To study temporal trends and geographic variations in the use of surgery for spinal stenosis, estimate short-term morbidity and mortality of the procedure, and examine the likelihood of repeat back surgery after surgical repair. Cohort study based on Medicare claims. Hospital care. All Medicare beneficiaries 65 years of age or older who received a lumbar spine operation for spinal stenosis in 1985 or 1989 were followed through 1991 (10,260 patients from the 1985 cohort and 18,655 from the 1989 cohort). Two outcomes were measured: (1) rates of operation for spinal stenosis by state and (2) on an individual level, operative complications (cardiopulmonary, vascular, or infectious), postoperative mortality, and time between first operation and any subsequent reoperation. Rates of surgery for spinal stenosis increased eightfold from 1979 to 1992 for patients aged 65 and older and varied almost fivefold among US states. Mortality and operative complications increased with age and comorbidity. Complications were more likely for men and for individuals receiving spinal fusions. The 1989 cohort experienced a slightly higher probability of reoperation than the 1985 cohort for the first 3 years of follow-up. A rapid increase in surgery rates for spinal stenosis was identified over a 14-year period. The wide geographic variations and substantial complication rate from this elective surgical procedure (partly related to patient age) suggest a need for more information on the relative efficacy of surgical and nonsurgical treatments for this condition. The risks and benefits of particular surgical procedures for specific clinical and demographic subgroups as well as individual patient preferences regarding surgical risks and possible outcomes should also be evaluated further. These issues are likely to become increasingly important with the aging of the US population.