Randomized trials and registries: a computer simulation to study the impact of surgeon/patient factors on outcomes

Background contextPatient factors (diabetes, osteoporosis, cardiopulmonary problems, previous surgery, smoking, worker's compensation, litigation) and surgeon factors (operative experience, patient selection, technical skill, setting) are known to significantly impact outcomes of spinal surgery. The impact of these factors is difficult to assess clinically given the volume of patients required to obtain statistically significant information, the costs involved, and ethical/equipoise considerations. Computer simulation offers a viable and useful alternative.PurposeTo establish a computer simulation for randomized trials (randomized controlled clinical trials)/registries and to examine the impact of surgeon and patient factors on surgical outcomes.Study designComputer simulation of randomized controlled trials and nonrandomized trials (registries).MethodsOn the basis of an extensive review of the literature regarding surgical outcomes (lumbar disectomy and decompression) and patient/surgeon factors affecting such outcomes, hazard functions were developed to model the distribution of relative outcome as a function of the risk profile of individual patients and surgeons. An iterative algorithm was used to randomly or nonrandomly pair patients and surgeons to create simulated randomized controlled clinical trials/registries encompassing 10,000 performed procedures per run.ResultsWhen fully randomized, outcomes were as expected with 80% of patients obtaining a satisfactory result. When the best surgeons were paired with the best patients, success rates approached 98%; and when the worst surgeons were paired with the worst patients, success rates dropped to 53%. Other nonrandom combinations were also assessed.ConclusionsThe computer simulation obtains expected outcomes for randomized controlled clinical trials and closely mirrors the range of outcomes seen in available case-series/registry data—a very useful model allowing assessment of the impact of patient/surgeon factors on surgical outcomes. Multiple patient/surgeon combinations are assessed and the implications of findings discussed.