Clinical assessment scales to evaluate motor abilities in stroke survivors could be used to individualize rehabilitation interventions thus maximizing motor gains. Unfortunately, these scales are not widely utilized in clinical practice because their administration is excessively time-consuming. Wearable sensors could be relied upon to address this issue. Sensor data could be unobtrusively gathered during the performance of motor tasks. Features extracted from the sensor data could provide the input to models designed to estimate the severity of motor impairments and functional limitations. In previous work, we showed that wearable sensor data collected during the performance of items of the Wolf Motor Function Test (a clinical scale designed to assess functional capability) can be used to estimate scores derived using the Functional Ability Scale, a clinical scale focused on quality of movement. The purpose of the study herein presented was to investigate whether the same dataset could be used to estimate clinical scores derived using the Fugl-Meyer Assessment scale (a clinical scale designed to assess motor impairments). Our results showed that Fugl-Meyer Assessment Test scores can be estimated by feeding a Random Forest with features derived from wearable sensor data recorded during the performance of as few as a single item of the Wolf Motor Function Test. Estimates achieved using the proposed method were marked by a root mean squared error as low as 4.7 points of the Fugl-Meyer Assessment Test scale.