Lessons from evaluating an automated patient severity index.

OBJECTIVE: To report lessons learned from evaluation of an automated interface between a hospital clinical information system and a severity of illness index. DESIGN: A system was developed to convert coded electronic patient findings from the HELP System at LDS Hospital into the attributes used by the Computerized Severity Index (CSI) to calculate a severity of illness score. Performance was assessed by comparing the automated CSI score with the manual CSI score (from paper chart review) and by evaluating changes introduced by augmenting the manual CSI score with verified patient data discovered by the automated CSI method. MEASUREMENTS: The strengths and weaknesses of each method are presented. RESULTS: The automated CSI score matched the manual CSI score in 61% of the cases. Sources of errors were analyzed. When the automated score was in error, two-thirds of the time it was due to the lack of codes in the HELP system representing CSI concepts; one-third of the time it was due to nurses not using established HELP system codes. Surprisingly, significant problems were also discovered in the manual system, making it difficult to define a "gold standard". CONCLUSIONS: Automated computerized severity indices have great potential for future applicability once their performance exceeds that of the time-consuming manual chart review method. Neither automated nor manual methods are adequate at the present time. This area remains a fertile ground for future research.

An automated Computerized Severity Index.

The Computerized Severity Index (CSI) is a commercially available scoring system for hospital inpatients. Trained abstractors review the patient's paper medical record and enter the diagnoses and relevant physiological attributes. The HELP (Health Evaluation through Logical Processing) System at LDS Hospital stores patient data in discrete codes. This paper describes the development of an automatic interface between the standalone, personal-computer-based severity system and the mainframe-based hospital information system. The interface scores patient severity without the need for manual chart review. Severity scores from the automated and manual methods were identical for 70% of 222 general medical patients scored retrospectively. An evaluation of the causes for differing scores between the two methods is presented.

Expression of the Pseudomonas gene coding for carboxypeptidase G2 in Saccharomyces cerevisiae.

The Pseudomonas gene coding for carboxypeptidase G2 was introduced into Saccharomyces cerevisiae on an Escherichia coli/yeast shuttle vector pROG5. The level of enzyme activity obtained was independent of the orientation of the gene within the pBR322-derived tetracycline resistance gene of the vector, indicating that expression can occur from a Pseudomonas promoter in yeast.