The adoption of interoperable EMR systems could produce efficiency and safety savings of $142–$371 billion.

by Richard Hillestad, James Bigelow, Anthony Bower, Federico Girosi, Robin Meili, Richard Scoville, and Roger Taylor

ABSTRACT: To broadly examine the potential health and financial benefits of health infor-mation technology (HIT), this paper compares health care with the use of IT in other indus-tries. It estimates potential savings and costs of widespread adoption of electronic medical record (EMR) systems, models important health and safety benefits, and concludes that ef-fective EMR implementation and networking could eventually save more than $81 billion annually—by improving health care efficiency and safety—and that HIT-enabled prevention and management of chronic disease could eventually double those savings while increas-ing health and other social benefits. However, this is unlikely to be realized without related changes to the health care system.

heu.s.healthcareindustryis arguably the world’s largest, most in-efficient information enterprise. However, although health absorbs more than $1.7 trillion per year—twice the Organization for Economic Coopera-tion and Development (OECD) average—premature mortality in the United States is much higher than OECD averages.1 Most medical records are still stored on paper, which means that they cannot be used to coordinate care, routinely mea-sure quality, or reduce medical errors. Also, consumers generally lack the informa-tion they need about costs or quality to make informed decisions about their care.

It is widely believed that broad adoption of electronic medical record (EMR) systems will lead to major health care savings, reduce medical errors, and improve health.2 But there has been little progress toward attaining these benefits. The United States trails a number of other countries in the use of EMR systems.3 Only

Richard Hillestad ( and James Bigelow are senior management scientists at RAND in Santa Monica, California. Anthony Bower is a senior economist there; Federico Girosi is a policy researcher; and Robin Meili is a senior management systems analyst. Richard Scoville and Roger Taylor are senior consultants at RAND Health—Scoville, in Chapel Hill, North Carolina, and Taylor, in Laguna Beach, California.

15–20 percent of U.S. physicians’ offices and 20–25 percent of hospitals have adopted such systems.4 Barriers to adoption include high costs, lack of certifica-tion and standardization, concerns about privacy, and a disconnect between who pays for EMR systems and who profits from them.

In 2003 the RAND Health Information Technology (HIT) Project team began a studyto(1) betterunderstandthe role andimportanceofEMRsinimproving health care and (2) inform government actions that could maximize the benefits of EMRs and increase their use. This paper summarizes that study’s results about benefits and costs. A companion paper by Roger Taylor and colleagues in this vol-ume describes the policy implications of our findings.5

Study Data And Methods

Here we summarize the methodologies we used to estimate the current adop-tion of EMR systems and the potential savings, costs, and health and safety bene-fits.Weuse theword potential to mean “assuming that interconnected and interoperable EMR systems are adopted widely and used effectively.” Thus, our es-timates of potential savings are not predictions of what will happen but of what could happen with HIT and appropriate changes in health care. We also provide a more thorough explanation of our data and methods in an online supplement.6

Estimation of current HIT adoption and related factors. Our primary data source was the Healthcare Information and Management Systems Society (HIMSS)–Dorenfest survey, which represents a broad canvassing of acute care hos-pitals, chronic care facilities, and ambulatory practices on their adoption and plans to adopt various HIT components.7 We included in the adoption category the pro-vider organizations that had contracted for but not yet installed an EMR system. To examine the factors related to differences in adoption, we merged additional data about the providers and then performed probit regression analysis. Our lower-boundestimateofHIT adoption assumedanintegratedsystemthathad an EMR, clinical decision support, and a central data repository—from the same vendor to ensure interoperability. We adjusted the estimates according to the known under-representation of smaller providers in this survey.
Estimation of potential HIT efficiency savings. We conducted a broad liter-ature survey to capture evidence of HIT effects. The survey was primarily from peer-reviewed literature, but it included some information from non-peer-reviewed liter-ature. Expert opinion was used to validate some of this evidence. In some cases, such as savings from transcription, reported results covered a broad range, and we used these ranges to estimate a possible distribution of savings. For effects supported by only a few useful articles, we superimposed the same degree of dispersion.8

In general, the currently useful evidence is not robust enough to make strong predictions, and we describe our results only as “potential.” However, we do not believethattheyrepresent the“best-casescenario,”for threereasons:(1) We have not included many other effects (such as transaction savings, reductions in mal

1104 September/October 2005

practice costs, and research and public health savings), and there may be more siz-able savings from HIT-motivated health care changes that we are not able to pre-dict: Modern EMR systems may be more effective than the legacy systems reporting evidence; (2) we have not included certain domains such as long-term care; and (3) we do not report possible values above the mean.

