Discussion

Among a large number of nursing home residents, no resident had the required documentation to meet the revised McGeer confirmed ACMS criterion, as defined using the CAM. This finding was surprising given that other long-term care and post-acute care settings have reported estimates of CAM-defined delirium ranging from 16% to 60%.^{Reference Inouye1,Reference Kiely, Bergmann, Murphy, Jones, Orav and Marcantonio2,Reference Voyer, Richard and McCusker6} Documentation of the individual core CAM elements was also lower than expected when compared with a study of CAM documentation in Canadian nursing homes, where the frequency of delirium symptoms ranged from 1.9% for inattention to 13.0% for altered level of consciousness.^{Reference Voyer, McCusker and Cole8} Notably, inattention was also the least frequently documented element (0.1%) in our survey. Because inattention is a required element in the CAM-based delirium definition, its absence had a large effect on the prevalence of residents meeting the confirmed ACMS criterion.

Although it is unclear why so few of the CAM elements were documented in our survey, explanations proposed in the literature include high resident-to-nurse staffing ratios, a preference of nursing staff to provide verbal reports of resident status, and time constraints for completing documentation.^{Reference Voyer, Richard and McCusker6–Reference Voyer, McCusker and Cole8} One study comparing delirium assessment between trained researchers and untrained clinical staff showed that due to the complexity of bedside assessments and the presence of resident factors such as dementia, specialized training is required for nursing home staff to adequately assess delirium.^{Reference Voyer, Richard and McCusker6} Additional explanations may include failure of clinical staff to perform formal CAM assessments or failure of staff to appreciate the severity of delirium symptoms.^{Reference Inouye, Foreman, Mion, Katz and Cooney7,Reference Voyer, McCusker and Cole8}

When creating the possible ACMS criteria, we evaluated all combinations of the CAM elements and ultimately selected the most sensitive—any 1 of 4 CAM elements. Our proposal for use of a possible ACMS criterion in infection surveillance is supported by the findings of Inouye et al,^{Reference Inouye, Leo-Summers, Zhang, Bogardus, Leslie and Agostini10} who validated a chart-based delirium detection method among older persons in acute care settings. In their study, chart documentation of any key descriptive CAM term yielded good sensitivity (74%) and specificity (83%) compared to direct delirium assessment using the formal CAM. These authors recommended use of this more sensitive, chart-based method for patient safety or quality improvement initiatives but not for clinical assessment or establishing a diagnosis of delirium.^{Reference Inouye, Leo-Summers, Zhang, Bogardus, Leslie and Agostini10} In a separate study, chart-based detection was also recommended for use in research settings to augment formal assessments and to identify additional persons with possible delirium.^{Reference Saczynski, Kosar and Xu11}

Using a possible ACMS criterion, the prevalence of ACMS in our survey increased from 0 to 1.9%. This change resulted in the identification of 21 additional HAIs (5.6% increase in HAIs), and most were respiratory infections (n = 17, 81.0%). Because other diagnostic or localizing criteria must also be present to meet the revised McGeer infection surveillance definitions,^{Reference Stone, Ashraf and Calder5} misclassifications of residents with HAIs were not anticipated. We also found that residents meeting the possible ACMS criterion were more recently admitted to nursing homes and were significantly older than those with no CAM elements documented. Characteristics associated with the prevalence of delirium symptoms in nursing home residents include recent hospitalization and increasing age.^{Reference Inouye1,Reference Kiely, Bergmann, Murphy, Jones, Orav and Marcantonio2} This finding increases our confidence that the possible ACMS criterion identifies a resident population of interest for delirium symptoms.

Notably, the assessment of ACMS documentation was not the primary objective of the prevalence survey, and our findings may be limited by incomplete medical record documentation in nursing homes.^{Reference Voyer, McCusker and Cole8} However, given the need for specialized CAM assessment training and evidence of suboptimal documentation of the core CAM elements in nursing homes,^{Reference Voyer, Richard and McCusker6–Reference Voyer, McCusker and Cole8} our findings suggest that use of the CAM to define confirmed ACMS in infection surveillance definitions is likely too stringent. Although investigators have emphasized the importance of formal CAM assessments to clinically diagnose delirium,^{Reference Inouye, Leo-Summers, Zhang, Bogardus, Leslie and Agostini10} a possible ACMS criterion may be better suited for retrospective infection surveillance. Additional research to assess barriers associated with incorporating CAM assessment and documentation practices in nursing homes is needed when evaluating residents suspected of infection to improve infection surveillance and resident safety.