Analytics: The Nervous System of IT-Enabled Healthcare
Transcription
Analytics: The Nervous System of IT-Enabled Healthcare
Institute for Health Technology Transformation Analytics: The Nervous System of IT-Enabled Healthcare Strategies for managing sophisticated analytic tools in the health care industry Acknowledgements Chad Brisendine VP & CIO St Luke’s Hospital and Health Network Jeffrey L. Brown Chief Information Officer Lawrence General Hospital Charles DeShazer, MD Chief Quality Officer BayCare Health System John McDaniel National Practice Leader, U.S. Healthcare Provider Market NetApp Jonathan Weiner, DrPH Professor of Health Policy & Management and Health Informatics; Director Center for Population Health Information Technology, Director PhD Program in Health Services Research & Policy, Director Public Health Informatics Certificate Program Program Johns Hopkins Bloomberg School of Public Health Author Ken Terry, Contributing Editor, InformationWeek Healthcare, and author of the book Rx For HealthCare Reform (Vanderbilt University Press, 2007), wrote this white paper. 2 Dear Colleagues As the healthcare industry addresses very challenging times ahead, with transitions to new delivery-of-care processes, contracted business, and reimbursement models, I believe successful organizations will harness and use clinical and financial information to make informed decisions around business and patient care. At the heart of many healthcare industry debates is what to do about data: how to realize its value for quality of care, how to use it to bend the cost curve, how to share it, and how to secure it. Healthcare providers face significant obstacles in implementing clinical analytics, business intelligence tools, and data warehousing technologies. Health data, diverse in nature, consists of structured and unstructured patient and business data in varying fragmented formats, rarely standardized or normalized. It is generated in legacy IT systems, distributed across hard-to-penetrate silos, and owned by a multitude of stakeholders with differing interests and business incentives. Although other industries are already leveraging their data to improve efficiencies and make more informed decisions, to our detriment the healthcare sector has lagged. A 2011 McKinsey report estimated that the healthcare industry can potentially realize $300 billion in annual value by leveraging patient and clinical data. The purpose of this paper, “Analytics: The Nervous System of IT-Enabled Healthcare,” is to help executives from hospitals, health systems, and other provider organizations identify and understand models for innovative uses of data that can enable them to reduce costs, improve gaps in care, stratify patient populations, improve quality, and provide more accessible care. Areas of focus include creating a nervous system and solid infrastructure foundation that leverages storage, processing, analysis, and data management to make better, evidencebased business and clinical decisions. The healthcare industry must identify and establish proven strategies and best practices to manage data and to conduct the advanced analysis necessary to generate real insights that can benefit the health system. Those healthcare organizations focused today on gathering patient and clinical data, decoupling the data from siloed applications and solutions, and determining which data points to measure will be well positioned for the evolving future state of the healthcare landscape. I’d like to thank iHT2 for its assistance in developing and coordinating the research paper “Analytics: The Nervous System of IT-Enabled Healthcare ,” which is composed of contributions from individuals in the provider, health system, health information technology, academic, and health policy domains. This group is well versed in data analysis, patient-centered care, health information technology, decision support systems, and the emerging—and urgent---imperative to transform healthcare delivery with innovative uses of health data. I’d like to personally thank each contributor for his time, contribution, and knowledge in compiling this research paper: • Jeffery L. Brown, Chief Information Officer, Lawrence General Hospital • Charles DeShazer, MD, Chief Quality Officer, BayCare Health System • Jonathan Weiner, DrPH, Professor of Health Policy & Management and Health Informatics;, Director of the Center for Population Health Information Technology, Director of the PhD Program in Health Services Research & Policy, Director of the Public Health Informatics Certificate Program, Johns Hopkins Bloomberg School of Public Health • Chad Brisendine, Vice President and Chief Information Officer, St. Luke’s Hospital & Health Network Respectfully, John P. McDaniel National Practice Leader—Provider Market NetApp Healthcare Institute for Health Technology Transformation 3 Table of Contents Executive Summary...........................................................................................5 Analytics Framework..........................................................................................6 Why Anayltics ...................................................................................................7-8 Rapid Evolution 7 Anayltics Take Center Stage 8 Collecting the Data............................................................................................10 -11 Claims Data 10 Clinical Intelligence............................................................................................13 -14 Risk Stratification and Predictive Modeling 13 Care Management 14 Business Intelligence.........................................................................................15 -17 Total Cost of Care 15 Cutting Edge Business Intelligence 16 C&BI Applications 16 Performance Evaluation.....................................................................................18 -19 Attribution of Patients 18 Utilization Management 18 Cultural Change 19 Conclusion .......................................................................................................20 Recomendations 21 Notes 22 Institute for Health Technology Transformation 4 Executive Summary The healthcare industry is moving from volume-based reimbursement to value-based reimbursement that is designed to achieve the Triple Aim of the Institute for Healthcare Improvement: higher quality, lower costs, and a better patient experience.4 To succeed in this new environment, healthcare providers are forming accountable care organizations (ACOs), preparing for payment bundling, and restructuring their care delivery systems. Organizations require sophisticated analytic tools so they can manage care within a budget while providing high-quality care. Population health management, a model designed to maximize the health and minimize the cost of caring for a defined population, requires clinical analytics to improve the quality of care and reduce avoidable hospital admissions. But to manage risk effectively, a provider organization such as an ACO must also apply financial analytics to measure the cost of care delivery and to use its healthcare resources wisely. Both kinds of analytics require a 360-degree view of patient care that is just starting to become available. Institute for Health Technology Transformation 5 Analytics Framework Data governance committee Distinguish between data and ‘actionable’ data Identify the single source of data Establish staff and your analytics team Identify data sources today and determine what data to move to the single source for analysis Pick a project and focus on data quality and timeliness that work on a particular disease state Determine benchmarks Identify a sample population then measure and adjust around individual patient