The results are not worst-case, either. We chose to interpret reported evidence of negative or no effect of HIT as likely being attributable to ineffective or not-yet-effective implementation. Characteristics of the provider organizations that re-ported the savings were used to scale the results for cases of broader EMR adop-tion. Assuming ten-and fifteen-year HIT adoption periods, we used Monte Carlo simulation to generate the range of savings that might be achieved at different points in the future, assuming that at least part of the reported benefit could be achieved by each newly adopting provider organization. We generally report the mean value of the potential savings.

Estimating the costs of adoption. For hospital adoption, we built a model of EMRsystemcosts basedonthe literature andoninformation supplieddirectlytous from hospitals. We included one-time implementation costs, such as provider downtime and hardware costs, and ongoing maintenance costs. Our data allowed us to relate hospital adoption costs to size and operating expenses of hospitals and generally represented the adoption of newer, more complete EMR systems, includ-ing clinical decision support and computerized physician order entry (CPOE). For the acquisition and setup costs of ambulatory systems, we used a publicly available database of commercial systems and excluded products that did not have most of the desirable features of an ambulatory EMR system.9 To these costs, we added a productivity loss of 15 percent for three months, $3,000 per physician for additional hardware costs, and yearly maintenance costs equal to 20 percent of the one-time cost. Starting with current adoption rates of EMR systems, we simu-lated ten-and fifteen-year adoption periods, in which physicians’ choices were ap-proximated by random selections from the ambulatory EMR list, and hospitals adopted systems and paid costs consistent with our data related to size and oper-ating expenses. From these simulations, we report the mean and show sensitivity to assumptions about the initial adoption rate and assumed adoption period.
Estimating potential safety benefits. Using medication error and adverse drug event rates from the literature, as well as limited evidence of CPOE’s reduction of medication error rates, we extrapolated these potential safety benefits to a future with broad national adoption of CPOE.10 Several databases—the Medical Expendi-ture Panel Survey (MEPS) 1999 Inpatient File (which tracks a large number of pa-tients and their interaction with health care), the American Hospital Association (AHA) 2000 Hospital Survey, and the Healthcare Cost and Utilization Project (HCUP) 2000 National Inpatient Sample—were used to distribute the errors across hospitals and patients.11 A spreadsheet model was then used to calculate the poten-tial adverse drug events and costs avoided as a function of hospital size and patient

HEALTH AFFAIRS ~ Volume 24, Number 5 1105

age. For ambulatory care, our model used error and adverse drug event reductions reported in the literature for ambulatory CPOE. Using the 2000 National Ambula-tory Medical Care Survey (NAMCS) database on office visits, we extrapolated the effects to full national adoption and show the likely distribution of possible savings and adverse drug events avoided as a function of practice characteristics and size.12

Estimating other potential health benefits. We considered two kinds of in-terventions—disease prevention and chronic disease management—that would ex-ploit key features of HIT. To estimate the potential effects enabled by EMR systems, we used several years of the MEPS data to develop a representative national patient sample, with its associated information on health care use, diagnosis, and self-reported health status. We applied recommended disease management and preven-tion interventions to appropriate members of that population. Then, given the liter-ature and clinicians’ opinions regarding the effect of the interventions, we calculated the differences in cost, use, health status, and other outcomes measured in MEPS, such as sick days in bed and workdays lost. We evaluated a representative sample of near-term (some effects within one or two years of intervention) prevention, near-term disease management, and long-term (most effects five or more years into the future) chronic disease management and prevention interventions. We report the health benefits and savings associated with various degrees of patient participation in these programs, as might be obtained with HIT support.

What Can We Learn From Other Industries?

We examined a range of industries to understand IT’s effects on productivity and related enabling factors. During the 1990s, many industries—most notably, telecommunications, securities trading, and retail and general merchandising— invested heavily in IT.13 Consumerssaw thefruitsofthisinvestmentinbar-coded retail checkouts, automated teller machines, consumer reservation systems, and online shopping and brokerages. During the late 1990s and continuing into this century, these industries recorded 6–8 percent annual productivity growth, of which at least one-third to one-fourth annually can be attributed to IT. But dra-matic productivity improvements did not follow automatically from IT invest-ments. For example, the hotel industry, which underused its IT investment in the late 1990s, did not see sizable productivity increases.

What if health care could produce productivity gains similar to those in tele-communications, retail, or wholesale? Exhibit 1 superimposes a range of produc-tivity improvements on a plot of estimated growth in national health care spend-ing from 2002 to 2016. The smaller improvement (1.5 percent per year) is similar to the productivity gains in retail/wholesale attributed to IT; the upper end (4 per-cent per year) is half the IT-enabled gains in telecommunications. Either level of productivity improvement could greatly reduce national health care spending. The lower improvement implies an average annual spending decrease of $346 bil-lion, and the upper end, $813 billion.

