Intensive care units (ICUs) currently represent the largest clinical cost centres in hospitals, with expenses estimated to reach up to 20% of a hospital's budget 1. The total cost per ICU patient highly depends on the severity of illness and the length of the ICU stay 2345. Complications and the need for prolonged mechanical ventilation lead to an increase in diagnostic procedures, invasive monitoring and the amount of drugs and blood products, and thus lead to an increase of the daily cost per patient 235678910. The prolonged length of stay in this resource and the personnel-intensive environment results in overall costs that, for example, in septic patients are two-fold to 11-fold higher compared with the general cost per patient 591112. Costs for personnel make up 30–69% of the total cost per patient 111314151617181920. Besides the high impact of fixed personnel and overhead costs, direct variable costs are very important to consider in order to understand the cost of an ICU patient. Depending on the therapeutic and diagnostic needs of the patient, these direct variable costs can vary a lot 2122232425. Knowing the cost of what we prescribe 2627 might lead to a more rational but not restricted intensive care 28 and is the basis for studies on cost-effectiveness 2930.

Despite the need for cost data, there are still few investigations on the real costs of intensive care therapy based on the individual patient's resource usage 2122253132. Moreover, the comparison of studies on the cost of ICU care is often difficult due to the use of varying methods of cost calculation 303334. Many studies are based on ICU annual expenditures or budgets, from which the costs are broken down for the patient numbers and the days spent in the ICU 1535363738, charges 639 or cost-to-charge ratios 610404142 to analyse the cost of ICU care. Furthermore, most of these studies were performed in university hospitals or maximal care facilities and only a few investigations include hospitals of a lower level of care 17204344. To our knowledge there are no studies directly comparing the costs of intensive care from hospitals of different levels of care.

The present study, endorsed by the German Interdisciplinary Association of Critical Care Medicine (DIVI), was performed by visiting a nationwide randomly selected representative sample of ICUs from various hospital levels of care. The study is a one-day cross-sectional survey, collecting data on all patients with a length of stay > 24 hours who were treated in the ICU. Based on these data we aimed to obtain detailed information on the costs of intensive care in German ICUs from different levels of hospital care. Part of the data obtained was presented at the congress of the European Society of Intensive Care Medicine in 2004 45.

After written consent from the hospitals' administration and from the head of the ICUs to take part in our study and with local ethics committee approval, two independent interviewers visited 51 representatively selected ICUs in hospitals all over Germany between January and October 2003. Hospital selection was based on a nationwide prevalence study on sepsis performed by the German Competence Network Sepsis, for which a representative hospital sample was randomly selected from the registry of German hospitals and stratified by size 46. From this larger study sample (454 ICUs in 310 hospitals out of 2,075 ICUs in 1,380 hospitals) a smaller sample of 51 ICUs, representing 2.5% of all ICUs in Germany, was randomly selected by the German Competence Network Sepsis administration. Four ICUs refused to take part in the study and thus were replaced by further randomly chosen ICUs according to their hospital size.

The ICUs included were defined by four levels of hospital care (Table 1). The allocation of hospital levels is based on the governmental mandate of medical care provision and mainly differs in terms of medical specialties and of diagnostic and therapeutic possibilities provided.

The present study aimed to estimate the current situation in German ICUs over all levels of care. This was achieved by visiting 51 representatively selected ICUs across Germany, covering all types of care (general, basic, maximum, main focus).

Mean total intensive care expenditure per patient per day in Germany was €791 ± 305, with 19.4% of the patients costing more than €1,000 per day and a maximum of €2,815 per patient-day. Studies from previous studies from different European countries found mean daily costs ranging between €1.125 and €1.590 per day 14163132353649. The majority of these studies were performed in university or teaching hospitals or did not break down the cost in order to compare different levels of hospital care 1722354349. Taking only university hospitals into account we found a mean cost of €1.132, which is well in line with the abovementioned recent findings. The overall lower mean daily cost of €791 compared with these studies can therefore be easily explained by the high number of hospitals of levels other than maximum care.

