Tuesday, February 2, 2016

The duration of depressive episodes - one of the reasons that screening may be a bad idea

Depression is common and sometimes undetected. This has led to a continued interest in screening for depression. The assumption behind screening is that there must be a lot of people who meet criteria for a depressive episode and who could be benefit from treatment, except that they have not sought treatment because they don't know that they are depressed. A screening scale such as the PHQ-9 could possibly assist with the identification of these episodes, leading to initiation of treatment and hopefully better outcomes. However, there are some unexpected drawbacks of depression screening in practice. One is that these scales produce false positives. For example, if a scale such as the PHQ-9 is roughly 80% specific, this means that 20% of people without depression will screen positive. These would need to be assessed along with other positive results, resulting in an inefficient use of resources. A related problem is that many depressive episodes are mild and self-limited. These episodes will resolve on their own. This wouldn't necessarily be a big problem for screening except that most new episodes (the ones that screening would presumably try to detect) ARE brief. This is an under-appreciated fact since in the population the average episode is about 3-4 months. However, this average is a mixture of many brief episodes and a smaller number of longer episodes - so that screening is likely to divert resources towards those with lower levels of need, or those with no treatment needs at all. This dynamic is hard to conceptualize, but I've made animation, part of a paper published in 2006, to illustrate it. In the animation the people depicted with lighter coloured shirts have brief episodes but those with darker shirts (longer episodes) predominate in the population of people with episodes because they stay longer in the population. The most effective intervention would be to help those with longer episodes recover faster through treatment as opposed to earlier detection through screening. You can see the animation here.