1106 September/October 2005

EXHIBIT 1 Possible Improved Productivity Effects Of Health Information Technology (IT) On Future National Health Spending, 2002–2016

Annual health spending (billions of dollars) 4,000

Effect of 1.5% annual productivity 3,500 improvement (like retail industry)


National health spending 2,500 projection



Effect of 4% annual productivity improvement 1,000 (half of telecom industry increase)

2002 2004 2006 2008 2010 2021 2014 2016

SOURCE: Authors’ analysis based on data from Centers for Medicare and Medicaid Services, “National Health Accounts,” 17 March 2005, (26 May 2005).

However, we believe that when thought leaders discuss transforming health care with HIT, they are talking about the kinds of benefits seen in the telecom and securities industries: gains of 8 percent or more per year, year after year. These sec-tors illustrate that it can be done. But our analysis found that the ingredients needed to achieve this growth (strong competition on quality and cost, substan-tial investments in EMR systems, an enhanced infrastructure that can accommo-date increased future demand or reduce costs without increasing labor, a strong champion firm or institution that drives change, and integrated systems) are mostly absent in today’s health care industry. Achieving savings at the upper end of the range will be limited by the degree of transformation that accompanies HIT.

What Are The Potential Efficiency Savings From HIT?

There are few comprehensive estimates of savings from HIT at the national level.14 Using a simulation model of HIT adoption and scaling literature-based HIT effects, we built a national estimate.15

At 90 percent adoption, we estimate that the potential HIT-enabled efficiency savings for both inpatient and outpatient care could average more than $77 billion per year (an average annual savings of $42 billion during the adoption period). Ex-hibit 2 shows the most important sources of the savings we estimated: The largest come from reducing hospital lengths-of-stay, nurses’ administrative time, drug us-age in hospitals, and drug and radiology usage in the outpatient setting.16

Thesepotential savings, whilequite large, areconsiderablylower than thean-nual IT-enabled productivity gains just described in other industries. Although achieving these more limited savings would not require radical changes in the

HEALTH AFFAIRS ~ Volume 24, Number 5 1107

EXHIBIT 2 Potential Efficiency Savings With Adoption Of Electronic Medical Record (EMR) Systems

Savings category Mean yearly savings ($ billions) Cumulative savings by year 15 ($ billions) Annual savings ($ billions) Year 5 Year 10 Year 15 (90% adoption)a
Outpatient Transcription Chart pulls Lab test 0.9 0.8 1.1 13.4 11.9 15.9 0.4 0.4 0.5 1.2 1.1 1.5 1.7 1.5 2.0
Drug usage Radiology Total outpatient savings 6.2 1.7 10.6 92.3 25.6 159.0 3.0 0.8 5.2 8.6 2.4 14.8 11.0 3.3 20.4
Inpatient Nursing time Lab test 7.1 1.6 106.4 23.4 3.4 0.8 10.0 2.2 13.7 2.6
Drug usage Length-of-stay Medical records 2.0 19.3 1.3 29.3 289.6 19.9 1.0 10.1 0.7 2.8 27.6 1.9 3.5 34.7 2.4
Total inpatient savings 31.2 468.5 16.1 44.5 57.1
Total 41.8 627.5 21.3 59.2 77.4

SOURCE: F. Girosi et al., Extrapolating Evidence of Health Information Technology Savings and Costs, Pub. no. MG-410 (Santa

Monica, Calif.: RAND, 2005), sec. 4.2.6.

NOTE: These savings have not been discounted, nor do they take into account inflation in health care expenditures.

a The potential savings at 100 percent adoption would obviously be larger, but the uncertainty about when and whether that

level can be reached is very high. We have assumed a fifteen-year adoption period, based on A. Bower, The Diffusion and

Value of Healthcare Information Technology, Pub. no. MG-272-HLTH (Santa Monica, Calif.: RAND, 2005).

health care delivery system, it would require process changes and, in some cases, resource reduction. Also, the potential savings would not be realized immediately. They would require widespread adoption of HIT by providers, and most of the savings would start only after a successful implementation period and associated process changes or resource reductions had taken place. Also, the efficiencies could be used to improve health care quality rather than to reduce costs.

Although the savings would accrue to different stakeholders, in the long run they should accrue to payers. If we allocate the savings using the current level of spending from the National Health Accounts (kept by the Centers for Medicare and Medicaid Services), Medicare would receive about $23 billion of the potential savings per year, and private payers would receive $31 billion per year. Thus, both have a strong incentive to encourage the adoption of EMR systems. Providers face limited incentives to purchase EMRs because their investment typically trans-lates into revenue losses for them and health care spending savings for payers.

What Are The Potential Safety Benefits Of EMR Systems?

Studies showing improved patient safety from EMR use in hospital and ambu-latory care largely focus on alerts, reminders, and other components of CPOE.17 CPOE makes information available to physicians at the time they enter an order— for example, warning about potential interactions with a patient’s other drugs.

