Diagnostic and management procedures for compensable back injuries without serious associated injuries. Modeling of the 1991 injury cohort from a major Michigan compensation insurer

STUDY DESIGN: A retrospective cohort of 2425 workers with compensated back conditions was created from the 1991 compensated cohort of the largest compensation insurer in Michigan by linking computerized information on social and demographic factors, diagnostic and management procedures, and cumulative missed worktime. OBJECTIVE: To describe medical care use from date of injury in 1991 to mid-1993 by type of back condition, to identify factors affecting this use, to determine the effect of alternative model selection strategies on identifying such factors, to investigate the timing of surgery and the use of diagnostic studies in patients with displaced or herniated discs with and without surgery, and to compare the use of medical care in the study group with that recommended by national expert panels. SUMMARY OF BACKGROUND DATA: Despite the enormous costs involved in medical care for patients with work-related back injuries, almost no information on the use of medical care is available for compensated back injuries by diagnostic and procedure code. METHODS: Use of medical care was grouped into 18 categories for tabulation. Factors affecting use of medical care were identified by logistic regression, supplemented by Cox analysis for time to first procedure. The Hosmer-Lemeshow chi-square statistic was compared with the Bayes Information Criterion for evaluating model fit. Overall model utility was evaluated by comparing receiver operating characteristic curves generated by the model. For patients with displaced or herniated discs, algorithms were used to identify the diagnostic procedures performed before and after the first surgery and the amount of time that passed before each procedure was performed. RESULTS: In patients with diagnoses of disc displacement or herniation or vertebrogenic neuritis, approximately 80% underwent radiography to obtain plain views of the spine, 75% underwent diagnostic imaging, 45% underwent electrodiagnostic procedures, and 24% underwent spinal surgery (29.3% had surgical procedures among those with disc conditions). In patients with diagnoses of back sprain or other symptomatic diagnoses, the percentages for the first three procedures are approximately 70%, 12%, and 12%, respectively. Diagnostic category, age, gender, and cumulative missed worktime predicted the receipt of diagnostic and treatment procedures. Women were 30% less likely to undergo computed axial tomography or magnetic resonance imaging and 50% less likely to undergo spinal surgery. Median time to spinal surgery was twice as long in the group that underwent diagnostic imaging and electrodiagnostic testing before surgery (134.5 days). Of the 622 patients with disc displacement/herniation, 510 (approximately 80%) had initial diagnostic imaging studies, and, of the 510, 162 (approximately 30%) had surgery. Thirteen (8%) were reoperated. Of the 162 patients who had surgery, 46 (approximately 30%) had follow-up diagnostic imaging, and, of the 46, nine (approximately 20%) were reoperated. Of the 348 managed conservatively, 96 (approximately 30%) had follow-up diagnostic imaging. CONCLUSIONS: The data suggest only modest differences in the use of medical care between this study group and noncompensated study populations from previous reports. The authors of this study estimate that 27% of diagnostic imaging studies and 43% of plain radiography of the spine could have been avoided if then available Canadian recommendations or current American, guidelines had been followed. The gender effect remains unexplained and needs to be investigated in additional studies. There was an increase of 6% in the use of surgery in compensated patients compared with the use of surgery in a recent American series involving traditional health insurance, but this may be associated with greater work disability in the compensated group. The results of the current study suggest that the use and reporting of model selection strategies and the use of receiver operati