Thursday, May 29, 2008

The May, 2008 CBO Report: Evidence on the Costs and Benefits of Health Information Technology (Publication number 2976).

The recently published report by the Congressional Budget Office summarizes well the dilemma faced by those who argue for adoption of health information technology based on formal analyses:

No aspect of health IT entails as much uncertainty as the magnitude of its potential benefits. Some analysts believe that the adoption of such systems could provide substantial savings by lowering the cost of providing health care, eliminating unnecessary health care services (such as duplicate diagnostic tests), and improving the quality of care in ways that might reduce costs (by diminishing the likelihood of adverse drug events, for example). Other analysts expect little effect on costs but some improve-No aspect of health IT entails as much uncertainty as the magnitude of its potential benefits. Some analysts believe that the adoption of such systems could provide substantial savings by lowering the cost of providing health care, eliminating unnecessary health care services (such as duplicate diagnostic tests), and improving the quality of care in ways that might reduce costs (by diminishing the likelihood of adverse drug events, for example). Other analysts expect little effect on costs but some improvement in the quality of care. Another school of thought holds that health IT could bolster the quality of care but also increase expenditures on health care services— because improvements in quality would stimulate demand for additional services.

CBO's critique of the RAND report is in part due to the difference in mission. RAND more or less measures an idealistic potential outcome. CBO must focus on the likely outcome and the true impact of Congressional intervention. Take gravity as an example. RAND would claim that the impact of a falling rock will depend on mass, resistance, and gravitational pull. CBO would claim that little Congress can do will impact the mass, the velocity, or the resistance. (CBO would perhaps explore the cost of measures mandating attaching a parachute to the rock and the resulting effect on impact.)The RAND research focused primarily on savings that the use of health IT could generate by reducing costs in physicians’ practices and hospitals and hence had a different scope. RAND also ignored neutral or negative reports. CBO questions their rationale for doing this.

The RAND analysis itself notes that its estimate is of health IT’s potential savings and costs depends on an optimal - but perhaps not likely - state : "We use the word potential to mean ‘assuming that interconnected and interoperable EMR systems are adopted widely and used effectively.’ Thus, our estimates of potential savings are not predictions of what will happen but of what could happen with HIT and appropriate changes in health care.”

RAND assumed a constant rate of adoption. CBO would assume the increasing rate of adoption currently observed. Since CBO's savings estimates are based in part on the extent to which legislation encourages adoption, savings due to adoption alone will be exaggerated by the RAND report relative to the CBO report.

RAND assumes that internal hospital costs are reduced to an extent proportional to hospital stay. That is, if the length of stay is reduced by 10%, so will the costs. Although this may be true for fixed costs (beds, personnel) as currently measured, the variable costs are unlikely to change significantly since in many respects throughput efficiencies are realized by performing the same tests and providing the same intensity of services as would have been provided anyway. If a patient is improving, variable costs of care decrease closer to discharge. If the patient is failing and subject to increasingly intense interventions in a effort to save a life, variable costs can be expected to increase until either death or clinical improvement ensues.

Savings incurred by replacing paper medical records with electronic health records exhibit the same dependency on fixed and variable costs. If an organization is large, changes in variable costs can translate to lower labor costs. If an organization is small - a solo practice, for example - personnel costs are essentially fixed and financial benefits are only realized if volume is increased and revenues are therefore enhanced using the same labor pool.

One criticism of the RAND study seems questionable to this reviewer. Using a traditional microeconomic argument, CBO assumes that lower costs will lead to lower prices and, consistent with economic theory, lower prices mean higher demand. This demand would offset some savings. It would be welcome news to see this theory play out where prescription drug adherence is concerned. Many reports suggest that individuals do not take beneficial cholesterol-lowering or other disease-prevention drugs in part because of the relative costs of these drugs; in general, rather than taking such medications monthly for a sometimes indefinite period, individuals take such drugs intermittently or cease taking them after several months. Here, lowered prices through greater efficiency may lead to greater adherence programs - particularly if other means of reminding and encouraging patients can be identified. But translating this to broader interventions is perhaps questionable.


The CITL study limited its scope to savings from achieving full interoperability of health IT, explicitly excluding potential improvements in efficiency within practices and hospitals. It essentially measured the long-term (15 year) savings when migrating from a theoretical state of complete paper to an equally theoretical state of idealized comprehensive use; it compares the potential savings from a bygone era with a Nirvana which may never be realized.

