ders for providing comprehensive services. This has tempted some providers to under-serve their patients to increase revenues, according to HCFA. Providers also have less incentive to provide states with accurate and timely data. States, in turn, have found themselves under pressure to gather and evaluate data supporting both the traditional fee-for-service program and managed care. That's because states need reliable data to understand how best to improve the quality of managed care, which means monitoring, evaluating and influencing a managed-care plan's performance.
A few years ago, states began building decision-support systems (DSS) to help make sense of the vast data they were collecting and to speed up the decision-making process. A DSS consists of hardware, analytical software and a database that gives the user timely access to data that hasn't been duplicated, is consistent, has been integrated and has passed through standard integrity controls.
In 1996, HCFA surveyed the states and, out of 42 responses, found that 17 states already had a DSS in place or were building one. Another 22 states planned to have a DSS within the next two years. The survey found that technical analysts were the most frequent DSS users, in terms of creating and analyzing reports, while senior staff were least likely to use the system.
The typical DSS contained about two to three years of claims history -- that's about 13 million to 15 million claim lines per year -- but some systems were gargantuan. California, with more than 5 million recipients, had nearly 185 million claim lines, units of measure for stored Medicaid data, for each year in storage.
Most states used that data to develop ad-hoc reports. But nearly every state used its DSS to look at patterns of treatment, to profile hospitals and doctors and to measure the performance of providers. They compared fee-for-service coverage with managed care, ran "what-if" analyses of potential policy changes and analyzed trends with cost and use of health care.
Only Arkansas, Illinois, Mississippi, Michigan, Minnesota, New York, Ohio and Wisconsin have opted to pursue decision support through a data-warehouse strategy, although other states have data-warehouse projects in the development or planning stages.
There's nothing whiz-bang about data warehousing. The typical warehouse consists of data stores, which are databases containing the operational data; data conversion and extraction procedures that translate and convert disparate data into a standardized format; the warehouse itself, which is a powerful database optimized for online analytical processing; business-intelligence tools, which provide the interface for users to query data-warehouse information; and warehouse-administration tools for monitoring and support.
Data warehouses and other DSS differ in scale, performance and price. Data warehouses are often hundreds of gigabytes in size -- some a terabyte or more -- often running on powerful servers, sometimes with parallel processors to boost their performance. None of this is cheap, however. The hardware, software and technical expertise needed to make a warehouse run can cost tens of millions of dollars. But, according to Medicaid consultant Smith, they are worth it. "Medicaid is a huge program that dwarfs almost all others in state government. It deserves the best technology you can bring to it."
Apparently that's the philosophy in Illinois. In July, Gov. George H. Ryan announced that the state was building a $28 million data warehouse to help manage the state's $6 billion medical-assistance programs. The warehouse, which will be built by Bull, will be used by the Department of Public Aid to administer Medicaid and a children's program called Kidcare -- which both provide health coverage to nearly 1.4 million people each month. The department processes 25.7 million claims per year and tracks nearly 50,000 medical providers.
To improve its ability to analyze trends, the department will store five years' worth of data in the warehouse. "We're going to use the warehouse to try and understand the characteristics of a problem," said Theron Aflaksen, deputy director of administrative operations. "It may lead us to alter how we reimburse or who we have for providers."
The system, which will begin operation next year, is expected to give the department a much more accurate and detailed picture of its services and what's being spent on them. While the data warehouse won't be used to set policy -- that's up to the Illinois Legislature -- it will help the department ensure it gets the right kind of funding to cover any policy changes.
Not having that information could prove costly. "If we don't understand the full impact of a change, we could underestimate the cost and have a definite problem on our hands," explained Aflaksen. The warehouse should make sure that scenario doesn't happen.