outcomes Measure patient behavior change and impact on quality and cost Institute for Health Technology Transformation 6 Why Analytics? Healthcare providers are going through an historic transition from volume-based to valuebased reimbursement. As a result, healthcare delivery is also changing. Accountable care organizations (ACOs) are being formed to improve care coordination and to enable providers to take financial responsibility for care. Bundled payments for acute and post-acute care, as well as Medicare penalties for avoidable readmissions, are compelling hospitals to collaborate with other providers. And healthcare systems and physician practices are building patient centered medical homes to ensure that patients receive appropriate preventive and chronic disease care and that their care is well coordinated. All of this requires an advanced health IT infrastructure, and the centerpiece of that infrastructure is analytics. “Analytics is the backbone and the nervous system and the learning center of the health IT-enabled healthcare system,” says Jonathan Weiner, a professor of health policy and management at the Johns Hopkins Bloomberg School of Public Health and director of the university’s Center for Population Health Information Technology. Analytics is the backbone and the nervous system and the learning center of the health IT-enabled healthcare system Jonathan Weiner, DrPH Professor of Health Policy and Management John Hopkins Bloomberg School of Publich Health Director, Center for Population Health Information Technology The literature refers to two types of healthcare analytic applications: “business intelligence” and “clinical intelligence.” Business intelligence (BI) applications address financial and operational aspects of healthcare systems, such as contract negotiations, facility management, measurement of resource utilization, and cost analysis. Clinical intelligence (CI) software supports activities such as quality improvement, care management, and population health management.2 BI and CI overlap in a number of areas. For example, when analytics are used to assess the staffing needs of an organization, both financial and clinical aspects come into play. Both kinds of applications may be also be used to evaluate the efficiency and quality of care delivered by an organization or an individual provider. In fact, Atrius Health, a large healthcare network that operates an ACO in eastern Massachusetts, views the reduction of health cost growth as one of the primary aims of clinical intelligence.3 So some experts use the acronym “B&CI” to describe the new healthcare analytics. In any organization that is taking financial risk for care delivery, analytics necessarily has a very broad scope. That is because risk-bearing organizations rely on population health management, which is a systematic effort to keep patients as healthy as possible so they have better outcomes and don’t end up in costly care settings such as hospitals and emergency rooms. Population health management requires analytics for everything from identifying care gaps and providing clinical decision support to predictive modeling, risk stratification, outcomes measurement, and performance measurement. Analytics are used to improve the quality of care, not only for individuals, but also for subpopulations of patients such as those who have diabetes or congestive heart failure.4 Rapid Evolution Until recently, most healthcare analytics focused on the financial and operational aspects of healthcare organizations. Those applications are still widely used, because the fundamentals of healthcare financing have not yet changed for most provider organizations. To the extent that these applications go beyond billing and collections, they are used to improve efficiency in areas like supply chain management. As healthcare systems and physician groups form accountable care organizations, however, they discover that they need entirely new forms of clinical and business intelligence related to the cost and quality of care. 7 Population health management requires analytics for everything from identifying care gaps and providing clinical decision support to predictive modeling, risk stratification, outcomes measurement, and performance measurement. Why Analytics? The rapid spread of electronic health records (EHRs) in recent years is one of several factors that have driven the adoption of clinical intelligence applications. Before EHRs, the clinical data available to feed these applications was limited mostly to hospital pharmacy, lab, and radiology systems. Since 2010, when the Meaningful Use EHR incentive program started taking applications, the use of EHRs by hospitals and physician practices has soared.5-6 As a result, the amount of data available for clinical intelligence has also grown exponentially. The rapid spread of electronic health records (EHRs) in recent years is one of several factors that have driven the adoption of clinical intelligence applications. Some healthcare organizations have acquired integrated information systems that cover hospital and ambulatory care. But the bulk of clinical data is still silo’ed in disparate systems that can’t communicate with one another or that can exchange only limited subsets of information. Hospitals and healthcare systems have been slow to develop the clinical data warehouses they need to aggregate and normalize the data from their multiple clinical and administrative systems. According to HIMSS Analytics, in 2011 only 30% of U.S. hospitals used a clinical data warehouse/mining solution. Nearly half of institutions with more than 500 beds did, but only 20% of hospitals with 200 or fewer beds had a data warehouse.7 Healthcare organizations’ adoption of and interest in clinical intelligence has increased significantly over the past few years. In a 2010 HIMSS Analytics survey, respondents said they were collecting and/or leveraging clinical and claims data to enhance patient care, cost, safety and efficiency. Requirements from government agencies and other parties such as the Leapfrog Group were responsible for much of this activity. To reduce costs, hospitals were tracking big-ticket items and areas such as cardiology, transplants, surgery, and obstetrics. For the most part, analytics were being used for retrospective review of data, rather than clinical decision support.8 According to a similar study done in 2011, providers were beginning to use analytics to identify trends in their patient populations and opportunities to improve quality and efficiency. There were also new initiatives aimed at improving preventive care to reduce the need for medical interventions and expensive services. However, the use of analytics continued to be mainly retrospective.9 Analytics Take Center Stage The real burst of interest in analytics arrived in 2012 with the advent of ACOs and the recognition by many healthcare systems that value-based reimbursement was just around the corner. The Centers for Medicare and Medicaid Services (CMS) introduced its valuebased purchasing programs for hospitals in October 2012, along with penalties for excessive readmissions of Medicare patients. A 2012 survey of seven healthcare systems—most of them advanced health IT users-showed that they viewed clinical and business intelligence as a competitive advantage and as a key component of the organization’s performance improvement efforts. These organizations all had data warehouses that allowed them to optimize the care process and understand business issues related to the cost of care. They all viewed C&BI as a strategic asset and a core business competency that healthcare organizations would need to deliver value in the future.10 The real burst of interest in analytics