Validation of the Salzburg Dementia Test Prediction (SDTP) as a cognitive screening tool in an English speaking in-patient medical setting

Background: The Salzburg Dementia Test Prediction (SDTP), developed using Artificial Intelligence and based on the Mini Mental State Examination (MMSE), was recently introduced as a brief cognitive screening tool for patients with various extent of cognitive impairment. Having a brief and valid cognitive screening tool which bypasses sensory or motorinvolvement (features commonly impaired in physically frail people) has a potential to be incorporated in routine screening for cognitive impairment of older adults in medical and primary care settings.Methods: 216 medically ill older patients who had completed the MMSE score (from which the SDTP scores can be calculated), with a subsample 58 patients who had also completed the ACE-R/ACE-III scores. Whether the individual had received a diagnosis of one of 4 dementia types (n = 127) and socio-demographic information were also collected. MMSE, SDTP ACE-R/ACE-III, and dementia diagnosis were used to examine the construct validity of the SDTP through assessments of the structural, concurrent, and convergent validity. Results: The SDTP demonstrated structural validity through the three items of the SDTP loading on one component both among the total sample and the dementia subsample. Constructvalidity was demonstrated for the STDP with a sufficient correlation size with MMSE scores against a benchmark correlation size for Alzheimer's Disease (AD), Lewy body diseases (LBD), and mixed dementia diagnosis only, but not those with a Vascular Dementia (VaD). Convergent validity was demonstrated for the STDP with equivalent correlations sizes with ACE-R/ACE-III as the MMSE across all samples, though for VaD sample the magnitude of thiscorrelation was not as strong as the other correlations.Conclusions: Using three validity criteria, we highlight the applicability of SDTP as a cognitive screening tool among medically ill patients diagnosed with dementia. Our findings support using it as a brief assessment tool among patients who have been diagnosed with AD, LBD, and mixed dementia; however, there is some statistical variability to overall MMSE scores and correlations with the ACE-R/ACE-III among patients diagnosed with VaD.