As shown in a number of studies, the severity of illness has a huge impact on ICU cost 23567891050. In our study population, 10% of all patients (45 patients; mean cost, €1,469) consumed about 19% of the total resources. In all levels of care the most expensive patients were those needing mechanical ventilation, those patients having a high severity of illness and/or severe sepsis as well as nonsurvivors. Patients admitted for unscheduled surgical procedures (that is, emergency cases) caused significantly increased cost.

In previous studies the following daily costs were found for septic patients: €1,318 (in 2001 from three German teaching hospitals) 5, and US$931 (in 1998 from one teaching hospital in the United Kingdom) 11.

Overall there was no difference in SAPS and SOFA scores on the study day between the hospital levels. We have to bear in mind, however, that these scores were evaluated during intensive care treatment. The lack of difference therefore only indicates a more or less stable situation during the treatment in general, not the primary severity of illness. Nonetheless, the patients treated in mcH were obviously more severely ill than those in smaller hospitals: cases needing mechanical ventilation were nearly twice as frequent in mcH as in pcH, and renal replacement therapies and other invasive procedures were more frequent in mcH. Emergency cases with unscheduled surgery requiring more intensive care interventions were also more frequent in mcH. Consequently, the TISS was higher in mcH and 84.4% of the most expensive patients (that is, the upper 90th percentile of costs) were treated there. Prediction of patients' average daily costs in intensive care, however, is only scarcely linked to descriptive criteria. Only 33.6% of the variation of daily costs (mean ± SD, €704 ± 422) in a monocentre analysis could be explained by criteria such as the Acute Physiology and Chronic Health Evaluation II score, gender, age, mechanical ventilation, emergency admission and others 21.

Resource consumption and the use of diagnostic procedures differed significantly between the hospital levels. Related to the level of performance measured by the TISS-28, however, the overall mean daily ICU cost per patient was €32.0/TISS point with only minor differences between the hospital levels (Table 3). This shows that mcH are not at all more expensive if matched against the level of performance. This profile of daily cost per TISS point is slightly less than the values of €34–37/TISS recently evaluated from a single university ICU in Finland 38 or of €36/TISS in Germany 16.

The differences between hospital categories are explained by the allocation of various hospital responsibilities. This is reflected by the patients transferred between hospitals of different levels of care. Patients transferred to maximum care ICUs were more severely ill and more expensive compared with those transferred to pcH or gcH. This partly confirms the recent findings of Golestanian and coworkers, who showed that patients transferred to tertiary care centre ICUs were more severely ill and more expensive 51. This reflects a common practice that patients who cannot be handled in primary or general care facilities due to limited diagnostic or therapeutic capabilities are usually transferred to fcH or mcH ICUs for more effective treatment 525354. On the other hand, patients who are successfully treated in fcH or mcH are often transferred back for further intensive care treatment in the lower level hospitals, often due to the lack of local intermediate care facilities.

The cost for staffing is the highest expenditure of intensive care treatment, with 56.1% on average overall (Table 3). Staffing of nurses is remarkably similar at all hospital levels. This is a consequence of official regulations on staffing for nurses, which is related to the number of intensive care beds. There are, however, no such strict regulations for physicians. In general, the larger ICUs in mcH are mostly run by full-time physicians, whereas in smaller hospitals the ICU allocation of physicians to the ICU is more reduced and they often have additional tasks (for example, in the operating theatre). Consequently, the mcH are burdened with the highest staffing cost.

To our knowledge this is the first study that compared the ICU cost nationwide in intensive care in a representative sample of 51 ICUs by analysing the resource consumption on an individual patient level. It must be mentioned that this bottom-up approach is very laborious and probably difficult to perform in studies analysing cost in a larger number of ICUs over the ICU stay. Alternatives such as cost blocks proposed Edbrooke and colleges 17243355, cost analysis based on the therapeutic score 1644565758 or cost prediction models 59 might be more applicable in daily practice. These methods should only be considered after carefully testing for accuracy on a national level, however, and are less reliable on the individual patient basis 13. With the increasing number of computerized patient data management systems in the ICU, the analysis of direct variable cost becomes easier 60. Besides the relatively large number of ICUs included in our study, there are further strengths one could consider. The ICUs were included based on a stratified random sampling strategy and the data were collected by two dedicated intensivists visiting the ICUs instead of sending out data sheets to collect probably inhomogeneous information.

