Bayesian methods: a useful tool for classifying injury narratives into cause groups.

To compare two Bayesian methods (Fuzzy and Naïve) for classifying injury narratives in large administrative databases into event cause groups, a dataset of 14 000 narratives was randomly extracted from claims filed with a worker's compensation insurance provider. Two expert coders assigned one-digit and two-digit Bureau of Labor Statistics (BLS) Occupational Injury and Illness Classification event codes to each narrative. The narratives were separated into a training set of 11 000 cases and a prediction set of 3000 cases. The training set was used to develop two Bayesian classifiers that assigned BLS codes to narratives. Each model was then evaluated for the prediction set. Both models performed well and tended to predict one-digit BLS codes more accurately than two-digit codes. The overall sensitivity of the Fuzzy method was, respectively, 78% and 64% for one-digit and two-digit codes, specificity was 93% and 95%, and positive predictive value (PPV) was 78% and 65%. The Naïve method showed similar accuracy: a sensitivity of 80% and 70%, specificity of 96% and 97%, and PPV of 80% and 70%. For large administrative databases, Bayesian methods show significant promise as a means of classifying injury narratives into cause groups. Overall, Naïve Bayes provided slightly more accurate predictions than Fuzzy Bayes.

The effect of waiting time on health and quality of life outcomes and costs of medication in hip replacement patients: a randomized clinical trial.

OBJECTIVE: To evaluate the effect of waiting time on health and quality of life outcomes and costs of medication in total hip replacement (THR) patients in a randomized clinical trial. METHODS: 395 THR patients were recruited into the study. When placed on the waiting list, patients were randomized into a short (< or =3 months) or a non-fixed waiting time (NFWT) (>3 months) group. In the final analyses 309 patients (179 women) with a mean age of 65 years were included. Health-related quality of life (HRQoL) (generic 15D), and pain and function (modified Harris Hip Score (HHS)) were calculated when placed on the waiting list, at hospital admission, and at 3 and 12 months postoperatively. The costs of disease-specific medication were calculated at the same measurement points. All analyses were performed using the intention-to-treat (ITT) principal. RESULTS: Of the recruited patients, 309 (78%) completed the follow-up (short group 140 and non-fixed group 169 patients). The mean waiting time was 74 days in the short and 194 days in the NFWT groups. In the ITT analyses there were no statistically significant differences between the groups in the weekly use and costs of medication, HRQoL or HHS at baseline, at admission, or 3 or 12 months after surgery. The only difference was in total medication costs during the waiting time period, at EUR 83 and 171, respectively. CONCLUSIONS: The length of the waiting time did not generate different effects on the studied health and quality of life outcomes of the randomized groups. However, those in short waiting time group reached earlier better HRQoL.

Assessment of mutilans-like hand deformities in chronic inflammatory joint diseases. A radiographic study of 52 patients.

OBJECTIVES: To evaluate patients with mutilans-like hand deformities in chronic inflammatory joint diseases and to determine radiographic scoring systems for arthritis mutilans (AM). METHODS: A total of 52 patients with severe hand deformities were collected during 1997. A Larsen hand score of 0-110 was formed to describe destruction of the hand joints. Secondly, each ray of the hand was assessed individually by summing the Larsen grade of the wrist and the grades of the MCP and PIP joints. When the sum of these grades was > or = 13, the finger was considered to be mutilated. A mutilans hand score of 0-10 was formed according to the number of mutilans fingers. Surgical treatment and spontaneous fusions were recorded. RESULTS: The study consisted of 22 patients with juvenile rheumatoid arthritis (JRA), nine with rheumatoid factor (RF) positive and 13 with RF negative arthritis, 27 patients with RF positive RA, and three adult patients with other diagnoses. The mean age of patients with adult rheumatic diseases was 27 years at the onset of arthritis. The mean disease duration in all patients was 30 years. The mean Larsen hand score was 93. Four patients had no mutilans fingers and in 15 patients all 10 fingers were mutilated. The Larsen hand score of 0-110 and the mutilans hand score of 0-10 correlated well (rs = 0.90). Fourteen patients showed spontaneous fusions in the peripheral joints. A total of 457 operations were performed on 48 patients. CONCLUSION: Both the Larsen hand score of 0-110 and the mutilans hand score of 0-10 improve accuracy in evaluating mutilans-like hand deformities, but in unevenly distributed hand deformities the mutilans hand score is better in describing deformation of individual fingers.

Motor vehicle crashes in roadway construction workzones: an analysis using narrative text from insurance claims.

Motor vehicle travel through roadway construction workzones has been shown to increase the risk of a crash. The number of workzones has increased due to recent congressional funding in 1991 for expanded roadway maintenance and repair. In this paper, we describe the characteristics and costs of motor vehicle crashes in roadway construction workzones. As opposed to using standard accident codes to identify accident types, automobile insurance claims files from 1990-93 were searched to identify records with the keyword "construction" in the accident narrative field. A total of 3,686 claims were used for the analysis of crashes. Keywords from the accident narrative field were used to identify five pre-crash vehicle activities and five crash types. We evaluated misclassification error by reading 560 randomly selected claims and found it to be only 5%. For each of four years, 1990-93, there was a total of 648,996,977 and 1,065 crashes, respectively. There was a 70% increase in the crash rate per 10,000 personal insured vehicles from 1990-93 (2.1-3.6). Most crashes (26%) involved a stopped or slowing vehicle in the workzone. The most common crash (31%) was a rear-end collision. The most costly pre-crash activity was a major judgment error on the part of a driver (n = 120, median cost = $2,628). An overturned vehicle was the most costly crash type (n = 16, median cost = $4,745). In summary, keyword text analysis of accident narrative data used in this study demonstrated its utility and potential for enhancing injury epidemiology. The results suggest interventions are needed to respond to growing traffic hazards in construction workzones.