CBO echoes criticisms already made on unrealistically high estimates for laboratory administrative expenditures. It made equally generous assumptions about high degree of redundant test elimination. CITL, for example, assumed virtually all telephone costs would be eliminated from e-prescribing transactions

General Methodology

CBO helpfully distinguishes between the dynamic of savings incurred "internally" from those that are "externalities". They state:
  • Internal savings are those that can be captured by the provider or hospital that purchases the system; they are most likely to be in the form of reductions in the cost of providing health care—that is, improvements in the efficiency with which providers and hospitals deliver care.
  • External savings are those that the provider or hospital that purchases the system cannot realize but that accrue to another such provider or perhaps the relevant health insurance plan or even the patient. Such savings might arise, for example, from the newfound ability of participants in the health care sector to exchange information more efficiently.

CBO points out the relative maturity of internal cost savings examples within highly integrated delivery systems like Kaiser but emphasize the relative immaturity of knowledge where more disparate settings are incurred. They recognize that where health information exchanges are concerned, it is early in the game. They state "estimating the impact of some potential sources of savings, especially those arising from greater exchange of information among providers, insurers, and patients, is especially difficult because health IT networks are in an early stage of development."

Yet CBO does mention the positive effects. Their argument is one of degree. Among the areas in which savings may be realized are:

  • Eliminating paper records. Savings are less if practices are small (i.e. high fixed personnel costs) or if systems are not used effectively.
  • Reducing transcription services
  • Avoidance of duplicate tests. Here the issues are those of actual availability of previous test results; awareness of the test results; and avoidance of revenues if providers have an incentive to repeat the tests. Most estimates of redundant tests are small. These may be larger if test results from a broader range of sites are available
  • Reduction of radiology services. CBO suggests that most evidence supports the actual type of test ordered but little in the way of test reduction. (Our experience in Memphis suggests that some high-cost tests are reduced using a regional exchange).
  • Prescription drug costs. Although CBO mentions lack of incentives, their most compelling argument is that because of "strong incentives," health plans and PBMs "may already be capturing a substantial portion of those savings."
  • Improved productivity. Most studies are restricted to isolate roles and not to overall system productivity. Demonstrations of productivity are more compelling in some hospital environments than in ambulatory environments. Transcription costs can be reduced significantly.
  • Reduction in hospital length of stay. CBO points out that slight reductions in length-of-stay are not commensurate with cost reductions. They also emphasize correctly that the current financial incentives for LOS reduction are sufficient to accelerate more effective adoption independent of legislative imperative. Adoption just makes good business sense.
  • Quality. The connection between quality and EHR use remains conjectural but presents many long-term opportunities. One cannot improve what one does not measure so the extent to which one measures the right thing affords promise. CBO discounts the potential savings primarily because they are conferred upon patients and intermediaries rather than those who make the initial investment.
  • Adverse drug events. CBO reviews the conflicting data on in-patient settings and the dearth of data on ambulatory settings. CBO recognizes that "avoiding even a fraction of the errors that now occur in inpatient and outpatient settings could yield significant savings" but claims that some of these errors may be addressed by current systems. One could argue, however, that although drug-drug interactions are common in pharmacy systems and there are multiple fail-safe mechanisms, comprehensive drug-disease, drug-allergy, and drug-lab interactions are in a relatively primitive state because of lack of standards (e.g., for allergies), lack of uniform clinical decision support mechanisms, and lack of widespread incorporation of these data in uniform ways.
Although lowering of administrative costs are mentioned in a systemic sense, CBO does not mention revenue cycle acceleration and coding accuracy - one of the principle drivers for some practitioners.

CBO brings into light the theoretical value of incorporating clinical decision support and evidence-practice into systems in a effort to improve quality on a systematic basis. Although results again are conflicting at this early stage, the potential seems intuitive.

The CBO report also emphasizes research and comparative effectiveness. The report states: "And some potential areas of research and analysis remain largely unexamined. They include the ways in which the delivery of health care services might change in response to the efficiencies that health IT offers and how the large amounts of clinical data available through EHRs could contribute to analyses of the comparative effectiveness and cost-effectiveness of different treatments." Similarly, public health reporting and disease surveillance are in their early stages; the potential is there but the jury is out.

