Monday, August 27, 2018

Evidence-Based Prognostication

by Christian Sinclair (@ctsinclair)

We are prognosticating beings. It is how we survive. Many everyday decisions begin with an estimation of likely future outcomes. If my first clinic appointment is at 9:15am, and my drive from the hospital to the clinic usually takes 25 minutes, then I need to leave by 8:50am at the latest to give myself time to spare for bad traffic light timing, lack of a good parking spot or some other problem that may delay my arrival. I make my estimates, and go with the safest choice. I could go with my gut and my experience or I could use Waze, an app where I can select where I am leaving from, where I am going and what time I want to arrive by and it will tell me the best time to go based on a much bigger data set from drivers on the roads at that moment. Not only that, I could also use Waze to send an updated map to the clinic team to let them know when I will be there, so they can plan ahead as well.

But when it comes to medicine, we often deal with prognosis of many different outcomes, but we rarely use data and technology (evidence) and we rarely share this information clearly with others (documentation.) I recently finished the chapter on Prognostication for the new edition of the Oxford Textbook of Palliative Medicine, so prognosis has been on my mind a lot. One key article stuck with me and made me reflect on my practice of documenting evidence-based prognosis. Andrew Bruggeman and his colleagues at UC San Diego reviewed 412 inpatient palliative care notes for oncology patients and only 5 contained documentation of evidence-based prognostication. The validated tools used included the MELD score, the Walter Index, and the Lee Index. They looked for the Palliative Prognostic Score and the Palliative Prognostic Index as well but never found either of those documented.

First of all, I want to say kudos for publishing this research that basically says at their own institution a potentially critical part of a good palliative care note is missing. They are in good company thought because from my discussions with colleagues around the country I have not heard anyone else doing this on a consistent basis. At the organizations where I have and currently work, we do document prognosis but the note template and the EMR are not designed to easily include any particular evidence-based tool. This is consistent with other research of palliative care notes which showed 72% of notes contain survival estimates.

Secondly, we are entering a new age where machine learning and big data are crunching numbers and relationships with more power and speed then our simplistic point-based prognostic tools are capable of. But the output of these tools has not been widely tested, and how they construct their findings can be pretty enigmatic and not inherently logical to clinicians. These new black box prognostic machines may help us identify patients who may benefit most from palliative care, but I think we have not quite dealt with the ethical ramifications of what improved prognostic accuracy means for our clinical work and how it is delivered at scale.

We are barely starting to make evidence-based prognostication part of our clinical work and there are strong forces already at play using potentially more accurate tools that we don’t comprehend. This will be an important part of not only palliative care and hospice delivery, but health care delivery, health care reform, and even the overall insurance-based payment structure. We need clinician eyes, ears, and voices involved with the advances in prognostic science, and it begins with all of us being willing to use the tools we currently have and include them in our daily work.

If you are interested in this topic, I will be hosting a conversation about prognostication at the #hpm chat on Twitter Wednesday, August 29nd, 2018 at 9pET / 6p PT.

Christian Sinclair is the co-founder of #hpm chat and Editor-In-Chief of Pallimed. He leads the outpatient palliative care oncology clinics at the University of Kansas Health System. When he is not thinking about predicting the future, he can be found watching the Back to the Future trilogy looking for plot inconsistencies. You can find him on Twitter at @ctsinclair.

References

1: Bruggeman AR, Heavey SF, Ma JD, Revta C, Roeland EJ. Lack of documentation of evidence-based prognostication in cancer patients by inpatient palliative care consultants. J Palliat Med. 2015 Apr;18(4):382-5. doi: 10.1089/jpm.2014.0331. Epub 2015 Jan 21. PubMed PMID: 25608220.

2: Zibelman M, Xiang Q, Muchka S, Nickoloff S, Marks S. Assessing prognostic documentation and accuracy among palliative care clinicians. J Palliat Med. 2014 May;17(5):521-6. doi: 10.1089/jpm.2013.0454. Epub 2014 Apr 10. PubMed PMID:24720384.

Pallimed | Blogger Template adapted from Mash2 by Bloggermint