arrived in 2012 with the advent of ACOs and the recognition by many healthcare systems that value-based reimbursement was just around the corner. Some organizations were feeding EHR, administrative, and claims data into their data warehouses in near real time and running analytics against that constantly updated Institute for Health Technology Transformation 8 Why Analytics? information. This kind of approach enables healthcare systems to provide actionable information for clinical decision support and care management while automating routine aspects of population health management. On the whole, however, the industry was still in the early stages of using analytics. Even a Medicare Pioneer ACO like Atrius Health, for example, was still using pre-formatted Excel spread sheets to run clinical intelligence reports in its clinics.11 Few organizations allowed physicians to query data warehouses or registries.12 Today, the C&BI field is very fluid and continues to evolve rapidly. Health insurers are giving some providers access to analytics such as predictive modeling applications; but their tools were developed for actuarial purposes, rather than population health management. Moreover, they were designed for use with claims data, not the clinical data in EHRs. Most EHR vendors do not supply analytics suitable for population health management. But other health IT vendors, including some “big data” firms, offer a wide range of analytics that can be integrated with EHRs and other data sources. Among these are applications for patient attribution, patient registries, prebuilt clinical protocols, risk stratification, predictive modeling, care gap identification, utilization management, benchmarking, clinical dashboards, and automated work queues. Some companies claim that they can predict who will be hospitalized and can measure the relationships between patient outcomes and cost.13 9 Most EHR vendors do not supply analytics suitable for population health management. But other health IT vendors, including some “big data” firms, offer a wide range of analytics that can be integrated with EHRs and other data sources. Collecting the Data Before an organization can use clinical or business intelligence, it must collect the necessary data and convert it into a homogeneous form that can be analyzed. Clinical data, to put it mildly, is full of holes. Even in a healthcare organization with a fully implemented electronic health record, much of the EHR data is unstructured, because physicians tend to dictate their reports and progress notes. (Overall, about 80% of electronic health information is said to be unstructured.14) While natural language processing (NLP) applications are being developed to extract that data from free text, the technology is still a work in progress. Another problem is that a lot of healthcare information is still on paper. Paper documents, which may be faxed, mailed or hand delivered to an organization, or generated within it, must be scanned into an EHR, or the data contained in them must be entered manually. Scanned documents are not easily searchable, and data entry is expensive and can lead to errors. Physicians and nurses may also neglect to enter some data elements, reducing the value of any analytics applied to the data. Recent studies found common diagnoses missing from a large portion of EHR problem lists in two different physician groups.15-16 In addition, it can be difficult to extract the necessary data elements from multiple systems and transform it into a useable format. A HIMSS Analytics study said: Several respondents noted that it can be complicated to map the data in the database to the appropriate field in the report, so that the analysis makes sense. This can be particularly challenging when the information is not captured at the source as discrete data elements, and may have to be manually abstracted or converted through an intermediate process.17 Another major challenge to data collection for analytic purposes is that the information systems in an enterprise contain only data generated in that enterprise or received from outside systems that are interfaced with the enterprise system. So, even if programmers can integrate all their in-house data into a single, searchable database, they might be lacking big chunks of relevant information on a patient who was treated outside the enterprise. Health information exchanges (HIEs) are supposed to solve the interoperability problem. But communitywide HIEs are still far from common; where they do exist, they don’t include all of the providers in their area; and the data exchanged between providers may not include all of the data elements required for analytics. Claims Data For all of these reasons, claims data is necessary to do population health management and to take financial risk for care delivery. Claims data can lag the provision of services by a month or more, and it has other flaws and deficiencies. But it offers the broadest view of the healthcare services that have been provided to patients both within and outside of an organization. Some providers that are working closely with payers on ACOs or that have their own health plans—such as Kaiser Permanente, HealthPartners, and Geisinger—already have access to claims data. Other providers are using claims data for their own employees to begin the journey toward population health management. 10 Clinical data, to put it mildly, is full of holes. Even in a healthcare organization with a fully implemented electronic health record, much of the EHR data is unstructured, because physicians tend to dictate their reports and progress notes. (Overall, about 80% of electronic health information is said to be unstructured.14) Collecting the Data To get real value out of claims data, providers must combine it with clinical data. This is not a simple matter, even if an organization has a data warehouse. To start with, the claims data must be “cleaned” so that it reflects the clinical record as closely as possible. For example, if a doctor ordered an MRI test for a patient to rule out a stroke, the diagnostic code for “stroke” might be in the claim, but that doesn’t mean the patient had a stroke. So the claims data must be subjected to rules that ensure that irrelevant information is excluded. To get real value out of claims data, providers must combine it with clinical data. Some non-EHR vendors—including those that call themselves “big data” firms—contract with providers to combine claims and clinical data. Currently, Weiner notes, the analytics these vendors bring to the table are usually those that insurers have historically applied to claims, so the clinical data must be brought into that framework. But in the future, healthcare executives agree, the approach will be more clinically based, with claims data used to round out the picture. One other data stream that will undoubtedly become more important is the input from home monitoring devices and mobile health applications. But today, partly because of the lack of standards for connecting these devices to EHRs, this is not a factor in analytic applications.18 Institute for Health Technology Transformation 11 Clinical Intelligence Roughly speaking, the analytic approaches required to do population health management fall into two buckets: retrospective review and the generation of actionable data. Both elements are necessary, but the need for actionable data dictates that the information be collected in near real time. The main purposes of clinical intelligence are as follows: • Assess population health needs in order to develop appropriate methods of • • • • • • • service delivery Stratify the population by level of health risk Predict which individuals are likely to become seriously ill