1. 2021;24;E191-E201Laboratory-Generated Urine Toxicology Interpretations: A Mixed Methods Study
   Cohort Study
   Endel J. Orav, PhD, Gordon D. Schiff, MD, Athena K. Petrides, PhD, Christiana A. Demetriou, PhD, Zain M. Virk, BS, Olga Ehrlich, PhD, RN, Ethan Katznelson, BA, Jaime R. Ransohoff, BA, Isaac S. Chua, MD, and Stacy E.F. Melanson, MD, PhD.

BACKGROUND: Clinicians frequently order urine drug testing (UDT) for patients on chronic opioid therapy (COT), yet often have difficulty interpreting test results accurately.

OBJECTIVES: To evaluate the implementation and effectiveness of a laboratory-generated urine toxicology interpretation service for clinicians prescribing COT.

STUDY DESIGN: Type II hybrid–convergent mixed methods design (implementation) and pre–post prospective cohort study with matched controls (effectiveness).

SETTING: Four ambulatory sites (2 primary care, 1 pain management, 1 palliative care) within 2 US academic medical institutions.

METHODS: Interpretative reports were generated by the clinical chemistry laboratory and were provided to UDT ordering providers via inbox message in the electronic health record (EHR). The Partners Institutional Review Board approved this study.

Participants were primary care, pain management, and palliative care clinicians who ordered liquid chromatography-mass spectrometry UDT for COT patients in clinic. Intervention was a laboratory-generated interpretation service that provided an individualized interpretive report of UDT results based on the patient’s prescribed medications and toxicology metabolites for clinicians who received the intervention (n = 8) versus matched controls (n = 18).

Implementation results included focus group and survey feedback on the interpretation service’s usability and its impact on workflow, clinical decision making, clinician-patient relationships, and interdisciplinary teamwork. Effectiveness outcomes included UDT interpretation concordance between the clinician and laboratory, documentation frequency of UDT results interpretation and communication of results to patients, and clinician prescribing behavior at follow-up.

RESULTS: Among the 8 intervention clinicians (median age 58 [IQR 16.5] years; 2 women [25%]) on a Likert scale from 1 (“strongly disagree”) to 5 (“strongly agree”), 7 clinicians reported at 6 months postintervention that the interpretation service was easy to use (mean 5 [standard deviation {SD}, 0]); improved results comprehension (mean 5 [SD, 0]); and helped them interpret results more accurately (mean 5 [SD, 0]), quickly (mean 4.67 [SD, 0.52]), and confidently (mean 4.83 [SD, 0.41]). Although there were no statistically significant differences in outcomes between cohorts, clinician-laboratory interpretation concordance trended toward improvement (intervention 22/32 [68.8%] to 29/33 [87.9%] vs. control 21/25 [84%] to 23/30 [76.7%], P = 0.07) among cases with documented interpretations.

LIMITATIONS: This study has a low sample size and was conducted at 2 large academic medical institutions and may not be generalizable to community settings.

CONCLUSIONS: Interpretations were well received by clinicians but did not significantly improve laboratory-clinician interpretation concordance, interpretation documentation frequency, or opioid-prescribing behavior.

KEY WORDS: Compliance monitoring, chronic pain, urine drug testing, opioid, liquid chromatography-tandem mass spectrometry, palliative care, primary care, substance use disorder, diagnostic error, clinical decision support