Cost analysis of computerized physician order entry using value stream analysis: a case study.

With costs increasing at double digit rates each year, the healthcare industry is Working on various methods to reduce costs. It is estimated that over 1 million serious medication errors occur every year in U.S. hospitals and that 20 percent of these errors are life-threatening. In addition, medication errors are believed to cause the deaths of up to 7,000 patients each year. This means that medication errors cause more deaths annually than all industrial accidents in the United States (Institute for Safe Medical Practices [ISMP], 2000). The total financial cost related to drug prescription and administration errors is estimated as high as $77 billion per year.

The vast majority of hospitals and other medical services still use the hand-written paper prescription process, developed years ago, as the backbone of the drug prescription and administration (medication) value stream. It is the nature of the handwritten prescription that is a main contributor to errors in the medication value stream (ISMP, 2000). To improve the medication value stream, an information technology solution has been proposed called Computerized Physician Order Entry (CPOE). In this system, hand-written prescriptions are replaced by computer entries made by the prescriber. The system not only eliminates or minimizes hand-written prescriptions but also informs the prescriber of potential adverse drug interactions as well as other medical information relative to the medication value stream. The CPOE process is in early development with only 58 locations in the U.S. having deployed some version of a system as of December 2001 (Metzger & Turisco, 2001). Usually hospitals that have adopted CPOE systems are large urban hospitals with the financial and human resources needed to address implementation issues required by CPOE (Scalise, 2002). The hospital under consideration is typical of the hospitals that have adopted CPOE systems. Selected financials from the hospital are shown in Exhibit 1 and indicate a well-managed and efficient operation.

The goal of this paper is to use the process of value stream mapping to evaluate the current medication value stream at a local hospital. With the current value stream state mapped, an estimated differential cost is generated for the medication value stream. A future value stream map is proposed relating to a CPOE process, and potential cost reductions are evaluated. In addition, a cost benefit model is constructed.

VALUE STREAM MAPPING OVERVIEW

A value stream represents all the actions, both value added and non-value added, that are executed in a given process (Baker, 2002). Value streams reflect not only the process flow, but also the information flow that is used in the process. Value stream mapping is the processes of documenting a given value stream to identify the process and information flows (Rother & Shook, 1998). One of the specific benefits of value stream mapping is that it focuses efforts on the flow of a process rather than discrete process steps. If focus is given to discrete process steps, issues addressed in a specific step may create non-value added changes to upstream or downstream processes in the value stream.

CURRENT STATE VALUE STREAM MAP

Exhibit 2 details the current state of the medication value stream. This value stream starts off with the prescriber, typically a physician, writing a paper prescription for a medication to be administered to a patient. The written order is then passed to the unit clerk. The clerk faxes a copy of the order to the pharmacy and then transcribes the order to a Medication Administration Record (MAR). The MAR is the control document to record when a medication was administered and also includes other relevant data.

For an order to be entered for a patient, the patient's information must be in the general hospital system. As indicated, 30 percent of the orders submitted to the pharmacy are lacking information or have some other issue that requires the pharmacist to contact the prescriber for clarification. These incomplete orders represent a non-value added operation identified in the medication value stream. Because of the amount of time required for order entry and for tracking and correcting incomplete orders, three persons are assigned to the order entry function during the day shift, two pharmacists and one pharmacy technician. During the second shift, two pharmacists are assigned to order entry. Night shift order entry is maintained with a single pharmacist.

After the order is entered, both the pharmacist and the pharmacy computer system evaluate the order for possible adverse drug reactions. As the order is accepted, the nursing station is informed through electronic means that the prescription may be administered. Inventory positions are maintained on the units designated as Patient Dose or Pyxis Dose on the value stream map. Patient Dose inventories are drugs that are specific to a given patient. The Pyxis Dose inventory represents a computerized distributions system in which selected high-use medications are stored on the units for administration. The medications stored in a Pyxis Dose inventory will vary from unit to unit depending on which medications are most highly used. Those medications that are most highly used on a cardiac unit will vary from the most highly used medications on a maternity unit.

