Psycho-Oncology Psycho-Oncology 23: 1079 (2014) Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/pon.3646
Letter to the Editor
Study provides little insight into routine screening for depression
Dear Editor, Screening for depression among cancer patients is not without controversy. Resolving this controversy depends on data being obtained and reported in as rigorous and transparent a manner as possible. Tu et al.[1] conclude that their data support screening among Taiwanese cancer patients to improve detection of depressive disorders. Moreover, they argue for a specific cut point on the Taiwanese Depression Questionnaire (TDQ) for use in this population. Tu et al. should be applauded for the transparency with which their data are presented, but readers should carefully examine their method, data, and analytic strategy before accepting their conclusions. Tu et al. obtained 4149 unique patient screens using the TDQ, of which 1087 screened positive. Of these, 322 individuals were referred for a psychiatric interview, which 298 completed. No diagnostic information is available for those screening negative (73.8% of the sample) or the 72.6% of the positive screening sample who were not referred or did not complete an interview. Usable diagnostic data were produced for 27.4% of those screening positive—just 7.2% of those initially screened. These data are then subjected to receiver operating characteristics analysis. Again, this analysis excludes all those who initially screened negative as well as almost 73% of those who screened positive. Diagnostic interviews were not performed for any negative screens, and no attempt is made to demonstrate a lack of bias among those who screened positive and made it through the interview procedure.
Copyright © 2014 John Wiley & Sons, Ltd.
To put it another way, there are no diagnostic data for 92.8% of the individuals screened. Tu et al. report that the TDQ was set such that false negatives were unlikely but provide no data to support this. Given the available data, most test characteristics cannot be calculated without extremely tenuous assumptions being made, such as perfect performance for negative screens and no biases in the selection of the minority of positive screens who went on to complete an interview. These data simply do not allow conclusions about either diagnostic accuracy or appropriate cut points to be drawn. What can be learned? Routine screening is difficult to implement—even when patients produced positive screens, almost three-quarters did not receive a diagnostic interview either because of their preference or that of their provider. Tu et al. also provide an example of the importance of transparency in reporting, without which we would have to rely solely on interpretations of data that are likely biased.
Reference 1. Tu C-H, Hsu M-C, Chi S-C, Lin H-Y, Yen Y-C. Routine depression screening and diagnosing strategy for cancer inpatients. Psycho-oncology 2014;23(9):1057–1067.
Dr Steven C. Palmer Abramson Cancer Center, University of Pennsylvania, 1103 Penn Tower, Philadelphia, PA, 19063, USA DOI: 10.1002/pon.3646
Letter to the Editor
Study provides little insight into routine screening for depression
Dear Editor, Screening for depression among cancer patients is not without controversy. Resolving this controversy depends on data being obtained and reported in as rigorous and transparent a manner as possible. Tu et al.[1] conclude that their data support screening among Taiwanese cancer patients to improve detection of depressive disorders. Moreover, they argue for a specific cut point on the Taiwanese Depression Questionnaire (TDQ) for use in this population. Tu et al. should be applauded for the transparency with which their data are presented, but readers should carefully examine their method, data, and analytic strategy before accepting their conclusions. Tu et al. obtained 4149 unique patient screens using the TDQ, of which 1087 screened positive. Of these, 322 individuals were referred for a psychiatric interview, which 298 completed. No diagnostic information is available for those screening negative (73.8% of the sample) or the 72.6% of the positive screening sample who were not referred or did not complete an interview. Usable diagnostic data were produced for 27.4% of those screening positive—just 7.2% of those initially screened. These data are then subjected to receiver operating characteristics analysis. Again, this analysis excludes all those who initially screened negative as well as almost 73% of those who screened positive. Diagnostic interviews were not performed for any negative screens, and no attempt is made to demonstrate a lack of bias among those who screened positive and made it through the interview procedure.
Copyright © 2014 John Wiley & Sons, Ltd.
To put it another way, there are no diagnostic data for 92.8% of the individuals screened. Tu et al. report that the TDQ was set such that false negatives were unlikely but provide no data to support this. Given the available data, most test characteristics cannot be calculated without extremely tenuous assumptions being made, such as perfect performance for negative screens and no biases in the selection of the minority of positive screens who went on to complete an interview. These data simply do not allow conclusions about either diagnostic accuracy or appropriate cut points to be drawn. What can be learned? Routine screening is difficult to implement—even when patients produced positive screens, almost three-quarters did not receive a diagnostic interview either because of their preference or that of their provider. Tu et al. also provide an example of the importance of transparency in reporting, without which we would have to rely solely on interpretations of data that are likely biased.
Reference 1. Tu C-H, Hsu M-C, Chi S-C, Lin H-Y, Yen Y-C. Routine depression screening and diagnosing strategy for cancer inpatients. Psycho-oncology 2014;23(9):1057–1067.
Dr Steven C. Palmer Abramson Cancer Center, University of Pennsylvania, 1103 Penn Tower, Philadelphia, PA, 19063, USA DOI: 10.1002/pon.3646
