We’re Not Getting Any Younger! Empowering Better Senior Care through Machine Learning

We're getting old at an increasing rate, and today's options for care are often expensive, inefficient and depressing. Can machine learning help?

We are getting older. Fast.

The population of American adults 85+ is projected to triple between 2015–2050[1], and more than 70% of people 65+ will require long-term care.[2] Most of the available care options are expensive, inefficient and depressing. The space is ripe for innovation.

CarePredict, a “people company” with the goal of improving quality of life “for our aging parents, grandparents and loved ones,”[3] is working towards change by pairing deep-machine learning (ML) and data analytics with a wearable device.

The oldest age group is project to dramatically increase while the younger population (under 65) is projected to only slightly increase.

What does CarePredict do?

CarePredict developed TempoTM, a wrist-worn wearable that detects a person’s location and activities of daily living (i.e. eating, drinking, bathing, walking, sitting and sleeping). They take the data and use analytics to detect behavioral pattern changes. When a meaningful change occurs, family and caregivers are alerted by text and email. They can also track behavior in real-time on web-based dashboards.[4]

Web-based analytics dashboard [3]
With more and more data collected, CarePredict applies ML to understand what different behavioral changes might mean. It combines data with the latest industry trends to understand the impact of patterns. For example, frequent bathroom visits at night may signal early signs of a urinary tract infection. This annoyance for young people could lead to sepsis or even death in the elderly if not contained.[5]

While CarePredict may note what a pattern may mean, the ultimate diagnosis is left to the caregiver.

Alerts to caregiver’s app upon behavioral changes [3]
Why does it matter?

Nursing homes are understaffed, low-tech, paper-filled environments. Vital signs are checked on a regular basis, but everything else – including patient feedback, behavior, and complaints – remain undocumented. “They’re not doing the kind of analysis that needs to be done, so they’re making the same errors again and again,” says Patricia McGinnis of California Advocates for Nursing Home Reform.[6]

Seniors who live on their own do not track their health or behavioral statistics. When a fall or fever occurs, they take costly (and often unnecessary) trips to the emergency room. Countless hours and dollars are wasted because of a lack of information.

CarePredict’s easy to use technology product closes the information gap. It collects information around the clock, whether or not a caregiver is present, portrays weeks of data in graphs, and most importantly, uses ML to understand which behaviors are unique and normal for each individual person.

Charles Turner, president of LifeWell Senior Living said, “If you can measure it, you can improve it.”[7] With the help of CarePredict, caregivers can act and improve situations. They can work to prevent illness, depression and falls by watching for when seniors move less, eat less, or sleep more.

This can lead to immense financial savings in trips to the ER and in trained staff being brought into care facilities unnecessarily. It can also reduce the stress on seniors’ families and friends. In essence, more information and understanding will lead to healthier seniors, cost savings, and peace of mind.

So, what’s next?

CarePredict has raised $10.2M since 2014.[8] In the short term, they will use the money to bring in more electrical and systems engineers to further build out the ML and artificial intelligence (AI) platform and its reporting applications. Additionally, they are creating formal documentation and guidance for customers. This will help the marketing team when trying to partner with care facilities.[9]

In the medium term, CarePredict plans to work with developers of senior living communities. They know from the data which rooms in care facilities have the heaviest usage and at what times – valuable information for developers and architects building efficient and intuitive building flows.[10]

I recommend that in the short term CarePredict work with caregivers to complete the information loop, and then use ML/AI to enhance the app and analytics systems. What did the caregiver do when she noticed Charlotte sleeping more? What happened after Max’s fall?

When there is a change in behavior, the app should alert caregivers and prompt them with a question asking what action was taken, and what the results were. CarePredict should use these responses to further enhance the data set and prediction algorithm. Over time, CarePredict may even detect patterns that are unfounded in current research.

In the long term, I recommend that CarePredict be conscious of differences between patients. Age, gender and pre-existing conditions may lead to varying recommendation outputs. With more patients and more data, these trends should become clear.

Furthermore, CarePredict should focus on usability in future development. Having feedback sessions and focus groups will help mature the product. Plus, the more goodwill CarePredict has in one care facility, the more likely another facility will be willing to join a trial.


Senior care desperately needs more care, attention, and innovation.