Made-to-order medications represent standard orders that, either because of low use or short shelf life, are only made when an order is given. Chemotherapy medications are typical of this type. Special order items are unique medications that may not be in stock or require significant additional work to complete. Roughly 65 percent of the unit dose orders can be delivered from the unit level Pyxis Dose inventory. Pyxis Dose inventories are restocked every 24 hours. Thirty-five percent of the unit dose orders cannot be filled from the local Pyxis Dose inventories. These orders are initially delivered by a pneumatic tube system to the patient's unit; subsequent doses are delivered by the standard cart delivery process.

There are two other types of orders that the medication system executes. "Stat orders" are extremely urgent orders that should be administered to the patient within 15 minutes of the order being written. "Now orders" are urgent orders that should be administered to the patient within 30 minutes of the order being written. Stat and Now orders can be either standard or specially prepared medications. Average turnaround times for orders are shown in Exhibit 3.

COMPUTERIZED PHYSICIAN ORDER ENTRY (CPOE) OVERVIEW

The concept of CPOE revolves around creating an information system that will help eliminate or minimize medication errors generated in the value stream. CPOE systems come in many forms with various functions. In the simplest form, the CPOE system simply addresses illegibility and incomplete orders by having the prescriber enter the information into a computer. The computer then prints out an order that is used by the rest of the value stream the same as a hand-written order. Other systems offer more features and more integration with the rest of the value stream. Due to the wide variety of potential options and packages that are available for CPOE, the minimum functional requirements of CPOE have been defined as a system that supports the following functions: First, the system must offer prompts that warn against possible drug interactions, patient allergies, and over- or under-doses of the drug. Second, the information in the CPOE system must be maintained and up to date. Third, drugspecific information must be available that will eliminate confusion about sound-alike or look-alike drug names. Fourth, the system must improve communication between the physicians and pharmacies. Finally, the system must reduce healthcare costs through improved efficiency (Leap Frog Group, 2000). To further define the requirements of a CPOE system, only those systems that will pass the evaluation method specified in "Overview of the Leapfrog Group Evaluation Tool for Computerized Physician Order Entry" will be considered to have met the specified minimum requirements (Classen, Kilbridge & Welebob, 2001). This type of system was considered for analysis.

CPOE systems require extensive network connections and IT support functions. Capital, software, installation, and start-up costs would be expected to be between $1.7 million and $2.2 million for a hospital between 350 and 400 beds in size. Development and implementation cost at Brigham and Women's Hospital were $1.9 million, with a yearly maintenance cost of $500,000 per year afterwards (Leap Frog Group, 2000).

COST DIFFERENTIAL COMPARISON

The activities that are different between the two value streams are those highlighted in bold on the Exhibit 4 value stream map. The first variance in activity is the computerized order entry compared to a hand-written prescription. After the initial training period, the computerized system is expected to require the same amount of time as a hand-written prescription because of editing functions and online reference information (Metzger & Turisco, 2001). In addition, physicians will charge the patients for the prescription activity as part of a separate value stream activity; thus any costs associated with the physicians' activity should be matched to the revenue received. The net effect is that, assuming the physicians take an active role in using the CPOE software, no cost differential is expected.

The pharmacist activity changes from order entry to order review. In addition, 25 percent of problem orders will be eliminated due to the CPOE system. Because of the reduction in the workload at the order entry operations, it is expected that one pharmacist and the pharmacy technician could be removed from the operations. However, the pharmacist would more than likely have to be redeployed to maintain the drug interaction database for the CPOE system as well as other IT functions. The first shift technician, a Full Time Employee (FTE), could be removed, resulting in a $40,000 yearly savings.

The majority of a unit clerk's time is spent in support functions. Roughly 60 percent of their time is spent transcribing orders, faxing orders, and preparing MARs. The balance of their time is spent on other support functions such as patient food orders, answering phone calls, and filing and maintaining patient charts. There are 17 unit clerks on day shift at the hospital, 14 unit clerks on second shift, and no unit clerks on night shift. It is estimated that the first and second shifts could be redeployed down to ten and nine per shift respectively, resulting in a reduction of 12 FTE's. Again, the yearly savings is estimated at $40,000 per year per FTE for a total savings of $480,000.

