sources. Most healthcare organizations have some form of business intelligence. We are working closely with them and EHR vendors to bring this insight to the point of care. Clinicians—and patients—have a lot of medical information that they need to assimilate and understand. They need help in applying and integrating that information within the context of patient-specific data. Driving meaningful improvements in care requires a collaborative effort to ensure that clinical analytics supports critical decision making within the natural flow of patient care—or else it will be not used nor be effective for decision making.
Payer-provider collaboration is increasingly important within value-based care models as all healthcare stakeholders need access to the full patient picture. In the past, claims data was primarily used to fuel analytics initiatives, but now healthcare organizations must expand beyond the limited scope ICD and CPT codes to fully utilize the additional clinical data held in EMRs for analytics.
Specifically, Health Language Solutions are addressing this need by providing the expertise and tooling needed to understand and leverage clinical terminologies such as SNOMED, LOINC, and RxNorm. The ongoing challenge is lack of standardization. We must continually identify the best methods for bringing together and standardizing large amounts of disparate codes and free text. Current estimates suggest that as much as 80% of healthcare data is still locked in free text and clinical notes.
There is cost in performing this standardization using services and tools like Health Language. But the return on investment is typically substantial in that healthcare organizations improve their foundation for analytics. Most can identify additional patients for reporting, for instance, leading to improved reimbursement within value-based arrangements.
The future, advancements, and new technology
Clinical analytics must be delivered seamlessly within the workflow so the clinical side doesn’t even know it’s there. Data sources will continue to get broader, and clinical analytics will continue to expand and incorporate nontraditional, patient-generated data. It will go far beyond what is captured in our traditional physician patient relationship today and has great potential to overcome “blind spots” in our current understanding of disease progression at both the individual and population health levels.
At Allscripts, we’re beginning the development of predictive algorithms, utilizing clinical and financial data from more than 50 million lives. In addition, we’re utilizing machine learning and artificial intelligence. One example is utilizing AI to help understand variant patterns of chronic diseases and then predict which patients will be most likely to benefit from different approaches to disease management.
There is a convergence of analytics, data sciences, and cognitive computing occurring to address the opportunities and challenges that the health industry is facing. In five years, the combination of clinical analytics, financial (or actuarial) analytics incorporating new datasets such as social determinants of health and genomic markers, will become more common in value-based care decisions. To make sense of all that “big data”, newer tools of the data scientist such as Hadoop, R, Spark, etc. will begin to replace the traditional relational databases and query techniques. This will require the health industry to invest in and train more data scientists. Moreover, visualization of clinical analytics will move from traditional approaches to those designed for an increasingly mobile decision-maker and will incorporate “design thinking” principles. Finally, we see a need to replace some traditional clinical analytics with cognitive insights to augment decision-making at the point-of-need. The amount of knowledge and data that is being accumulated related to health is simply beyond the human grasp and we must recognize that providing more data alone is not necessarily going to lead to making the best decision given the practice variation that we see.
There are three areas of technology and advancement that we are working on in the field of clinical analytics. One area is identifying and utilizing sources of traditional data and nontraditional data to help (1) understand the health status and behavior motivation of a person, whether a patient, health plan member, or employee; and (2) understand clinical practice variation and care team member actions. Examples of traditional data could be electronic medical records, administrative claims, etc. Nontraditional data could include patient-generated data from wearables and portals, genomic markers, and social determinants of health. Another area is our use of Watson technology to read accumulated knowledge in the health domain (textbooks, journal articles, clinical studies, etc.) and incorporate specific patient information to produce cognitive insights such as suggestions or recommendations to decision-makers. Watson will then learn and adapt because of those interactions and outcomes. In addition to applying Watson in the care process, we are using Watson during the drug discovery process and soon to monitor safety. For example, the collaboration, Watson for Patient Safety, between IBM Watson Health and Celgene, will be designed to use clinical analytics and cognitive computing to improve the pharmacovigilance process.
Some people think about analytics in stages. The first stage is descriptive—what happened in the recent past based on the data you collected. The second is predictive—using your collected data to predict what will happen next. The third stage is prescriptive—prescribing the next steps to take, given the data available. There has been a lot of discussion about the potential progress of predictive analytics and using, for example, various forms of artificial intelligence solutions. However, I think we are going to see predictive analytics move forward slowly while prescriptive analytics becomes more prominent.
Prescriptive analytics will allow us to take information that care teams should have access to but often do not and make it available at the point of care in a prescriptive way. For example, currently, a PCP will see a patient, record information and conduct lab tests, then review the patient’s record again and propose more testing or perhaps some kind of therapy. As analytics grow in sophistication, we will see PCPs with access to data collected outside of the clinic (for example, wearables or home health monitors) that will help prescribe the next best steps to take—without the patient setting foot in a clinic or diagnostic facility. Prescriptive analytics, supported by trusted broad-based data, will change where and how patients and care teams interact.
Care teams are conservative when it comes to adopting information and new practices. Changes tend to be incremental. Prescriptive analytics offer the promise of incremental improvements and familiar care enhancements rather than requiring an entirely different set of skills than what most clinicians have received. Because of this, care teams will eagerly adopt prescriptive analytics as a way of enhancing patient outcomes and satisfaction.
First, we are working on data acquisition and enlightenment—getting good at understanding different types of data sources, their locations, and how to gather them—an area where we have in-depth experience achieving clinical and claims information analytics and interoperability. This includes identifying the various silos that have made our healthcare ecosystem so complex and getting those silos to communicate.
Second, we are working on different ways to deliver analytics. For some models, it is a matter of us collecting clients’ data, analyzing it, and sending them a report with which they can take action on to improve health outcomes. For others, there is more of a real-time analytics feedback loop that includes the cloud-based installation of an analytics solution to keep clients from waiting on a report—receiving responses to analytics requests as they need them and when and where they need them.
Third, we want to push prescriptive analytics forward, where we can give care teams the next best step to take, given the information that is available. Most care teams have the information they need but not in one place at the point of initial prescribing, for example. This causes therapeutic delays or initial efforts that end up getting revised. We can move that forward by organizing information and presenting it in a prescriptive, transparent way that makes sense to care teams.
