Abstract
Gill SJ, Lukmanji S, Fiest KM, Patten SB, Wiebe S, Jetté N. Epilepsia 2017;58(5):695–705. doi: 10.1111/epi.13651. Epub 2017 Jan 8.
OBJECTIVE: Depression affects approximately 25% of epilepsy patients. However, the optimal tool to screen for depression in epilepsy has not been definitively established. The purpose of this study was to systematically review the literature on the validity of depression-screening tools in epilepsy. METHODS: MEDLINE, EMBASE, and PsycINFO were searched until April 4, 2016 with no restriction on dates. Abstract, full-text review and data abstraction were conducted in duplicate. We included studies that evaluated the validity of depression-screening tools and reported measures of diagnostic accuracy (e.g., sensitivity, specificity, and negative and positive predictive values) in epilepsy. Study quality was assessed using the Quality Assessment of Diagnostic Accuracy Studies Version 2. Medians and ranges for estimates of diagnostic accuracy were calculated when appropriate. RESULTS: A total of 16,070 abstracts were screened, and 38 articles met eligibility criteria. Sixteen screening tools were validated in 13 languages. The most commonly validated screening tool was the Neurological Disorders Depression Inventory for Epilepsy (NDDI-E) (n = 26). The Mini International Neuropsychiatric Interview (MINI) (n = 19) was the most common reference standard used. At the most common cutpoint of >15 (n = 12 studies), the NDDI-E had a median sensitivity of 80.5% (range 64.0–100.0) and specificity of 86.2 (range 81.0–95.6). Meta-analyses were not possible due to variability in cutpoints assessed, reference standards used, and lack of confidence intervals reported. SIGNIFICANCE: A number of studies validated depression screening tools; however, estimates of diagnostic accuracy were inconsistently reported. The validity of scales in practice may have been overestimated, as cutpoints were often selected post hoc based on the study sample. The NDDI-E, which performed well, was the most commonly validated screening tool, is free to the public, and is validated in multiple languages and is easy to administer, although selection of the best tool may vary depending on the setting and available resources.
Commentary
I suppose it is tempting, if the only tool you have is a hammer, to treat everything as if it were a nail.
—Abraham H. Maslow; The Psychology of Science, 1966
Certainly it is true that with the right tool, any job will be markedly easier. Tools for depression screening may represent the Holy Grail in the world of psychiatric comorbidity in epilepsy. Depression is a dangerous condition and, unfortunately, is widely prevalent in persons with epilepsy. The presence of depression has a huge impact upon quality of life, and convincing data show that outcomes for depression are intertwined with outcomes related to seizure disorder (1, 2). Thus, identifying depression is critical for comprehensive care and to improve epilepsy itself in many cases. The problem is that depression is often hidden from direct view and is a difficult subject to discuss for any clinician. In fast-paced clinical care environments, depression may be especially elusive. A measure or tool to identify depression would be extremely worthwhile. Not surprisingly, many rating scales have been developed to address this concern.
The universe of depression screening tools runs the gamut from clinician administered structured interview questionnaires to brief online self-reports. The goal is the same, to efficiently identify clinical depression that requires prompt treatment. The recent paper by Gill and colleagues comprehensively looks through the toolshed and evaluates the utility of screening instruments for depression. The authors highlight 16 measures that have been used in a wide variety of clinical settings. All tools selected for analysis had been validated with diagnostic benchmarks and had reasonable reliability ratings.
The good news is that many high-quality tools are at our disposal. Most of them involve a questionnaire format where either the patient or the caregiver rates items on a Likert scale. The questionnaire may specify a certain recent time period and usually asks raters to select a 0, 1, or 2 to indicate whether symptoms are never present, sometimes present, or always present. Some questionnaire tools have dozens of items and some have only a few, even less than 10. However, all result in a scaled score that involves a cutoff point above which depression is most likely present. The cutoff scores are determined by comparisons with normative data obtained from a nondepressed population and may have been further evaluated with test-retest analyses or correlations to gold standards in the field.
Gill and colleagues present a powerful analysis assessing the utility of many well-established rating scales that are widely used and well validated. The analysis is precise and detailed, and the conclusions are well reasoned. In that sense alone, this article will be frequently referenced by investigators. Ultimately, the authors suggest that the Neurological Disorders Depression Inventory for Epilepsy (NDDI-E) is the most useful of the many rating scales that were assessed. The NDDI-E was designed for use in an epilepsy population and is geared toward eliciting pertinent information regarding depressive symptoms, so it may be no surprise that the NDDI-E emerges as the best tool for the job.
However, the bad news is that even with the comprehensive analysis, the premise of the article still falls short in one important sense. In many cases, rating scales are insufficient. Even the NDDI-E has sensitivities and specificities in the low 80% range; this range is very respectable for screening tools, but that leaves at least one out of every five persons with depression hidden from view—too much to accept for clinical purposes. The authors were limited to studying existing questionnaire measures, but the larger question is whether any questionnaire measures are sufficient. Put another way, Gill and colleagues thoroughly and well assessed all the hammers that are available. The problem is that we are not always finding the right nails.
Depression is heterogeneous and may differ in an epilepsy population as opposed to a population without epilepsy. Even in persons without epilepsy, depression is heterogeneous and results from multiple etiologies, even if the scores from rating scales are similarly above cutoff points. For example, rating scales may not differentiate complex bereavement from bipolar depression from melancholia, yet all of those are managed much differently.
Ultimately, we in epileptology must accept what is begrudgingly accepted in the field of mental health: that diagnosis of depression may be unpleasantly arbitrary. Rating scales are still tied to categorical diagnostic criteria either directly or indirectly, so by design can be no more valid than those criteria. In making an official diagnosis of major depressive disorder according to the Diagnostic and Statistical Manual of Mental Disorders, no scientific evidence affirms that five symptom criteria are superior to four criteria or even six (3). Symptom criteria are rough approximations, descriptive categories to use because there are no better options. Even the NDDI-E, the best tool of the bunch, may be insufficient for the purpose of finding depression.
In our own Neuropsychiatry in Epilepsy Program at Kennedy Krieger Institute, we use rating scales that are more general in nature. We do not use specific tools for depression alone. We use comprehensive tools that offer screening for a wide range of mental illness. Broad-based questionnaires can yield an index of suspicion for a variety of illnesses, including anxiety disorders and attention or cognitive problems. Still, if the scores are not consistent with our clinical impressions, then we ignore the ratings.
Unfortunately, we have not found a shortcut to identifying depression. It has to be identified the hard way, by asking pointed questions in a bold and stepwise manner while providing comfort and safety for patients who may be frightened about the gravity of their depressive symptoms. The rating scales may be helpful but do not compare to the power of a trusted clinician simply asking the question, “Are you depressed?”
Although it is tempting to resign oneself to nihilism regarding the repertoire of our tools, I doubt that Maslow would have agreed. We have identified the best hammers possible for our time. Gill and colleagues have helped us to do that. Now let's dedicate ourselves to learn more about the nails.