The CPOE process is also expected to reduce the number of adverse drug reactions that occur in the hospital. Exhibit 5 indicates the actual reductions achieved in medication errors that were a result of the application of a CPOE system at Metropolitan Hospital in New York (Igboechi, Ng, Yang & Buckner, 2003). Periods 1 and 2 represent the data from the two years prior to CPOE launch. Period 3 represents the year after CPOE launch. The numbers reflected in Exhibit 5 indicate dramatic reductions in errors in the medication value stream with the exception of drug therapy problems. Further investigation showed that the low reduction in drug therapy problems was being caused by prescribers ignoring the computer alerts that were generated by the system. Further training and feedback to the prescribers could improve this issue and bring reduction more inline with the rest of the results.

Exhibit 6 shows the medication errors experienced during the previous year in the local hospital and the expected reduction in errors if a CPOE system was installed. The size and type of hospitals, Metropolitan and the local hospital, are very comparable, with Metropolitan having 350 beds and the local hospital having 400.

Two types of medication value stream errors are found in Exhibit 6. Illegible and incomplete orders are documentation errors that result in losses of administration efficiency. Cost reductions for these types of errors are achieved through the productivity improvements that have been previously calculated. Incorrect orders and drug therapy problems can result in Adverse Drug Reactions (ADR) of various degrees. Evaluating the cost savings for a reduction in ADRs is more complicated. The Leapfrog Group, a consortium of more than 90 Fortune 500 companies, calculates the average cost of an ADR at $2,000 (Classen et al., 9.001). The Institute for Safe Medication Practices also sites a value of roughly $2,000 per ADR, although other literature assigns average ADR costs of up to $4,685 (Gardner, 2003). However, reports on actual applications of CPOE systems at hospitals have uniformly commented that projected savings found in the literature were not achieved. This would indicate that the previously sited cost estimates include economic opportunity costs that would be realized by society, but not by the hospital applying CPOE. Mercy General, part of the 45 hospital Trinity Health organization headquartered in Michigan, estimated an internal ADR cost of $800 (Gardner, 2003). The hospital that is being evaluated had no internal cost evaluated for an ADR.

In order to determine which of these cost estimates should be used to evaluate potential savings for a CPOE system, the severity of ADRs was investigated. Exhibit 7 shows the distribution of severity for all medication errors identified in a one-year period. Level O severity represents a medication error where the patient was not impacted. An example of a Level O error would be an incorrect dosing level that was never given to the patient. A Level VI error represents a medication error that resulted in the death of the patient. The Type I through Type VI errors represent various levels of ADRs. Type III errors and above are errors that would more than likely have additional costs associated with them, while Type Il errors and below are errors that have minimum impact on a patient and a very low probability of any added cost. Based on this distribution, the average cost of an ADR was estimated at $800, the same value used by Mercy General (Gardner, 2003). Exhibit 8 shows the cost savings that are estimated for a reduction in ADRs. The CPOE system is estimated to result in a reduction of 170 ADRs at an average savings of $800 per ADR. The total estimated savings are $135,686.

Exhibit 9 shows the net savings, project NPV (Net Present Value), and project IRR (Internal Rate of Return) based on the cost estimates that have been established. The capital cost and IT/MIS support costs are assumed to be equivalent to those costs reported by Brigham and Women's Hospital (Leap Frog Group, 2000). Cost reductions projected for the CPOE system are estimated at $655,686 per year. Added costs for system maintenance are expected to be $500,000, resulting in a net savings of $155,686 per year. Project NPV equates to a negative $1,270,112 based on an interest rate of prime +3%. Based on these projections, investing in a CPOE system would not be recommended.

The preceding analysis is based on the best guess estimates for costs and savings. Exhibit 10 shows a sensitivity analysis based on varying levels of capital costs and savings. The cost range was selected between +10% and -30%. This range was slanted downward because experience curve forces would likely decrease the costs of CPOE systems rather than increase the costs. In addition, the costs of technological components that support CPOE systems such as computers, wireless networking equipment, and network systems are also likely to decrease in the future, based on the past price performance of these products.

The variation for the cost savings was selected between +40% and -40%. Total cost savings are based on the three components of labor, ADR reduction, and system maintenance costs. New drugs, off-label applications of existing drugs, new drug interactions, and continually changing drug protocols would tend to keep system maintenance costs static although advances in database architecture may reduce these costs over time. This cost reduction range includes plus or minus 2 labor FTEs. By far the highest potential variation to cost savings depends on the type and number of ADRs that have been avoided. This range represents roughly plus or minus 75 ADR events; all other things held constant. As can be seen from Exhibit 10, positive NPV values have not been reached. Roughly a 50 percent reduction in capital costs, coupled with a 50 percent increase in cost reductions, would result in a $0 NPV for the system.

