Dr. Kuno is affiliated with the Department of Veterans Affairs Health Services Research and Development Center of Excellence, Roudebush Veterans Administration Medical Center, Indianapolis. Dr. Kuno is also with the Department of Electrical and Computer Engineering, Purdue University School of Engineering and Technology, Indiana University-Purdue University Indianapolis. Dr. Hadley and Dr. Rothbard are with the Department of Psychiatry, University of Pennsylvania, Philadelphia. Send correspondence to Dr. Kuno at the Roudebush Veterans Administration Medical Center, 1481 W. 10th St., 11-H, Indianapolis, IN 46202 (e-mail: firstname.lastname@example.org).
As the practice of health care becomes increasingly more complex, incorporation of computer systems into practice presents the opportunity to enhance efficiency and effectiveness in all areas of health care (1). Computerized prescription systems are a promising technology that can reduce medication errors and improve operational efficiency and quality of prescribing (2,3,4,5).
However, despite their widely recognized benefits, actual implementation of such systems has been slow. Previous research on randomly selected hospitals in the United States in 2002 found computerized ordering systems were partially available at 6.5% of hospitals and completely available at 9.6% of hospitals (6). The National Ambulatory Medical Care Survey of office-based physicians found that when information technology was used, it was used most frequently for billing (73%) and much less frequently for maintaining medical records electronically (17%) and ordering prescriptions electronically (8%) (7).
A study of information management systems in ambulatory mental health care settings in the state of New York found that few provider agencies utilized a computer system to manage clinical care for patients (8). The reported barriers to the use of computer applications include physician and organizational resistance, as well as substandard functionality and reliability of the technology (9,10). Cost is also seen as a significant obstacle to implementing information technology in practice (11). Although computerized systems are not without problems, as seen in some studies in which they are reported to introduce errors of their own (12,13), this risk is unlikely to outweigh the other benefits of computerized systems (14).
This case report examined the cost of implementing an electronic prescription system for psychiatrists and other authorized personnel in a large public-sector mental health agency in 2004. Although the actual cost would differ across organizations, the method of determining cost on the basis of the identified categories is generalizable. Because cost is always a concern in publicly funded agencies, which are generally struggling with a high volume of patients and budgetary constraints, it is important to inform decision makers about costs associated with adding a computerized prescription component to their existing information system so they can weigh the relative costs and benefits.
As part of a study that looked at quality of care in relation to reducing racial disparities in mental health treatment, a computerized prescription system was implemented in four not-for-profit specialty mental health agencies in an urban setting. The number of psychiatrists working at the agencies ranged from five to ten full-time equivalents (FTEs). We chose the Web-based system that provides the ability to electronically input and track medication data and create medication orders that can be submitted electronically to the pharmacy of choice or printed locally. The system also provides clinical resources at the point of prescribing that offer extensive drug interaction information. Each agency already had its own information infrastructure for billing purposes. Because the new electronic prescribing system was Web based, agencies were required to have a computer network with an Internet connection that was fast, secure, and reliable as a prerequisite to being a study site. An electronic interface connection between the prescription system and existing management information applications needed to be installed to import information on existing patients and to continue adding information on new clients. The costs associated with the implementation process were covered by the research grant. This included the initial start-up costs, as well as monthly fees for prescribing and nonprescribing users.
This brief report describes the implementation costs at one agency with ten FTE psychiatrists for which we had information on time devoted by the management information system (MIS) personnel to implementing the computerized prescription system, in addition to cost estimates in other categories described below. A cost scenario is also presented to show how costs would vary on the basis of the number of prescribers.
There were four major areas of cost associated with the implementation of the computerized prescription system: first, costs for preimplementation or start-up activities; second, technical or capital costs for system upgrade and integration; third, labor costs related to training and support; and fourth, the cost of ongoing user fees and technical support. Cost items are presented in Table 1.
The preimplementation phase activities included internal planning meetings, vendor and system selection, and workflow analysis of staff at the agency sites to determine location of computers and printers. For the purpose of the research project, the research team examined and compared existing prescription systems, presented candidate systems to a steering committee consisting of executive directors and medical directors of the participating agencies, negotiated with a vendor, and made final contracting decisions.
The technical costs of implementation included the capital costs associated with upgrading computer hardware and MIS staff time to incorporate the new system into the existing systems and to integrate information into the new system on new patient's sociodemographic characteristics, diagnoses, and other comorbid conditions. Training costs involved the clinical and administrative staff involved in the prescribing process. The vendor provided group training to the four participating agencies designed for several kinds of groups: a full day of training for in-house trainers, a three-hour training session for prescribers, a full day of training for administrative and data management staff, and a 1.5-hour training course for nonprescribing users. In addition, each agency had a two-day individualized training session in which the data management staff was shown how to activate the system, assist physicians by showing them how to test data input using their own computers, and discuss with the in-house implementation team strategies to facilitate the implementation process. The vendor charged $36,000 for training, and we assigned $9,000 as the cost of individual agency.
Ongoing costs to the agency included user fees paid to the vendor, lost productivity for training of new users, and staff time for maintenance of electronic data interfaces.
