Comparing continuous versus categorical measures to assess and benchmark intensive care unit performance.

Abstract

To compare categorical and continuous combinations of the standardized mortality ratio (SMR) and the standardized resource use (SRU) to evaluate ICU performance. We analysed data from adult patients admitted to 128 ICUs in Brazil and Uruguay (BR/UY) and 83 ICUs in The Netherlands between 2016 and 2018. SMR and SRU were calculated using SAPS-3 (BR/UY) or APACHE-IV (The Netherlands). Performance was defined as a combination of metrics. The categorical combination was the efficiency matrix, whereas the continuous combination was the average SMR and SRU (average standardized ratio, ASER). Association among metrics in each dataset was evaluated using Spearman's rho and R2. We included 277,459 BR/UY and 164,399 Dutch admissions. Median [interquartile range] ASER = 0.99[0.83-1.21] in BR/UY and 0.99[0.92-1.09] in Dutch datasets. The SMR and SRU were more correlated in BR/UY ICUs than in Dutch ICUs (Spearman's Rho: 0.54vs.0.24). The highest and lowest ASER values were concentrated in the least and most efficient groups. An expert focus group listed potential advantages and limitations of both combinations. The categorical combination of metrics is easy to interpret but limits statistical inference for benchmarking. The continuous combination offers appropriate statistical properties for evaluating performance when metrics are positively correlated.