PHQ-9 Overestimates Depression Prevalence

The Patient Health Questionnaire-9 (PHQ-9) is often used in research to estimate the prevalence of depression in a population. However, a meta-analysis in the Journal of Clinical Epidemiology suggests that using this common screening tool this way can lead to substantial overestimates when compared with diagnostic methods such as the Structured Clinical Interview (SCID) for the Diagnostic and Statistical Manual of Mental Disorders.

“Disease prevalence estimates have important implications for interpreting medical research, understanding disease burden, and making decisions about health care resource utilization,” wrote Brooke Levis, Ph.D., who conducted the research during her doctoral training at McGill University in Montreal, and colleagues. “Administering diagnostic interviews in large enough samples to estimate prevalence, however, is resource intensive. Thus, researchers sometimes use self-report depression symptom questionnaires or screening tools instead and label the percentage of participants scoring above a screening cutoff as depression prevalence.”

The researchers analyzed data from 44 studies that included a total of 9,242 participants. They estimated the percentage of participants who scored at least 10 on the PHQ-9, the percentage of participants who were classified as having major depression based on the SCID, and the differences between the two estimates. The researchers found the prevalence of depression using the PHQ-9 to be 24.6%, compared with 12.1% using the SCID. This means that using PHQ-9 data more than doubles the estimate of depression compared with using SCID data.

“We weren’t surprised [by the results] because theoretically, one would expect this. Screening tools are supposed to cast a wide net to find lots of people to examine, not to estimate disorder prevalence,” Levis told Psychiatric News. In other words, the true prevalence may be different from what PHQ-9 estimates suggest because many people who report depressive symptoms on the PHQ-9 do not meet the diagnostic criteria for deression when they are further evaluated with tools such as the SCID.

“The message is simple: Researchers who want to estimate depression prevalence should use appropriate methods, not the PHQ-9,” Levis said. “It’s good for other things, such as monitoring depression symptom levels, but not for this.”

“Although the finding is not surprising, given that the PHQ is a screening tool, it is good to see this quantified,” said APA Director of Research and Deputy Medical Director Nitin Gogtay, M.D. “Screening tools are to be taken as [such] and not to bypass a thorough clinical interview and the use of semi-structured interviews, which tend to be more thorough.”

Some policymakers may interpret the results to mean that fewer people have depression than previously thought and attempt to allocate fewer resources to mental health care. That would be a mistake, said study co-author Brett Thombs, Ph.D., a professor in the Department of Psychiatry at McGill University. “Anybody who cares about mental health knows that many people with mental health conditions [already] can’t access the kind of care they need, when they need it, and the amount they need,” Thombs said. “We don’t need to exaggerate estimates of prevalence to make that clear.”

In their conclusion, the researchers noted that although they looked at data from one only depression screening tool, the PHQ-9, they “expect that other [screening] tools would similarly exaggerate depression prevalence.” ■

This study was funded by the Canadian Institutes of Health Research. “Patient Health Questionnaire-9 Scores Do Not Accurately Estimate Depression Prevalence: Individual Participant Data Meta-Analysis” is posted here.