1108 September/October 2005

Once the order has been entered, the system can track the steps involved in exe-cuting the order, providing an additional mechanism for identifying and eliminat-ing errors. In the longer term, CPOE provides the information needed to redesign the order-execution process so that errors become even harder to make. To pro-vide these benefits, CPOE must be an integrated component of a more compre-hensive health care information system that is designed and used well.18 We ad-dressed the safety benefits of CPOE by using models to extrapolate existing evidence to the national level and estimated separately the potential to reduce ad-verse drug events in inpatient and outpatient settings.

Reducing adverse drug events in the inpatient setting. The measures—ad-verse drug events avoided, and bed days and dollars saved—all follow the same pat-tern, which suggests that CPOE could eliminate 200,000 adverse drug events and save about $1 billion per year if installed in all hospitals. But the bulk of the savings couldberealizedbyinstallationinhospitals with more than 100beds. Abouttwo-thirds of the CPOE benefits are attributable to adverse drug events avoided for pa-tients age sixty-five and older. Although this group comprises only 13 percent of the population, it accounts for a much larger fraction of hospital bed days, and its mem-bers are more susceptible than others to adverse drug events.
Reducing adverse drug events in the ambulatory setting. Medication er-rors and adverse drug events in ambulatory settings have been studied much less than in hospitals. The available data suggest that roughly eight million outpatient events occur each year, of which one-third to one-half are preventable. About two-thirds of preventable adverse drug events might be avoided through widespread use of ambulatory CPOE. Each avoided event saves $1,000–$2,000 because of avoided office visits, hospitalizations, and other care.19 Scaling these numbers to the national level, we estimate that two million such events could be avoided, generating annual savings of $3.5 billion.20 Avoided adverse drug events in patients age sixty-five and older account for 40 percent of the savings.

Our models also show that to obtain the benefits of ambulatory CPOE, one can-not ignore small providers. About 37 percent of the potential savings and error avoidance would come from solo practitioners. Recent estimates suggest that CPOE systems can be cost-effective even for small offices.21

What Are The Potential Health Benefits Of EMR Systems?

Beyond safety, the literature provides little evidence about EMR systems’ ef-fects on health. We must, therefore, hypothesize about both mechanisms and magnitudes of effects. We considered two kinds of interventions intended to keep people healthy (or healthier): disease prevention measures and chronic disease management.

These interventions are key to understanding HIT’s potential. First, they would exploit important features and capabilities of EMR systems: communication, co-ordination, measurement, and decision support. Second, they are potentially high-

HEALTH AFFAIRS ~ Volume 24, Number 5 1109

leverage areas for improving health care. Physicians deliver recommended care only about half of the time, and care for patients with chronic illnesses absorbs more than 75 percent of the nation’s health care dollars.22 Third, evidence from re-gional health information network (RHIN) demonstrations suggests that these are key applications of HIT.23

Using HIT for short-term preventive care. EMR systems can integrate evi-dence-based recommendations for preventive services (such as screening exams) with patient data (such as age, sex, and family history) to identify patients needing specific services. The system can remind providers to offer the service during rou-tine visits and remind patients to schedule care. Reminders to patients generated by EMR systems have been shown to increase patients’ compliance with preventive care recommendations when the reminders are merely interjected into traditional outpatient workflows.24 More systemic adaptation—for example, by Kaiser Perma-nente and Group Health Cooperative—appears to achieve greater compliance.25 We estimated the effects of influenza and pneumococcal vaccination and screening for breast cancer, cervical cancer, and colorectal cancer, using data about the current compliance rate, the recommended population, and the costs.26 We assumed that the services are rendered to 100 percent of people not currently complying with the U.S. Preventive Services Task Force recommendation.27 We also applied the health benefit estimates from the literature to this population (Exhibit 3). We conclude that all of these measures, except for pneumococcal vac-cination, will increase health care use and spending modestly. But the costs are not large, and the health benefits are significant: for example, 13,000 life-years gained from cervical cancer screening at a cost of $0.1–$0.4 billion.
Using HIT for near-term chronic disease management. The U.S. burden of chronic disease is extremely high and growing. In one study, fifteen chronic condi-tions accounted for more than half of the growth in health care spending between 1987 and 2000, and just five diseases accounted for 31 percent of the increase.28 Dis-ease management programs identify people with a potential or active chronic dis-ease; target services to them based on their level of risk (sicker patients need more-tailored, more-intensive interventions, including case management); monitor their condition; attempt to modify their behavior; and adjust their therapy to prolong life, minimize complications, and reduce the need for costly acute care interventions.

