Can persistent outreach reduce the risk of suicide attempt in people who report frequent thoughts of death or self-harm? Our MHRN Suicide Prevention Outreach Trial is testing the effectiveness of persistent outreach programs, using decision support software built into the electronic health record.

In one of those outreach programs, care managers use data from electronic health records to watch over thousands of at-risk patients and identify those who appear to have disengaged from outpatient care. A suicide prevention registry harvests data from patients’ records regarding key indicators: the date and result of the most recent suicide risk assessment, the most recent care manager outreach contact, and the next scheduled follow-up visit. Those data then pass through an algorithm to generate patient-specific outreach and follow-up recommendations. For example: “If ((last suicide risk score =3) and (days since last risk score < 28) and (date of next mental health appointment > 14 days)) then send outreach message to arrange follow up visit with 14 days.” Over 60 other scenarios cover every possible permutation of those key indicators.

The decision support software is a necessary tool for supporting thousands of patients, but algorithms certainly don’t replace human caring. The software can only generate guideline-based recommendations. Care managers’ actual outreach contacts (via online messaging and telephone) are completely personalized. Each outreach message is shaped by everything the care manager knows about an individual patient’s history, current situation, and preferences. We have no illusions that a chat bot or some other software tool will replace the human part of this work. Instead, we count on software tools to do what they do best – to consistently and persistently remind us when follow-up care has fallen off the recommended path.

In one of our weekly meetings, our lead care manager Deborah King was describing the challenge of following a decision support program that consistently and persistently reminds us about guideline-based follow-up. Care managers are rarely prompted to reach out to patients who are doing well - those who recover or stay engaged in care. Instead, care managers are constantly directed back to situations where caring can be stressful or difficult. The care manager’s daily “to do” list focuses attention on those at highest risk, typically those who are most hopeless and alienated.

That discussion reminded me of the classic children’s book by Margaret Wise Brown and Clement Hurd, The Runaway Bunny. Over and over, the little bunny declares his determination to run away. Each time, his mother’s responses are patient, compassionate, flexible, and incredibly persistent: “If you become a bird and fly away, then I will be a tree that you come home to.”

While algorithms alone can’t produce caring, neither is compassionate caring alone sufficient to restore hope and connection with the most alienated. Left to our own devices, we tend to direct our caring to the places where we are most comfortable or where our outreach is the most gratefully received. Algorithms and reminders are important, because they consistently and persistently push us back toward the places where caring is most necessary - even if those places are the least comfortable.

When I re-read my family’s well-used copy of The Runaway Bunny, I noticed something I hadn’t seen when reading it to my children. The mother’s patient responses are actually written as a series of “If-Then” statements. They read almost like the conditional expressions in our decision support software. Perhaps Margaret Wise Brown was really coding the first care management algorithm: “If you cancel all of your appointments with your therapist, then I’ll keep trying to reach you - to see if you are safe.”