Drug-induced, dementia-associated and non-dementia, non-drug delirium hospitalizations in the United States, 19982005: An analysis of the national inpatient sample

Background: The incidence and pattern of delirium recorded in a broad spectrum of American hospitalizations has not been well described. The National Inpatient Sample (NIS) of the Healthcare Cost and Utilization Project is an administrative database of hospitalizations in the US that affords an opportunity to examine for International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9) codes relating to delirium. Objective: To examine the prevalence of delirium diagnoses and associated clinical factors, including adverse drug effects, in a broad spectrum of hospitalizations in the US. Delirium was grouped into three categories: drug-induced delirium, dementia-associated delirium, and non-dementia, non-drug (NDND). Methods: Hospitalizations during the years 19982005 in the NIS databases were examined. These databases represent samples of hospitalizations that allow for national prevalence estimates. ICD-9 codes for drug-induced, dementiaassociated and NDND delirium were identified in the hospitalizations for each year. Delirium tremens was not considered in this classification, and paediatric and psychiatric admissions were excluded. Yearly prevalence for drug-induced, dementia-associated and NDND delirium were tabulated, and time trends were analysed with negative binomial regression. A hospitalization subset cohort with urinary tract/kidney infection, pneumonia, heart failure and lower extremity orthopaedic surgery diagnosis-related group categories was also analysed for clinical associations with the presence of the three categories of delirium using multinomial logistic regression. ICD-9 E codes (external causes of injury) constituting adverse drug effects were identified and considered as clinical predictors. Results: Delirium was recorded in 1 269 185 (0.54%) non-psychiatric adult hospitalizations during the study years. Whereas the overall prevalence of dementia-associated delirium and NDND delirium decreased over time, drug-induced delirium prevalence increased (p < 0.0001). As expected, the presence of dementia and adverse drug effects had the strongest associations with dementia-associated and drug-induced delirium, respectively, in the cohort hospitalizations. Drug-induced delirium and NDND delirium had the strongest associations with lower extremity orthopaedic surgery hospitalizations and urinary tract/kidney infection hospitalizations, respectively. Among the NDND co-morbid conditions, volume depletion and sodium imbalance had the strongest, albeit modest, associations with delirium. The association between decade of age and delirium was strongest for NDND delirium (adjusted odds ratio 1.53; 95% CI 1.52, 1.53), but age had significant associations with drug-induced and dementia-associated delirium as well. In the cohort, the most frequent adverse effects codes were for opioids and for benzodiazepines or other sedatives, which were noted in 21.3% and 15.2% of drug-induced delirium hospitalizations, respectively. Conclusions: Drug-induced delirium is being increasingly identified in hospitalized patients. Administrative hospitalization databases constitute a resource to explore factors and trends associated with delirium. The findings suggest that interventions focusing on adverse drug effects have the greatest potential for preventing delirium.