Identify individual care gaps Measure intermediate and long-term outcomes Evaluate performance of providers and organizations on quality measures Drive quality improvement programs Measure and analyze reasons for variations in care As previously mentioned, provider organizations use data warehouses to aggregate and normalize data for analytic purposes. Some enterprise data warehouse solutions include patient registries, and some physician groups use standalone registries that incorporate data from EHRs and practice management systems. Registries list all of the members of a patient population, their health problems, what healthcare services they have received, and when and from whom they received them. They also contain information about the health status of each patient, based on their lab results, medications, and other information.19 Analytics can be applied to registries for a variety of purposes, ranging from analyzing the health status and needs of subpopulations to generating alerts at the point of care. When combined with imbedded clinical protocols, registry-based analytics can trigger automated messaging to patients who need preventive or chronic disease care. And care managers can use this kind of intelligence to prioritize the patients who need their interventions the most. To some provider executives, it makes sense to generate the registry out of the enterprise data warehouse, where the information from diverse systems already has been mapped to a common format. That works well for the purposes of retrospective review, including measuring patient outcomes, provider performance, and variations in care. But there is some disagreement over whether a registry based in a data warehouse can generate actionable information quickly enough to be effective in clinical decision support. Charles DeShazer, MD, chief quality officer of Baycare Health System, based in Clearwater, Fla., says that he wants his data warehouse to include a near-real-time registry that receives updates on all clinical data within 24 hours. For example, test results or information on outreach to a patient should be available within a day of it being documented so that nearreal-time alerts can be transmitted to clinicians. Chad Brisendine, vice president and CIO of St. Luke’s University Health Network, based in Bethlehem, Pa. agrees on the need to provide actionable alerts to physicians in a way that fits into their workflow. But in St. Luke’s environment, which includes a variety of ambulatory care EHRs being used by employed and community doctors, he doesn’t believe that St. Luke’s data warehouse is the right location for a registry, because it can’t generate registry data and get alerts back to providers fast enough. 12 Analytics can be applied to registries for a variety of purposes, ranging from analyzing the health status and needs of subpopulations to generating alerts at the point of care. When combined with imbedded clinical protocols, registry-based analytics can trigger automated messaging to patients who need preventive or chronic disease care. And care managers can use this kind of intelligence to prioritize the patients who need their interventions the most. Clinical Intelligence So St. Luke’s is considering the possibility of installing a registry in its internal health information exchange, which connects its hospitals, its ambulatory-care practices, and other providers who aren’t part of the system, he says. The HIE’s links to the disparate EHRs could be used to send alerts to physicians at the point of care. Risk Stratification and Predictive Modeling Population health management also requires the classification of patients by their health risk, which includes their current health status and their chance of becoming sick or sicker in the future. Risk stratification, as this process is known, is used to assign patients to groups that receive different kinds of interventions. Low risk patients, for example, might receive information on preventive care and wellness; medium risk patients might receive coaching and education on how to manage their chronic diseases; and high-risk patients might be referred to care managers who can help them prevent their conditions from getting worse. Population health management also requires the classification of patients by their health risk, which includes their current health status and their chance of becoming sick or sicker in the future. Part of risk stratification is an analytic technique known as predictive modeling or health forecasting. Originally developed by health plans to predict which members are most likely to get sick and incur high costs, predictive modeling has also been used by health plans and integrated delivery systems as part of disease management programs.20 For any organization that plans to bear financial risk, predictive modeling is very important, because high-risk patients generate the majority of health costs. Only 30% of high-risk patients today were in that category a year ago.21 So the ability to manage costs is closely tied to the ability to predict risk. Health forecasting in health plans uses statistical models based on claims data. Some insurers are supplying this kind of software to ACOs and other provider organizations. In addition, advanced organizations such as the Colorado Beacon Community are beginning to apply sophisticated algorithms to registry data. One such model, which comes from an outside vendor, matches registry data against very large public databases to predict an individual’s chances of developing a serious chronic disease or having acute events such as a heart attack or a stroke.22 Other healthcare providers are seeking or using analytics that will accurately predict a patient’s chance of being hospitalized. Heritage Provider Network in Northridge, Calif., recently offered a $3 million prize for the best such application. Health Management Associates, a for-profit chain with 70+ hospitals, is working with outside vendors to develop an algorithm to predict readmissions.23 Meanwhile, Dallas’ Parkland Health and Hospital System and Indianapolis’ Community Health Network are already applying predictive modeling to EHR data to identify patients who are likely to be readmitted.24 Other healthcare providers are seeking or using analytics that will accurately predict a patient’s chance of being hospitalized. Some provider organizations use not only clinical and claims data, but also patient-supplied information to do risk stratification and predictive modeling. A common method of collecting this data is to have patients complete a health risk assessment (HRA) survey that asks them to rate their own health and supply information about their health behavior and family history. Patient activation measures have also been shown to have predictive value.25 Another close relative to risk stratification is risk adjustment. Risk adjustment uses algorithms to make data sets comparable by adjusting for the severity of illness of the patients referenced in each data set.26 On the clinical analytics side, risk adjustment must be applied to the evaluation of provider performance and variations in care. On the financial side, it is essential to any sophisticated analysis of costs or utilization, Weiner points out. Institute for Health Technology Transformation 13 Clinical Intelligence Care Management From the viewpoint of managing population health, care management encompasses preventive care, wellness activities, chronic disease management, care coordination, and transitions of care. This kind of care management focuses on the entire patient population, not just the sickest patients or those who make contact with the healthcare system. Clinical analytics plays a key role in care management. We’ve