Neuropsychiatric Inventory (NPI)
Setting the Standard for Alzheimer Research
Below is a list of questions concerning the Neuropsychiatric Inventory. We will be adding more questions as we get them.
If you can not find your answer here, please contact us with your question.
The NPI-C was developed for expert clinicians and can be considered when the interviewer is an expert psychiatric/behavioral interviewer. It has not been used in clinical trials and its performance in the trial setting is not known. Clinician input may increase the validity of the ratings and could decrease the reliability by increasing variability. The NPI is the preferred instrument for most trials where the rater is not an expert clinician. With further experience, the NPI-C might be considered for trials of psychotropic agents where experts are involved.
Analytic strategies will depend on the specific hypotheses and the design of the trial. For studies, where the NPI is a secondary outcome, the following sequence of steps is suggested:
Total NPI score analysis
Determine the drug-placebo difference at each assessment time and with LOCF if that is the specified method
Intent to treat (ITT) and observed case (OC) populations
Drug-placebo difference for each domain at each assessment time
ITT LOCF and OC
Total NPI distress score analysis
Drug-placebo difference at each assessment time and with LOCF if that is the specified method
ITT and OC
Domain distress analysis
Drug-placebo difference for distress in each domain at each assessment time
ITT LOCF and OC
Symptom improvement analysis
Determine the drug-placebo difference at end of study of domains in which patients had symptoms at baseline (e.g., all patients who had non-zero agitation scores at baseline); these patients can show an improvement by decreased symptom severity.
Determine drug placebo difference at end of study for domains in which patients did not have symptoms at baseline (e.g., patients with agitation scores of zero at baseline); these patients cannot improve (zero scores at baseline) but there may be less emergence of new behaviors in the course of the trial in the treatment compared to the placebo group.
How is the NPI calculated in the situation where one of the 12 pages has YES checked for the page, but the frequency and/or severity rating is missing?
There is no good way to accommodate this type of missing data. Since there may have been a score (marked "yes"), using 0 probably underestimates the pathology. It would be better to eliminate the entire NPI and note that the total number of NPIs in the calculation is one less than might have been present in other time points.
How do you rate behaviors at the baseline assessment. Should they be rated if they are chronic and unchanged?
A behavior that has been present for the life of the patient is not scored and is assumed to be unrelated to the dementia; a behavior that is new since the onset of cognitive impairment and has been present in the past four weeks is scored.
The NPI does not have a cut off score. Some phenomena such as delusions or hallucinations are always abnormal and a score of 1 or more in these domains indicates the presence of an abnormality. On the other hand, symptoms such as irritability, anxiety and depression occur in non-diseased individuals and low scores in these domains may be normal. Therefore, no cut off score can be calculated.
For example, how much large should the dug-placebo difference be to be clinically meaningful?
The answer to this FAQ will very according to baseline severity of the behavioral change and the domain of behavior affects. For example, as a general rule, a decrease in 4 points or a 30% reduction in baseline score would be regarded as clinically meaningful. However, a smaller change in some symptoms such as agitation might be viewed as important by caregivers or might indicate an important drug effect. This must be considered on a case-by-case basis and is optimally prespecified in the analysis.
How do I handle the situation where a patient cannot answer a question and I am unsure of the scoring?
Use the not-applicable (NA) indication for this situation. One cannot score the response without reasonably informative data.
Analytically, the NA response must be treated as missing data. It should be coded separately for operational analysis to distinguish between NA (where the item was considered but no response was possible) and missing data that resulted from an oversight.
No, the NPI was validated based on the report of a caregiver. Patients (even those very mildly affected) tend to under-estimate their own psychopathology and give artificially low scores.