  1. What are other applications for ML/AI/data analytics in senior care?
  2. Should the data captured be regulated?
  3. Is the industry ready for technological advances?

(797 words)



1) https://www.youtube.com/watch?v=8AglUMCKyns

2) https://www.youtube.com/watch?v=_bZjKC0EaY0


[1] Flinn, B. and Houser, A. (2017). Capped Financing for Medicaid Does Not Account for the Growing Aging Population. [online] Aarp.org. Available at: https://www.aarp.org/content/dam/aarp/ppi/2017/01/Capped-financing-for-Medicaid-Does-Not-Account-For-The-Growing-Aging-Population.pdf [Accessed 12 Nov. 2018].

[2] Longtermcare.acl.gov. (2017). How Much Care Will You Need? – Long-Term Care Information. [online] Available at: https://longtermcare.acl.gov/the-basics/how-much-care-will-you-need.html [Accessed 12 Nov. 2018].

[3] CarePredict. (2018). CarePredict™ Elderly Monitoring Systems to Improve Senior Care & Living. [online] Available at: https://www.carepredict.com/about-us/ [Accessed 12 Nov. 2018].

[4] CarePredict. (2018). Improving Senior Care | The CarePredict Difference. [online] Available at: https://www.carepredict.com/why-carepredict/ [Accessed 13 Nov. 2018].

[5] Bhattacharya, A. (2016). Researchers have developed sensors that can detect which senior citizens are at risk of falling. [online] Quartz. Available at: https://qz.com/769867/sensors-to-keep-senior-citizens-from-falling/ [Accessed 13 Nov. 2018].

[6] Schwartz, A. (2014). How Can We Reduce Adverse Events in Long-Term Care Settings?. [online] Scienceofcaring.ucsf.edu. Available at: https://scienceofcaring.ucsf.edu/research/how-can-we-reduce-adverse-events-long-term-care-settings [Accessed 13 Nov. 2018].

[7] CarePredict (2017). CarePredict Benefits for Developers of Assisted Living Communities. Available at: https://www.youtube.com/watch?v=9FEE9UJWpbw [Accessed 13 Nov. 2018].

[8] crunchbase. (2018). Company Overview. [online] Available at: https://www.crunchbase.com/organization/carepredict#section-overview [Accessed 13 Nov. 2018].

[9] CarePredict. (2018). CarePredict Careers. [online] Available at: https://www.carepredict.com/careers/ [Accessed 13 Nov. 2018].

[10] CarePredict. (2018). Actionable Insights for Independent Senior Living | CarePredict Benefits. [online] Available at: https://www.carepredict.com/independent-living/ [Accessed 13 Nov. 2018].

[11] Thumbnail photo: First Plymouth (2018). Smiling seniors. [image] Available at: http://www.firstplymouth.org/firstplymouth-events-page/2018/10/3/state-of-senior-care-in-our-community [Accessed 13 Nov. 2018].



San José Tackles Open Innovation for Smart Cities


Steam-powered ideas: a market for open innovation

Student comments on We’re Not Getting Any Younger! Empowering Better Senior Care through Machine Learning

  1. What an exciting product! I think the industry is more than ready for technological advances, for all the reasons you laid out above. My question would be whether this will have to fall under privacy regulation? That would be a big ask for a small start up.

    1. Absolutely. Nursing homes are the most regulated industry, second only to nuclear energy (!). DataPredict will definitely need to work with the facilities to follow whatever privacy laws are needed. I would hope/assume they are using some of the funding to build a secure system that has the ability to annonymize data.

      For a little more info on some steps care facilities need to take to meet regulations: http://mcgowanprograms.com/blog/ensuring-senior-care-facilities-maintain-proper-hipaa-and-data-breach-compliance/

  2. Interesting article! I recall an article from a while ago that mentioned that, specifically in Japan, a primary issue in the care-giving industry is that there are simply too few caregivers, i.e. not enough employees. If this is at all a parallel issue in the U.S., then my concern with this product is the potential for false alarms. If the caregivers are overworked, then the product would quickly lose credibility if it constantly pulled the employees away from their work to run unnecessary checks. I think a big challenge here will be tuning the algorithms to balance frequency of alert with potential severity; really hoping they can pull it off!

