THE CAPITAL COST AND PRODUCTIVITY OF MRI IN A BELGIAN SETTING *

Against the Belgian background of national planning of MRI units and a detailed reimbursement scheme, a study was undertaken to investigate the cost and productivity of MRI units in different investment scenarios and for various operational schedules. This article focuses on capital costs, not operating cost. Using data from a hospital survey and manufacturers, total capital costs per examination were simulated. The survey revealed considerable variation in operational hours, with on average 66 hours per week, resulting in 6 300 examinations per year per unit. Whilst operational hours remained approximately unchanged over the last 8 years, the number of examinations per unit grew by on average 6% per year. Correspondingly, average examination time declined from 45 to 31 minutes. The findings suggest that, mainly due to the increased productivity, capital costs per examination decreased considerably in the examined period. In 2008, the average capital cost per examination is estimated to vary from 23 Euros to 45 Euros for 1.5 Tesla units and from 32 Euros to 62 Euros for 3 Tesla units, assuming an equal examination speed for both types of units.

The supply of medical imaging services may not only be influenced by the medical know-how of the prescribing physicians and radiologists or by the demand of the patients, but also by other factors such as the reimbursement and national infrastructure planning policies.Concerned about the increase in diagnostic imaging expenditures, health care policy makers aim to use the health care financing and national infrastructure planning policy (if any) to encourage efficient delivery of high quality medical imaging.Financing and planning are used to discourage over-or under-use of services but also to avoid providers giving preference to one technique over another for other reasons than those that can be justified by evidence based medicine.
In Belgium, the number of MRI units is subject to a national supply constraint.End 2008, 92 units were accredited.This is equivalent to 8.6 units per one million inhabitants.As there is no data available on the appropriateness of current prescriptions and examinations compared to the clinical guidelines, it is not possible to demonstrate under-or overuse of this imaging technique, nor is Sources A number of data sources were used, notably a hospital survey and face-to-face meetings with manufacturers.Table I in appendix gives an overview of the sources used per cost item.More details can be found in Obyn et al. (1).

Hospital survey
In search of reliable MRI cost data, separate questionnaires were sent to the general and financial management and the head of radiology departments of 56 Belgian hospitals with at least one accredited MRI scanner.Table I in appendix provides more detailed statistics on the response rates on the different cost items.The questionnaire sent to the heads of the radiology department contained a question relating to the operational hours, while the questions posed to the financial managers related to financial investment aspects of the services.

Manufacturers
The three manufacturers active on the Belgian MRI market were contacted: Philips, Siemens and General Electric.Questions posed related to the capital costs and the operational lifetime of the equipment as well as to the technical evolution in MRI over the last decade.Two of the mentioned manufacturers provided precise information that could be used for the analyses.

Capital costs
Capital costs cover the initial purchase (including installation) of the MRI unit, the upgrades, the building adaptations and the financing of it possible to determine the number of units needed at national level that can be scientifically justified.When jurisdictions want to install a national infrastructure planning, it is beyond their ability to fully base it on clinical evidence.What policy makers are able to do, however, is to determine an appropriate financing basis.Therefore detailed data on the investment as well as operational costs of MRI are needed.Against this background, a cost analysis was conducted to assess the total costs of purchasing and running an MRI unit.This article focuses on the evolution of the capital costs and the productivity of the units over the last decade, in order to estimate the evolution in average capital cost per examination over time.The results on the operational costs can be found in the report by Obyn et al. (1).

Materials and methods
The costing methodology applied in the current cost study is historical costing, as opposed to standard costing, in that it is based on historical and actual cost data of the hospitals instead of standards defined for qualitative, efficient and safe care.

SPECIAL ARTICLE THE CAPITAL COST AND PRODUCTIVITY OF MRI IN A BELGIAN SETTING*
Against the Belgian background of national planning of MRI units and a detailed reimbursement scheme, a study was undertaken to investigate the cost and productivity of MRI units in different investment scenarios and for various operational schedules.This article focuses on capital costs, not operating cost.Using data from a hospital survey and manufacturers, total capital costs per examination were simulated.The survey revealed considerable variation in operational hours, with on average 66 hours per week, resulting in 6 300 examinations per year per unit.Whilst operational hours remained approximately unchanged over the last 8 years, the number of examinations per unit grew by on average 6% per year.Correspondingly, average examination time declined from 45 to 31 minutes.The findings suggest that, mainly due to the increased productivi ty, capital costs per examination decreased considerably in the examined period.In 2008, the average capital cost per examination is estimated to vary from €23 to €45 for 1.5 Tesla units and from €32 to €62 for 3 Tesla units, assuming an equal examination speed for both types of units.

Initial purchase
Data on purchase and building adjustment costs were obtained through the hospital survey.For the initial purchase costs, a distinction was made between the purchase of a 1.5 versus a 3 Tesla unit.

Building adaptations
For building adaptation costs two scenarios were analysed: major versus minor building adjustment requirements.Building adjustment requirements were considered to be major when a new building place was needed, generally in case of a purchase of a first or extra unit or a switch to a higher field strength unit.They were considered to be minor when limited refurbishment of an existing building was sufficient, generally in case of replacement of an older unit without switching to a higher field strength.Based on data from the hospital survey and contacts with manufacturers, a probability distribution was defined for the initial purchase and building adaptation costs, serving as input for the investment cost simulations.

