At a Glance
An effective business intelligence (BI) system includes four different information strategies:
- Population health analytics
- Risk-based cost analysis
- Performance analytics
- Care management
Healthcare reform is spurring hospitals and health systems, physician groups, and other provider organizations to collaborate to deliver more coordinated care through the creation of clinically integrated networks (CINs) with a focus on population health management. These CINs require meaningful information they can use to proactively manage patient care, utilization, and costs for a defined population. This trend is causing business intelligence (BI) to quickly become a critical component of healthcare financial management.
Yet several obstacles stand in the way of developing effective BI capabilities. First, obtaining complete data is a perennial challenge. The necessary data usually exists in multiple information systems owned by different stakeholders. Second, the data needs to be organized before it will yield useful insights. In particular, most organizations still struggle to tie cost data to clinical data. Third, information alone will not improve care or reduce costs; programs will be required that make use of BI to impact patient care by reducing the cost and improving the quality of that care. Using BI in this way ultimately will provide a basis for CINs to negotiate risk-based contracts that reward participants for controlling costs and improving care quality and patient outcomes.
Finance leaders can play a major role in using BI to reduce care costs. The first step is to understand the four different information strategies that make up an effective BI system.
The Four Levels of Healthcare BI
Effective population health management begins with a broad view of community health and gradually narrows its scope to individual patient care. From a BI standpoint, four information goals and strategies are key.
Population health analytics. The first objective is to understand the health trends within your community and how they affect costs. What is the chronic disease footprint within the population you serve? Population health analytics is the process of quantifying risk factors within your service population and using predictive modeling to understand the cost implications. For example, what percentage of your population is obese? How does that compare with national figures? Based on the literature, what are your predicted costs in terms of cardiology, endocrinology, and other services?
Population health analytics also can identify referral patterns. One issue is leakage. If a CIN’s assigned beneficiaries receive 40 percent of their services outside of the network, it will be difficult to manage that population effectively. Identifying referral patterns also is critical to effective care management and efforts to influence care quality. If a CIN has established a heart care program, for instance, it should be able to identify patients who are good candidates for these services.
Risk-based cost analysis. Once you understand the chronic disease drivers within your population, the next step is to stratify the population into risk cohorts. Based on health status metrics and other indicators, you can segment the population into groups by risk level. High-risk patients are generally responsible for the greatest costs, particularly when they are admitted to the hospital.
Case in point: Analysis of a CIN’s diabetic population might reveal that 80 percent have well-controlled disease (low risk), 15 percent have uncontrolled diabetes (medium risk), and 5 percent have uncontrolled disease complicated by multiple comorbidities (high risk). Segmentation analysis provides guidance on where to invest resources to reduce costs and improve outcomes.
Performance analytics. Once you construct a detailed picture of your population and your risk groups, the next step is to establish metrics for tracking patient health, engaging providers, planning interventions, and evaluating results. For example, an organization focusing on diabetic health might track the percentage of patients who have undergone A1C testing and have received foot and eye exams (a process quality metric) and the percentage of patients with A1C below a defined threshold (a clinical outcome).
Organizations should establish performance measures cooperatively with a broad range of clinical stakeholders, not just the hospital CMO or department chiefs. Successful CINs have established clinical governance bodies that coordinate input from all physicians within the network. In turn, physicians receive regular provider scorecards that quantify their performance on the network’s measures.
Care management. Once you understand patient health at the granular level, the next step is to create an infrastructure for driving patient interventions. This infrastructure includes clinical coordinators who use IT tools to manage patient care according to defined care pathways and clinical protocols.
Many care management programs focus on coordinating care for patients with complex, high-risk conditions. For instance, an organization might create a special program for patients with chronic obstructive pulmonary disease (COPD). A nurse navigator may use care gap reports and patient management tools to coordinate patient outreach, education, and referrals in this population.
Assembling the Data
BI requires extensive data, and assembling it can be a technical challenge. It is also a political challenge, requiring healthcare leaders to establish trust and shared goals among many different stakeholders. The initial focus should be on two separate data sources: claims and clinical data.
Claims data. Healthcare claims provide financial information, but they also yield important clinical information on encounters, diagnoses, costs, and services provided. These details are an important source of information on population health.
Claims data can be obtained directly from participating hospitals, physicians, and post-acute providers in a CIN. Unfortunately, this data describes only what is happening inside the CIN. To the extent that your population receives services outside of your CIN, you lose sight of care, costs, and outcomes. The solution is to obtain claims data directly from payers.
Clinical data. Claims data only goes so far. CINs need clinical metrics to explain what is driving services so clinicians can implement effective interventions. For example, many organizations try to capture hypertension metrics for network patients. These metrics alert the organization when a patient becomes high risk and enable clinical coordinators to intervene appropriately with additional targeted services.
