Sampling rate causes bias in APACHE II and SAPS II scores

Objective: To study the effect of sampling rate of laboratory and haemodyna mic data on severity scorings and predicted risk of hospital death. Design: Prospective study. Setting: Medical-surgical intensive care unit (ICU) wit h 23 beds in a university hospital. Patients: Sixty-nine consecutive emerge ncy admission patients. Interventions: Blood samples were drawn from indwel ling arterial lines for the laboratory tests of all variables contained in the APACHE II and SAPS II scores at 2-hourly intervals from the time of adm ission up to 24 h or earlier discharge or death of the patient. Haemodynami c data and temperature were collected either manually by the attending nurs e once an hour or as 2-min median values automatically using a Clinical Inf ormation Management System (CIMS, Clinisoft, Datex-Ohmeda, Helsinki, Finlan d). Three sets of severity scores were obtained. (1) "Traditional" scores ( haemodynamic data from manual records and laboratory values from tests take n at admission and subsequently on clinical basis only). (2) "CIMS" scores (haemodynamic data from 2-min median values and laboratory values prescribe d on clinical indication) and (3) "High rate" scores (haemodynamic data fro m 2-min median values and laboratory values at 2-hourly intervals). Probabi lity of hospital death was calculated using the SAPS II and APACHE II score s, respectively. Results: Increasing the sampling rate of haemodynamic moni toring interval to 2-min from once per hour resulted in 7.8% and 11.5 % inc reases (p < 0.001) in the APACHE II and SAPS II scores, respectively. The c ombined effect of increased sampling rate of haemodynamic and laboratory te sts on the APACHE II and SAPS II scores was 14.4 % and 14.5 % compared to t raditional scores (p < 0.001), respectively. The probability of hospital de ath increased from 0.23 and 0.21 ("traditional" SAPS II and APACHE II) to 0 .31 and 0.25 ("high rate" SAPS II and APACHE II), respectively, and, becaus e eight patients died, standardised mortality ratio (SMR) decreased from 0. 53 to 0.41 (SAPS II) and from 0.60 to 0.50 (APACHE II). Conclusions: Increa sed sampling rate results in higher scores and lower SMR. Comparisons betwe en hospitals using severity scores are biased due to differences in the sam pling rates.