This research aimed to create and evaluate an algorithm for detecting steps in people with lumbar spinal stenosis (LSS) using accelerometer data. The first goal was to describe a method for step detection from accelerations measured at the wrist, hip, lower back, thigh, and ankle. The second goal was to evaluate the method’s accuracy while walking with and without walking aids and while performing other activities. Patients with LSS often experience a decline in their walking abilities; however, there is currently no validated test to evaluate this aspect of their daily functioning. About 30 people with LSS wore accelerometers at 5 distinct wear sites and completed a standardized mobility protocol that includes walking with and without walking aids and non-walking activities. As a follow-up to the aforementioned walk tests, a system was developed for optimal step detection and compared to an observed gold standard of step count. By applying the step detection method to accelerations felt in the lower back, hip, thigh, and ankle, they could obtain a reliable step count even when walking for extended periods without using any assistive devices. When using mobility assistance, accuracy decreased everywhere except the ankle. Accelerometers on the thigh and ankle were prone to inaccurately detecting steps during cycling, while the wrist delivered the highest inaccurate step count overall. The most precise step count came from the accelerometer worn on the ankle, although it incorrectly recorded steps taken while doing other activities. Further development and testing under free-living conditions are warranted because the developed step detection method shows promise as a measure of walking activity.