Kamuzu Central Hospital (KCH), located in Lilongwe, Malawi, is a government-operated tertiary referral hospital with 710 beds. Care is available to patients at no cost. In 2010, it treated 275 880 patients, with an average bed occupancy of 83%.11
KCH employs 176 physicians and clinical officers (similar to nurse practitioners) and 262 nurses. The hospital has long-standing shortages of both staff and supplies,12
factors that affect how fully the benefits of an EMR system are realized. Resource constraints are relevant to this analysis because they affect the ability of the hospital to capture the financial benefits; non-personnel savings are naturally more easily realized than personnel-related savings, and this difference is compounded by the existing staff shortages. For example, the hospital complement is supposed to have 532 nurses, indicating a shortage of over 50% in 2010.13
These issues are compounded by similar shortages in surrounding health centers, which increase the number of patients seeking treatment at KCH and the amount of primary care that is provided.
In 2001 the non-governmental organization Baobab Health Trust implemented a prototype of the EMR in operation at KCH.14
Over the following decade the EMR was iteratively developed, beginning with a pediatric module and expanding to include patient registration, radiology, laboratory specimen management, inpatient admission/discharge, and chronic disease management for HIV/AIDS and non-communicable diseases.15
Additional modules currently under development include a laboratory management system, and computerized provider order entry for ordering medications. The KCH EMR has approximately 70 touchscreen clinical workstation appliances (workstations) deployed at the point of care in the ambulatory setting that enable health workers to retrieve medical records using barcode scanners.16
Two central servers within KCH store the EMR data on a local network supported by a centralized power backup system. We estimated a one-time cost for the full implementation of all modules of the EMR that included initial software configuration to accommodate clinical workflow, followed by routine operating costs such as electricity and consumables.
A 2010 internal analysis identified 13 potential areas of cost savings from the EMR system at KCH.11
We attempted to re-analyze in greater detail the four areas with the largest potential benefit: length of stay; transcription time; adverse drug events; and laboratory use. We collected data from several departments to quantify a baseline under the paper-based system. The hospital administration provided data on patient volume and length of stay, the laboratory supplied data on test volume and supply orders, and interviews with staff such as nurses and clinical officers gave insight into the daily workflow. We chose to exclude the category of adverse drug events from this analysis, as we did not have the data to support this initial study; we are currently independently studying this category in more detail. Costs and projected savings are modeled in US$.
We chose the hospital as the unit of analysis to capitalize on the returns to scale of the EMR system and account for the relationships among the different hospital departments. Furthermore, because the decision to invest in an EMR system was made at the hospital level, and its sustainability is the responsibility of the Ministry of Health, it seemed relevant to analyze the effect of the EMR system on the hospital's budget. However, it is also possible to extend the theory that economies of scale are important in realizing the financial benefit of EMR and analyze potential financial benefits at a higher level of organization, such as at the district, zonal, and national levels. For example, one of the natural consequences of an EMR is the collection of large amounts of data. These data can easily be aggregated to generate routine reports required by the Ministry of Health, and thus additional cost savings may also be realized at higher levels.
Length of stay
In the USA, a business case analysis conducted by Kaiser Permanente estimated that approximately 35% of net benefits identified from the introduction of an EMR are attributable to a reduction in the average length of stay resulting from efficiency gains.17
Electronic communication of orders and results between clinical and ancillary departments (laboratory, radiology, pharmacy) are cited as a driver of efficiency gains.18
Evidence of the impact of EMR on hospital length of stay varies widely. An early study conducted on an internal medicine service in a large urban hospital in the USA showed a 0.89 day (10.5%) reduction in length of stay resulting from electronic ordering.19
The Government Accountability Office benchmark for length of stay reduction associated with an EMR is 30%.18
Thus far, we have found no study measuring the impact of EMR on length of stay in a low-income setting.
Length of stay at KCH averaged 4.82 days in 2010–11, but varied significantly by ward. For example, the male surgical ward averaged 18.6 days per stay, while it was only 2.18 days in the maternity ward. Our analysis includes all non-paying inpatient wards at KCH. While there are challenges that contribute to length of stay that the EMR system may not directly benefit, such as pharmaceutical and blood shortages, communication lapses also contribute to longer stays in the hospital; for example, if a patient's laboratory results are ready but have not been picked up in time for rounds, the patient stays admitted until the next opportunity for rounds, which may be 2–3 days due to limited rounding at the weekend (B. Gondwe, personal communication, 2012).
To maintain the conservative nature of this model, we assumed a 10.5% reduction in length of stay for inpatients and consider only cost savings associated with meal provision and time saved for clinical staff. The hospital provides meals to a patient at a cost of Malawi kwacha 350, or approximately US$1.75, per day. In calculating a reduction in nurse and clinician time required we assume that approximately 60% of their time is spent managing inpatients (T. Bui, personal communication, 2012).
Transcription activities are those that involve transferring information from one place to another. In the USA, EMR systems have been found to reduce transcription time by 28–50%, largely through the ‘partial elimination of dictation’.2
Dictation is not a relevant part of the KCH workflow, but medical staff spend a significant amount of time engaged in ‘indirect care’, defined as discussing cases with colleagues, reading and writing patient charts, and prescribing medications.20
Time is also spent in the compilation of mandatory reports for the Ministry of Health, admission registration, and chart creation. A study of three HIV clinics in Uganda found that providers spent approximately 34% less time during the workday engaging in indirect patient care after the implementation of an EMR system.20
A similar study at a rural health center in Kenya found a two-thirds reductions in inter-clinic personnel contact time as a result of the implementation of the Mosoriot Medical Record System.21
Anecdotal evidence of the time burden of indirect care under a paper-based system is provided for hospital antiretroviral therapy clinics in Ethiopia, where providers cite the excessive time spent in activities such as error checking with pharmacists and consultations with laboratory technicians.22
Based on interviews conducted in 2010, we estimated that doctors, laboratory technicians, nurses and some administrative staff spend approximately 1 h per working day on transcription.11
The integration of barcodes and label printing into the EMR system reduces transcription time by eliminating the need to write information that has previously been entered elsewhere in the system. We attribute a reduction of 17 min per employee per working day (28%) in transcription time to the EMR system.