already mentioned the use of registries to identify care gaps and alert clinicians and patients about them. Risk stratification is also essential in care management, because it enables organizations to plan interventions for groups with varying health risks. When combined with workflow automation programs, analytics can help care managers provide appropriate programs to far more patients than they could on their own, because it saves them the time of going through charts to identify patient needs. Clinical intelligence can also help managers and physicians decide how to improve the quality of care for subpopulations. For example, Banner Health Network (BHN) in Arizona is working closely with Aetna to develop its ACO infrastructure. An Aetna subsidiary has given BHN physicians access to point of care clinical decision support services and a desktop-based workflow tool to track, monitor, and coordinate care and report on patient health outcomes.27 Using these tools, the BHN doctors can see the breakdown of their patient panels by disease cohorts such as diabetes and asthma. This enables the providers to risk-stratify those subpopulations and see how they’re doing with patients who have specific conditions. While this part of the solution is being driven by the need to report on CMS’ Pioneer ACO quality measures, it can also be used in quality improvement programs. Massachusetts’ Atrius Health, which consists of several multispecialty groups, has an ACO that contracts with area hospitals for inpatient care. Atrius uses clinical intelligence to measure the quality and efficiency of care being provided to its patients in these hospitals. The ACO has an analytic program that uses data from a variety of sources to score the hospitals in areas ranging from ED care and care management to discharge planning and readmissions.28 14 From the viewpoint of managing population health, care management encompasses preventive care, wellness activities, chronic disease management, care coordination, and transitions of care. Business Intelligence As discussed earlier, the financial and clinical aspects of healthcare are intertwined, so the analytics used in these two spheres are often described as clinical and business intelligence (C&BI). Nevertheless, there are distinct areas in which risk-bearing organizations need financial analytics. Among those areas are the ability to identify and analyze costs; comparison of those costs with payments or prepayments; allocation of reimbursement to providers; referral leakages out of network; risk stratification from a cost standpoint; and utilization management. Today, most healthcare organizations either own or contract out for revenue cycle management applications that improve their billing and collection processes. Today, most healthcare organizations either own or contract out for revenue cycle management applications that improve their billing and collection processes. These applications include analytics and workflow automation software for determining insurance eligibility and benefits, ensuring that bills are correctly prepared, posting and reconciling payments, reworking denied claims, and collecting from individuals and third party payers. An accountable care organization may bill Medicare or a commercial payer for each service that its providers perform. In a shared savings program, however, the ACO’s goal is not to maximize fee-for-service revenues, but to generate savings that it can divide with the payers. If the ACO takes financial risk for all or part of patient care, it will get a bonus if the cost of providing that care is less than the budget for a particular time period. Conversely, the organization’s bottom line will be smaller than expected if the cost of providing care exceeds the budget. Because revenue cycle management maximizes volume-based reimbursement, it will not be helpful to an ACO that creates savings by lowering volume and improving quality. What that ACO needs is analytics that can help it use healthcare resources “as efficiently as possible in a population perspective,” Weiner says. Total Cost of Care Hospitals have patient cost accounting systems that enable them to increase the efficiency of certain areas of the hospital: for example, managers can use these systems to analyze whether patients’ lengths of stay are too long in relation to fixed payments known as diagnosis-related groups (DRGs). They can see how much is being spent on supplies and whether some of those supplies are being wasted. And they can analyze staffing levels to determine whether they have too many nurses working at certain times when the hospital census is low. But none of this gives an accountable care organization the ability to measure the total cost of providing care. One reason is that many staff physicians bill for their own services, so their charges are not reflected in the hospital billing system. In addition, the ACO’s total spending includes the costs of caring for a population across all care settings, including ambulatory, acute and post-acute care. If a patient is discharged early from the hospital, the hospital’s bottom line might benefit, but the cost of caring for that patient in a skilled nursing facility might be higher. Charles DeShazar of Baycare notes that the currently available financial analytics is not adequate for determining healthcare costs and utilization of services. “That’s where the systems need to evolve,” he says. “For the most part, in the short term this is going to require custom development and custom design. The standard systems are really oriented toward a volume type of reimbursement.” Charles DeShazar of Baycare notes that the currently available financial analytics is not adequate for determining healthcare costs and utilization of services. “That’s where the systems need to evolve,” he says. “For the most part, in the short term this is going to require custom development and custom design. The standard systems are really oriented toward a volume type of reimbursement.” Institute for Health Technology Transformation 15 Business Intelligence Ideally, he said, “you’d use an activity-based costing approach” to analyze the costs of population health management. “But currently, those systems don’t exist in healthcare, so organizations are forced to draw inferences from billing data.” Cutting Edge Business Intelligence St. Luke’s University Health Network uses its patient cost accounting system to measure variations in care and to look at “high-level service-line P&Ls [profit and loss statements],” says Chad Brisendine. Now St. Luke’s is in the process of combining cost accounting for ambulatory care with that of the hospital, he says. This will be essential, he notes, when the organization begins its participation in Medicare’s payment bundling demonstration next summer. But St. Luke’s still lacks a sophisticated, episode-based accounting system for bundled payments, he points out. It will need that to understand how the costs of care are broken down between the hospital and other providers, including physicians and post-acute-care providers. So St. Luke’s has hired an outside vendor that has experience in working with health plans to help it learn how to do that. In addition, the organization has contracted with another vendor that has financial analytics for budgeting, forecasting, and costing. “Long term, we want to be able to cost across an episode, but then we need to be able to marry what we get paid against that,” Brisendine says. “That’s regardless of how we get paid – bundled payments or whatever it is. We need to be able to run that against our cost data, and then within that cost data, we need to see what the variability of that cost is.” To analyze cost data across