Medical language processing with SGML display.

The paper demonstrates several ways that medical language processing can be combined with emerging display technologies to facilitate the extraction of data from free-text patient documents. The techniques allow rapid review via highlighting of the results of processing. Coupling of text markup with further procedures is envisioned.

Recovery of mycobacteria from blood in mycobacteria growth indicator tube and Lowenstein-Jensen slant after lysis-centrifugation.

Five hundred five blood samples for culture were processed in the Isolator lysis-centrifugation system and were then inoculated into a Mycobacteria Growth Indicator Tube (MGIT) and onto a Lowenstein-Jensen (L-J) slant. Forty-nine isolates of Mycobacterium avium complex and three isolates of Mycobacterium tuberculosis were recovered from 50 of the blood culture specimens. Forty-five isolates from 43 specimens were recovered in the MGIT, with a mean time to detection of 21 days. Forty-one isolates from 40 specimens were recovered in the L-J slants, and the mean time to detection was 36 days. Nine specimens were positive in the MGIT alone, while seven specimens were positive only in L-J medium.

Medical language processing: applications to patient data representation and automatic encoding.

A linguistic approach is presented to develop a representation of patient data. Semantic categories developed for computer processing of narrative clinical reports are shown to be similar to the Medical Concepts used manually to extract data from narrative in Exercises of the Computer-based Patient Record Institute. Clinical statement types composed of these categories are used in the Linguistic String Project (LSP) medical language processing (MLP) system to convert narrative information into relational database tables of patient information. A procedure for mapping the output of the LSP MLP system into SNOMED International codes was developed. Preliminary results and further requirements are discussed.

Natural language processing and the representation of clinical data.

OBJECTIVE: Develop a representation of clinical observations and actions and a method of processing free-text patient documents to facilitate applications such as quality assurance. DESIGN: The Linguistic String Project (LSP) system of New York University utilizes syntactic analysis, augmented by a sublanguage grammar and an information structure that are specific to the clinical narrative, to map free-text documents into a database for querying. MEASUREMENTS: Information precision (I-P) and information recall (I-R) were measured for queries for the presence of 13 asthma-health-care quality assurance criteria in a database generated from 59 discharge letters. RESULTS: I-P, using counts of major errors only, was 95.7% for the 28-letter training set and 98.6% for the 31-letter test set. I-R, using counts of major omissions only, was 93.9% for the training set and 92.5% for the test set.

Automatic encoding into SNOMED III: a preliminary investigation.

The Linguistic String Project (LSP) medical language processing (MLP) system converts narrative clinical reports into database tables of patient data. A procedure for mapping the output of the LSP MLP system into SNOMED III codes was developed. Preliminary results and further requirements are discussed.

Natural language processing of asthma discharge summaries for the monitoring of patient care.

A technique for monitoring healthcare via the processing of routinely collected narrative documentation is presented. A checklist of important details of asthma management in use in the Glasgow Royal Infirmary (GRI) was translated into SQL queries and applied to a database of 59 GRI discharge summaries analyzed by the New York University Linguistic String Project medical language processor. Tables of retrieved information obtained for each query were compared with the text of the original documents by physician reviewers. Categories (unit = document) were: (1) information present, retrieved correctly; (2) information not present; (3) information present, retrieved with minor or major error; (4) information present, retrieved with minor or major omissions. Category 2 (physician "documentation score") could be used to prioritize manual review and guide feedback to physicians to improve documentation. The semantic structuring and relative completeness of retrieved data suggest their potential use as input to further quality assurance procedures.

Analyses of the FlashTrack DNA probe and UTIscreen bioluminescence tests for bacteriuria.

Five hundred urine specimens were selected at random and screened for bacteriuria by a DNA probe method, FlashTrack (Gen-Probe, San Diego, Calif.), and an automated bioluminescence method, UTIscreen (Los Alamos Diagnostics, Los Alamos, N.M.), and the results were compared with those of the semiquantitative plate culture method. The performance of each test versus culturing was evaluated at colony counts of greater than or equal to 10(4), greater than or equal to 5 x 10(4), and greater than or equal to 10(5) CFU/ml. Since the interpretive breakpoint of each test was user selectable, the results were reported as receiver operator characteristic curves. Optimum interpretive breakpoints were determined for each test at each colony count by calculating a performance index that emphasized sensitivity over specificity in a 70:30 ratio. Although both tests had less-than-optimal sensitivities and specificities, the performance of FlashTrack was significantly better than that of UTIscreen at two of the three colony counts (10(4) and 10(5) CFU/ml); however, FlashTrack costs more and is a labor-intensive procedure. Neither method was evaluated for the detection of colony counts of less than 10(4) CFU/ml.

TEXTINFO: a tool for automatic determination of patient clinical profiles using text analysis.

The clinical data contained in narrative patient documents is made available via grammatical and semantic processing. Retrievals from the resulting relational database tables are matched against a set of clinical descriptors to obtain clinical profiles of the patients in terms of the descriptors present in the documents. Discharge summaries of 57 Dept. of Digestive Surgery patients were processed in this manner. Factor analysis and discriminant analysis procedures were then applied, showing the profiles to be useful for diagnosis definitions (by establishing relations between diagnoses and clinical findings), for diagnosis assessment (by viewing the match between a definition and observed events recorded in a patient text), and potentially for outcome evaluation based on the classification abilities of clinical signs.