We launched this Blog for the Grief study in February. By March 2013, we were beginning to to get feedback and discussion via Twitter and e-mail about our plans for the project. One particularly interesting e-mail came from Professor Ronald Pies, psychiatrist and lecturer at The State University of New York. Professor Pies’ name was familiar to us from reading the contributions he has made, along with his colleague at the University of California, San Diego, Professor Sidney Zisook, on the merits of removing the bereavement exclusion from the new Diagnostic and Statistical Manual of Psychiatric Disorders.
Professor Pies raised some important questions about the design of the Grief Study. We thought other people might be interested in this correspondence and considering how our approach compares to alternative approaches available for the study of grief. Thus, with Professor Pies’ permission, an edited version of our discussion appears below. Initial pleasantries and introductions have been removed and the emails have been merged together to create a discussion.
If you have any comment or questions about the project that you would like to discuss publicly or privately, then please contact a member of the Grief Study team via e-mail, twitter, or by using the comment box at the bottom of the blog.
– Mark
RP:
Dear Grief Study Team,
Dr. Allen Frances recently made me aware of your intriguing study of bereavement, and I wanted to offer just a few thoughts for your consideration. First, I have some concerns re: “Our proxy measure for poor mental health is prescription of anti-depressant or anxiolytic drugs in the post-bereavement period.” From my perspective as a mood disorder specialist, I am not convinced that such prescription—though an easy and convenient proxy to track in data bases—is a useful or valid “marker” of psychopathology or “poor mental health.” In the U.S., at least, prescription of sedatives or antidepressants—frequently by general practitioners or non-psychiatric physicians—may indicate anything from lack of access to psychotherapy; to over-zealous prescribing; to a simple complaint of insomnia, during the bereavement period—not necessarily a patient’s “poor mental health.” (Perhaps things are different in the UK?). I would encourage you and your team to consider some more revealing measure of psychopathology, such as use of the GAF scale, Beck Depression Inventory, or the Symptom Checklist 90 (SCL-90). But if you are limited to searching data bases, and cannot actually assess the bereaved patients, psychiatric hospitalization would probably be a better proxy for “poor mental health”, in my view.
MMcC:
Dear Professor Pies,
Thank you very much for your e-mail and your interest in the study.
We would expect to find some patterning of prescribing such as you have described, and we hope to account for that in analysis. Hypnotics or anxiolytics may be prescribed to help with insomnia in the days and weeks post-bereavement. We would expect antidepressant prescribing to be much less common until six months after bereavement (i.e. after the period of the bereavement exclusion. {Edit – Here’s another article about bereavement exclusion}. We can investigate our research question with antidepressant prescribing after six months as our outcome measure, we can also account for the fact that some people may or may not have taken medication to help with sleep disturbance earlier in the course of the grief process.
In Northern Ireland, the vast majority of antidepressant and anxiolytic prescribing is by general practitioners (GPs). Everyone is registered with a GP, and their services and prescriptions are free at the point of use. There may be variation in prescribing due to differential access to psychotherapy, or difference in custom and practice of GPs (this includes under- or over-zealous prescribing). We will account for prescriber-level variation using multilevel modelling approaches – as such, our results will present the average risk estimate of prescription use, accounting for the this variation.
(Note – our next blog post features a discussion of using prescription information as an indicator of poor health)
As this study is based on linkage to administrative records, we cannot administer scales to bereaved individuals. The use of this proxy measure is a limitation, but this is offset against having information on such a large sample. As to the suggestion of using psychiatric hospitalisation as an alternative proxy, this would be another useful mental health threshold to measure. Under the current data linkage agreements this outcome is not currently available, although it may be possible to attempt linking such records for a future project.
RP:
On a more theoretical note, my colleagues and I have been developing a self-completed questionnaire, called the PBPI, which I believe may help discriminate “ordinary grief” (of bereavement) from major depressive disorder. However, the PBPI has not been “field tested” for validity. I attach the scale and the associated article for your perusal; and, you are more than welcome to use the PBPI in your work and/or research, with no restrictions. I would be especially interested in your experience with the PBPI as a “predictor” of post-bereavement grief, versus a course more consistent with major depressive illness. I believe this could be done fairly easily by tracking post-bereavement hospitalization as a good “proxy” for severe psychiatric illness.
In any case, I appreciate your work in this area and will be eager to learn the results of your research.
MMcC:
Thank you very much for the article which I and the rest of the team read with interest. As mentioned previously, we won’t be able to deploy any scales in fieldwork for this administrative data linkage project, but we are keen to develop further research following on from this study. We will keep in contact about developments and future possibilities for using the PBPI.
Thanks again for your interest in the study, please keep checking in with the Blog (subscribe to the RSS feed would be an easy way to keep up to date) as we’ll be posting about progress and findings over the course of the study.
RP:
Dear Mark,
I do understand the limits of a records-based, statistical methodology in this study. I hope that you can gain access to hospitalization records, since I believe that would be a much better proxy for “severity” than medication prescription (for all the reasons we have discussed). I will try to “tune in” to your website from time to time; and, I do hope you find the PBPI of use at some time in your work, and I would be happy to hear any feedback on it. Many thanks for the informative note. You certainly have my permission to post my comments, either verbatim or edited, on your website.