MANAGED CARE July 2005. ©MediMedia USA
Pay-for-performance programs imply improved patient care, but are frustrated by fragmented data collection and reporting systems. Think big.
Our health care system is a patchwork quilt of separate — often competing — data repositories, data standards, disease management programs, high-risk patient management programs, and carved-out mental health and pharmacy management programs. Each insurer has data systems and care management programs that are different from the next insurer, and often even a single insurer has differing and incompatible programs and systems.
Although the goal of all these programs is to improve care through greater knowledge, efficiency, and consistency, the actual effect of all these programs and data systems is often to further confuse an already fragmented health care system. Unless we make renewed efforts to integrate the analysis and management of providers and all their patients, our opportunities to improve care will be severely limited.
Pay-for-performance programs will require us to do a far better job of profiling providers. Physicians will ignore incentive programs and outcomes will remain poor if profiling is not done with valid methods of measurement.
First, we need to undo the lessons of the past. The typical physician receives reports from several sources, but each profiles a small percentage of the physician's practice and suffers from small-number variability, inconsistent data definitions, and poor statistical validity. It is very common for reports to contradict each other, and for physicians to draw the reasonable conclusion that most of the performance information they receive is seriously flawed and not worth their time to evaluate.
To get physicians to trust information again, we need to aggregate data across insurers and pharmacy benefit managers and to develop consistent, statistically valid profiles. We need, quickly, to move beyond profiling that is based only on claims, and to integrate laboratory values and chart extraction into our profiling systems.
Profiling of specialists will expand under pay-for-performance programs. Their profiles are usually based on episodes of care, but determining who is responsible for a given episode is often difficult. Thus, a significant number of patient visits are often excluded from analysis (the number can be as high as 40 percent), which increases the number of patients needed to detect valid patterns of care compared with primary care profiles. Without data aggregation across payers, the statistical credibility of specialist profiling will be less than current primary care profiles. If we don't address this, we will not enlist specialists in care improvement, regardless of the incentives.
Successful models
Although we have a formidable task, there are successful models of aggregated care analysis. State-sponsored efforts, such as the New York State Cardiac Surgery Reporting System (CSRS), have long records of success. In addition, nongovernmental statewide collaboratives, such as Massachusetts Health Quality Partners, have created credible physician group profiling for groups of insurers. More and more vendors are providing reasonably-priced technology that gives providers and employers of all sizes the tools they need to aggregate patterns of care for all their insurers. Organizations that have put this technology to use have significantly improved efficiency and quality of care.
These examples demonstrate not only that we can aggregate data successfully, but that providers will cooperate if the profiling is statistically credible. In fact, physician groups are willing to take leadership roles if they see a real opportunity to improve data analysis and performance profiling. The California Association of Physician Groups' statewide data aggregation effort could be replicated in many states if leaders of local health care systems accepted the challenge.
Recognize limitations
Insurers, who often use data to compete, need to recognize the limitations of their data sets and to seek broad-based collaborations for analysis and profiling. Employers need to push their insurers to not only engage in these efforts but to help fund them. State medical societies, large statewide employers, and local physician leaders should be talking about and building such initiatives.
The time to do this is now. Experience suggests that statewide efforts can be implemented over a couple of years and drive real success. They are at least a reasonable interim step, if not necessarily a permanent solution.
Consolidating data is a place to start, not to end. The ever-increasing number of performance metrics and standards being introduced by insurers, regulators, and accrediting bodies is overwhelming our providers. The cost of reporting for hospitals, insurers, and physician groups is quickly becoming unsustainable. Successful improvement efforts usually start with a focus on a small number of key factors influencing performance. At present, we are actually working against this focus.
We need to define consistent metrics and standards for performance for the entire health care system. The long-standing success of the Institute for Clinical Systems Improvements (ICSI) in Minnesota and the growing influence of the Leapfrog Group are worth examining. State hospital associations, state medical societies, employer associations, and insurers should come together now to define standard sets of measurements and common guidelines.
Care management programs also need to be consistent, at least within a state. Identification of high-risk patients should be based on standard methodologies, not on the tools of competing vendors. Assistance with care coordination is valuable to physicians as well as to patients, but this is difficult for physicians to see because of the sheer number and variety of programs they work with.
Insurers could retain separate treatment programs, but case-finding methodologies and protocols for intervention should be standardized. Quality should be declared a noncompete zone for insurers, where all have to meet consistent standards and work collaboratively, not only with each other but with physicians. We should not allow wasteful competition in the area of health care improvement.
We also need to revisit mental health carve-out programs, which ignore the reality that mental health disorders are a major part of routine primary and specialty care. The last programs to improve primary care physicians' treatment of depression were shelved with the advent of mental health carve-outs, despite the fact that most depression is treated by primary care physicians. Employers have a key role to play here, defining what is acceptable for their insurers.
In the realm of prescribing, we need to take advantage of the consolidation of pharmacy benefit managers nationwide. Blue Cross Blue Shield of Massachusetts and Tufts Health Plan are working together to give real-time access through standard handheld devices to one centralized database that contains information on all medications prescribed for the plans' members, regardless of the prescribing physician.
Physicians in Massachusetts are finally moving into e-prescribing. With a complete prescribing history, often for the first time, physicians are finding new opportunities to improve patients' drug regimens. Insurers in other states should be actively considering similar collaborations.
Economies of scale
The failure of insurers to adopt common technology platforms raises the costs for all programs. We can increase the cost efficiency of care management programs by spreading their high-fixed systems costs over a larger number of patients. Clinical systems become more efficient and effective when they can go to one source for patient information.
The federal government has recognized this and has put out, in the last year, a request for information on the development of a national health information network. The responses to this RFI, published in June, clearly support the necessity of broad-based collaboration.
However, local initiatives should not wait for federal action. One of the overarching conclusions of the RFI responses is that there will still be a "need for implementation and harmonization at a local level." Local medical leaders should pursue collaborations to support common databases and fund broad-based EMR solutions.
We do not yet have national models for collaborative programs, and I don't know if we will in the foreseeable future, but we do already have several good models at the state level. It is clearly time for all health care leaders to learn from these groundbreaking efforts and to initiate and support these efforts in their own states. Without change, our current shift to pay for performance will fall far short of what can be achieved.