Recently, Ketterer et al1 reported that a score of 10 or greater on the Patient's Health Questionnaire-9 (PHQ-9)2 retrospectively predicted age of initial diagnosis of coronary artery disease (CAD) among patients in a cardiac rehabilitation program. Specifically, they found that a score of 10 was the lowest of all possible PHQ-9 cutoff thresholds for which there was a statistically significant difference in age of CAD diagnosis between patients above and below the cutoff point. Ketterer et al concluded that screening for emotional distress is necessary and that the PHQ-9 should be used for this purpose. They noted that this should be done to improve quality of life, compliance, and, possibly, morbidity and mortality.

The issue that Ketterer and colleagues address is an important one. There is a high prevalence of distress among CAD patients, including comorbid major depression in 1 of 5 patients,3,4 and many, although not all, studies have reported evidence for an association between measures of psychosocial distress and adverse outcomes, such as cardiac morbidity and mortality.5-8 Thus, it would be of great use to identify an accurate and easily implemented screening process to identify patients with high levels of psychosocial distress who might benefit from intervention. Furthermore, Ketterer et al are correct in their assertion that the decision to screen CAD patients for psychosocial distress does not depend on whether or not treatment of depression improves cardiac outcomes or overall mortality. Depression is a chronic, disabling condition with a major impact on quality of life, and, for many patients with CAD, quality of life is as important as survival.9

In general, screening is indicated if a disease or condition is an important health problem, if its presence would not be readily detected without screening, if it is prevalent in the population of interest, if screening mechanisms with good performance characteristics (eg, sensitivity and specificity) exist, if effective treatments are available, and if failure to identify and treat would have important negative consequences. In addition, screening methods should carry a minimal risk of false-positive results that might lead to unnecessary diagnostic testing, adverse effects and costs of inappropriate treatment, and the sequelae of being incorrectly labeled.10-12

The Ketterer et al study falls short on 2 counts. First, the authors did not provide any evidence that the PHQ-9 cutoff they reported is useful to detect cases of high psychosocial distress. The logic of using retrospective prediction of age of initial CAD diagnosis as a criterion to judge the screening accuracy of measures of psychosocial distress is suspect. Although early diagnosis of CAD may be associated with an earlier age of death, age of death is not an outcome variable of interest in clinical cardiology studies. Rather, time to death from a discrete event is an important variable of interest, and a robust predictor of that outcome is age at the time of the event, not age of initial diagnosis. Even more important, independent risk factors of mortality in CAD are just that-risk factors. They are statistically associated with a higher likelihood of mortality but are not typically robust enough to accurately identify which patients will die over a given period of time. The authors used a cut point of CAD diagnosis before or after age 56 to calculate PHQ-9 sensitivity and specificity, but there is no evidence that age of CAD diagnosis can be successfully used as a sensitive and specific measure of mortality risk among cardiology patients regardless of the cutoff level used. Even if age of CAD diagnosis were a useful predictor of adverse outcomes, the PHQ-9 was not a good predictor of initial age of CAD diagnosis in their study. The sensitivity for identifying patients younger than 56 years was only 46%. Probabilistically, Ketterer et al would have been as successful identifying these patients by flipping a coin. On the other side, the 66% specificity they reported is only slightly better than a coin flip in correctly identifying patients who were 56 years and older at CAD diagnosis (assuming a median-split).

The cutoff of 10 or greater on the PHQ-9 that Ketterer identified in this study does, in fact, match the recommended cutoff established for primary care.2 This, however, most likely involved some combination of chance related to nuances in the PHQ-9 score distribution that were reflected in the score dichotomizations performed by Ketterer et al. The only study that has used the PHQ-9 among patients with cardiovascular disease to identify patients with major depression found a score of 10 to be a poorly sensitive (54%) predictor of patients with major depression, suggesting that a much lower cutoff might be more accurate.13

The second flaw in the global recommendation for screening endorsed by Ketterer et al lies in their assumption that this will necessarily lead to improved patient outcomes if an accurate screening tool is identified. As demonstrated in primary care settings, screening alone does not result in improved psychosocial outcomes. Rather, patients benefit from screening when integrated systems are in place that ensure accurate diagnosis, effective treatment, and follow-up.14 Although no studies have investigated the effect of screening for depression on outcomes in cardiovascular care, these caveats make sense to consider here as well.

Thus, before we recommend that screening be done by "[horizontal ellipsis]non-mental health personnel in the clinic or at the bedside[horizontal ellipsis]" for all patients in cardiovascular care, care providers should have a readily implementable and proven protocol and resources to effectively integrate screening, accurate diagnosis, treatment planning, intervention, and follow-up. Recommendations that are not accompanied with the knowledge and resources necessary to facilitate implementation and achieve results are not likely to be used by clinicians. Indeed, the American College of Cardiology/American Heart Association practice guidelines for AMI recommend that the psychosocial status of patients be evaluated, "including inquiries regarding symptoms of depression,"15 but screening for depression does not seem to be part of standard cardiac care.16 Even worse, well-intentioned cardiology teams that implement psychosocial screening without the training and support resources necessary to provide benefits for patients may become discouraged and be less receptive to effective screening, diagnosis, treatment, and follow-up programs that are surely forthcoming.

In summary, the data reported by Ketterer et al showed that the PHQ-9 was not a reasonably sensitive or specific screening tool for retrospectively assessing age of initial diagnosis of CAD. Ketterer et al justified their selection of age of diagnosis of CAD as a "gold standard" criterion based on a mortality risk argument, but presented no evidence that age of initial CAD diagnosis, the "gold standard" used here, is at all sensitive and/or specific in identifying patients at greater risk of adverse outcomes. Furthermore, they did not provide any evidence that age of initial CAD diagnosis is useful to identify patients who have high levels of distress or recognized diagnoses, such as major or minor depression, which is the most fundamental problem of all. After all, what good is a proxy "gold standard" that is notable to identify patients who would be cases based on the actual "gold standard?" Even if it could, however, why would we need a screening tool to predict when patients were originally diagnosed with CAD? Shouldn't we just ask them?

REFERENCES