As mentioned, clinical data typically resides in several different acute and ambulatory EHRs and ancillary information systems. These platforms usually do not share a common “language” for recording and exchanging information. Aggregating clinical data from these disparate systems requires a technology infrastructure that integrates well with providers’ clinical workflow, and a plan for ensuring that data entry results in clean, standard data sets.
Assembling the Technology
BI programs use a multilayer technology infrastructure that takes raw data and converts it into usable information. Basic components include health information exchanges (HIEs), disease registries, and reporting systems.
HIEs. An HIE is a system that extracts clinical data from multiple EHRs and aggregates it along with claims data. An HIE usually includes a data repository that stores clinical data and documents.
Disease registries. Once clinical and financial data have been aggregated from all sources, they should be harmonized and organized. A disease registry is essentially a data warehouse for healthcare information.
The disease registry aggregates data from various systems and assigns it to individual patients through an enterprise master patient index (EMPI) solution. The disease registry also establishes a common format for categorizing information. For instance, raw clinical and claims data may include information on a dozen different statin drugs. But from a population health point of view, you want to be able to track and manage all statin patients as a group. A properly configured disease registry can sort all drug data by defined drug classes.
Reporting systems. Disease registries can be configured with multiple subregistries, which are customized data sets that feed various reports. Key reports include population health indicators, provider performance scorecards, care gap reports for planning interventions, and financial reports for tracking costs related to care.
Case Study: ACO Uses BI to Lower Costs
In 2012, a hospital system in the Southwest partnered with a large primary care medical group to form an accountable care organization (ACO). The organization focused initially on a Medicare population of 15,000 beneficiaries, and participating providers had an opportunity for shared savings. Because ACO participants were independent, there was little existing integration between physicians and hospital providers.
The first step for the ACO was to aggregate data from multiple sources. The Medicare Provider Analysis and Review (MedPAR) file of the Centers for Medicare & Medicaid Services (CMS) provided a source for complete claims files for all beneficiaries on a monthly basis. The ACO pulled clinical data from the hospital inpatient EHR, specialty clinic EHRs, hospital laboratory and e-prescribing systems, and the medical group EHR. To aggregate data within an HIE, the ACO used a patient ID solution. The HIE served as a BI gateway, providing the ability to export data based on various criteria.
Leaders used a clinical disease registry to evaluate and analyze the ACO population by costs and various clinical indicators. They stratified the population into several groups: low risk, moderate risk, high risk, and end of life. In addition to CMS quality measures, the disease registry aggregated Physician Quality Reporting System and meaningful use data as well as a variety of homegrown metrics to track performance.
ACO leadership used multidimensional analytics to craft care management strategies tailored to specific risk groups.
Standard risk. The focus for standard-risk patients was prevention. Using care gap reports generated by the BI system, clinical coordinators identified beneficiaries who had not received recommended screenings and referred them to the appropriate services.
Moderate risk. ACO leaders used clinical data to identify patients with controlled chronic diseases—for example, heart failure patients whose condition was stable and being well managed. Nurse navigators maintained ongoing contact with these patients to help them maintain their health. Close monitoring helped ensure these patients did not progress to high-risk status.
High risk. Most costs within the ACO were driven by patients with multiple complex medical issues or complicated diabetes. These patients were referred to a special extensivist clinic staffed by physician-led teams that coordinated care across multiple settings. Key tools included extended patient visits, comprehensive treatment plans, timely access, and preventive services. The extensivist team placed special emphasis on care transitions and monitoring patients post-discharge.
The entire care management effort was coordinated with the use of a customer relationship management (CRM) system. Fed by the disease registry, the CRM tool enabled clinicians and care coordinators to identify patients within specific programs, initiate risk-specific interventions, engage patients, and track process outcomes.
One-year results were encouraging. The ACO reduced costs for high-risk patients by 7 percent. Now in its second year, the organization is on track to realize higher savings across all risk groups. It also plans to expand its ACO program to private payers.
BI and Financial Strategy
Building a high-performing BI system requires extensive collaboration with other provider organizations. But one of the advantages of an effective BI program is that it can allow cooperating providers to remain independent.
BI enables healthcare organizations to pool their resources, provide orchestrated care, and track outcomes. For example, three Philadelphia-area health systems recently joined forces to develop a shared population health strategy. The consortium is using BI to improve outcomes and reduce costs. Its starting point is managing the health benefits of the systems’ combined 32,000 self-insured employees. The three participating systems remain independent but are positioned to share any savings generated by their combined efforts.
Ultimately, BI capabilities may support efforts by this and other provider partnerships to contract jointly for value-based payment. By enabling providers throughout a region to cooperate on reducing costs, strong BI systems may help organizations meet the challenges of healthcare reform.
Daniel J. Marino is president and CEO, Health Directions, LLC, Chicago, and a member of HFMA’s First Illinois Chapter.
Publication Date: Monday, March 03, 2014