the organization, he adds, St. Luke’s has already connected the practice management systems of its physician groups to its enterprise financial system. That system is sending data on physician billing codes and RVUs to the data warehouse of the costing vendor. C&BI Applications There are hardly any activities in an ACO that don’t have financial implications. Here are a few examples of C&BI applications that can help ACOs and other risk-bearing organizations reduce costs. Out of network utilization. Patients often seek care or are referred outside of an organization— and in the Medicare shared savings program, they must be allowed to do so. Since ACOs generally pay more to providers with whom they have no contractual relationships than to those in their networks, they want to minimize the use of out-of-network services. But, unless they have a way to track these services, it’s difficult to work with their physicians to keep referrals within the ACO network. So they need referral management software to reduce the “leakage” of patients to outside providers. This kind of application helped one healthcare organization increase the rate of in-network services from 40% to 90%.29 Another kind of analytic software uses claims data to detect the provision of services by outside providers. That can be useful to ACOs if network managers factor in the time lag of claims information. Admissions and readmissions. A financial risk-bearing organization must focus on preventing avoidable ER visits, admissions, and readmissions. While this is a very complex 16 St. Luke’s University Health Network uses its patient cost accounting system to measure variations in care and to look at “high-level service-line P&Ls [profit and loss statements],” says Chad Brisendine. Now St. Luke’s is in the process of combining cost accounting for ambulatory care with that of the hospital, he says. This will be essential, he notes, when the organization begins its participation in Medicare’s payment bundling demonstration next summer. Business Intelligence subject that goes beyond the scope of this paper, some of the Beacon Communities funded by the Office of the National Coordinator for Health IT are using analytics in this area. For example, the Greater Cincinnati Beacon Collaboration (GCBC) is working with 18 hospitals to reduce readmissions with the help of health risk assessment tools. Another GCBC project is designed to reduce readmissions of adults with asthma by doing root cause analyses of their original ED visit or admission. A web-based clinical decision support tool then incorporates these results and suggests appropriate follow-up activities.30 Two other Beacon communities, in Tulsa, Okla., and Colorado, are using a tool from an outside vendor that creates risk profiles of patients and shows an individual’s chances of having a heart attack or stroke or of developing diabetes or cancer. The profiles are displayed on a patient portal, and doctors use them to educate patients.31 Variations in care. As the Dartmouth Atlas of Healthcare has amply documented, there are widespread variations in how various conditions are treated across the country, including variations within local areas.32 Those differences in care delivery, which are frequently at odds with evidence-based-medicine guidelines, can have an enormous impact on the cost of care. Leading integrated delivery systems such as Kaiser Permanente, Cleveland Clinic, and Geisinger are all striving to identify and reduce variations in procedures and other services. Risk adjustment is a prerequisite to analyzing variations in care. Current risk adjusters are designed for use with claims data. But John Hopkins University, which developed the Adjusted Clinical Groups (ACG) system for risk adjustment,33 is working on a new generation of tools that incorporate EHR data such as lab results, body-mass index, and blood pressure, says Weiner. “That’s the direction to go from an analytic standpoint,” says DeShazer. “You’ve got to be able to understand and track episodes of care and be able to identify variations and best practices in treating those conditions.” Brisendine notes that St. Luke’s is already looking at the variability within orders in its computerized physician order entry (CPOE) system. The goal is to determine which tests are being ordered that shouldn’t be done. St. Luke’s has built analytics for that into the CPOE system itself, he says. Risk adjustment is a prerequisite to analyzing variations in care. Current risk adjusters are designed for use with claims data. But John Hopkins University, which developed the Adjusted Clinical Groups (ACG) system for risk adjustment,33 is working on a new generation of tools that incorporate EHR data such as lab results, body-mass index, and blood pressure, says Weiner. Institute for Health Technology Transformation 17 Performance Evaluation Clinical and business intelligence is required to evaluate the performance of both healthcare organizations and individual providers. If an organization is bearing financial risk, it needs to monitor how well it is managing population health and controlling costs. And to ensure that its providers are delivering high-quality care as efficiently as possible, while satisfying their patients, the organization must continuously track their performance on such indices as utilization management, quality of care, patient outcomes, and patient experience. Clinical and business intelligence is required to evaluate the performance of both healthcare organizations and individual providers. All of this requires analytics that can evaluate performance according to a variety of clinical guidelines from specialty societies and other sources. While ideally based on the best available evidence, these guidelines or protocols are often tweaked to fit the practice parameters of each physician practice or healthcare organization. Attribution of Patients The first step in performance evaluation—and in population health management--is to “attribute” each patient to a particular provider or care team. This process is much more complicated than it might seem. Many patients do not have a primary care physician, and some people receive the bulk of their care from a subspecialist, such as a cardiologist, a gastroenterologist or an endocrinologist. Asaf Bitton, MD, a researcher at Harvard Medical School’s Center for Primary Care, writes: Attribution happens with about 60-90 percent fidelity, so some patients fall through the cracks. It is a key starting point for knowing generally who your clinicians care for, and getting to near-100 percent attribution within your EHR is an important milestone at the outset of your journey toward population management.34 Today, the information required for patient attribution is mostly contained in claims data. So the patient’s insurance company usually does the attribution, although organizations that encompass most of their patients’ providers can also do it with EHR data. In either case, analytics can assist attribution, but algorithms can only go so far; patients must also be asked who their primary provider is, Bitton notes. Utilization Management Utilization management or “stewardship of resources,” as some in the industry now refer to it, is inseparable from a value-based approach to healthcare. Because the resources available to care for a population are limited, organizations must assess how efficiently and judiciously providers order tests, perform procedures, and prescribe medications and other treatments. Health plans use sophisticated applications in their utilization reviews. But these aren’t necessarily the same analytics that one might want to use for performance evaluation in a population health management context. Health