Internal And External Validation Of Vulnerability Indices: A Case Study Of The Multivariate Nursing Home Vulnerability Index

As the frequency of natural disasters increases, there has been an emphasis on vulnerability index creation studies. To test the validity of the most common models used in these studies, an interdisciplinary approach is used to assess the vulnerability of nursing homes throughout the Southeast (U.S). Using an inductive-hierarchical index structure; underlying community characteristics, natural hazards frequency, and nursing home facility data are combined to create the Multivariate Nursing Home Vulnerability Index (MNHVI). To internally validate these indices; a manual construction method and Monte Carlo simulations are used to create multiple unique versions of the MNHVI. External validation is used to determine if the indices are accurately predicting harm and mortality caused by storm events. Harm is identified at the county scale as reported by emergency management personnel, damage surveys, and local news outlets. Additionally, ICD codes from North Carolina death certificates are evaluated at the Census Tract scale for 30/90-days pre/post-storm. Identifying accuracy and precision for vulnerability indices provides additional assurance on the appropriate identification of at-risk regions. Internal validation processes show that none of the indices were appropriately precise. The external validation processes show the hazard level analysis to be the most accurate predictor of harm/mortality.