Haylie M DeMercy

  • Designation: Department of Psychology, Loma Linda University
  • Country: USA
  • Title: Examining Differences Across Several Latent Factor Structures for the Neuropsychiatric Inventory—Questionnaire (NPI-Q)


Haylie M. DeMercy is a third-year Ph.D. student in the Clinical Psychology doctoral program at Loma Linda University. Before graduating from college, Haylie received her B.A. in psychology from Utah State University. Haylie’s research is focused on investigating the utility and quality of diagnostic and therapeutic interventions for cognitive impairment in aging individuals or clinical populations with neurodegenerative disorders. Haylie is passionate about increasing the accessibility and quality of mental health care for these populations. Thus, her clinical interests involve serving older adult populations and American Veterans through the U.S. Veterans Administration in therapy and neuropsychological assessment.


Background: Dementia often includes behavioral and psychological symptoms such as behavioral excitement, mood disorders, and psychosis. The Neuropsychiatric Inventory Questionnaire (NPI-Q) is a measure designed to assess the occurrence and intensity of 12 behavioral and psychological symptoms among dementia patients, along with the distress experienced by the informants. Despite the utility of the NPI-Q in research and clinical practice, there is a lack of consensus in the literature regarding its latent factor structure. In addition, previous research indicates that models with three to five factors have a good fit, but none of our knowledge has reflected on what these factor structures truly mean. Thus, the current study aimed to compare differing bifactorial models using confirmatory factor analysis to help determine the best latent model of the NPI-Q.

Methods: The current study utilized data compiled from a convenience sample of patients (N = 703) from the Alzheimer’s Disease Neuroimaging Initiative (ADNI). The study population consisted of individuals aged ≥ 50 or older who were diagnosed with dementia and BPSD symptoms. We compared three different models of BPSD using confirmatory factor analysis.

Results: The confirmatory factor analysis revealed that all three models fulfilled the criteria of excellent fit. The factor structure includes three factors of Activity, Psychosis, and Affective. The fit indices indicated this model, indicated that it was a good fit for the data, χ² (35) = 43.063, p = .164; RMSEA = .008; CFI = .993; SRMR = .012. The structure includes four factors: Activity, Psychosis, Apathy, and Affective. The fit indices indicated this model, indicated that it was a good fit for the data, χ² (40) = 54.908, p = .058; RMSEA = .010; CFI = .988; SRMR = .012. The structure includes five factors: Activity, Psychosis, Apathy, Mania, and Affective. The fit indices indicated that this model fits well for the data, χ² (31) = 36.675, p = .137; RMSEA = .008; CFI = .993; SRMR = .010. In each of the models, all subscales loaded positively onto their respective factors (p’s < .05), and all direct effects were significant (p’s < .001).

Conclusion: Since all three models provided an excellent fit for the data, the chosen factor model depends on the goals of the investigation. The three-factor model fits the data well and is most parsimonious. The four-factor model separated Affective and Apathy symptoms, which may have clinical relevance separately. The five-factor model further separated Activity into Hyperactivity and Mania. While Psychosis is a unified construct for every model, the Activity and Affective domains can be broken down further depending on the construct one is interested in measuring. The source of the variability across these models seems to depend on how sleep and appetite items are distributed. Measuring BPSD subsyndromes is valuable for tracking the progression of symptoms longitudinally and assessing interventions' impact. Thus, it is important for clinicians and researchers to carefully consider which constructs they are investigating and how they interpret sleep and appetite items before choosing a factor model. 

Don't miss our future updates!

Get in Touch