Joseph L. Fleiss was an American professor of biostatistics at the Columbia University Mailman School of Public Health, where he also served as head of the Division of Biostatistics from 1975 to 1992. He is known for his work in mental health statistics, particularly assessing the reliability of diagnostic classifications, and the measures, models, and control of errors in categorization.
While still a college student, Fleiss began his career at the Biometrics Research Unit of the New York State Psychiatric Institute, first as a statistical clerk and later as a research scientist and biostatistician. He was affiliated with the Psychiatric Institute until 1986. In 1975, Columbia University recruited Fleiss to be a professor and head of the Division of Biostatistics at the School of Public Health. He remained in that capacity until 1992. Under his leadership, the Division increased in size and stature. Fleiss transformed the Division from a small program consisting chiefly of New Yorkers into a department with international prestige. He instituted a Ph.D. program in 1977. He recruited top faculty from major institutions around the world. The Division trained students, performed independent research, and supported clinical research associated with Columbia University's health sciences divisions.
Field of expertise
One of Fleiss's chief concerns was mental health statistics, particularly assessing the reliability of diagnostic classifications, and the measures, models, and control of errors in categorization. He was among the first to notice the equivalence of weighted kappa and the intraclass correlation coefficient as measures of reliability in categorical data. In an influential 1974 paper co-authored with Robert Spitzer, Fleiss demonstrated that the second edition of the American Psychiatric Association's Diagnostic and Statistical Manual of Mental Disorders was an unreliable diagnostic tool. They found that different practitioners using the DSM-II were rarely in agreement when diagnosing patients with similar problems. In reviewing previous studies of 18 major diagnostic categories, Fleiss and Spitzer concluded that "there are no diagnostic categories for which reliability is uniformly high. Reliability appears to be only satisfactory for three categories: mental deficiency, organic brain syndrome, and alcoholism. The level of reliability is no better than fair for psychosis and schizophrenia and is poor for the remaining categories."
Publications
Fleiss wrote two textbooks that are considered classics in biostatistics. His first book, Statistical Methods for Rates and Proportions, concerned the issue of using proportions to summarize counts and frequencies. According to professors Patrick E. Shrout and Melissa D. Begg, the book "attracted a wide readership with its many engaging examples and a thorough, but accessible discussion of esoteric statistical principles". It continues to be particularly influential in the fields of psychiatry and epidemiology. Fleiss's second book, Design and Analysis of Clinical Experiments, was equally influential among a different group of medical researchers. It reviewed the principles and challenges concerning the planning and interpretation of studies involving people. Human subjects present experimenters with challenges not seen in other scientific experiments: they may fail to take the treatment, they may drop out of a study, and they may lie or misrepresent the outcome. In his book, Fleiss addressed these problems. Fleiss also contributed chapters to more than two dozen books and authored or co-authored more than 200 statistical and scientific papers concerning the application of statistics in fields ranging from psychiatry and cardiology to dentistry.
Fleiss suffered from Lewy-body Disease, which led to his retirement as department head in 1992, when he was 55 years old. By the age of 58, he was no longer able to work. He died in 2003.