STRATEGIC AND ENVIRONMENTAL ISSUES

The market environment for CPOE systems is very dynamic at this point. Several new companies have started to develop systems, and established companies are continuing to improve their systems. In addition, new technological innovations, such as wireless networking and hand-held computers, have the potential to significantly reduce capital requirement for system launches. Only 2 percent of the hospitals in the U.S. have any type of CPOE system, indicating a vast growth market opportunity for the future (Leap Frog Group, 2002). In the next several years, it is likely that system standards and market leaders will emerge, resulting in reduced implementation costs and simplified maintenance costs. Considering these facts, it is likely that both the capital and operating cost side of the financial evaluation will decrease while the cost savings side will increase. One cost that was not considered in this evaluation was insurance premiums paid by the hospital for malpractice. As CPOE systems develop a wider deployment, it is possible that insurance companies will require CPOE.

Additional considerations must be given to overall healthcare network effects. Similar to CPOE systems are E-prescribing systems used in physicians' offices. While these systems do not have as much functionality as CPOE systems, their proliferation can have an impact on CPOE implementation. The power of E-prescribing to influence CPOE deployment comes from the large size that these networks are achieving. The gateway "SureScripts" environment is tying together multiple retail pharmacy systems to create the foundations of a national pharmacy system (Vaczek, 2003). Not only will physicians be able to write prescriptions electronically, they will also be able to see all the medical conditions that a patient has received prescriptions for from other prescribers, regardless of the physician that wrote the prescription. As this system expands, hospitals will have more pressure and incentive to digitize their medication systems to gain access to patient information as well as update patient information.

CONCLUSION

Clearly CPOE systems are in the early developmental stages and there are a number of technical, structural, cultural, and behavioral issues that must be addressed in order for the benefits of CPOE systems to outweigh the implementation and maintenance expenses associated with the system. Rapid developments in CPOE systems and other network issues that are occurring in the healthcare system indicate that this CPOE evaluation should be interpreted as a "not now" answer rather than a "don't do" answer. Future costs are expected to decrease for CPOE systems as network costs and competition drive capital costs down, while future savings are likely to increase as CPOE functions are integrated with other software based services. The best possible recommendation would be to set up a data collection and evaluation system that will collect data and help aid in future evaluations of CPOE deployment. Annual evaluations are recommended to examine the financial position of CPOE. It is likely that in the next two to three years, the financial evaluation of the CPOE system will become positive. Additional avenues of research on CPOE systems can focus on these developmental issues including specific and detailed case studies of CPOE implementations at various hospitals or healthcare facilities and behavioral analysis of physicians during the transition from paper prescription systems to CPOE systems accessing many of the cultural and behavioral implication of CPOE adoption.

The CPOE system that was modeled in this case study generated a negative net present value of roughly $1.3 million. Sensitivity analysis also indicated that the system net present values stayed negative even with a 40 percent increase in cost savings coupled with a 30 percent reduction in system expenses. In addition, value stream mapping indicated that the current process had a limited number of non-value added activities that could be addressed for cost reductions. The analysis indicates that CPOE systems have a limited financial basis for adoption at this time.

REFERENCES

Baker, W. (2002). Eliminate non-value-added costs. Industrial Management, 44(3): 22-27. Classen, D., Kilbridge, P., & Welebob, E. (2001, Dec.). Overview of the Leapfrog Group Evaluation

Tool for Computerized Physician Order Entry. Retrieved Feb. 22, 2003 from www.leapfroggroup.org/CPOE/CPOE%20Evaluation.pdf.

Gardner, E. (2003). A measured close of safety. Modern Healthcare, 33(8): 32-36.

Igboechi, C., Ng, C., Yang, C., & Buckner, A. (2003). Impact of computerized prescriber order entry on medication errors at an acute tertiary care hospital. Hospital Pharmacy, 38(March): 227-231.

Institute for Safe Medical Practices. (2000). A Call to Action: Eliminate Handwritten Prescriptions Within 3 Years. Retrieved Mar. 25, 2003 from www.ismp.org/MSAarticles/WhitepaperPrint.htm.