EMR systems can be instrumental throughout the disease management proc-ess. Predictive-modeling algorithms can identify patients in need of services. EMR systems can track the frequency of preventive services and remind physicians to offer needed tests during patients’ visits. Condition-specific encounter templates implemented in an EMR system can ensure consistent recording of disease-specific clinical results, leading to better clinical decisions and outcomes. Connec-tion to national disease registries allows practices to compare their performance with that of others. Electronic messaging offers a low-cost, efficient means of dis-tributing reminders to patients and responding to patients’ inquiries. Web-based

1110 September/October 2005

EXHIBIT 3 Summary Of EstimatedResults ForIncreasingFivePreventiveServices

Program description Influenza vaccination Pneumococcal vaccination Screening for breast cancer Screening for cervical cancer Screening for colorectal cancer
Target population (age) 65 and older 65 and older Women 40 and older Women 18–64 50 and older
Frequency 1 per year 1 per lifetime 0.5–1 per year 0.33–1 per year 0.1–0.2. per year
Population not 13.2 million 17.4 million 18.9 million 13.0 million 52.0 million
currently compliant backlog; 2.1
million new
per year
Financial impacts
Program cost (with 100% compliance) $134–$327 million per year $90 million per year $1–$3 billion per year $152–$456 million per year $1.7–$7.2 billion per year
Financial benefits Health benefits $32–$72 millioper year n $500–$1,000 million per year $0–$643 millioper year n $52–$160 million per year $1.16–$1.77 billion per year
Reduced workdays missed 180,000– 325,000 per year 100,000– 200,000 per year a a a
Reduced bed days 1.0–1.8 million per year 1.5–3.0 million per year a a a
Deaths avoided 5,200–11,700 per year 15,000–27,000 per year 2,200–6,600 per year 533 per year 17,000–38,000 per year
Life-years gained a a a 13,000 per year 138,000 per year

SOURCE: J. Bigelow et al., Analysis of Healthcare Interventions That Change Patient Trajectories (Santa Monica, Calif.: RAND,2005), 74, Table 5.1.NOTE: Assumes 100 percent participation.a Not applicable.

patient education can increase the patient’s knowledge of a disease and compli-ance with protocols.

For higher-risk patients, case management systems help coordinate workflows, including communication between multiple specialists and patients. In what may prove to be a transformative innovation, remote monitoring systems can transmit patients’ vital signs and other biodata directly from their homes to their providers, allowing nurse case managers to respond quickly to incipient problems. Health information exchange via RHINs or personal health records promises great bene-fits for patients with multiple chronic illnesses, who receive care from multiple providers in many settings.

We examined disease management programs for four conditions: asthma, con-gestiveheart failure(CHF),chronic obstructivepulmonary disease(COPD), and diabetes (Exhibit 4) and estimated the effects of 100 percent participation of peo-pleeligiblefor each program.29 By controlling acute care episodes, these programs greatly reduce hospital use at the cost of increased physician office visits and use

HEALTH AFFAIRS ~ Volume 24, Number 5 1111

EXHIBIT 4 Potential Annual Effects Of Near-Term Disease Management Programs For Four Diseases: Asthma, Congestive Heart Failure (CHF), Chronic Obstructive Pulmonary Disease(COPD), AndDiabetes

Effect Change
Use (millions)
Inpatient stays –4
Hospital outpatient visits –5
Physician office visits 33
Spending (billions)
Hospital –$30.1
Physician –$0.0
Rx drugs $1.9
Total –$28.5
Outcomes (millions)
Workdays lost –28
School days lost –13
Bed days –245

SOURCE: J. Bigelow et al., Analysis of Healthcare Interventions That Change Patient Trajectories (Santa Monica, Calif.: RAND,2005), 137, Table 6.17.NOTE: Assumes 100 percent participation.

of prescription drugs. As shown, the programs could generate potential annual savings of tens of billions of dollars. Keeping people out of the hospital is, of course, a health benefit, but we can also expect important outcomes such as re-ductions in days lost from school and work and in days spent sick in bed.

Exhibit 4 also highlights an important disincentive for health care providers to offer these kinds of services or to invest in HIT to effectively perform them: The savings come out of provider receipts, as patients spend less time in acute care. This key misalignment of incentives is an important barrier to EMR adoption and, more generally, to health care transformation.

� Using HIT for long-term chronic disease prevention and management. A program of EMR-enhanced prevention and disease management should change the incidence of chronic conditions and their complications. We considered cardiovas-cular diseases (hypertension, hyperlipidemia, coronary artery disease/acute myo-cardial infarction, CHF, cerebrovascular disease/stroke, and other heart diseases), diabetes and its complications (retinopathy, neuropathy, lower extremity/foot ulcers and amputations, kidney diseases, and heart diseases), COPD (emphysema and chronic bronchitis), and the cancers most strongly associated with smoking (can-cers of the bronchus and lung, head and neck, and esophagus, and other respiratory and intrathoracic cancers). Using our MEPS-based model, we estimated how com-binations of lifestyle changes and medications that reduced the incidence of these conditions would affect health care use, spending, and outcomes (Exhibit 5).

