"There was a sense of urgency since she had immediate thoughts of suicide and access to her parents’ firearms.”
Quanah Walker, Director of Behavioral Health Services at HealthPartners, recalls how our behavioral health case managers reached out to help a health plan member.
Algorithm tracks 313 characteristics to predict suicide risk
"Our case managers were alerted of this member via an algorithm that uses real-time updates to electronic health records to assess suicide risk. Developed by our HealthPartners Institute, the algorithm examines 313 demographic and clinical characteristics from records for up to 5 years to determine the potential for suicide.
The data indicators tracked include:
- sociodemographic characteristics (such as age, race, ethnicity and gender)
- past suicide attempts
- mental health, substance use and medical diagnoses
- psychiatric medications dispensed
- mental health outpatient, inpatient and emergency department care
- questionnaires about depression
"Because we were already working with this individual in our case management program, the algorithm was able to scan her record to determine if any warning signs were present,” explains Walker. “That prompted one of our case managers to call her.”
Using the Colombia-Suicide Severity Rating Scale (C-SSRS), the case manager asked the member a series of questions that revealed she did have current thoughts of suicide.
“We then took immediate action to create a crisis plan to help her remain safe,” describes Walker. "The plan included a lockbox for firearms that only someone else could access. We were also able to connect her with a therapist.”
According to Walker, the suicide algorithm enabled his team to quickly reach out and respond in the right way, saving her life, and it has also helped save the lives of many others.
Algorithm identified 500 health plan members at risk for suicide
In a year, from August 2019-2020, the algorithm alerted our case managers to 500 members at a rising risk for suicide.
“Using the C-SSRS tool in our outreach calls to these individuals, we identified those who had current thoughts of suicide, and we swiftly provided crisis support and interventions,” Walker says. “The suicide algorithm gives us a much better way to assess risk, and I’m not aware of other health plans that provide this capability in a real-time, data-driven way.”
Rebecca Rossom, MD, Senior Research Investigator at HealthPartners Institute, helped create the algorithm. "Our HealthPartners Institute collaborated with the Center for Health Research at Kaiser Permanente,” she says. “We studied 20 million visits by three million patients in seven different care systems from 2009-2015 to develop a machine learning algorithm. It uses electronic health records, claims data and patient-reported outcome questionnaires to estimate a patient’s risk of suicide over a 90-day period.”
Mental health diagnoses key to assessing suicide risk
Our researchers previously determined that two-thirds of people surviving suicide attempts had a mental health diagnosis and received treatment during the prior year. For those reasons, Rossom indicates, it made sense for HealthPartners’ behavioral health group to be the first to use the algorithm to identify at risk health plan members with existing behavioral health concerns. Our institute’s research also revealed that while self-reported depression questionnaires such as a PHQ-9 can reasonably predict suicide, they are based on a limited scoring system which is unable to consolidate risk in a sophisticated way across diagnoses, care visits, prescriptions and sociodemographic characteristics. That’s why our researchers developed the algorithm to track a comprehensive set of known suicide risk data indicators.
Pandemic’s constant crisis fatigue increases already elevated suicide rates
And with American workers coping with heightened levels of stress amid COVID-19, Walker feels it’s vital for employers to act now by asking if their health plan can quickly respond to help people in crisis. Although we have yet to find out how the pandemic will affect suicide rates, we know that Americans' elevated anxiety, stress and depression, can create mental health crises that may lead to suicide. Also, suicide rates have already increased in the U.S. by 30% in six short years - from 2010-2016.
Suicide prevention is critical in the workplace
In addition, nearly 80% of all people who die by suicide are of working age (18-65). These factors make the workplace a crucial environment for crisis prevention and intervention.
“It’s so important that we help people if they’re feeling this way because they may not be comfortable talking about it with anyone else,” adds Walker. “They could be worried about being judged. That’s why campaigns like HealthPartners’ Make It OK are helpful for employers trying to change workplace culture. Make it OK helps break down the stigma, getting the message out there that mental health and behavioral health conditions are treatable and should be addressed just like medical diagnoses.”
Make It OK also helps employees focus on other elements of health and well-being, such as exercise, stress management and effective medicines that can make them feel and live better.
“By utilizing Make It OK’s materials and resources, employers can normalize the conversation around mental health,” explains Rossom. "Mental health conditions are so common, and yet talking about them remains taboo. Employers should encourage their employees to seek care when needed, and if they don't feel comfortable going to see their doctor in- person, letting them know that virtual care is available can be life-saving."
Aligning health and data expertise to save more lives
Further developing the algorithm’s life-saving abilities is also a key goal of our institute’s researchers, according to Rossom.
“We are continuing to explore ways to expand the algorithm’s data tracking beyond medical records,” she says. “Specifically, we are looking at how to incorporate social determinants of health such as job loss, homelessness, death of a spouse or divorce into the algorithm, which can be highly predictive of suicide risk.”
For Walker’s behavioral health case managers, the suicide algorithm currently integrates well with other data-driven informatics that identify individuals in need of additional support. For example, another algorithm the team uses helps prevent hospital re-admissions for individuals with behavioral health diagnoses.
“Our team has a lot of experience providing support and health care advocacy to individuals with behavioral health conditions,” says Walker. “Our case managers are a resource of invaluable expertise to our health plan members. Their extensive knowledge of motivational interviewing techniques help members determine what their goals are and how to achieve them. Their experience combined with our algorithms’ real-time alerts, prove that even one simple intervention can make a crucial difference in preventing suicide.”