Uncategorized

The ePrognosis App: How Calculating Life Expectancy Can Influence Healthcare Decision-Making

Last month an intriguing new decision support app launched, created by experts in geriatrics and palliative care. It’s meant to help with an important primary care issue: cancer screening in older adults.

Have you ever asked yourself, when considering cancer screening for an older adult, whether the likely harms outweigh the likely benefits?

Maybe you have, maybe you haven’t. The sentence above, after all, is a bit of wonky formulation for the following underlying questions:

  • How long is this person likely to live, given age and health situation?
  • Given this person’s prognosis, does cancer screening make sense?

The first question seems like one that could easily occur to a person — whether that be a patient, a family member, or a clinician – although I suspect it doesn’t occur to people perhaps as often as it should.

As for the second question, I’m not sure how often it pops up in people’s minds, although it’s certainly very important to consider, given what we now know about the frequent harms of cancer screening in the elderly, and usually less frequent benefits.

Furthermore, there is abundant evidence that “inappropriate” cancer screening remains common. “Inappropriate” meaning the screening of people who are so unwell and/or old that they’re unlikely to live long enough to benefit from screening.

For instance, one astounding study found that 25% of physicians said they’d order colon cancer screening for an 80 year old with inoperable lung cancer. So it’s clear that improving the decision-making around cancer screening would help improve healthcare safety, quality, and value.


Enter the ePrognosis Cancer Screening app, which is the first app created by the UCSF-based ePrognosis team. It’s free in the iTunes store, and according to co-creator Eric Widera, a geriatrician and colleague of mine, it was designed to be used by the public as well as by clinicians.

The app basically works like this:

  1. You tell it whether you’re interested in screening for breast cancer, colon cancer, or both.
  2. The app asks you 15 questions about yourself, your health, and whether or not certain activities are difficult for you. (The questions are a combination of the Lee and Schonberg mortality indices; see here for the questions and relevant scholarly citations.)
  3. You are then presented with a statement on whether cancer screening is recommended, along with a nifty speedometer graphic illustrating the balance of harms vs. benefits.
  4. If you click “Learn more” you are shown a series of pictographs illustrating information on expected harms and benefits of screening, along with information on what proportion of similar people will die over the next 10 years whether or not they get screened for cancer.
  5. You can also click share and reportedly send the report to another person. (Note: I wasn’t able to get this to work for me.)

For more information on the app, including some good screenshots, you can read the GeriPal post introducing the app.

Using prognosis: From research to ePrognosis to…app?

So will the app work as its creators hope? Meaning, will it help patients and clinicians make better decisions about cancer screening in older adults? And will it reduce inappropriate cancer screening in older adults?

Prognosis is, after all, a thorny topic. Even if you can get decent prognostic information at the time you need it, going through discussions and decision-making based on this information is not easy. For instance, this qualitative study of older adults found that although most participants were open to discussing life expectancy, about a third perceived such discussions as not useful, or even harmful. (Interestingly, 64% of participants felt that their physicians could not correctly estimate their life expectancy.)

The ePrognosis project itself was meant to address the first part of the challenge: the fact that practical prognostic information is hard for front-line clinicians – and regular people — to find. Consider this: even today, UpToDate has no topic page on how to estimate life expectancy in older adults, despite many guidelines now urging clinicians to factor prognosis into medical decisions. (For a useful slideshow on why we should use prognosis, see here.)

Instead, UpToDate has a page titled “Communication of Prognosis in Palliative Care”, authored by UCSF’s Alex Smith, a co-creator of ePrognosis. This resource has terrific content on how clinicians can discuss prognosis, but because of its title, may not be easily spotted by doctors searching for a quick way to figure out life expectancy. (“How to talk to patients about how long they’re likely to live” might catch more notice among generalists.)

I myself like the ePrognosis website in concept. That said, I recently found myself sheepishly admitting to Alex that I’ve hardly ever used it. Why? Mainly because my patients tend to be quite elderly and frail, and I feel that I’m already incorporating a sense of their limited life expectancy into my recommendations and conversations with them.

And herein lies the rub: the clinicians who think of using ePrognosis, or of incorporating life expectancy into their decision-making, are probably not the ones who most urgently need to use something like ePrognosis. (Not that we’re perfect – I’m sure using more decision support would help me — but at least we’re not the type to order cancer screening on an 80 year old with inoperable lung cancer.)

In general, the ePrognosis story illustrates a common challenge in improving healthcare quality: the problems that experts see at the population level (excess cancer screening in frail elders) are experienced very differently by the clinicians and patients on the ground (clinicians and patients have historically had enthusiasm for screening).

But changing the behavior of individuals – which is what you need to improve a quality problem – is very hard, especially if people remain embedded in their usual environments. Although expert guidelines and advice do have an important role to play, it’s usually not nearly enough to counter the habits and attitudes of the people in the trenches.

Still, when it comes to better healthcare for older adults, creating the tools to calculate and communicate life expectancy is a good start. Once a tool exists, it becomes possible to see how people respond to it, refine it, improve it, and also think of how to make it more available to the right people at the right time. One could even imagine certain tools being embedded into the clinic processes that nudge clinicians and patients towards (or away from) certain actions.

Now, where do you think we should go from here, if we want to use prognosis to improve healthcare, and healthcare value, for older adults? The creators of ePrognosis are looking for feedback and suggestions as to next steps. So check out the ePrognosis website, try out the app, and share some thoughts in the comments.

