The use of pedometers in stroke survivors: are they feasible and how well do they detect steps?

I used one for a while and I think it severely undercounted.
http://www.hubmed.org/display.cgi?uids=22373934&utm_source=feedburner&utm_medium=feed&utm_campaign=Feed%3A+nih%2Fbxxu+%28Stroke+rehabilitation%29&utm_content=Google+Feedfetcher
To determine (1) the feasibility of pedometers for stroke patients and (2) the level of agreement between pedometers and actual step count.Observational agreement study.Six stroke units.Independently mobile stroke patients (N=50) ready for hospital discharge.Patients were asked to apply 3 pedometers: 1 around the neck and 1 above each hip. Patients performed a short walk lasting 20 seconds, then a 6-minute walk test 6MWT. Video recordings determined the criterion standard step count.Agreement between the step count recorded by pedometers and the step count recorded by viewing the criterion standard video recordings of the 2 walks.Five patients (10%) needed assistance to put on the pedometers, and 5 (10%) could not read the step count. Thirty-nine (78%) would use pedometers again. Below a gait speed of about 0.5 m/s, pedometers did not generally detect steps. Agreement analyses showed that even above 0.5 m/s, pedometers undercounted steps for both the short walk and 6MWT; for example, the mean difference between the video recorder and pedometer around the neck was 5.93 steps during the short walk and 32.4 steps during the 6MWT.Pedometers are feasible but generally do not detect steps at gait speeds below about 0.5 m/s, and they undercount steps at gait speeds above 0.5 m/s.