insurers’ utilization reviews have traditionally tilted toward cost savings; in contrast, ACO performance criteria must balance efficiency against quality. Risk-bearing organizations look at which providers utilize the most resources for a patient population and which ones follow evidence-based guidelines for care. With this risk-adjusted information, they can construct reliable cost and quality profiles and steer patients to the most efficient, highest quality providers. Alternatively, organizations can identify physicians who have opportunities to practice more efficiently and have medical directors work with them. 18 Utilization management or “stewardship of resources,” as some in the industry now refer to it, is inseparable from a value-based approach to healthcare. Performance Evaluation Cultural change Both the data collection and the analytics required for population health management necessitate a cultural change, especially among physicians. DeShazer cites the frustration of many doctors who are being required to document certain information in their EHRs for billing purposes while their organizations are trying to use the same data in quality improvement programs. In some cases, a particular data element might be documented in various places in the chart, making it difficult to collect data. “We need to standardize those kinds of things while minimizing the impact on the clinician’s time,” he says. Another kind of standardization, however, is likely to encounter a cultural barrier when analytics are used to measure physicians’ deviation from clinical guidelines. To some physicians, these “care pathways” look like a type of manufacturing process that negates their medical judgment.35 The right approach is to introduce this change gradually as a form of quality improvement, without overburdening doctors with too many quality improvement programs, says Brisendine. DeShazer cites the frustration of many doctors who are being required to document certain information in their EHRs for billing purposes while their organizations are trying to use the same data in quality improvement programs. Another big challenge is physician compensation. Many healthcare organizations already use analytics that measure physician work RVUs and some quality indicators to calculate each physician’s yearly bonus. As they move into value-based reimbursement, healthcare systems and physician groups will have to factor quality and efficiency into doctors’ compensation to a greater extent than they do now. Today, however, the old volume-based reimbursement method still holds sway. Healthcare organizations don’t want to change their physician compensation methods radically until the entire healthcare system switches over to value-based reimbursement. But until they do, their efforts to become high-quality, low-cost providers will encounter resistance from physicians who are being incentivized to produce volume, rather than value. Institute for Health Technology Transformation 19 Conclusion Clinical and business intelligence applications are the linchpin of ACOs and other risk-bearing organizations that do population health management. The two kinds of analytics are already closely connected, and they may eventually merge into a single, multi-faceted approach to analyzing data and making it actionable. The most obvious gap in the current analytics lineup is the lack of business intelligence software that can help organizations measure the true costs of care delivery across care settings. While such applications are being developed and tested, ACOs will find it difficult to manage financial risk until they have a better understanding of what their costs are and where they’re coming from. On the clinical intelligence side, the biggest challenge lies in data collection. Until interoperability among EHRs is widespread, or health information exchanges are ubiquitous, healthcare organizations will have to rely on claims data to fill gaps in their information about patient care. Even then, the prevalence of unstructured data will be a problem until natural language processing is perfected. Clinical and business intelligence applications are the linchpin of ACOs and other risk-bearing organizations that do population health management. The two kinds of analytics are already closely connected, and they may eventually merge into a single, multi-faceted approach to analyzing data and making it actionable. Because of the rapid evolution of analytics and the variability in the situation of each healthcare organization, there is no single roadmap to achieving analytics excellence. But the following steps are critical to any organization going down the path to accountable care: • Risk-stratify the patient population so that care teams can intervene with the high-risk patients who generate the majority of costs • Identify the care gaps of all patients and provide actionable data to remedy them • Build partnerships with health plans in order to obtain claims data • Prepare physicians for big changes in how they practice medicine. Finally, we’d be remiss not to mention the new capabilities that big data offers in healthcare. Big data firms are promising to use their enormous data crunching power to combine, not only claims and clinical information, but also genomic, environmental, and other kinds of data. With the proper analytics, these huge data pools could supply new insights into disease causes and treatments. But for now, most healthcare organizations just need analytics to ensure that patients receive appropriate care and stay as healthy as possible.36 Institute for Health Technology Transformation 20 Recommendations • Construct a data warehouse that is the single source of truth for all the data your organization aggregates. Ensure consistency of data and terminology in addition to establishing a robust data mapping and cleaning process. • Track process information such as patient outreach efforts and patient compliance with physician recommendations. • Change the analytic perspective from episode-based or procedure-based analyses to patient-based and population-wide views. Manage the population for the individual with a longitudinal care approach. • Ensure data availability is real time and accurate so that the information is timely enough to help clinicians intervene with patients. • Integrate claims and administrative data with clinical data from EHRs to provide a 360-degree view of patient care. • Makes sure to engage the right skill sets to enable data schemes and models for actionable data. • Data governance policies are critical to success – Consider appointing a chief information management officer • Use predictive modeling and other “big data” decision support tools with the expanded source data. Base this on best practices and establish what to focus on. i.e. diabetes, heart disease, obesity and patient engagement. • Make sure that all stakeholders within the organization help define the goals of health IT. For analytics to have the desired result, it must meet the financial, care delivery, and operational business needs of the organization. • Create a culture of using data to treat patients so that the organization consistently collects data and applies analytics to all of the information it needs to manage population health successfully. • Don’t boil the ocean. Big things have small beginnings. The entire process will need piloting, evaluation and ongoing improvement. Institute for Health Technology Transformation 21 Notes 1. Donald M. Berwick, Thomas W. Nolan and John Whittington, “The Triple Aim: Care, Health and Cost,” Health Affairs, May/June 2008, 759-769. 