Estimating the Minimal Clinically Important Difference (MCID) of the Neuropsychiatric Inventory Questionnaire (NPI-Q) in Dementia Patients

Background and purpose：About seventy percent of dementia patients have at least one behavioral and psychological symptoms of dementia (BPSD). The BPSD not only can significantly aggravate caregiver’s distress, decrease quality of life of caregivers and patients but also associated with multiple medical and psychiatric needs. After review many measurements, researchers find out that the NPI-Q appears to be a valid and reliable clinical tool which is a brief, informant-based assessment of BPSD and associated caregiver stress. However, there are few studies discuss about the minimal clinically important difference (MCID) of the instruments of BPSD. Being no exception, the reliability and the MCID of the Chinese version of the NPI-Q was little revealed, which limited the interpretation of clinical change of noncognitive symptoms of dementia patients. Therefore, our objective was to establish the reliability and MCID of the NPI-Q in dementia patients.
Methods：The study included 45 dementia patients from a dementia institution at Taipei city, Taiwan. The NPI-Q was assessed by primary caregivers in dementia institutions every month from 2012.1. to 2012.6. We determinated the test-retest ability from 35 randomly selected patients within 5-8 days; inter-rater reliability from 24 patients after observation for one month by two raters. Intraclass correlation coefficient (ICC) and weighted Kappa were used to estimate the reliability of total score and each item of the NPI-Q.
The MCID of the NPI-Q were estimated by following three ways: (1) global rating of change by 7-point Liker scale; (2) optimal cutoff point by ROC curve with anchor-based method; (3) standard error of measurement (SEM) with distribution-based method. The range of these three values will become the MCID of the NPI-Q.
Result：The test-retest ability of the NPI-Q was good (the ICC of the severity and distress subscales were 0.95 and 0.96, respectively) and the inter-rater reliability was acceptable (the ICC of 2 subscales were 0.73 and 0.67, respectively). The MCID by first method for severity subscale was 3.33 point, and distress subscale was 4.15 point; by second method for severity subscale was 3.5 point, and distress subscale was 5.5 point; by last method for severity subscale was 2.77 point, and distress subscale was 3.1 point. Thus, the MCID of severity subscale of the NPI-Q was ranged from 3-4 point, for distress subscale was 4-6 point. We can claim that patients’ change have clinically important change when the difference of the NPI-Q over those range.
Conclusion: The NPI-Q showed good test-retest reliability and the acceptable inter-rater reliability. The MCID of the NPI-Q not only can assist clinicians in explaining about the clinical changes of BPSD, but also can help to determine the effects of treatment methods.