Savings are evenly divided between the populations under age sixty-five and those age sixty-five and older, despite the fact that the older population consti

1112 September/October 2005

EXHIBIT 5 Estimated Long-Term Effects Of Prevention And Management Of Selected Chronic Conditions, By Age Group

Under age 65 Age 65 and older Total
Population (millions) 244.8 37.3 282.1
Utilization (millions)
Inpatient stays –3.2 –3.9 –7.1
Inpatient nights –18.6 –30.6 –49.2
Hospital outpatient and ER visits –8.8 –3.7 –12.5
Office visits –63.2 –54.8 –118.0
Expenditures (billions)
Hospital –$31.8 –$39.9 –$71.7
Physician –$11.7 –$11.4 –$23.1
Rx drugs –$16.2 –$13.4 –$29.6
Other –$4.4 –$9.9 –$14.3
Total change –$64.1 –$74.6 –$138.7
Outcomes (millions)
School days lost –1.6 0.0 –1.6
Workdays lost –39.4 –2.5 –41.9
Total bed days –132.1 –125.1 –257.3
Deaths (thousands) –119.4 –280.4 –399.8

SOURCE: J. Bigelow et al., Analysis of Healthcare Interventions That Change Patient Trajectories (Santa Monica, Calif.: RAND,2005), 160, Table 7.6.NOTE: Assumes 100 percent participation. ER is emergency room.

tutes only 13 percent of the total. Since chronic diseases are, by and large, diseases of the elderly, a large fraction of the long-term savings attributable to prevention and disease management would accrue to Medicare. Yet, to realize these benefits, people would have to begin participating in these programs as relatively young adults.

We combined the effects of the reduced incidence attributable to long-term prevention and management and reduced acute care due to disease management. We estimated the potential combined savings, again assuming 100 percent partici-pation, to be $147 billion per year.30

Realizing the potential of these interventions. Realizing the benefits of pre-vention and disease management requires that a substantial portion of providers and consumers participate. Since, on average, patients comply with medication regi-mens about half the time, it is plausible to assume that about half of the chronically ill would participate in disease management programs and, therefore, the health care system would reap about half of the estimated short-term benefits, assuming that EMR systems and community-based connectivity were operational.31

Patients comply with their physician’s lifestyle recommendations only about 10 percent of the time.32 We assumed that in a future with EMR-based reminders and decision support and patient-physician messaging, we could realize at least 20 percent of the long-term benefits shown in Exhibit 5. Under these assumptions,

HEALTH AFFAIRS ~ Volume 24, Number 5 1113

the net savings would be on the order of $40 billion per year. We varied the partic-ipation in disease management and prevention activities parametrically to show the potential beyond these estimates.33

What Will It Cost To Implement EMR Systems?

There are a few published estimates of the costs of widespread implementation of EMR systems in the United States. Samuel Wang and colleagues have provided a model for estimating the cost and return on investment for a physician office practice.34 Jan Walker and colleagues have estimated the costs ($28 billion per year during a ten-year deployment, $16 billion per year thereafter) and net savings ($21.6–$77.8 billion per year, depending on the level of standardization) of a broadly adopted, interoperable EMR system.35 The Patient Safety Institute esti-mated the initial cost of widespread connectivity of EMR systems (not of the EMR system itself) to be $2.5 billion.36

Adoption costs for hospitals. From cost data obtained from the literature, as well as from direct discussions with providers, we used simulation to estimate that the cumulative cost for 90 percent of hospitals to adopt an EMR system is $98 bil-lion if 20 percent of hospitals now have such a system. Average yearly costs for the fifteen-year adoption period are $6.5 billion—about one-fifth of our earlier-described estimate of potential efficiency savings in hospitals.37
Adoption costs for physicians. Our models for adoption by physicians show that the cumulative costs to reach 90 percent adoption are $17.2 billion, almost equally split between one-time costs and maintenance costs. The average yearly cost during the adoption period is about $1.1 billion. In comparison, we estimated the po-tential annual average efficiency and safety benefits from ambulatory EMR systems during thesameperiodtobe$11 billion.

What Are The Potential Net Savings From EMR Systems?

Exhibit 6 plots the net cumulative and yearly potential savings (benefits over costs) from EMR systems in hospital and outpatient settings over time. Because we do not take credit for savings from providers already in the adoption process and because process changes and related benefits take time to develop, net savings are initially low and then rise steeply. Over fifteen years, the cumulative potential net efficiency and safety savings from hospital systems could be nearly $371 bil-lion; potential cumulative savings from physician practice EMR systems could be $142 billion. This potential net financial benefit could double if the health savings produced by chronic disease prevention and management were included.

