Embed Size (px)
Citation preview
Matthew ThomasChristopher BourdeauxZoe EvansDavid BryantRosemary GreenwoodTimothy Gould
Validation of a computerisedsystem to calculatethe sequential organ failureassessment score
Accepted: 21 October 2010Published online: 9 December 2010� Copyright jointly held by Springer andESICM 2010
Dear Editor,The sequential organ failure assess-ment (SOFA) score is a commonlyused severity score [1]. It can be useddaily as well as on admission [2].Previous trials have suggested crea-tion and validation of an electronicmethod of assessing severity scores[2]. We have designed and validated amethod of creating a computerisedSOFA score.
The local research ethics commit-tee accepted that ethical approval wasnot required.
Fifty patients were selected byusing a stratified sampling system.Data from 1 day on the ward (chosenat random) for each patient wereincluded in the final sample.
A computer score was calculatedas follows: raw data from the Inno-vianTM system were extracted,cleaned and processed into holdingtables in a reports database. To gen-erate SOFA scores a series of SQLqueries, updates, inserts and stored
procedures interrogated the sampledata to produce a final output forviewing and analysis.
A physician score was calculated;two intensivists independently calcu-lated a SOFA score for each set ofdata using electronic records. Thetime taken to generate these scoreswas determined in blocks of 10 with astopwatch. The two physiciansreviewed the raw data, with theirindividual scores, to determine anintegrated score.
Areas of difference were identifiedbetween the integrated human scoreand the computer score and werereviewed independently by an expe-rienced ICU physician. Theindependent unblinded reviewerdetermined the correct score and thecause for the error was identified.
The groups were tested for nor-mality and their correlation tested byusing Pearson correlation coefficients.SPSS was used for all data.
A mean time of 4.9 min per patientwas required for physician scoring.The computer processes the raw datafrom the Innovian system at a speedof 8 s per patient and can then gen-erate 169 SOFA scores per second.No manual input of data is requiredfor the computer-generated score.
SOFA scores were all normallydistributed, allowing us to do the fullrange of parametric analysis. Themean SOFA score was 8.1 and thestandard deviation was 2.9. Thecomputer was more accurate withthan either doctor and was similar tothat of the integrated score.
Several recurrent errors werenoted. The physician errors includedpoor arithmetic, misinterpretation ofthe rules and incorrect data identifi-cation. Common errors were seen inthe respiratory component as a
mathematical calculation of eachindividual P/F ratio had to be done.The computer errors included venousblood gases labelled as arterial, oxy-gen percentage mislabelled in litres/minute and pre-sedation GCS recor-ded as text. We have subsequentlyrewritten the query to eliminate theseerrors
This trial has shown that a com-puterised scoring system based onrules can be more accurate and rapidthan that performed by a physician.
Using our eSOFA score we arenow able to deliver daily SOFAscores on each patient to the multi-disciplinary team prior to the morningward round. The total unit SOFA is agood indicator of the acuity of theunit and the individual scores assist inthe organization of nursing allocation.
References
1. Vincent JL, Moreno R, Takala J, WillattsS, De Mendonca A, Bruining H, ReinhartCK, Suter PM, Thijs LG (1996) TheSOFA (sepsis-related organ failureassessment) score to describe organdysfunction/failure. On behalf of theWorking Group on Sepsis-RelatedProblems of the European Society ofIntensive Care Medicine. Intensive CareMed 22:707–710
2. Arts DG, de Keizer NF, Vroom MB, deJonge E (2005) Reliability and accuracyof sequential organ failure assessment(SOFA) scoring. Crit Care Med33:1988–1993
M. Thomas ()) � C. Bourdeaux �Z. Evans � D. Bryant � R. Greenwood �T. GouldBristol Royal Infirmary – Intensive Care,Upper Maudlin Street,Bristol BS2 8HW, UKe-mail: [email protected].: ?44-117-3422089Fax: ?44-117-3423089
Intensive Care Med (2011) 37:557DOI 10.1007/s00134-010-2083-2 CORRESPONDENCE