The costs of the computer network hardware and user fees were based on purchase orders initiated by the research team. Costs of staff time by discipline were derived from the project's activity logs and training schedule. Information on MIS staff time was derived from agency records gathered by the research team documenting the time spent for system implementation and support for clinical users. To generalize labor costs we used the metropolitan area occupational wage estimates for administrative, clinical, and MIS personal time, rather than actual salary of agency staff. Median hourly wages used for these estimates were $100 per hour for psychiatrists, $27 per hour for nurses and other clinical staff, $28 per hour for MIS personnel, and $60 per hour for senior management staff.
A cost scenario was done to show the variation in cost categories based on the number of prescribers, which is a proxy for the size of the agency. In this analysis, we assumed a personal computer and printer would be purchased for each user. The cost components that varied by the number of prescribers include lost productivity of psychiatrists associated with preimplementation planning, lost productivity of psychiatrists associated with training, monthly fee for each prescribing user, and cost of computers and printers. The remaining cost components—for example, lost time of other staff personnel and the training fee paid to the vendor—were considered fixed.
As part of preimplementation planning and buy-in efforts, a series of meetings were held by the research team with the executive directors, medical directors, psychiatrists, and the MIS staff. Information on the existing computer network and hardware environment was gathered in the meeting with the MIS staff. The estimated cost of the preimplementation efforts at the single agency analyzed (that is, lost productivity) was $3,720. The university research team devoted 40 hours to the preimplementation activities. The cost of the research staff time for this effort was estimated using an hourly wage equivalent to data management staff personnel ($28 per hour) (Table 1).
The system is Web based and requires broadband Internet access for all users. Because the agency already had the required wiring for the Internet connection, no additional cost was incurred. However, ten desktop computers ($600 per unit) and printers ($80 per unit) were purchased. The MIS staff logs showed a total of 116 hours spent on system integration ($3,248). In total, the estimated cost for technology and system integration was $10,148.
In addition to the $9,000 vendor fee for training (as described above), the cost for physician time in lost productivity was $2,100 for ten FTE psychiatrists, followed by in-house trainer costs ($1,024). The subtotal of costs for lost time for these two groups was $13,739. This includes the on-site individual training and system activation.
Ongoing costs are dependent on the numbers of FTE prescribers who use the system and new users who need to be trained. The annualized cost of user fees for prescribers was based on the number of FTE physicians at the agency (ten physicians, which led to a cost of $9,600). The amount of staff time spent on support or maintenance was estimated at four hours per month. The cost of training one new psychiatrist and one nurse was included in the ongoing support category. The annualized costs were $14,725.
The start-up cost of implementation and first-year use of a computerized prescription system was $42,332. Future annual costs would be $14,725, on the basis of the same staffing configuration used for implementation.
A cost scenario presents the cost estimates, corresponding to varying numbers of psychiatrists (Figure 1). The cost of components with costs that varied by the number of prescribing psychiatrists was estimated at $2,140. The cost estimate of remaining components was $21,726. Thus the cost per prescribing user would be $6,500 for an agency with five psychiatrists and $4,300 for an agency with ten psychiatrists.
This brief report provides evidence that it is feasible to implement a computerized system in public mental health agencies at a reasonable cost, with annual costs after implementation of around $14,000 per year for ten full-time prescribing users. In the four agencies in which the system was implemented, this is generally less than 1% of medical staff costs for each agency. A positive cost benefit should also be seen for these systems, given that prior studies have shown reduced medication errors (2,3,4,5). In addition, these Web-based systems can provide clinical and management staff with reports on prescribing patterns of their own physicians and identify potential cases of underuse or overuse of medications at an agency. They can also show users how patients with similar diagnoses are treated in general by using their entire data bank to create profiles of prescribers and patients. Finally, patient medication patterns can be compared with best practice information, which can be used by prescribers and management to enhance quality of care.
There are several limitations to the generalizability of the cost estimates in this study. Undoubtedly, the technological sophistication of agencies will differ considerably. In this study, our case site was appropriately wired for Internet use and required very little hardware support. In the case of another agency with poor technology it was necessary to wire an entire building with costs of $2,500. However, implementation costs related to improved technology for electronic prescribing should be reduced in the future as agencies upgrade their general computing capabilities for all activities, allowing them to meet the minimum requirements of broadband Internet access.
Another limitation to generalizability of these analyses was that training costs were shared by several agencies, which likely reduced the cost to some degree for the agency reported on here. However, if a county mental health system or managed care organization required prescribing data from its providers, these costs might be shared among groups of agencies.
This brief report demonstrates that the implementation of a well-designed computerized prescription system is a viable policy option that can be done at a relatively small cost to the agency and can bring substantial long-range benefits for purposes of monitoring and improving quality of care for public-sector clients. Once computerized prescription writing becomes routine, the prescription data system can be linked to clinical information from electronic records and progress notes that contain client outcomes. This allows the possibility of monitoring medication patterns and outcomes for individual clients, psychiatrists, programs, and specific cohorts of patients.
This research was supported by grant ME-02-382 from the Commonwealth of Pennsylvania, "Reducing Disparity for SMI African Americans."
The authors report no competing interests.