Upgrading costs and lifetime of equipment
As the MRI technology continues to advance, machines that are up-todate at the time of installation may be considered obsolete within a number of years.Regular upgrades of soft-and hardware are therefore desired.By upgrading an MRI unit, scanning speed and consequently operating efficiency can be increased, imaging quality can be enhanced and clinical capabilities may be expanded.Upgrading costs were obtained from the hospital survey.As most of the units of the responding hospitals were still operational, no average lifetime, nor average upgrade costs, could be determined.Therefore, 4 scenarios were examined: 14 years lifetime of the equipment with a 50% or a 70% upgrade cost after 7 years, 10 years lifetime with a 50% upgrade cost after 5 years and a 7 years lifetime without upgrade investments.Each of these scenarios was analysed for both a 1.5 and a 3 Tesla unit and for both minor and major building adjustment costs.charge and idle time in between two patients.Examination time is not only influenced by the type of the MRI unit (mainly magnetic field strength), but also by case complexity (type of examination and patient), and by the personal preferences of the radiologist with regard to the trade-off between sharpness of images, signal-noise ratio and scanning time.Belgian figures show that the average case-mix in terms of body parts examined and examination setting (hospitalized versus ambulatory) did not change much over the considered period (Table III).These data hence do not support the hypothesis that the increased speed is linked to lower case complexity.The available data rather suggest that the increase in speed is driven by technology advancements (and the accompanying switch to higher magnetic field units) and likely also by improvements in general workflow efficiency.

Capital costs
Figure 1 shows the cost of purchase over time for 50 MRI units at 28 Belgian hospitals.The figure shows that the responding hospitals most frequently installed a 1.5 Tesla unit, which has currently a lower investment cost than the 3 Tesla units installed by some hospitals more recently.None of the responding hospitals acquired a 1 Tesla unit after 2004.Based on this dataset, the average initial investment costs for a 1.5 Tesla MRI unit slightly decreased from 1999-2000 to 2006-2008 from around €1 300 000 to €1 200 000, second hand units omitted.The small sample size of purchased units (n = 16 respectively n = 5) however precludes firm conclusions.
Price information was also obtained from 2 manufacturers for a 1.5 and 3 Tesla MRI unit in 2008.The average sales price for a basic or standard configuration of an MRI unit, which can be used for routine MR imaging in the whole body (neuro, orthopedics, abdomen and angio) and including software and coils for it, was €1 027 0001 for a 1.5 Tesla and € 1 581 000 for a 3 Tesla.With extra options for hardware (mainly specialized coils for specific body parts) and software (such as software for advanced neuro imaging, spectroscopy, soft tissue motion correction etcetera), the upper price would be about €1 378 000 for a

Financing costs
An interest rate of 4.61% was used to estimate financing costs.This rate corresponds with the average 10 year-OLO rate over the period 1998-2007 (4.46%), increased with 15 basis points as risk premium.

Scanning speed
The historical evolution of the scanning speed was estimated by combining data from the NIHDI on the number of examinations in Belgium with data on the number of operational MRI units and the average number of operating hours as derived from the hospital survey.
Uncertainty and variability of cost inputs (within each of the scenarios) was taken into account by fitting probability distribution functions to the data and incorporating the distributions instead of mean or median point estimates in the simulations.By running 1000 Monte Carlo simulations, during which input values are drawn at random from the distributions, a probability distribution was obtained for the output, the total capital costs.For this probabilistic analysis, the software package @risk 5.0 (Palisade, London, UK) was used.Appendix Table II shows the distribution functions applied for capital cost inputs.

Productivity of the units
Through the hospital survey, data was obtained on the current and historical weekly operating hours of a Belgian MRI service.The data show that the average operating hours did not change considerably over the last eight years (from 64.7 hrs on average in 1999/2000 to 65.7 hrs in 2007/2008).Summary statistics of the survey results are show in Table I.The number of examinations per unit grew by on average 6% per year (Table II).Correspondingly, the inferred examination time decreased from on average 45 minutes in 1999/2000 to 31 minutes in 2007/2008.This examination time includes imaging, patient positioning, computer set-up, patient dis-1.5 Tesla unit and about €1 945 000 for a 3 Tesla unit.These prices include first year maintenance and training of nurses and physicians.This data is in line with the data derived from the hospital survey.
Table IV shows the results for the building adjustment costs.Within the two categories considered, major versus minor building adjustments, large differences in costs were observed between hospitals.These may be explained by, amongst others, the size and suitability of the location, the used materials, the variations in cages of Faraday and the number of walls needing extra shielding.
Figure 2 plots the probabilistic outcomes for the resulting total yearly equivalent capital costs for 1.5 and 3 Tesla units.These costs vary from on average €160 000 (in the scenario 14 yrs -50% upgrade -minor building adaptations) to €240 000 (7 yrs -0% upgrade or 10 yrs -50% upgrade -minor building adaptations) for 1.5 Tesla units.For 3 Tesla units, these average costs vary from €230 000 to €330 000 respectively.These capital costs are one-time, fixed costs.The associated cost per examination therefore depends on the utilization of the unit.Combining the annual capital costs with the different scenarios for operational hours, whilst taking into account an average examination speed of 31 minutes, an average capital cost per examination was obtained ranging from €23 to €45 for 1.5 Tesla units and from €32 to €62 for 3 Tesla units in 2008 (Table V).In 1999/2000, when the average examination speed was still 45 minutes, the average capital cost per examination is estimated to range from €40 to €71 for 1.5 Tesla units 2 .

Discussion and conclusion
As the demand for MRI examinations increased over the last years and as the number of MRI units was restricted by the government, Belgian hospitals have been spurred to make the most efficient utilization of the limited resources.The hospital survey highlights that operating hours ranged from 53 to 86 hours per week with an average of 66 hours, resulting in on average 6 300 examinations per year per unit.In a previous Belgian study (3), this productivity appeared the

Table I .
-Number of operational hours per week.Table II.-Evolution of time required per MRI examination.64.7 hours/week * 50 weeks/year ** 65.7 hours/week * 50 weeks/year.
*Source: based on national statistics from NIHDI.