The Costs of a National Health Information Network

Context The United States needs a national health information network (NHIN). To build one, we need realistic estimates of costs. Contribution An expert panel conceptualized a model NHIN and determined the costs of implementing the model throughout the United States. The model NHIN would require $156 billion in capital investment over 5 years and would incur $48 billion in annual operating costs. Cautions The authors used expert opinion to estimate some costs and assumed fixed prices for hardware and software and no major new technological developments. Implications The United States probably needs to spend more now if we want to implement an NHIN in the next decade. The Editors The Institute of Medicine (IOM) drew attention to the importance of patient safety through its landmark report, To Err Is Human (1), and highlighted the central role that information technology (IT) must play in improving the quality of our health care system (2). Information technology can help achieve the IOM's goals of more safe, effective, patient-centered, timely, efficient, and equitable health care. However, health care lags behind other industries in investment and use of IT in its frontline processes. As suggested in the IOM report Crossing the Quality Chasm (2), a national health information network (NHIN) is critical for advancing IT in health care today. However, despite much useful dialogue, the structure of the NHIN is still being defined and its costs are even more uncertain (3). Increasing public policy attention is being directed toward an NHIN. In the 2004 State of the Union address, President George W. Bush noted that by computerizing health records, we can avoid dangerous medical mistakes, reduce costs, and improve care (4). In late April 2004, he created a new position of National Health Information Technology Coordinator at the U.S. Department of Health and Human Services, subsequently naming Dr. David Brailer to this role. In July 2004, the second annual NHIN conference was convened, culminating in testimony to the National Committee on Vital Health Statistics. This issue has received bipartisan support. For example, Senators Bill Frist and Hillary Clinton coauthored an editorial in The Washington Post calling for more IT in health care (5). Other governments, including those of the United Kingdom and Canada, have recently made major investments in health care information infrastructure (6, 7). The United Kingdom's government has allocated 8 billion, and Canada's government has invested $1.5 billion Canadian (although it is expected that much more will be necessary). The government of the United Kingdom plans to fund nearly all the required investment, whereas the government of Canada plans to catalyze investment by providing some central support but also requiring matching funds. Because the United States may make a similar investment, we sought to make projections about the potential technical structure and costs of a model NHIN. Our specific aims were to define the structure of a model NHIN in terms of functionality and interoperability, to estimate its costs, and to determine how much more would be required to achieve a model NHIN than that likely to be spent if historical spending trends persist. Methods Overview An expert panel delineated a model NHIN, defined as an achievable and desirable NHIN in 5 years rather than an ideal infrastructure. We then estimated the costs of achieving a model NHIN, defined as the costs of moving from current levels of IT investment to a model NHIN in 5 years. An NHIN must have 2 components: the ability to perform key functions, such as computerized physician order entry (CPOE), and interoperability, such as linking providers for the purpose of data exchange. We elected to separately assess the functionality and interoperability costs. We based the functionality costs on expert panel estimates of the percentage of key providers that currently have specified IT functionalities, the percentage of providers that the expert panel anticipated will have those functionalities within 5 years without a major influx of either money or policy pressure, and the cost of implementation in each domain. Supplementing the expert panel's estimates with data on the number of facilities within each provider group, we extrapolated costs to the national level by determining the cost to evolve from the current level of functionality to a model NHIN. We also sought to estimate the amount that would be expended by our health care system if current trends persist over the next 5 years and no policies are implemented to change these trends. To estimate the national costs of achieving interoperability among key providers, we used the technical approach and experience of the Santa Barbara County Data Exchange (SBCDE) as a template (8). The SBCDE is a network of health care providers within Santa Barbara County, California, linked through a central host to allow data exchange. Although this is only one of many potential methods for clinical data exchange, the network is likely to be effective and good data on costs are available. To extrapolate the SBCDE experience to the national level, we estimated the cost of replicating the SBCDE nationwide to create regional networks, then added a layer of super and national hosts to allow data exchange between the regional networks. Providers and Functionality Domains We convened a panel of IT experts (listed in the Acknowledgments section) to develop a model of an achievable NHIN within 5 years given current technology. This panel consisted of health care IT experts from industry, academia, and government. We identified the most important providers and the critical functional domains of an NHIN. The expert panel reached consensus by using a modified Delphi approach (9-11). The experts' opinions generally converged. After considering a wide range of providers, the expert panel identified the key providers as physician offices, hospitals, skilled nursing facilities, home health agencies, clinical laboratories, payers, and pharmacies. The panel focused on functionalities available to health care providers rather than patients, because health care providers make most decisions on investments for health care IT. Physician offices were segmented into small practices with 1 to 4 providers, medium practices with 5 to 20 providers, and large practices with more than 20 providers. Similarly, hospitals were divided into small hospitals with 300 beds or fewer and large hospitals with more than 300 beds. Providers or Facilities We estimated the numbers of facilities within each provider group by using data from the U.S. Census Bureau's County Business Patterns, 2000 (12). We supplemented these data with data from the National Center for Health Statistics (13), the U.S. Census Bureau's 1997 Economic Census (14), and other published sources to augment and adjust provider data (15). Functionalities The panel also identified a set of critical functional domains for a model NHIN: inpatient and ambulatory result viewing, inpatient and ambulatory electronic health record (EHR), inpatient and ambulatory CPOE, electronic claims submission, electronic eligibility verification, secure electronic patient communication, and electronic prescription acceptance by pharmacies. In general, each of these functional domains was relevant for only a subset of the key providers. We describe these functional domains in more detail in another paper (16). Briefly, results viewing allows electronic viewing of test results, such as laboratory tests or radiologic examinations. Electronic health records are computerized systems that maintain relevant health information, including electronic charting. Computerized provider order entry refers to an application that allows all medical orders to be entered electronically. Electronic claims submission and eligibility verification are methods of computerizing communications with third-party payers. Secure electronic patient communication refers to computerized e-mail or messaging systems that allow private communication between patients and their health care providers. Finally, pharmacy electronic prescribing refers to the ability of a pharmacy to accept electronic prescriptions. The expert panelists achieved very good consensus during the development of a model NHIN. Functionality Estimates The expert panel estimated the current state of IT functionality and the expected state in 5 years if we continue on our current trajectory of investments, along with 95% CIs around their projections. To simplify the response burden, the expert panel estimated an overall average by provider type. However, since physician offices and hospitals were divided into subgroups, we used the expert panel's average and 95% CIs to impute functionality percentages for these subgroups. We constrained the imputation so that a weighted average of the subgroup functionalities would equal the overall average estimated by the expert panel. The expert panel's consensus surrounding the estimates of IT functionalities was good, except for 1 member who felt unable to provide informed estimates. Cost Estimates The expert panel determined average estimates of functionality cost after a presentation of a summary of the published, peer-reviewed literature on health care IT costs. The literature has many limitations. We assumed annual operation and maintenance costs to be 25% of capital costs on the basis of expert consensus. Most physician office costs were estimated on a per physician basis, while other capital costs were estimated on a per facility basis. We based the costs of claims and eligibility processing, patient communications, and pharmacy acceptance of electronic prescriptions on the numbers of transactions. We estimated the volume of transactions as a function of outpatient visits, hospitalizations, and similar events that were likely to trigger a transaction. We multiplied the volume of t