CBA - Electronic Medical Records
CBA - Electronic Medical Records
CBA - Electronic Medical Records
METHODS
Study Design
We performed a cost-benefit analysis of electronic medical record usage by primary care physicians in an ambulatory-care setting. The primary outcome measure was
net financial costs or benefits per provider during a 5-year
period. The model was framed from the perspective of the
health care organization, and the reference strategy was
the traditional paper-based medical record. All costs and
benefits were converted to 2002 U.S. dollars (27).
Data on costs and benefits came from primary data
collected from our electronic medical record system,
from other published studies, and from expert opinion.
When data were not available, expert opinion was obtained using a modified Delphi (28) technique to arrive at
group consensus with a 7-member expert panel. Primary
data were obtained from several internal medicine clinics
using our internally developed electronic medical record
0002-9343/03/$see front matter 397
doi:10.1016/S0002-9343(03)00057-3
Table 1. Costs of Electronic Medical Record System Used in the Model (Per Provider in 2002 U.S.
Dollars)
Base Case
System costs
Software (annual license)
Implementation
Support and maintenance
Hardware (3 computers network)
Induced costs
Temporary productivity loss
$1600
$3400
$1500
$6600
$11,200
Sensitivity Analysis
(Range)
$ 800$3200
Reference
*
$ 750$3000
$3300$9900
*
*
$5500$16,500
Costs
There are two categories of costs associated with electronic medical record implementation: system costs and
induced costs (Table 1). System costs include the cost of
the software and hardware, training, implementation,
and ongoing maintenance and support. Induced costs are
those involved in the transition from a paper to electronic
system, such as the temporary decrease in provider productivity after implementation.
The software costs of $1600 per provider per year were
based on the costs for our electronic medical record system at Partners HealthCare on an annual per-provider
basis (as an application service provider model); this
figure includes the costs of the design and development of
the system, interfaces to other systems (e.g., registration,
scheduling, laboratory), periodic upgrades, and costs of
user accounts for support staff. Although these software
costs were based on our internally developed system, they
are consistent with license fees for commercially available
systems, which have been estimated at between $2500
and $3500 per provider for the initial software purchase,
plus annual maintenance and support fees of 12% to 18%
(K. MacDonald, First Consulting Group, Lexington,
Massachusetts, written communication, 1999). In sensi398
April 1, 2003
Volume 114
Benefits
Financial benefits included averted costs and increased
revenues. We obtained figures for average annual expenditures for a primary care provider at our institution before the implementation of an electronic medical record,
and applied to this the estimated percentage cost savings
after implementation (Table 2). For each item, the estimated savings was varied across the indicated range of
values in the sensitivity analysis. Benefits were divided
into three categories: payer-independent benefits, benefits under capitated reimbursement, and benefits under
fee-for-service reimbursement (32 40).
Payer-independent benefits, which apply to both capitated and fee-for-service patients, come from reductions
in paper chart pulls and transcription. The average cost of
a chart pull at our institution is approximately $5, accounting for the time and cost of medical records personnel to retrieve and then re-file a paper chart. After con-
Table 2. Annual Expenditures Per Provider (in 2002 U.S. Dollars) before Electronic Medical Record System Implementation and
Expected Savings after Implementation
Annual Expenditures before
Implementation
Payer independent
Chart pulls
Transcription
Capitated patients
Adverse drug events
Drug utilization
Laboratory utilization
Radiology utilization
Fee-for-service patients
Charge capture
Billing errors
Amount
Reference
Base Case
Estimated Savings
Sensitivity Analysis
(Range)
Reference
$5 (per chart)
$9600
*
*
600 charts
28%
3001200
20%100%
*
*,32
$6500
$109,000
$27,600
$59,100
3336
34%
15%
8.8%
14%
10%70%
5%25%
013%
5%20%
$383,100
$9700
2% (increase)
78%
1.5%5%
35%95%
3739
25,40
* Primary data from the Partners HealthCare Electronic Medical Record System, Boston, Massachusetts.
