STRATEGIES FOR A TIGHT MARKET
orientation, and even their turnover risk,” Wilkins writes. One client reduced first-year turnover by
43 percent. Another found that the “ideal hires” identified through the system were 64 percent more likely to make it past the 90-day mark.
Using machine learning Another method uses machine learning to make predictions about employees. It’s a complicated topic, but a simple explana- tion serves: A computer is fed large amounts of foundational data, and continuously fed more data. It uses its own artificial intelli- gence to “learn” what the data means. It looks for patterns and elements that stand out from patterns, matches them up to pre- vious and current patterns in hiring or re- tention, and draws its own conclusions. The more data, the more it learns. Arena, a health care hiring company, uses
these methods. Founder and CEO Michael Rosenbaum’s previous lives include one as the John M. Olin Fellow in Law and Eco- nomics at Harvard and as an economist at the White House. But for the past 10 years, he’s been interested in health care hiring and machine learning.
pany became Arena, Rosenbaum had made a goal to reduce turnover at a hospital by 10 percent. At first, turnover decreased by 58 percent. As the machine learned more, turnover decreased by 77 percent. Arena also predicts who might do best
in specific positions. Beyond the basic skills, hiring managers will typically look for some- one who is compassionate, who cares. “But the reality is that someone who really thrives in memory care may not thrive in assisted living,” Rosenbaum says. Job applicants don’t always know where they’ll be happi- est, nor may those hiring, so the predictions “give each side a clearer sense of what’s likely to happen.”
Potential drawbacks The consideration raised about predictive methods is the concern about bias. In the early days of machine learning, for instance, systems reflected their designers—the typi- cally younger, white males of the tech inno- vation world. Their implicit biases and back- grounds could then throw off the prediction and perpetuate more bias, creating a cycle. Rosenbaum, of Arena, got into the se-
nior living hiring field to break that cycle; he wanted to create a hiring system that
As anyone in hiring and staff development knows, gathering and interpreting data about people is complex and ever-changing—because people themselves are.
The hiring suggestions that result from
machine learning can be counterintuitive or even puzzling. For instance, for a client that operated both a hospital and a long-term care service, Arena found that a CNA who was also a leader in a neighborhood organi- zation was slightly more likely to be retained in the hospital—and slightly more likely to leave, when working in long-term care. Or when a candidate was asked about
working for a faith-based organization, the answer itself didn’t matter as much as the number of words an applicant used in writ- ing the answer: more words, better job fit. The results can be uncannily accurate. In one of its first forays, even before the com-
could vastly reduce bias. In 2010, he tested it out on a hospital system after he heard the turnover rate in that industry could reach 40 percent. As he learned more about health care hiring, he started exploring the senior living area. “Today, we’re almost entirely in hospitals, skilled nursing, and CCRCs,” he says, not least because they like the people. Bias isn’t confined to machine learning:
Even an automated response system for a mobile phone could potentially be problem- atic, if a person blocks calls or if the permis- sion to text isn’t given. A company takes the chance of overlooking someone without a mobile or who is less tech-savvy but would still make a good hire.
10 SENIOR LIVING EXECUTIVE SEPTEMBER/OCTOBER 2019
Have the conversation State and federal regulators have been eye- ing predictive hiring techniques as well, with Illinois this year passing limits on algorithms interpreting facial expressions in video in- terviews, and U.S. House Democrats intro- ducing the Algorithmic Accountability Act of 2019 (H.R. 2231), which calls for “auto- mated decision system impact assessments.” “I think for the most part it’s good that
states and the federal government have this conversation,” Rosenbaum says. “We want as a society to have a conver-
sation about what we should and shouldn’t use. When people don’t understand how machine learning works, it can have a lot of unintended consequences.”
Change is a constant Whatever automated methods you use, they’ll need to be updated over time, as the industry and community needs change. For example, HealthcareSource in June
added additional competencies to its staff assessment after extensive research, includ- ing measurements for “emotional evenness.” The new competencies “provide addi-
tional insights into applicants and correlate strongly to job performance ratings and customer service behaviors,” a company statement says. “If there’s a new executive director, al-
most always our algorithms need to change,” says Rosenbaum of Arena’s systems.
Balancing the costs Woodka has found two more hurdles that can make providers reluctant to use more automation. One is senior living’s reputation for the
personal touch. But automated responses early in the process can be personalized enough that a candidate will feel acknowl- edged, and that time saved can be applied to in-person interviewing. The second is common to any product or service: price. While OnShift sells such ser- vices, Woodka gave advice that works across the board: Get solid data on turnover and on who is being lost during the application process, then work out the costs of these, so you can make an informed comparison and decide what you need.
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52