It Starts with Fraud
If Illinois is like other states with data warehouses, it will use the system to tackle fraud and abuse as much as anything else. Medicaid experts estimate that the nation's Medicaid program loses about $20 billion annually to fraud. To combat the problem, experts have urged everything from empanelling more grand juries to look into the problem to mailing recipients and providers detailed explanations of services paid so honest individuals can help spot fraudulent billing. But the top solution is widespread use of computers that can search for patterns of abuse.
Texas and New York have developed high-profile systems. The Texas Health and Human Services Commission uses a neural network. New York's Attorney General's Office operates a 350GB data warehouse that more than 300 users can access.
In Michigan, data warehousing at the Department of Community Health began in 1994 as a surveillance tool to combat fraud. Today, with 750,000 out of 1.1 million Medicaid beneficiaries in managed care, the use of data warehousing has broadened. "We have gone from a reimburser to becoming a value purchaser of health care," explained Michigan's Olszewski. "We now set the standards and activities for providers, as well as manage contracts with them to ensure that the right services are provided."
To do that, analysts need to profile the providers through queries and computer reports. Prior to data warehousing, it could take a week or longer to receive a response to a query, according to Smith. "Within 30 minutes of receiving the report, the analyst often would have another question to ask. It was very frustrating." With data warehousing, however, answers come back in minutes.
Today, Michigan's data warehouse, also built by Bull, targets questions posed by the analysts ranging from tackling fraud and abuse to expanding child immunizations or cutting pharmaceutical costs. "With data warehousing, we can pose a question, look at the response and see whether we are on the money," explained Olszewski.
In Wisconsin, ad hoc queries rule. The state, which has the dubious distinction of having to build a warehouse twice, used to rely on canned reports from its mainframe that strictly limited the number of queries on a weekly basis. After its first effort at building a DSS failed in 1994, the state brought in EDS to finish the job properly.
Today, the Medicaid Evaluation and Decision Support (MEDS) system can handle hundreds of complex queries annually. More than 90 percent of the queries are ad hoc in nature, according to Diane Evenson, EDS' MEDS manager. "The warehouse has really improved the state's ability to budget and monitor Medicaid every month," she said. "It also helps them to make better decisions about programs, policies or changes in benefit levels. With rising costs, it's always a challenge to keep within budget."
The state has used the warehouse to show that recipients voluntarily enrolled in a managed-care program would not be the most needy and costly patients, as one managed-care provider asserted. When the data showed the opposite to be true, the state was able to negotiate better rates. MEDS has also helped the state force HMOs to provide more accurate information and to determine the cost-effectiveness of new types of health-care programs.
Knowing What Questions to Ask
The wealth of information pouring out of Medicaid data warehouses has its dark side. Officials talk of analysts who over-use the system, don't know how to pose questions to get the right answer, or don't understand the answer they have received. "We have found that several people will ask the same question and get different answers," observed Olszewski.
Lack of knowledge about the data inside a warehouse is the biggest issue in Wisconsin. "Some people ask questions without understanding what the response might be," said Evenson.
The solution is to create a help desk with someone in charge who's well- trained and can help users understand what the warehouse can and cannot do. That person can also show users how to best extract the data using the computer's query tools.
A more strategic problem states run into with their Medicaid data warehouses is the silo effect. Medicaid agencies need to share their warehouse with select agencies outside of social services and, conversely, integrate data from other agencies -- such as vital statistics from the secretary of state, or information from welfare services -- with health data to create a more comprehensive database.
According to Smith, states must look strategically at their health-care issues if they want to get a handle on costs while improving care. To do that, they need to collect and analyze data on everything from diabetes and heart failure to AIDS and asthma. "With the right kind of case management, costs can be reduced while care improves," he said, adding that states will have to look more carefully at this in the future.
So far, states have shown themselves to be very adept at processing information, Bull's Ginsburg pointed out. Now, they have to do a better job of analyzing that information. Since Medicaid is the largest item on two-thirds of all state budgets, that's not a trivial issue. With their use of data warehouses, states have taken a step in the right direction.