CBO report also includes a good summary of costs incurred to incorporate health care technology into ambulatory practice. Total initial costs are said to range from $25 - $45 per physician; ongoing costs range from 12-20%. New subscription based systems may eliminate much of the up-front costs and lower annual operating costs. Smaller practices incur higher per practitioner costs. (Even ASP models will not lower the per-clinician training costs and the considerable practice transformation opportunity costs).

Explaining Low Rates of Adoption
The CBO Report poses many explanations. These include the significant investments in choosing and implementing systems, the lack of identifying measurable financial returns. Although the authors are skeptical of adoption based only on quality or revenue improvements, they suggest that physicians "might change their thinking if they knew that they would be directly compensated for implementing a health IT system or if they could report data on the quality of care that they provided—data for which they were being compensated— only by using such a system."

The CBO report mentions the "free rider" effect of support through intermediaries. Because technology should help practitioners provide care for al of their patients, once a technology is adopted, all intermediaries will benefit whether such intermediaries made an initial investment or not. Only uniform means of enforcing certain outcomes or performance measures would seem to provide an incentive for all intermediaries to participate. This degree of coordination should not be expected and may, in the eyes of some, be viewed as collusion."

CBO identifies two major roles played by the federal government. The first is as a payer; federal contributions to Medicare and Medicaid account for approximately 25% of the total American health car spend. The second role is as a market-maker. Health care may be a public good that cannot be optimally conferred without a federal-government that eliminates the "free-rider problem. Combined, federal roles provide a network effect that, properly managed, could create new opportunities for lower costs and higher clinical values. The returns - one could argue - are to society as a whole.

But the government is not neutral in all of this. By virtue of the CBO charge to examine the marginal benefit of federal intervention to accelerate a trend that is already underway, one has difficulty identifying those who really change their behavior on the basis of a federal incentive from those who would change their benefit anyway. If one wishes to subsidize motive rather than results, one must end with the treacherous logic that favors punishment over incentive. Indeed, the only way to make sure one is not providing a causative reward for uniquely good behavior is to avoid all rewards for good behavior and instead focus on punishing for undesired behavior. But this response is also flawed. One could argue that some who do not adopt technology do so for good reasons but all would be punished without respect to motive.

The CBO report provides an excellent summary of the current challenges in identifying the unique value of health information technology. Left with this conundrum, it is easy to fall into the trap of accelerating mandates and punitive schemes rather than accelerating the need to address the many unanswered questions raised by the current state of technology in health care. The CBO report plainly demonstrates that we do not yet understand how a federal effort can be ideally structured because we do not clearly understand the cause and effect of the various interventions. But the CBO does not refute an obvious conclusion: making this system work and moving from clearly confusing and possibly dangerous paper-based approaches is both essential and inevitable.

This is a report worth reading carefully. The CBO is to be congratulated for their public contribution.

In a talk delivered to the Markle Foundation Connecting for Health Steering Group on June 19, Dr. Blackford Middle noted the report gave a "reasonable review and summary of the literature on HIT value." Some CITL report aspects "were not represented well." These include:
  • A fundamentally different purpose for the report - the impact of funding
  • Value of un-standardized (level 3) vs standardized (level 4) interoperability
  • They failed to note how CITL accounted for the current HIT context; CBO factored in existing provider-payer data exchanges, and existing lab and pharmacy integration
  • CBO treated costs of providers information systems but inadequately treated the internal benefits; their notion of data exchange was relatively restricted.
  • CITL differed with CBO on lab administration costs; phone call rates - but these were not key determinants of overall value
  • CBO's critique did not discuss some of the limitations of HIEI model limiations. The CITL model was more expansive and included realizing savings through quality improvements and the potential clinical benefit.
In the same discussion Zoe Baird noted that it isn't so much the report as how it is depicted to the public. Initial reports were rather unitarian in the view declaring that benefits are not there. CBO - recruiting a broad and talented array of health care economists - has emphasized a broader and more constructive mesage.


Blogger Kevin.johnson said...

Wonderful summary, Mark. I agree that the CBO report really gets at the heart of where we as a field need to focus. To me, the merging of health services research and biomedical informatics has never been more necessary. I hope we can see that happening soon.

June 11, 2008 7:05 AM  

Post a Comment

<< Home