Laboratory data are involved in 70% of medical diagnoses and can significantly affect the success and cost of patient treatment.23
In the USA, EMR have been found to reduce test orders by 7–15%.2
Some of this reduction is attributed to improved access to health records; it has been estimated that 13.7% of laboratory tests are ordered because of lack of access to earlier results.25
The HIV/AIDS pandemic has triggered investment in laboratory capacity building with some emphasis on information systems to improve laboratory workflow in several low and middle-income countries, including Côte d'Ivoire, Haiti, Malawi, Peru, Rwanda, South Africa, Vietnam and Zambia.26–29
However, for the benefits of information systems in this setting to be fully realized the management of laboratory information must extend beyond the walls of the laboratory, starting at the point the clinician requests the laboratory investigation and finishing at the time the clinician makes a decision based on the test result.27
Smaller health centers must send specimens to centralized laboratories for tests to be completed, further complicating the process and creating additional delays. The increasing dependence on expensive PCR testing in HIV management in Malawi means that more laboratory testing needs to be centralized.30
The 11-step process and associated information management challenges of centralized laboratory testing for early infant diagnosis of HIV has been described in this context.26
Benefits resulting from the use of information systems to support the management of laboratory information have been demonstrated in low-income countries such as Zambia and Rwanda, commonly finding a reduction in turnaround time and fewer results that do not reach clinicians.28
A 65% reduction in the delay of accessing tuberculosis results as well as the detection of duplication of laboratory testing has been demonstrated in Peru.32
In this model we limit cost savings from the EMR system to benefits realized in two areas: fewer samples redrawn due to better labeling, and fewer tests ordered due to improved access to past test results. KCH data quantified the magnitude of incorrect sample labeling and test ordering. A study examined 3549 samples sent to the KCH laboratory for testing during a 4-week period, identifying whether they were correct and complete or needed to be redrawn; if the latter, the reason(s) the sample was inadmissible were recorded.33
Overall, 18% of samples sent to the laboratory were discarded due to paperwork and labeling errors. Similar challenges associated with incomplete laboratory request forms and associated wastage and delays in processing laboratory specimens have been described in South Africa.34
While an EMR system should completely eliminate documentation and labeling errors, the model assumed a 50% reduction in wasted samples due to documentation and labeling problems. Improving clinicians’ access to medical history, specifically recent laboratory tests, should reduce the number of duplicate tests ordered. We make the conservative assumption that an EMR would reduce the number of tests conducted at KCH by 7%. Based on information provided by the KCH laboratory, we estimated the cost of a sample to be US$1, and the cost of the test to be US$3; both figures account for employee time and materials.
The investment cost of the EMR system is calculated at US$337 847, and includes hardware, software configuration, project management, installation and training; these costs are outlined in . In representing the costs of the comprehensive EMR system implementation, we excluded software development costs because the software is freely available and open source.16
While the level of EMR deployment at KCH is significant (approximately 70 workstations), costs described here are not incremental and assume no existing infrastructure. Costs are extrapolated from the actual installation of the EMR system at KCH, and include 151 workstations and two servers. Each workstation included a touchscreen computer, label printer, and barcode scanner. The hardware, network electronics and power backup system are described in detail in a previous publication.16
Project management cost is an estimated 1 year salary for a local project manager who coordinates the system implementation. Training includes the costs for providing half-day training sessions for clinical staff who will use the EMR, and salary for local software support staff who provide training full time for an estimated 6 months during EMR implementation. Installation includes the cost of a team of three local hardware technicians who install all system hardware over a period of 4 weeks. Software configuration includes 1 month of salary for one local software support staff and one software developer who configure the software modules to support the clinical workflow within each point-of-care setting. The yearly operating costs are estimated to be US$29 824, and include consumables (eg, labels) (US$9258), electricity (US$9656), hardware and software maintenance (US$5910), and recurrent training (US$5000). In year 3, the operating cost is higher (US$47 424) due to the cost of replacing the batteries (US$17 600).
Investment costs for the EMR system at KCH
Discounting and sensitivity analysis
The benefits and costs outlined here are summarized over time in , and the 5-year financial outlook is shown in . To maintain the conservative nature of this analysis, we assume that no benefits are realized in the first 6 months of use. This choice recognizes that benefits of the EMR will accrue gradually, and that any initial efficiency improvements are likely to be offset by productivity losses. Costs and benefits are discounted on a yearly basis at a 5% annual percentage rate, the assumed cost of capital for the hospital.
Annual savings due to EMR system
Five-year return on EMR implementation
The estimated reductions in the various categories are determined based on a mixed-methods approach, using the existing (western) literature, data from KCH, and interviews with KCH administration and staff. In we present variations on the assumptions in to examine how the financial outlook changes when the benefits vary. We examine two different sets of alternative assumptions. First, we examine how the outlook changes when we assume only half of Western the reductions identified in EMR in the model. Second, because savings associated with consumables are more quickly realized than those associated with personnel, we distinguish between personnel and non-personnel cost savings by removing personnel cost savings from our original analysis.
Robustness of financial gains to varying benefit assumptions