OBJECTIVE: To explore the use and interpretation of self reported on/off diary data for assessment of daily motor fluctuations in Parkinson's disease.

METHODS: 26 consecutive non-demented patients with fluctuating Parkinson's disease received standardised training on how to fill out the four category CAPSIT-PD on/off diary, followed by four hours of clinical observation and four weeks of daytime on/off diaries every 30 minutes at home.

RESULTS: Overall patient-clinician agreement in diary entries was good (kappa = 0.62; weighted kappa = 0.84). Agreement for individual diary categories was good for "off" and "on with dyskinesias" (kappa = > or =0.72), but moderate for "partial off" and "on" (kappa = 0.49). The overall validity of patient kept diaries was supported by expected symptom severity variability across diary categories, as assessed in the clinic. One day's home diary data failed to predict outcomes from the full four weeks for all diary categories, and data from three days failed to yield good prediction (predefined as R(2) = > or =approximately 0.7) for the time spent in "off" and "partial off". Data from one week yielded good prediction (R(2) = > or =0.74) in all instances except "partial off", which could not be well predicted even when two weeks' home diary data were considered (R(2) = 0.52).

CONCLUSIONS: The data provide support for the overall accuracy and validity of the four category CAPSIT-PD on/off diary, but suggest that a three category diary format may improve accuracy and validity. Interpretation of diary data beyond the assessed time frame should be made with caution unless diaries have been kept for sufficiently long periods.