Abstract Introduction: The positive predictive value of a low morning serum cortisol to diagnose adrenal insufficiency (AI) is reported to be >90%, which is the basis for guidelines recommending morning cortisol as the first-line test for central AI. A level <3 µg/dL is considered diagnostic. However, studies that established this cut-off were conducted primarily in outpatients, not hospitalized patients in whom diurnal variation may be disrupted. Studies suggest that opioids suppress cortisol levels acutely, may disrupt diurnal cortisol rhythm, and cause adrenal insufficiency in 8–50% of chronically-exposed patients. The impact of hospitalization, particularly in the setting of opioid use, on the accuracy of diagnostics for AI is unknown. We hypothesized that low morning cortisol values in hospitalized patients, especially those prescribed opioids, do not accurately diagnose adrenal insufficiency, as determined by corticotropin stimulation test (CST) peak cortisol <18 µg/dL. Methods: We performed a retrospective analysis of CSTs in hospitalized patients in the Mass General Brigham health system from 5/2015 to 9/2020. Opioid-exposed adults who underwent CST were included if they had a morning cortisol (5–9 AM) result. Control patients were matched by age, gender and race and had no opioid prescriptions within 30 days of testing. Patients were excluded if tested in the outpatient or ICU setting, had a history of cirrhosis or pituitary disease, had an elevated ACTH, were prescribed oral estrogen, or received oral, IV or intraarticular glucocorticoids within 30 days. Results: The analysis included 124 opioid-exposed and 322 control patients, mean (±SD) age 60.8±14.4 and 63.8±15.3y, and 55.6% and 45.0% female, respectively. Twenty-two (17.7%) opioid-exposed and 33 (10.2%) control patients were diagnosed with AI by CST (p=0.031). Nineteen opioid-exposed (15.3%) and 22 control (6.8%) patients had morning cortisol of <3 µg/dL (p=0.005). A morning cortisol <3 µg/dL had a positive predictive value of 36.8% (CI 19.1–59.0%) for AI in opioid-exposed and 50.0% (CI 30.7–69.3%) in control patients. In opioid-exposed patients with low morning cortisol levels, mean daily morphine milligram equivalent and duration of opioid exposure did not differ between those with AI confirmed by CST and those with normal CST (p=0.13 and 0.38, respectively). Conclusion: Among hospitalized patients with suspected AI, those prescribed opioids have a higher prevalence of central AI than controls. Serum morning cortisol <3 µg/dL is not an accurate test for AI in hospitalized patients, including those prescribed opioids in whom low morning cortisol is more prevalent than in controls. Given the risks associated with misdiagnosis of AI, caution should be exercised in relying on morning cortisol values for the diagnosis of AI in hospitalized patients.