Leap Frog Group. (2000, Nov.). Fact Sheet: Computer Physician Order Entry (CPOE). Retrieved Feb.

22, 2003 from www.leapfroggroup.org/FactSheets/CPOE_FactSheet.pdf. Metzger, J., & Turisco, F. (2001, Dec.). Computerized Physician Order Entry: A Look at the Vender Marketplace and Getting Started. Retrieved Feb. 22, 2003 from www.leapfroggroup.org/CPOE/CPOE%20Guide.pdf.

Rother, M., & Shook, J. (1998). Learning to See: Value Stream Mapping to Add Value and Eliminate Muda. Brookline: The Lean Enterprise Institute.

Scalise, D. (2002). CPOE: Is it worth it? Hospitals & Health Networks, 76(1): 50. Vaczek, D. (2003). Preparing for E-Prescribing. Pharmacy Times, 69(1): 48-52.

Mehmet C. Kocakulah

University of Southern Indiana 

James Upson

University of Southern Indiana

Address for correspondence: James Upson, School of Business, University of Southern Indiana, 8600 University Boulevard, Evansville, IN 47712 USA, [e-mail].

EXHIBIT 1
SELECTED FINANCIAL RESULTS FOR 2003

Cash                             $14,654,506
Current Assets                    78,022,142
Net Fixed Assets                 174,892,810
Total Assets                     395,044,351
Working Capital                   46,966,215
Total Liabilities                134,070,117
Total Revenue                    280,547,839
Net Income                        18,272,415
Net Cash Flow                    (4,952,047)
Profit Margin                          6.51%
Accounts Receivable Turnover           5.02
Current Ratio                          2.51
Quick Ratio                            2.46
Total Asset Turnover                   1.41
Return on Assets                       4.63%

EXHIBIT 5 MEDICATION ERRORS BY CATEGORY

                                          Average
                        Period   Period   Periods
Error Category             1       2        1&2

Illegible orders           155      183    169
Incomplete orders          616      750    683
Incorrect orders          1853     1884   1868.5
Drug therapy problems     1238     1127   1182.5

                                             Average
                        Period    Percent    Dollar
Error Category            3      Reduction   Impact

Illegible orders            38      78%        Low
Incomplete orders          200      71%        Low
Incorrect orders          1009      46%         Low
Drug therapy problems     1081      996      Moderate

EXHIBIT 6 PROJECTED REDUCTION IN MEDICATION ERRORS

Error                   Period    Expected    Projected
Category                  1      Reductions    Levels

Illegible orders         111        78%          25
Incomplete orders        161        71%          47
Incorrect orders         323        46%         174
Drug therapy problems    245         9%         224
ADR Totals               568                    398

EXHIBIT 8
ADR COST SAVINGS ESTIMATE

Pre-CPOE ADR Errors       568         ADR COST ESTIMATE
Post-CPOE ADR Errors      398        1       2           3
Reduction in ADR Errors   170       $800    $2,000      $4,685
Total Cost Estimate             $135,686   $339,216   $794,613

EXHIBIT 9
PROJECT RETURNS ANALYSIS

Capital Costs                      $1,900,000
Interest Rate (Prime + 396)                 7.5%
Cost Reductions
  13 FTE positions @ $40K each       $520,000
  170 ADR's @ $800 each              $135,686
Total Reductions per year            $655,686
Added Costs
  IT & MIS support costs             $500,000
Net Cost Savings (Loss)              $155,686
5 Year NPV                         ($1,270,112)
5 Year IRR                                 -24%

EXHIBIT 10
SENSITIVITY ANALYSIS OF NPV
(MILLIONS)

                                 Cost Sensitivity
Saving
Sensitivity     +10%         0%       -10%       -20%       -30%

+40%          ($1.208)   ($1.018)   ($0.828)   ($0.638)   ($0.448)
+20%          ($1.334)   ($1.144)   ($0.954)   ($0.764)   ($0.574)
0%            ($1.460)   ($1.270)   ($1.080)   ($0.890)   ($0.700)
-20%          ($1.586)   ($1.396)   ($1.206)   ($1.016)   ($0.826)
-40%          ($1.712)   ($1.522)   ($1.332)   ($1.142)   ($0.952)