For more information, please contact us.
Estimating the Likelihood of Two Raters Assessing the Same Patient Exactly the Same Using the PANSS Dementia
Alan Kott, MUDr and David Daniel, MD
The proportion of identical scorings (all 30 PANSS items identical) across consecutive visits has been estimated to be approximately 5% in schizophrenia clinical trials (Daniel and Kott, 2014). In the current analysis, Dr. Kott and Dr. Daniel estimated the likelihood two random raters would agree on all 30 PANSS items rating the same subject from 73 datasets of raters rating the same videotaped subjects used for certification purposes. The results indicate that it is extremely unlikely for two raters to agree exactly on all 30 PANSS items even when rating the same videotaped subject and the authors conclude that identical scorings should be considered as a marker of poor data quality and investigated further.
Frequency and Regional Distribution of Errors Scoring the CGI Detected by Blinded Data Analytics
David Daniel, MD and Alan Kott, MUDr
Blinded data analytics seeks to unobtrusively identify patterns of clinical trials ratings associated with poor quality data. Measures of function, such as the CGI, are particularly vulnerable to error. The problem is compounded in schizophrenia trials by the multidimensional nature of schizophrenia symptoms. The analysis of this poster sought to determine the frequency and regional distribution of errors in usage of the CGI in a data base derived from 10 global schizophrenia studies.