ObjectivesAgitated behaviors (behaviors) are common in nursing home (NH) residents with Alzheimer's disease and related dementias (ADRD). Pragmatic trials of behavior management interventions rely on routinely collected Minimum Data Set (MDS) data to evaluate study outcomes, despite known underreporting. We describe a method to augment MDS-based behavioral measures with structured and unstructured data from NH electronic medical records (EMR). DesignRepeated cross-sectional analyses of EMR data from a single multistate NH corporation. Setting and ParticipantsLong-stay residents (at least 90 days in NH) with ADRD from January 2020 through August 2022. MethodsQuarterly and annual assessments of NH residents with ADRD during the study period were identified. For MDS, any behavior was defined as a score of 1 or higher on the Agitated and Reactive Behavior Scale. For structured EMR data, any behavior was defined as increased resident agitation, verbal aggression, or physical aggression on the Interventions to Reduce Acute Care Transfers, Change in Condition form (INTERACT). For unstructured EMR data, any behavior was defined using keyword searches of free-text orders. ResultsA total of 77,936 MDS assessments for 19,705 long-stay residents with ADRD in 322 NHs were identified; 14.8% (SD 35.6) of residents had behaviors per month using MDS alone, 16.2% (SD 36.9) using MDS and INTERACT, and 18.6% (SD 38.9) using MDS, INTERACT, and orders. Supplementing MDS with EMR data increased behavior identification by 3.8 percentage points (a 25.7% relative increase). Less than 0.5% had behaviors noted in all 3 sources consistently across study months. Conclusions and ImplicationsUsing EMR data increased detectable behaviors vs the MDS alone. The 3 sources captured different types of behaviors and using them together may be a more comprehensive identification strategy. These results are important for better targeting of interventions and allocation of resources to improve the quality of life for NH residents with ADRD-related behaviors.
Read full abstract