Barriers To Realizing The Health Benefits And Savings

Our analysis shows that moving the U.S. health care system quickly to broad adoption of standards-based EMR systems could dramatically reduce national health care spending at a cost far below the savings. Further, these potential sav-ings would outweigh the costs relatively quickly during the adoption cycle. But

1114 September/October 2005

EXHIBIT 6 Net Potential Savings (Efficiency Benefits Over Adoption Costs) For Hospital And Physician Electronic Medical Record (EMR) Systems Adoption During A Fifteen-Year Adoption Period (2004–2018)

Dollars 350

300 Cumulative inpatient 250 Cumulative outpatient Yearly inpatient 200 Yearly outpatient




0 2004 2006 2008 2010 2012 2014 2016 2018

key barriers in the HIT market directly impede adoption and effective application of EMR systems; these include acquisition and implementation costs, slow and uncertain financial payoffs, and disruptive effects on practices.38 In addition, pro-viders must absorb the costs of EMR systems, but consumers and payers are the most likely to reap the savings. Also, even if EMR systems were widely adopted, the market might fail to develop interoperability and robust information exchange networks.

Given our analysis, we believe that there is substantial rationale for government policy to facilitate widespread diffusion of interoperable HIT. Actions now, in the early stages of adoption, would provide the most leverage. Taylor and colleagues discuss several alternatives for government action to remove barriers, correct mar-ket failures, and speed the realization of EMR system benefits.39

We have shown some of the potential benefits of HIT in the current health care system. However, broad adoption of EMR systems and connectivity are necessary but not sufficient steps toward real health care transformation. For example, adoption of EMR systems and valid comparative performance reporting would en-able the development of value-based competition and quality improvement to drive transformation. HIT also should facilitate system integration for broader op-timization, and comparative benchmarking should encourage development of market-leading examples of ways to better organize, pay for, and deliver care.

It is not known what changes should or will take place after widespread EMR

HEALTH AFFAIRS ~ Volume 24, Number 5 1115

system adoption—for example, increased consumer-directed care, new methods of organizing care delivery, and new approaches to financing. But it is increasingly clear that a lengthy, uneven adoption of nonstandardized, noninteroperable EMR systems will only delay the chance to move closer to a transformed health care sys-tem. The government and other payers have an important stake in not letting this happen. Thetimetoact is now.

This report is a product of the RAND HIT Project. It benefited from the guidance of an independent Steering Committee, chaired by David Lawrence, and was sponsored by Cerner, General Electric, Hewlett-Packard, Johnson and Johnson, and Xerox.


  1. Organization for Economic Cooperation and Development, “Health at a Glance—OECD Indicators 2003,” 17 September 2003,,2340,en_2649_201185_16502667_1_1_1_1,00.html (20 July 2005).

  2. The term EMR systems as used here includes the electronic medical record (EMR), containing current and historical patient information; clinical decision support (CDS), which provides reminders and best-prac-tice guidance for treatment; and a central data repository (CDR), for the information. It also includes IT-enabled functions, such as computerized physician order entry (CPOE). We use the terms health informa-tion technology (HIT) and EMR systems interchangeably.

    1. H. Taylor and R. Leitman, “European Physicians, Especially in Sweden, Netherlands, and Denmark, Lead

    2. U.S. in Use of Electronic Medical Records,” Harris Interactive 2, no. 16 (8 August 2002), www.harris (20 July 2005).
  3. K. Fonkych and R. Taylor, The State and Pattern of Health Information Technology Adoption (Santa Monica, Calif.: RAND, 2005).

  4. R. Taylor et al., “Promoting Health Information Technology: Is There a Case for More-Aggressive Govern-ment Action?” Health Affairs 24, no. 5 (2005): 1234–1245.

  5. The online supplement is available at The RAND Web site provides a comprehensive description of our methods, data, and models. See publications/MG/MG408; MG409; and MG410.

  6. HIMSS AnalyticsSM Database (formerly the Dorenfest IHDS+TM Database), second release, 2004.

  7. See J.H. Bigelow, K. Fonkych, and F. Girosi, Technical Executive Summary in Support of “Can Electronic Medical Record Systems Transform Healthcare?” This online summary of our methods includes a table listing the most important literature findings and some measures of their quality; see Note 6.

  8. See K.C. Voelker, “Electronic Medical Record (EMR) Comparisons by Physicians for Physicians,” www (26 May 2005).

  9. In R. Koppel et al., “Role of Computerized Physician Order Entry Systems in Facilitating Medication Er-rors,” Journal of the American Medical Association 293, no. 10 (2005): 1197–1203, it was reported that the medica-tion error rate actually increased because of computer-and interface-induced errors. We have assumed that this is not the case for a carefully redesigned medication process supported by modern CPOE.

  10. See Agency for Healthcare Research and Quality (AHRQ), Medical Expenditure Panel Survey (MEPS) (multiple years of data and documentation), at (24 February 2005); American Hos-pital Association, AHA Annual Survey Database (a survey conducted since 1946; data must be purchased); and AHRQ, Nationwide Inpatient Sample (NIS), part of the Healthcare Cost and Utilization Project (HCUP), at (24 February 2005).