From the Department of Finance, Brigham and Womens Hospital, Partners HealthCare System.
izing the encounter form process can improve the capture of in-office procedures that were performed but not
documented. Based on other studies (25,40), we projected a 2% improvement in billing capture (range, 1.5%
to 5%). By using an electronic medical record system that
either supplies or prompts for certain required fields, billing error losses can be reduced. The expert panel estimated that computerizing the encounter form could decrease these errors by 78% (range, 35% to 95%).
Statistical Analysis
We assumed that initial costs would be paid at the beginning of year 1 and that benefits would accrue at the end of
each year (Table 3). We assumed a phased implementation, in which only basic electronic medical record features were available in the first years (e.g., medicationrelated decision support), and more advanced features
were added in subsequent years (e.g., laboratory, radiology, and billing benefits). The primary outcome measure
was net benefit or cost per primary care provider. A discount rate of 5% was used in the base case and varied
from 0% to 10% in the sensitivity analysis.
One-way and two-way sensitivity analyses were performed using the ranges shown in Tables 1 and 2. Twoway sensitivity analyses were performed using all combinations of the five most important variables identified in
the one-way sensitivity analysis, and with pairwise combinations of one benefit variable with each of the three
primary cost variables (software, hardware, and support).
A five-way sensitivity analysis was performed using the
most and least favorable conditions for the five variables.
The time horizon was also varied from 2 to 10 years.
April 1, 2003
Table 3. 5-Year Return on Investment Per Provider for Electronic Medical Record Implementation
Initial Cost
Costs
Software license (annual)
Implementation
Support
Hardware (refresh every 3 years)
Productivity loss
Annual costs
Present value of annual costs*
Benefits
Chart pull savings
Transcription savings
Prevention of adverse drug events
Drug savings
Laboratory savings
Radiology savings
Charge capture improvement
Billing error decrease
Annual benefits
Present value of annual benefits*
Net benefit (cost)
Present value of net benefit (cost)*
$1600
$3400
$1500
$6600
Year 1
Year 2
Year 3
Year 4
Year 5
$1600
$1600
$1600
$1600
$1600
$1500
$1500
$1500
$6600
$1500
$1500
$14,300
$13,619
$3100
$2812
$9700
$8379
$3100
$2550
$3100
$2429
$3000
$2700
$3000
$2700
$2200
$16,400
$3000
$2700
$2200
$16,400
$3000
$2700
$2200
$16,400
$2400
$8300
$7700
$7600
$3000
$2700
$2200
$16,400
$2400
$8300
$7700
$7600
$5700
$5429
$(8600)
$(8190)
$24,300
$22,041
$21,200
$19,229
$24,300
$20,991
$14,600
$12,612
$50,300
$41,382
$47,200
$38,832
$50,300
$39,411
$47,200
$36,982
Total
$11,200
$13,100
$13,100
$(13,100)
$(13,100)
$46,400
$42,900
$154,900
$129,300
$108,500
$86,400
To account for variations in functionality among different systems, we constructed two additional models in
which only subsets of the full functionality were included
(Table 4). The light electronic medical record system
included savings from chart pulls and transcriptions
only, whereas the medium system also included benefits from electronic prescribing (adverse drug event prevention and drug expenditure savings).
RESULTS
In the 5-year cost-benefit model (Table 3), the net benefit
of implementing a full electronic medical record system
was $86,400 per provider. Of this amount, savings in drug
expenditures made up the largest proportion of the benefits (33% of the total). Of the remaining categories, almost half of the total savings came from decreased radiology utilization (17%), decreased billing errors (15%),
and improvements in charge capture (15%).
Sensitivity Analyses
The model was most sensitive to variations in the proportion of patients in capitated health plans; the net benefit
varied from $8400 to $140,100 (Figure). The model was
least sensitive to variations in laboratory savings, in which
the net benefit ranged from $82,500 to $88,300.