    1. Thanks! Yes the staffing problem is all to real in many countries. You bring up a great point about false alarms. They’re going through alarm fatigue now in many hospitals (https://whyy.org/segments/beep-beep-beep-hospital-alarms-sound-mostly-without-real-cause/) – so many beeps all the time, it’s hard to tell which ones are for real emergencies.

      I believe that with more data and more tech and ML, eventually we can train systems to set constraints and send different types of alerts (i.e. passive, active) within those constraints.

  3. Very interesting/high potential product in a very attractive area. As you noted, technological adoption in senior care is considerably low in both the “front-office” and the “back office”. The back office has made significant progress in recent years with the adoption of Electronic Health Record and Practice Management software from companies like Kinnser Software, but the provision of care or the “front office” continues to lag and i wonder if this is due to the high sensitivity towards removing the human from the decision-making. These are highly fragile situations and the cost of a mistake (as you note above as a risk) can be massive. Would hypothesize that this product works for lower acuity patients that can be tracked with less intervention. What would be the right go to market strategy here – would it be through the senior living care communities or the health plans (particularly Medicare Advantage)? Would imagine the latter is more sensitive to these issues and has more capital to spend. Additionally, from a machine learning perspective you can pair this data with other internal data from the insurance plans to enrich the algorithms and the product further.

  4. Interesting read. I would be concerned with data protection and privacy, especially how this data is transmitted and stored in the context of the caregiver and their association. Also, I would interested to see this as a progression from something like smartwatch functionality and health stats to gradually move into this space.

    Ideally there would be a way to incorporate this tech and the ML into the current series of wearables for younger generations, which would hopefully slow down the need for this type of palliative care.

  5. Super relevant read, especially given US population demographics. Seems like this product would be highly relevant to insurers, who are increasingly focused on tracking and/or behavioral health solutions for their elderly members. Humana in particular comes to mind – they have a Medicare-heavy membership base and recently spent ~$4bn to acquire Kindred’s hospice operations.

  6. “They’re not doing the kind of analysis that needs to be done, so they’re making the same errors again and again,” says Patricia McGinnis of California Advocates for Nursing Home Reform. I found this quote to be powerful. As someone who worked in nursing homes in a different life, the feedback loop is incomplete in nursing care. These are due to overworked and underpaid staff, poor communication between the caregivers and patients, and the lack of focusing on quality over quantity. I think any way you can provide electronic sources to combat against problems associated with quality of life for our aging population will be advantageous over the long run. The key is working with regulators early to prevent violations with HIPPA compliance, while still moving fast into the market space. Further, I would work with Medicare to subsidize this product to make it a viable solution for all types of nursing home facilities.

  7. I like the idea of these tools being used to augment caregivers abilities to diagnose changes In their patients. It’s clear that we are not doing enough to use data and analytics to monitor our elderly patients. It is clear for my own experiences with my grandparents in nursing homes, That data and attentiveness are two things that the industry could sorely use. I am however a bit leery of generating all of this data, and the questions that ensue about who owns, retains, and gets to use this data. I think we would all feel uncomfortable if we knew that our data was not deindividualized, Or that this information was used to violate privacy. I agree with the comment above regarding HIPAA compliance. I would draw a parallel with the use of genetic data in medical care- (https://www.genome.gov/27026050/president-bush-signs-the-genetic-information-nondiscrimination-act-of-2008/), and the GINA act of 2006, which prevents discrimination on the basis of one’s genetic data. I think a similar standard should apply to nursing home and other sensitive health data.

  8. Super fascinating read. On your 2nd question, I am concerned with the use of data here and the rights the CarePredict has to use this data. Over the past few years, data privacy and regulation has become increasingly more regulated, particularly in the EU with GDPR (General Data Protection Regulation), so I would be concerned that in the long run CarePredict needs to be extra careful with their privacy policy and that customers know what they are opting in to.

    On your last question, I think the industry is ready for these technological advances in the next 5-10 years as our aging population becomes more comfortable and adept with using technology. On a personal note, I think of my grandma who right now is 98 and would (a) have no desire to wear a device like this and (b) would be unable to likely use it if it had to sync with a cell phone (she still doesn’t have one)! All that being said, I do think this adoption is more likely than other technology for the elderly population because caretakers will want to push this on to their loved ones more than they otherwise might like.

Leave a comment