There are also certain limits caused by the study design. The study was performed as a 1-day prevalence investigation that may provide accurate actual information. Owing to this design, however, we cannot draw conclusions on the total cost per patient. Moreover, the quality of care provided or its effectiveness cannot be estimated since important information on the course of ICU therapy is lacking.

In our study we only included patients with an ICU LOS longer than 24 hours. In Germany, an ICU's task within the hospital differs highly between the different levels of care. In smaller hospitals there are no intermediate care wards; therefore, postoperative recovery supervision and care in pcH and gcH is routinely provided by the ICUs. The higher level of personnel costs in maximum care facilities is mainly caused by the fact that critically ill patients are treated round the clock in maximum care ICUs.

ICUs are not only responsible for critically ill patients, however – especially in gcH and pcH – but also take care of so-called intermediate care patients. After regular working hours the ICU staff takes care of postoperative recovery patients. To avoid the inclusion of non-ICU patients, therefore, a LOS > 24 hours was defined as an inclusion criterion. We might therefore have missed extremely severely ill patients who died within the first hours after admission.

Recent studies have shown that increasing the ICU size 43 but also increasing the adequate ICU staffing can be considered cost-effective 616263. From our study it cannot be deduced that the higher resource usage and higher fixed cost in mcH may also be comparatively cost-effective in terms of outcome improvement. Such presumptions should be avoided because of differences in case mix between the hospital levels of care and because of the one-day prevalence study design. For this purpose, a matched-pairs study with comparable patient groups analysed over the whole period of the ICU stay is required.

For cost calculation, a cost catalogue based on averaged resource information from the participating ICUs was used. Owing to the confidentiality of such data, however, it was impossible to collect complete specific cost information on every item from each ICU. An averaged cost catalogue such as we used, then, might underestimate some differences in daily cost in such situations. For example, the purchasing price for a venous canula may vary by about 40% between different ICUs due to different brand and price conditions. Nevertheless, we suppose that the overall average cost catalogue may provide a sufficient basis for general cost calculations.

The present study demonstrates that a considerable degree of variation exists between ICUs according to the hospitals' level of care. These differences are mainly caused by the case mix and by the need to provide a higher level of resource consumption for the cost of diagnostic procedures and of staffing in mcH. There are common cost patterns for certain patient groups independent of ICU or hospital categories, such as those with unscheduled surgical procedures. The need for prolonged mechanical ventilation as well as the occurrence of sepsis results in significantly increased cost per day.

• A high fraction of costs result from staffing (56.1% on average overall).

• Reason for admission, severity of illness and the occurrence of severe sepsis are directly related to the level of ICU cost.

• The case mix and workload (reflected by TISS score) significantly differs between different levels of hospital care.

• Specialized hospitals and mcH treat a higher proportion of the most expensive patients.

fcH = focused care hospitals; gcH = general care hospitals; ICU = intensive care unit; LOS = length of stay; mcH = maximum care hospitals; pcH = primary care hospitals; SAPS = Simplified Acute Physiology Score; SOFA = System Organ Failure Assessment; TISS = Therapeutic Intervention Scoring System.

The study was supported by the German Interdisciplinary Association of Critical Care Medicine (DIVI), Lilly Deutschland GmbH, and departmental funds. Neither the German Interdisciplinary Association of Critical Care Medicine (DIVI) nor Lilly Deutschland GmbH has been involved in any part of the study or preparation of the manuscript. The authors declare that they have no competing interests.

OM participated in conceiving and designing the study, carried out the hospital visits, data collection and data analysis, and drafted the manuscript. EP carried out the hospital visits and data collection. UM participated in the hospital visits and data collection. AS participated in data analysis and programming of the database. HS participated in the design of the study and data analysis. MBW performed the statistical analysis, and participated drafting the manuscript. HB conceived the study, participated in its design and coordination, and helped to draft and revise the manuscript.