In two-way sensitivity analyses, the pair of input variables that yielded the least favorable outcome was a low
proportion of capitated patients and a high discount rate;
Table 4. Effect of Electronic Medical Record Feature Set Variations on Net Benefits
Feature
Online patient charts
Electronic prescribing
Laboratory order entry
Radiology order entry
Electronic charge capture
Benefit
Light EMR
Medium EMR
Full EMR
($18,200)
$44,600
$86,400
April 1, 2003
Volume 114
Figure. Tornado diagram showing the one-way sensitivity analysis of net 5-year benefits per provider. Each bar depicts the overall
effect on net benefits as that input is varied across the indicated range of values, while other input variables are held constant. The
vertical line indicates the base case.
DISCUSSION
Our analysis indicates that the net financial return to a
health care organization from using an ambulatory electronic medical record system is positive across a wide
in the model did not include costs of malpractice settlements, injury to patients, or decreased quality of life for
patients, so the actual savings may be higher. We may
have also underestimated future cost savings because the
model did not account for the annual growth rate of expenditures, which may outpace inflation in some categories, such as in drug and radiology costs.
Other potential areas of savings were not included in
the model because adequate data were not available.
These include savings in malpractice premium costs (40),
storage and supply costs (47), generic drug substitutions
(48), increased provider productivity (19,23,24), decreased staffing requirements (23,24,49), increased reimbursement from more accurate evaluation and management coding, and decreased claims denials from inadequate medical necessity documentation.
Although we accounted for a temporary (3-month)
loss of productivity in our model, some providers may
have a longer period of reduced productivity. To measure
this effect, we performed a sensitivity analysis that included a prolonged 10% productivity loss for 12 months
and found that there was still a 5-year net benefit of
$57,500 per provider.
This study has several limitations. The cost-benefit
model was based on primary data from our institution,
estimates from published literature, and expert opinion.
The effectiveness of some of these interventions has been
demonstrated in the inpatient setting, but outpatient effectiveness is less certain. There may be other costs associated with implementation of an electronic medical
record. For example, system integration costs may be
greater at other institutions, depending on the number
and complexity of system interfaces that are required.
However, the majority of benefits in this model can be
obtained even with a minimal number of interfaces (i.e.,
registration, scheduling, and transcription). In addition,
there may be other unforeseen expenses associated with
clinic workflow process redesign, reassignment of clinic
staff, or productivity loss during unscheduled computer
system or network outages.
In most cases, clinical decision support features will
decrease utilization by suggesting more appropriate testing. This leads to cost savings among capitated patients,
but it could also have an adverse effect on revenues from
fee-for-service patients that may offset billing improvements. The overall net effect would depend on the mix of
capitated versus fee-for-service patients.
Our cost-benefit model was geared toward primary
care providers. Diagnostic test utilization may be higher
for specialists, so there may be more opportunities for
cost-saving interventions. On the other hand, specialists
may be less likely to comply with computer reminders
recommending alternative medications or tests.
This study was framed from the perspective of the
health care organization to aid in making decisions about
402
April 1, 2003
Volume 114
ACKNOWLEDGMENT
We would like to thank Marc Overhage, MD, Homer Chin,
MD, Barry Blumenfeld, MD, and Tejal Gandhi, MD, who
joined three of the coinvestigators to serve on our expert panel.
REFERENCES
1. Elson RB, Connelly DP. Computerized patient records in primary
care. Their role in mediating guideline-driven physician behavior
change. Arch Fam Med. 1995;4:698 705.
2. Shea S, DuMouchel W, Bahamonde L. A meta-analysis of 16 randomized controlled trials to evaluate computer-based clinical reminder systems for preventive care in the ambulatory setting. J Am
Med Inform Assoc. 1996;3:399 409.
3. Hunt DL, Haynes RB, Hanna SE, et al. Effects of computer-based
clinical decision support systems on physician performance and
patient outcomes: a systematic review. JAMA. 1998;280:1339
1346.
4. Bates DW. Medication errors. How common are they and what can
be done to prevent them? Drug Saf. 1996;15:303310.
5. Institute of Medicine Committee on Quality of Health Care in
America. Crossing the Quality Chasm: A New Health System for the
21st Century. Washington, DC: National Academy Press; 2001.
