PubMed1978
The case histories of more than 5000 outpatients with respiratory diseases were stored in free text from 1974 untill 1977. The retrieval efficiency was tested using STAIRS-IBM program product with the syntax operators "or", "and", "not", "with", and "adj". Chronic bronchitis in lung cancer patients was chosen as an example. The first word constellation was compiled as a retrieval argument for chronic bronchitis. The second word constellation was used as a retrieval argument for chronic bronchitis subsisting for more than 5 years before the diagnosis of lung cancer. The third word constellation retrieved documents without clear time relationship between chronic bronchitis and lung cancer. The review of each history of the 473 lung cancer patients issued a first group of 219 cases composed of lung cancer patients with histories of chronic bronchitis for more than 5 years. A second group of 180 patients had no history of preexisting chronic bronchitis. This manual selection of cases (reading of all 473 case histories) took 3 months compared to 4 h by computer dialogue selection. 87% of the first group (preexisting chronic bronchitis for more than 5 years) and 74% of the second group (no history of chronic bronchitis) were correctly obtained by the brief computer dialogue. This efficiency of the computer dialogue was further improved without difficulty by slight modifications of the dialogue strategy, so that only approximately 10% of documents remained falsely classified. It can be seen that free text analysis of case histories is possible by the STAIRS-IBM program and enables a doctor to observe new clinically important correlations in a very short time. Final deductions, however, must be followed by careful checking of all individual histories.