I just plugged information on a current elderly patient into the ePrognosis Gagne index and it predicts a 30% chance of death in the next year. Hm… perhaps I’ll reread that UpToDate page on discussing prognosis, as this patient is due for an advance care planning conversation. That page describes a five step process for discussing prognosis.

Should that be in an app?

Leslie Kernisan, MD MPH, is a practicing geriatrician, cautious techno-optimist, and enthusiastic caregiver educator. She hopes to someday be surrounded by cool tools and innovations that will make great geriatric care totally doable for all, especially primary care providers and family caregivers. She is a regular THCB contributor, and blogs at Geritech.org and at drkernisan.net.

6 replies »

  1. Enjoyed reading this article and comments!

    Some responses:

    Perry -completely agree that this does not replace the physician patient relationship. That relationship is central. Our hope is that this app will stimulate helpful conversations between clinicians, patients, and caregivers. We also worry about doctors or nurse practioners using this app as a kind of end-run around having difficult conversations. We need to do more work on the science of communication about prognosis to address this problem.

    John – yes, lots of ideas! Major question now is what SHOULD the next step be? So many possible directions, and only limited time/resources. What are your suggestions? We’re having an ePrognosis: Next Steps conference in January to plan for the future.

    Unfortunately, we have not heard much from users of the cancer screening app – so I can’t say we know what the response is yet. The app has not even been available for 2 full months yet, so it’s early.

  2. Sounds like you are citing Pew’s recently released report on Americans’ views on end-of-life. Yes, a substantial minority favor prolonging life even in the face of incurable disease and suffering. Still, it would seem reasonable to give people tools to be better informed, since research has shown that a certain proportion of people change their preferences or behavior after becoming better informed as to what lies ahead, and what their options might be.

    As for whether to reveal prognostic information or other info to patient & family, Paula Span of the NYT New Old Age blog queried her readers and many of them thought this information should be shared with patients. And in general, the movement in medicine is towards giving patients more access to the information doctors use, including access to their clinical notes. Still, it would be nice to study the impact of sharing prognosis or other info with patients and families.

  3. Thanks for these comments.

    The app itself is not about whether or not to treat; it’s about whether or not cancer screening is likely to be beneficial vs harmful.

    The app does generate this information based on one of the ePrognosis life expectancy calculators. Those calculators themselves are an estimate (i.e. educated guess) of a risk: the risk (or likelihood) that a person will die within a certain time frame.

    So neither the app nor the ePrognosis website tell people when they’ll die. But quite possible that some people may interpret the information as such, and as you point out, perception is very important!

    On the flip side, a month ago I heard Katy Butler, author of Knocking on Heaven’s Door, give Grand Rounds at UCSF. She quite emphatically asserted that PCPs and others need to let patients know when they’re likely to be within their last few years of life. Easier said than done, of course!

  4. Eric or Alex – are there any plans to broaden the scope of this app?

    And overall – quite curious – what kind of response have you seen? I imagine you’re getting a lot of the “App playing God” questions.

    How do you respond to them?

  5. Nothing is going to replace the physician-patient relationship with regard to workup and treatment. A recent poll shows at least 30% of Americans favor prolonging life, even in the face of lingering pain, suffering and imminent death anyway.
    While it is educational for physicians and scientists to look at statistics for reasonable approaches to testing and treatment, you most certainly cannot reveal this to a patient or family, you will get a lawyer’s letter immediately.
    The other problem is going to be how to make these decisions with declining financial resources, because whether we like it or not, some type of rationing is coming. Our only hope is to continue to educate the public on what is appropriate and hope some common sense wins out.

  6. Very interesting read and I am wondering if advertising/legal has not taken too much control out of the hands of doctors in terms of practicing say “pragmatic” medicine.

    The JAMA enthusiasm article underscores the difficulty in lessening unnecessary tests and related costs. The following comment at the end of the article was very interesting….

    “In the case of unproven tests such as total-body CT, excessive enthusiasm makes it extremely easy for exaggerated marketing tactics to succeed. Consequently, some have suggested that the government replicate what it has done with pharmaceutical advertising and regulate the direct-to-consumer advertising of unproven tests, requiring that the advertisements are factually true and provide balanced information about benefit and harm.”

    The problem is advertising is too effective and regulating pharma ads has done nothing. There’s always some guy at the end of the ad talking faster than the old Fedex spokesman saying the drug may not work, your leg may fall off, you may die, etc.

    Doesn’t put a dent in sales. The same applies to advertising the latest and greatest scanner/test. The minute people are aware of it many will want it – even the elderly.

    Isn’t it the doctor’s function to recommend course of treatment rather than some Madison Avenue ad agency or am I wrong?

    As per the ePrognosis app – I doubt there are many people that would want to know that you decided not to treat them based on an iPhone app that actuarially decided that they will die in 2.175 years.

    No one wants to be treated as a statistic. As a doctor, I would imagine you would know based on experience what makes sense and what doesn’t in terms of testing/treatment. So perhaps it’s bedside manner rather than coming at it like an insurance bean counter that would soften the blow when not recommending a test.

    Legally one would be torn apart on using such an app. They’d do a big dog and pony show in court challenging the algorithms, a bug gave the wrong result, etc. and say some greedy VC in the Valley funded an app that says you should die.

    In truth, maybe the app is a great tool, but perception is reality and if the average person knew this unclear that they would be too happy about it.