Certain Death by Pat Conrad MD
Don’t ya hate when a patient or family member asks how long they have before the Big Departure? Yeah, me too. But not to worry: “A New Algorithm Identifies Candidates for Palliative Care by Predicting When Patients Will Die.”
Some smart boys and girls at Stanford think they have identified “a deep neural network that can look at a patient’s records and estimate the chance of mortality in the next three to 12 months.” Better still, it finally justifies all those expensive EHR’s you’ve all been paying for through the nose. “The system works by training on several years’ worth of electronic health records and then analyzing a patient’s own records. It generates a prediction about the patient’s mortality, as well as a report for doctors to review about how it came to its conclusion… As we have noted before, doctors are much more likely to trust and accept an automated system if they understand its reasoning.”
Now I can hear you skeptics already grumbling out there but have an open mind. Taking the guesswork out of medicine should be the greatest thing since the hypodermic needle. Grumpy family in the ICU waiting room? No problem! Look here folks, check the algorithm…see this little dippy spot on the graph means it was Granny’s time. Another 3-pack/day wheezer back with flash pulmonary edema? No intubation necessary, but have some morphine dude, the chart says it’s cool. Oh, that’s your lawyer with you? Doesn’t this cover me?
Yes, any such system could be massaged for selective customers. All emergency departments would love a printout giving them free rein to shower the narc seekers with candy, just to shut them up and get them out the door. An algorithm leading to such conclusions more frequently could certainly be generated from the “right” data, for the right price. Nursing homes could be fun again, without having to sweat all those chronic skin breakdowns and combative sundowners. Check the algorithm and tape it over the patient’s bed! Enough fentanyl patches could cure any of these situations.
And we all know those trouble patients and families that want “everything” done. But time and resources are tight, futile care is supposedly unethical, and we’ve got all this data that makes it all comfortably clear. You same skeptics probably worry that the government would use this to legitimize some sort of death pa-, I mean Palliative Interpreting Surveillance Systems to trim the end-of-life crowd who, face it, weren’t having any fun anyway. But there is no reason to suspect that a society so compassionate as to empower its government to provide health care for all the old folks would ever allow that government to start denying diagnosis and treatment options. Come on, have a little faith! Quit being so anti-technology! The right algorithm, a smartphone or watch that can monitor your health status, a really cool app to tie it all together, and *BOOM*, you are plugged in as a nurtured, cared for part of the greater community health. And if the little app on your phone lights up – like those palm crystal thingys in “Logan’s Run” – then head on over to your nearest med tech’s office for some free narcotics that even the latest administration’s busybodies would be hard pressed to criticize.
Pitfalls? Maybe. It is conceivable that Medicare might regard adherence to the PISS as a quality indicator, and dock docs if they fell below the average of patients reasonably expected to be palliative only. Obviously, your local hospital CEO will need to have you in for a counseling session if you start hogging up too many beds with patients who could benefit from some higher doses of oxycodone, but that’s only fair. A patient not receiving as much opioid love as he or his family might think necessary, if they fall within PISS algorithm parameters, might be a wonderful financial opportunity for a compassionate lawyer. And, of course, keeping up with the latest algorithms would require, nay, demand updated MOC requirements from an eager ABIM and AAFP. In fact, these PISS tests could usher in a new era for medical competency. If algorithms can predict the date of death well enough to use as a basis of care, then why couldn’t they improve physician competency? Take the target, uh, candidate physician’s last board scores, number of patients they see, their adherence to quality indicator reporting requirements via their EHR’s over the previous five years, and a really sharp algorithm could determine exactly how competent that physician really is. And the low performers could be encouraged to take more frequent board exams, just to reassure a deserving public.
Totally agree with the overall point, but have an overwhelming response to the idea that 3 to 5 years of EHR data would be reviewed. I can’t even access my patients weights from 6, 12 and 18 months ago, and the EHR managers can’t help. GIGO!
You wouldn’t even really need many, if any doctors in the PISS era. Just let the LELTs plug in the data and away they go!
Although we should never let an algorithm do our thinking (think those lovely and often inaccurate interpretations on EKGs) we are not paying adequate attention to the palliative needs of our patients with life limiting, terminal illnesses. There are already a host of predictive systems. Those of use who practice in the hospice setting use these all the time. Attention to palliative and end of life needs doesn’t mean a free trip to the opioid candy store. For any physician to suggest this is irresponsible, unprofessional and, frankly, unscientific. The dirty little secret not adequately disclosed in med school and residency is 100% of our patients will die. We now have unimaginable tools at our disposal to prolong that process and prolong suffering if patients are poorly selected for “life sustaining” measures. I would argue that the 3 PPD O2 dependent and multiply intubated COPDer would very much benefit from a palliative care consult to discuss goals of care and fears with the dying process from their INCURABLE TERMINAL illness. Take a tour through a long term acute care hospital or a ventilator nursing home and tell me that our society is over utilizing palliative care. Examine what end of life procedures look like in this country for no improvement in quality time with huge emotional expense to the patient and family and tell me a robust palliative care system is a death panel. Care for a “destination” left ventricular assist device patient with albumin of 1.8, drive line infection, repeated septic episodes (not one but a steady stream of them) and tell me we dont need a rational, compassionate, SCIENTIFICALLY based approach to deployment of these technologies and guidance for our patients as to when use of tech can no longer defy our inevitable biological end. To overdramatize use of opioids in these conversations and care plans is inflammatory and irresponsible.
I normally find your posts to be on point. This one, however is a delusional fail.
FYI, I am absolutely PRO-hospice and agree with rational palliative care. The point of this column is that such algorithms will be immediately misused and viewed by too many as an easy substitute for physician judgment in a collaborative approach with an informed, suffering patient. PC.