Key Takeaways
Extended Takeaways
- In this referred inpatient sample, 730 patients were true positives and 160 were true negatives, with only 11 false positives and 26 false negatives, suggesting the algorithm may help standardize consult decisions while keeping missed intervention opportunities low.
- The NHHSRA’s positive predictive value was 98.5%, so when the tool flags an inpatient, addiction evaluation is very likely to confirm that an intervention is appropriate; the lower negative predictive value of 86% means a negative screen should not replace clinical judgment when suspicion remains high.
- Patients with borderline personality disorder had lower adjusted odds of receiving a clinician-recommended SUD intervention than patients with depression (aOR=0.36; CI, 0.13–0.99; P=.045), yet the discussion notes this difference was not seen for NHHSRA screening, implying an objective algorithm may reduce diagnosis-linked variation in access to addiction services.
- The referral population was relatively young and heavily weighted toward high-acuity substance use patterns: mean age was 36.48 (SD = 11.0) years, 66.9% were younger than 40 years, and the most common primary SUDs were opioids (37.9%) and alcohol (23.9%), which can inform staffing and medication readiness on psychiatric units.
- Because the algorithm uses objective admission data rather than patient self-report, it is particularly relevant in involuntarily hospitalized patients with SMI who may be acutely decompensated or unable to provide reliable substance histories at the time of psychiatric admission.
