A pressing issue in medicine today is the rising cost of care, particularly in intensive care medicine where costs were estimated to represent up to one-third of all hospital costs [1, 2]. Usually a small number of critically ill patients (roughly 10%) consume disproportionate resources and generate higher charges. In our healthcare environment the efficient utilization of expensive resources has become a priority. On the other hand, the monitoring of ICU performances on both clinical and economic dimensions is commonly required in many countries either by payers or by accreditation programs.

Evaluation methods should combine indicators of both clinical outcomes and resource utilization and must be constructed and presented in a way that provides enough information to guide managers toward specific approaches to performance improvement. Severity of illness scores, mortality, and length of stay are the variables most frequently used to assess the quality of care and performance in ICU. Severity of illness in the ICU setting is typically quantified using models relating risk of death to physiological variables within 24 h of admission to the ICU. Such models include the Simplified Acute Physiology Score [3, 4], the Acute Physiology and Chronic Health Evaluation [5, 6], and the Mortality Probability Models [7]. These scores are statistically built from large ICUs databases and generate predicted outcomes that are case-mix or risk-adjusted benchmarks which can be compared with observed outcomes, usually standardized mortality ratio (SMR). Length of stay is a common proxy measure for utilization of ICU care and comparison of hospital performances [8]. It has been reported in more than 10,000 publications in the past decade [8]. It is available on a consistent basis for large groups of hospitals and is less subject to problems of comparability than cost data. It statistically explains approx. 85–90% of between patients variation in total cost of hospitalization [9], and case-mix plays a major role in the length of stay of an ICUs patient population [10].

Risk-adjusted and length of stay benchmarks have several potential applications. They can be used to highlight the weaknesses and strengths of individual ICUs, and accreditation organizations, government and private insurers may use such benchmarks to evaluate the performance of hospitals and particularly ICUs [11]. They also have the ability to compare the performance of one ICU to another to enable individual ICUs to trend their own performance over time and to also study mortality and outcomes trends among specific groups of similar patients. Performance data are also management tools [9], when they identify factors potentially amenable to modification and allow individual ICUs to undertake quality improvement measures. For example, Higgins et al. [12] found that one amenable factor is the organizational structure of the ICU: the presence of a full-time ICU physician reduces the likelihood of excess ICU length of stay, possibly via effects on management of mechanical ventilation, infection, and end-of-life decisions.

The important study by Rothen et al. [13] assesses the performance of ICUs across the world through the analysis of both outcome and resource use. It takes advantages of the SAPS III project, which demonstrated good quality of the data and includes information about the structure and organization of ICUs [4, 13]. This study explores the relationship between the effectiveness of an ICU relative to its resource utilization, classifies the ICUs in terms of efficiency and tries to associate this with elements of unit infrastructure [13]. Moreover it combines such relationship with a graphic approach that could be very useful when implemented in a monitoring system [9, 14]. The goal of the monitoring system is to automatically show figures that are relevant to current activity and to display them in a friendly way. It may effectively allow ICU physicians to quickly analyze their performance for better decision making, as our healthcare word is characterized by rapid changes.

However, the Rothen et al. study identified only a few points that could be improved dynamically, in part because they chose a nonoperational indicator as “region” to enter the model. Indeed, this variable probably captures nonobserved and nonobservable phenomena that could influence performance, but in the same time it is not actionable unless patients can move in and out regions. Furthermore, the method is potentially flawed because it ignores the essential hierarchical nature of the data in the sense that individuals ICUs are clustered within region and may not constitute independent observations.

Nevertheless, and beyond the use of described indicators and their structural limitations, such evaluation will best be achieved through automated information technology to assure high quality data input, timeliness, and to minimize the burden of manual data collection effort [14]. Moreover, and because the robustness of case mix adjustment is necessarily limited, for practical applications economic comparisons should be performed among similar types of ICU. Under the pressure of purchasers and regulators, hospital managers, and physicians are increasingly pressured to monitor and improve efficiency. The literature on ICU performance provides a selection of indicators that could become part of operational dashboards used by ICU managers. Not all published indicators fulfill the set of “SMART” requirements (specific, measurable, achievable, realistic, timely) used in the business management literature which allows for real-time monitoring of business activity. This probably implies for ICU managers that they should rethink the “conceptual framework for ICU performance appraisal and improvement” [1] into directional and actionable indicators. The former reflect the improvement in the quality of care (outcomes such as standardized mortality ratios) and the latter describe potentially modifiable processes such as those described in the study by Rothen or others authors [1], either at the ICU level (number and nature of staff, interdisciplinary clinical rounds, screening for antimicrobial resistant bacteria) or at the hospital level (presence of an emergency room department, computerized order entry, dedicated pharmacist). Such information is important not only as a management tool but also because healthcare professionals, be they providers, purchasers or regulators are accountable to the public.