Epidemiologic Analysis of Opioid Use Disorder in Hospitalized Adults: A Multilevel Modeling Approach Using North Carolina State Inpatient Data, 2000-2020.

Objectives

The opioid epidemic has increasingly affected the U.S. population, posing significant public health challenges. We aimed to examine factors associated with opioid use disorder (OUD) among hospitalized adults in the United States using a multilevel modeling approach and to identify distinct patterns of co-occurring clinical conditions using latent class analysis (LCA).

Methods

We analyzed a random sample of 100,000 hospital discharge records from the 2000-2020 North Carolina State Inpatient Database, including equal numbers with and without OUD. Adult patients (18 years and older) with complete sociodemographic and clinical data were included. Diagnoses were identified using International Classification of Diseases, Ninth Revision, Clinical Modification and International Classification of Diseases, Tenth Revision, Clinical Modification codes, grouped via Clinical Classifications Software. Mixed-effects logistic regression models assessed associations with OUD, incorporating random effects for area-level income. LCA classified hospitalizations into two distinct clinical profiles based on cooccurring conditions related to OUD.

Results

Patients diagnosed as having OUD were younger, more frequently male, and predominantly White compared with those without OUD. Mental health diagnoses ranged from 2% for attention-deficit/hyperactivity disorder/adjustment/eating disorders to 22% for depression/bipolar disorder, whereas comorbidities such as hypertension was present in nearly half of hospitalizations. LCA identified two distinct clinical profiles: class 1, characterized by less obesity and psychiatric comorbidities (63.3% of records), and class 2, with higher obesity and psychiatric morbidities (36.7%). Mixed-effects logistic regression showed that male sex, White race, emergency department admission, and insurance type (Medicaid/self-pay) were independently linked to OUD, whereas hepatitis B or C was the strongest (odds ratio [OR] 5.62) and tobacco use was the weakest (OR 1.28) clinical correlate. In a mixed-effects logistic regression model, patients in class 2 versus class 1 had a nearly threefold higher odds of OUD (OR 2.83), after adjustment for sociodemographic factors.

Conclusions

This study highlights important sociodemographic and clinical factors associated with OUD, including higher odds among male, White, and publicly insured patients. LCA identified distinct clinical profiles with varying psychiatric and obesity-related comorbidities linked to OUD risk. Key comorbidities such as hepatitis B/hepatitis C strongly predicted OUD, underscoring the need for integrated care. These findings inform targeted interventions for hospitalized populations facing complex medical and psychosocial challenges.