2. HIMSS Analytics, "Atrius Health: Pioneer ACO Clinical Intelligence & Business Intelligence Approaches," Jan. 2013, accessed at http://www.himss.org/content/files/HIMSSAnalyticsPioneerACOAtriusHe althwhitepaper-Final.pdf. 3. Ibid. 4. Institute for Health Technology Transformation, "Population Health Management: A Roadmap for Provider-Based Automation in a New Era of Healthcare," 2012, accessed at http://ihealthtran.com/pdf/ PHMReport.pdf. 5. Dustin Charles, Michael Furukawa, and Meghan Hufstader, "Electronic Health Record Systems and Intent to Attest to Meaningful Use among Non-federal Acute Care Hospitals in the United States: 2008-2011," ONC Data Brief, no. 1, February 2012, accessed at http://www.healthit.gov/media/pdf/ONC_Data_Brief_AHA_2011.pdf. 6. Department of Health & Human Services, "More doctors adopting EHRs to improve patient care and safety," news release, Dec. 12, 2012. 7. HIMSS Analytics, "Clinical Analytics in the World of Meaningful Use," Feb. 2011, accessed at http://www.himss.org/content/files/20110221_Anvita.pdf. 8. HIMSS Analytics, "Clinical Analytics: Can Organizations Maximize Clinical Data?" June 7, 2010, accessed at http://www.himss.org/content/files/Clinical_Analytics.pdf. 9. "Clinical Analytics in the World of Meaningful Use." 10. James E. Gaston, "Clinical and Business Intelligence Survey," HIMSS Analytics, June 2012, accessed at http://www.himss.org/content/files/ClinicalandBusIntelWPSponOracle-Final.pdf. 11. "Atrius Health: Pioneer ACO Clinical Intelligence & Business Intelligence Approaches," op. cit. 12. "Clinical Analytics in the World of Meaningful Use." 13. Based on the websites of Explorys, MedVentive/McKesson, Phytel/Verisk, and Humedica/Optum. 14. Ken Terry, "New Query Tool Searches EHR Unstructured Data," Feb. 14, 2013, accessed at http://www.informationweek.com/healthcare/electronic-medical-records/new-query-tool-searches-ehrunstructured/240148634. 15. Amanda Parsons, Colleen McCullough, Jason Wang, Sarah Shih, "Validity of electronic health record-derived quality measurement for performance monitoring." J Am Med Inform Assoc doi:10.1136/ amiajnl-2011-000557. 16. Adam Wright, Justine Pang, Joshua C. Feblowitz, Francine L. Maloney, Allison R. Wilcox, Karen Sax McLoughlin, Harley Ramelson, Louise Schneider, David W. Bates, "Improving completeness of electronic problem lists through clinical decision support: a randomized, controlled trial." J Am Med Inform Assoc doi:10.1136/amiajnl-2011-000521. 17. "Clinical Analytics: Can Organizations Maximize Clinical Data?" 18. Frost & Sullivan, "Clinical Trial-based Evidence Vital to Obtain Venture Capital Funding for Wireless Patient Monitoring Technologies," press release, Feb. 14, 2013. 19. "Population Health Management: A Roadmap for Provider-Based Automation in a New Era of Healthcare." 20. J. Frank Wharam and Jonathan P. Weiner, "The Promise and Peril of Healthcare Forecasting." Am J Manag Care. 2012;18(3):e82-e85. 21. Ian Duncan, Healthcare Risk Adjustment and Predictive Modeling (Winstead, CT: ACTEX Publications, 2011. 22. Office of the National Coordinator of Health IT, Factsheet, Colorado Beacon Consortium (Grand Junction, CO), Oct. 25, 2012. 23. Shara Tibken, "Numbers, Numbers, and More Numbers," Wall Street Journal, Feb. 14, 2013. 24. Alicia Caramenico, "Hospitals Cut Readmissions With Predictive Modeling," FierceHealthcare, April 30, 2012, accessed at http://www.fiercehealthcare.com/story/hospitals-cut-readmissions-predictivemodeling/2012-04-30. 25. Health2Resources, "Predict, Prioritize, Prevent: Nine things practices should know about risk stratification and panel management," Issue Brief, vol. 2, issue 2, accessed at http://origin.library. constantcontact.com/download/get/file/1101292888704-1346/CBC-v2%232.edits2.pdf. 26. "The Promise and Peril of Healthcare Forecasting." 27. HIMSS, "Banner Health Network: Pioneer ACO Clinical Intelligence & Business Intelligence Approaches," October 2012, accessed at http://www.himss.org/content/files/PACOBannerWhitePaper11-052012-FINAL.pdf. 28. "Atrius Health: Pioneer ACO Clinical Intelligence & Business Intelligence Approaches." 29. Office of the National Coordinator for Health IT, Factsheet, Tulsa Beacon Community (Tulsa, OK). 30. Office of the National Coordinator for Health IT, Factsheet, Greater Cincinnati Beacon Collaboration (Cincinnati, OH). 31. ONC factsheets for Tulsa and Colorado Beacon Communities, op. cit. 32. Dartmouth Institute for Health Policy and Clinical Practice, The Dartmouth Atlas of Health Care, accessed at http://www.dartmouthatlas.org/. 33. DST Health Solutions, the Johns Hopkins ACG System, white paper, accessed at http://dsthealthsolutions.com/resourcecenter/JHU_ACG_System.pdf. 34. "Predict, Prioritize, Prevent: Nine things practices should know about risk stratification and panel management." 35. Michelle McNickle, "Doctors Push Back Against Health IT's Workflow Demands," Feb. 14, 2013, accessed at http://www.informationweek.com/healthcare/interoperability/doctors-push-back-againsthealth-its-wor/240148522. 36. Ken Terry, "Is Healthcare Big Data Ready for Prime Time?" InformationWeek Healthcare, Feb. 18, 2013, accessed at http://www.informationweek.com/big-data/news/big-data-analytics/is-healthcare-bigdata-ready-for-prime-time/240148371?utm_campaign=iht2-digest&utm_source=hubspot_email_marketing&utm_medium=email&utm_content=6992916&_hsenc=ANqtz-9L4wkFXtFIoLOGCGPn2i7LzYy UkpvocrBP85FEhp2-zg4euZ4vF02546WeQuZH2XNOp7E9S27HaK6GWHv3wrdez_4vyA&_hsmi=6992916 22 About The Institute for Health Technology Transformation The Institute for Health Technology Transformation (IHT2) is the leading organization committed to bringing together private and public sector leaders fostering the growth and effective use of technology across the healthcare industry. Through collaborative efforts the Institute provides programs that drive innovation, educate, and provide a critical understanding of how technology applications, solutions and devices can improve the quality, safety and efficiency of healthcare. The Institute engages multiple stakeholders: • Hospitals and other healthcare providers • Clinical groups • Academic and research institutions • Healthcare information technology firms • Healthcare technology investors • Health plans • Consumer and patient groups • Private sector stakeholders • Public sector stakeholders complex, under utilized and often misunderstood process. Stakeholder collaboration underscores the Institute’s focus working to ensure technology has a transformative effect at all levels of the healthcare sector. What We Do The Institute for Health Technology Transformation (iHT2) provides programs that drive innovation, educate, and provide a critical understanding of how technology applications, solutions and devices can improve the quality, safety and efficiency of healthcare. We do this though a number of vehicles including: educational workshops, access to industry thought leaders, peer reviewed research, high level conferences, webinars, focus groups, topic specific committees, and other unique initiatives allowing individuals and organizations access to resources that will enable them to leverage the full value of healthcare technology. Mission and Vision The mission of the Institute for Health Technology Transformation: to drive improvement and the effective use of technology throughout the continuum of care through education and collaboration among multiple stakeholders. Technology in-and-of itself will not solve the deep challenges facing our healthcare system nor will it alone ensure more accessible and higher quality care. Realizing the benefits of technology across the healthcare continuum is a Institute for Health Technology Transformation 23 Institute for Health Technology Transformation 244 5th Ave #2150 New York, NY 10001 © 2013 Institute for Heatlh Technology Transformation. All rights reserved