  11. National Center for Health Statistics, National Ambulatory Medical Care Survey (NAMCS)—multiple years of data and documentation available at (24 Febru-ary 2005).

  12. A. Bower, The Diffusion and Value of Healthcare Information Technology, Pub. no. MG-272-HLTH (Santa Monica, Calif.: RAND, 2005).

  13. Recently, Jan Walker and colleagues quantified the value of full adoption of interoperable EMR systems. J. Walker et al., “The Value of Health Care Information Exchange and Interoperability,” Health Affairs, 19 Janu-

1116 September/October 2005

ary 2005, (2 May 2005). Laurence Baker ar-gued that these savings were overestimated. L.C. Baker, “Benefits of Interoperability: A Closer Look at the Estimates,” Health Affairs, 19 January 2005, (26 May 2005).

  1. F. Girosi et al., Extrapolating Evidence of Health Information Technology Savings and Costs, Pub. no. MG-410 (Santa Monica, Calif.: RAND, 2005).

  2. Other factors and savings, mentioned in the Study Data and Methods section, could increase this total po-tential.

  3. D.W. Bates et al., “Effect of Computerized Physician Order Entry and a Team Intervention on Prevention of Serious Medication Errors,” Journal of the American Medical Association 280, no. 15 (1999): 1311–1316.

  4. Koppel et al., “Role of Computerized Physician Order Entry Systems.”

  5. D. Johnston et al., Patient Safety in the Physician’s Office: Assessing the Value of Ambulatory CPOE, April 2004, (26 May 2005).

  6. J.H. Bigelow et al., Analysis of Healthcare Interventions That Change Patient Trajectories (Santa Monica, Calif.: RAND, 2005).

  7. S.J. Wang et al., “A Cost-Benefit Analysis of Electronic Medical Records in Primary Care,” American Journal of Medicine 114, no. 5 (2003): 397–403.

  8. E.A. McGlynn et al., “The Quality of Health Care Delivered to Adults in the United States,” New England Journal of Medicine 348, no. 26 (2003): 2635–2645; and Centers for Disease Control and Prevention, ”Chronic Disease Overview,” 15 October 2004, (26 May 2005).

  9. See, for example, Institute for Healthcare Improvement, “My Shared Care Plan,” topics/chronicconditions/diabetes/tools/my+shared+care+plan.htm (26 May 2005).

  10. R.C. Burack and P.A. Gimotty, “Promoting Screening Mammography in Inner-City Settings: The Sus-tained Effectiveness of Computerized Reminders in a Randomized Controlled Trial,” Medical Care 35, no. 9 (1997): 921–931.

  11. B. Kaplan, “Evaluating Informatics Applications—Clinic Decision Support Systems Literature Review,” International Journal of Medical Informatics 64, no. 1 (2001): 15–37.

  12. Bigelow et al., Analysis of Healthcare Interventions.

  13. We make the 100 percent assumption to provide an upper bound on the net costs and the health effects of their service. We do not suggest that 100 percent participation can be realized in practice.

  14. K.E. Thorpe, C.S. Florence, and P. Joski, “Which Medical Conditions Account for the Rise in Health Care Spending?” Health Affairs, 25 August 2004, (26 May 2005).

  15. Bigelow et al., Analysis of Healthcare Interventions.

  16. This is less than the direct sum, because reduced incidence implies a lesser requirement for disease man-agement.

  17. R.B. Haynes, H.P. McDonald, and A.X. Garg, “Helping Patients Follow Prescribed Treatment: Clinical Ap-plications,” Journal of the American Medical Association 288, no. 22 (2002): 2880–2883.

  18. D.L. Roter et al., “Effectiveness of Interventions to Improve Patient Compliance: A Meta-Analysis,” Medical Care 36, no. 8 (1998): 1138–1161.

  19. Bigelow et al., Analysis of Healthcare Interventions.

  20. Wang et al., “A Cost-Benefit Analysis.”

  21. Walker et al., “The Value of Health Care Information Exchange.”

  22. Because there is not much experience with regional connectivity, cost estimates fall within a wide range. Our own scaling of data provided by the Santa Barbara Care Data Exchange indicates $2.4 billion for a non–standards-based system.

  23. Girosi et al., Extrapolating Evidence.

  24. R. Miller and I. Sim, “Physicians’ Use of Electronic Medical Records: Barriers and Solutions,” Health Affairs 23, no. 2 (2004): 116–126; and D.W. Bates et al., “A Proposal for Electronic Medical Records in U.S. Primary Care,” Journal of the American Medical Informatics Association 10, no. 1 (2003): 1–10.

  25. Taylor et al., “Promoting Health Information Technology.”

HEALTH AFFAIRS ~ Volume 24, Number 51117