Dr. Blashfield Replies

To the Editor: Pincus (1) noted that MEDLINE is constantly adding and deleting journals from its database. He worried that the results Mr. Intoccia and I reported about changes in the personality disorder literature over time might be influenced by these decisions. Dr. von Knorring et al. make the plausible suggestion that searching another literature database, such as PsycINFO, might help avoid this problem.

We decided not to report the results from PsycINFO in our search because this database, even more than MEDLINE, appears to have undergone substantial changes regarding the journals it includes. We generated a plot from the PsycINFO database of all articles, chapters, and such that were indexed per year. The resulting plot was scalloped in shape. There were years of clear growth, followed by a sudden drop, followed by more growth. In contrast, the plot of total number of articles on MEDLINE per year showed little deviation from a straight line. To us, this result suggests that the PsycINFO database has undergone major changes in the journals it was indexing, especially in the late 1970s and mid-1980s.

Second, Dr. von Knorring et al. suggest that using “personality disorders” as a general search term would be preferable to the tactic that we used of searching on all individual personality disorder terms (e.g., “schizotypal”) and summing the results. Wells and I (2) performed a detailed analysis of the 1985 personality disorders literature and reached a similar conclusion. The reason that Mr. Intoccia and I used a more detailed search tactic was that a major goal of our analysis was to compare how the literatures for individual personality disorders were changing relative to the total literature.

Dr. Pincus (1) and Dr. von Knorring et al. are correct in expressing concern about potential sources of systematic measurement error when performing empirical studies using journal article databases. Wells and I (2) demonstrated surprising differences in the literatures generated by MEDLINE and PsycINFO, as well as with a literature sample found from a manual search of the same journals. Leininger (3) showed that there are substantial reliability issues when using major categories as search terms in PsycINFO. Assessing the growth of scientific literatures involves complicated methodological decisions that can have dramatic effects on the results.

A minor point: the Journal of Personality Disorders is indexed in MEDLINE but only since 1997. Back cataloging may lead to the inclusion of all articles from this journal in the future.