Efficiency measures both clinical outcomes and cost, such that efficient intensive care units (ICUs) achieve optimal clinical results with minimal resource utilization. The COVID-19 pandemic placed unprecedented pressure on health care systems, with a significant increase in hospital care demand that may have affected the quality of care provided. Identifying the resulting health care resource needs allows projection of future demand during crises and anticipation of their impact on patient survival.

In this study, clinical outcomes were compared with resource utilization to evaluate the relative efficiency of a private-sector ICU in Mexico City (Mexico) over three periods: pre-pandemic (2018–2019), pandemic (2020–2021), and post-pandemic (2022–2023). To determine whether the COVID-19 pandemic caused significant changes in performance across these periods, the modified Rapoport–Teres method¹ was applied using data from a prospective database. The Rapoport–Teres method cross-tabulates the standardized clinical performance index (SCPI) and the standardized resource utilization index (SRUI) on the x- and y-axes, respectively, allowing comparison of an ICU’s outcomes with the mean values of ICUs participating in the IMPACT project (United States.), as well as across different time periods for the same ICU. The SCPI reflects survival at hospital discharge vs expected survival (in this study, the SAPS-3 model was used), whereas the SRUI is primarily based on each patient’s hospital length of stay, both in and out of the ICU setting, assigning a specific value to each day according to the level of interventions required (more points for the first ICU day). Detailed calculations for both indices can be found in the original publication.1 Additionally, we conducted a descriptive analysis of demographic variables and a comparative analysis of care workload for each period. Mean comparisons were carried out using ANOVA, and differences in proportions were tested using the chi-square test.

A total of 2,309 patients were included. The mean age was 63.9 years, with a significant age difference observed in the post-pandemic period (Table 1). During the pandemic, the proportion of male patients treated was higher than in the other two periods, which is consistent with former studies showing a greater incidence rate of severe disease and mortality among men with COVID-19, although this finding alone does not confirm higher susceptibility. The number of patients treated was also higher during the pandemic, which is consistent with ICU saturation and health care system overload.2

The SCPI and SRUI values were 2.31 and –0.30 in the pre-pandemic period, 0.49 and –2.49 during the pandemic, and 2.26 and –1.89 in the post-pandemic period, respectively (Table 1, Fig. 1). These results indicate that observed mortality was lower than expected across all three periods, exceeding the adjusted Rapoport–Teres standard by more than two standard deviations before and after the pandemic—a performance advantage that was not maintained during the pandemic, when mortality worsened by >1 standard deviation (SD) vs the other periods. Another notable finding is that resource utilization during the pandemic was significantly higher, exceeding two standard deviations, and never superior to the Rapoport–Teres benchmark in any period. Taken together, both indices indicate efficient performance, albeit at the cost of high resource use.

Figure 1.

Rapoport–Teres graph graduated in standard deviations (SD). Coordinates within ± 2 SD indicate ICU performance consistent with peer units participating in the IMPACT project, from which the method was derived. Graph quadrants may be designated as “more efficient” (upper right), “less efficient” (lower left), “efficient but at the expense of high resource use” (lower right), and “low performance” (upper left). SCPI and SRUI values are shown for the pre-pandemic (circle), pandemic (triangle), and post-pandemic (square) periods, as well as for the average of the three periods (diamond).

The results reflect the complex interaction among clinical factors, available resources, and workload across the 3 periods studied. Observed ICU mortality during the pandemic did not exceed the expected mortality predicted by the SAPS-3 model, suggesting that despite increased demand for beds and resources, health care teams maintained effective management of critically ill patients. This finding aligns with former studies indicating that ICU care quality, in terms of mortality, was not compromised during the pandemic despite care overload.3 Nevertheless, as illustrated in the Rapoport–Teres graph, a clear decline in clinical performance was observed during the pandemic.

On the other hand, resource utilization remains a major concern. During the pandemic, the length of stay was substantially longer due to increased case complexity. According to the Rapoport–Teres method, hospital stay duration is the primary resource indicator, given its impact on bed occupancy and staffing requirements. This method, which has been applied in various study populations,4 provides an easily interpretable visual representation of the cost-effectiveness relationship.

Although length of stay is a strong predictor of ICU costs, it does not directly measure actual economic expenses, since factors such as staffing levels, medication costs, and equipment use also influence total cost. Furthermore, reducing length of stay has been shown to yield limited savings, as ICU costs are highest during the initial days of care.5 Therefore, shortening hospital stay should not be the sole strategy to reduce costs. Similarly, patient mortality serves as a surrogate for clinical performance and fails to capture many other aspects of ICU functioning.

The stability in patient age between the pre-pandemic and pandemic periods suggests that demographic profiles did not change substantially, consistent with other analyses, although the higher mean age in the post-pandemic period may reflect delayed care for vulnerable patients during the pandemic.

The retrospective design of this study and its single-center data represent its main limitations. Due to hospital saturation during the pandemic, some critically ill patients may not have received ICU care—an important factor, since associated costs for severe cases outside the ICU were not analyzed. Current mortality prediction models, including SAPS-3, tend to overestimate mortality. Moreover, after 2 decades since its revision, the Rapoport–Teres method requires recalibration to align with current standards. The scarcity of studies comparing ICU efficiency before, during, and after the pandemic highlights the distinctiveness of this work.

In conclusion, although the COVID-19 pandemic did not increase the difference between observed and expected ICU mortality, it did cause a significant rise in resource utilization, particularly the length of stay. These findings underscore the importance of efficient resource management, especially during periods of high demand. The experience gained during the pandemic may prove invaluable in enhancing preparedness for future health crises.