31. United States Census 2000. Ranking Tables for States. Population in
2000 and Population Change from 1990 to 2000 (PHC-T-2). Washington, DC: U.S. Census Bureau; 2000.
32. Ambulatory EMR. Establishing a business case. Available at: http://
www.medicalogic.com/emr/business_benefits/index.html. Accessed
April 20, 2002.
33. Medicare RBRVS Physician Fee Schedule 2002. Washington, DC:
Centers for Medicare and Medicaid Services; 2001.
34. Prescription Drug Expenditures in 2001. Washington, DC: National
Institute for Health Care Management; 2002.
35. Jha AK, Kuperman GJ, Rittenberg E, et al. Identifying hospital admissions due to adverse drug events using a computer-based monitor. Pharmacoepidemiol Drug Saf. 2001;10:113119.
36. Honigman B, Lee J, Rothschild J, et al. Using computerized data to
identify adverse drug events in outpatients. J Am Med Inform Assoc.
2001;8:254 266.
37. Tierney WM, Miller ME, McDonald CJ. The effect on test ordering
of informing physicians of the charges for outpatient diagnostic
tests. N Engl J Med. 1990;322:1499 1504.
38. Tierney WM, McDonald CJ, Martin DK, et al. Computerized display of past test results. Effect on outpatient testing. Ann Intern
Med. 1987;107:569 574.
39. Tierney WM, McDonald CJ, Hui SL, et al. Computer predictions of
abnormal test results. Effects on outpatient testing. JAMA. 1988;
259:1194 1198.
40. Renner K. Cost-justifying electronic medical records. Healthc Financ Manage. 1996;50:6364.
41. Bates DW, Boyle DL, Rittenberg E, et al. What proportion of common diagnostic tests appear redundant? Am J Med. 1998;104:361
368.
42. Bates DW, Kuperman GJ, Rittenberg E, et al. A randomized trial of
a computer-based intervention to reduce utilization of redundant
laboratory tests. Am J Med. 1999;106:144 150.
43. Harpole LH, Khorasani R, Fiskio J, et al. Automated evidence-based
critiquing of orders for abdominal radiographs: impact on utilization and appropriateness. J Am Med Inform Assoc. 1997;4:511521.
44. Rothschild JM, Khorasani R, Bates DW. Guidelines and decision
support help improve image utilization. Diagn Imaging. 2000;22:
959799,101.
45. Reinhardt UE. Reforming American healthcare: an interim report.
J Rheumatol. 1999;26(suppl):6 10.
46. Jones DL, Kroenke K, Landry FJ, et al. Cost savings using a steppedcare prescribing protocol for nonsteroidal anti-inflammatory
drugs. JAMA. 1996;275:926 930.
47. Evans JC, Hayashi AM. Implementing on-line medical records. Doc
Manage. 1994;Sep/Oct:1217.
48. The quantum experience: a case study from Ernst & Young.
Available at: http://www.allscripts.com/ahcs/news_2.asp?S
2000&ID2. Accessed April 10, 2002.
49. Kian LA, Stewart MW, Bagby C, et al. Justifying the cost of a computer-based patient record. Healthc Financ Manage. 1995;49:58
60.
50. Essex D. Skip the song & dance. Healthc Inform. 1999;16:49 52
54,56.
51. Balas EA, Weingarten S, Garb CT, et al. Improving preventive care
by prompting physicians. Arch Intern Med. 2000;160:301308.
52. Bates DW, Teich JM, Lee J, et al. The impact of computerized physician order entry on medication error prevention. J Am Med Inform Assoc. 1999;6:313321.
53. McDonald CJ, Hui SL, Smith DM, et al. Reminders to physicians
from an introspective computer medical record. A two-year randomized trial. Ann Intern Med. 1984;100:130 138.
54. McDonald CJ, Overhage JM, Tierney WM, et al. The Regenstrief
Medical Record System: a quarter century experience. Int J Med Inf.
1999;54:225253.
April 1, 2003