Psychometrics

The RBANS manual contains extensive information regarding the psychometric properties of the test, which was normed on a United States population-representative sample. The manual includes information on measures of internal consistency, test-retest reliability, inter-rater reliability, andconcurrent validity with other neuropsychological tests. The test is scaled on the basis of age, and information is provided regarding educational effects on performance. Base rates of discrepancies among the various RBANS index scores are also provided for the normative sample within the manual. Below are some additional psychometric data for the RBANS:


Subtest means and Standard Deviations (SD)

The following table contains subtest information from the standardization sample (N=540), describ ed in the manual. The data are from Form A, as only a portion of the standardization sample (N=100) also received Form B. The Form B sample was collected primarily to ensure form equivalency at the index level (see manual), and this sample was not sufficiently large to provide age-based subtest means and SDs. From the existing data, however, it would appear that performance is comparable at the subtest level for both forms (note that there is one necessary subtest adjustment on Form B- four points are added to the semantic fluency subtest in the record form to ensure equivalency with Form A. This is already included in the test form).

RBANS Subtest Means (SD), per age group




Notes for The Interpretation of Data at The Subtest Level

The RBANS was designed with a few specific goals in mind. These included keeping the total test time under 30 minutes, as well as the utilization of a range of difficulty that would allow tracking of cognitive decline down to the level of moderately severe dementia.

In order to meet these objectives, it was necessary to construct certain subtests in a way that would obviously result in a restriction of range in normals (particularly young normals). This restriction of range is almost inevitable with certain types of tests (e.g., confrontation naming, figure copying), and the restriction of range problem is compounded when time is limited and one of the test goals is to capture as low a range of performance as possible.

The net result of this is that certain index scores can be significantly affected by relatively minor changes on certain subtests, and this is particularly true for scores in the normal average range for young patients. Because there is very little range, particularly in young normals, on measures such as picture naming, word list recognition, and figure copy, changes of a few points on any of these can result in a rapid drop of the associated index score. It is worthwhile to routinely examine the subtest scores underlying index score performance for additional interpretive information, particularly if the index score appears to be unusually low in the context of a patient’s presentation or other test scores.

Because certain subtests have a limited range and skewed distribution of scores in normals, caution should be exercised in attempting to interpret individual patient performance on the basis of the normal mean and standard deviation for these subtests. It is recommended therefore, that the subtest data be used primarily to interpret index score performance, and not as “stand-alone” measures.


Additional Normative Data

The RBANS underwent a United States population-based normative standardization, and data are scaled using age-based index scores. The manual also details the effects of education on performance. Some post-publication studies have reported RBANS data from additional normative samples. These include:

Gontovsky et al. (2004): This study reported internal consistency data from a sample of 631 geriatric patients with no documented or reported history of central nervous system dysfunction. The authors reported that, consistent with the data in the RBANS manual, the test exhibited “a high degree of internal reliability consistency”. They also concluded that the index score intercorrelations suggested that the index scores measured relatively distinct neurocognitive domains and that their results provided additional evidence supporting the internal psychometric validity and reliability of the RBANS.

Gontovsky et al., (2002): This study involved 781 non-demented elderly individuals, and reported that education accounted for 1.9-7.6% of variance in RBANS index scores. The magnitude of the education effect was similar to that reported in the RBANS manual.

Patton et al., (2003): This study examined the performance of 61 cognitively normal older African-Americans on the RBANS, identifying differences between these subjects and Caucasion subjects from the same geographic region. The magnitude of the difference between these groups was found to be very similar to that observed on other neurocognitive measures (e.g., WAIS-3/WMS-3).

Beatty et al., (2003): This study reported data from normal older adults, including 278 men and 353 women, exploring the effects of sex and education. The educational effects were similar to those described above. Women slightly out-performed men on the memory and language index scores, and men slightly out-performed women on the Visuospatial/Constructional Index score. These differences are commonly observed in studies of gender differences in neurocognitive abilities, and were not of a sufficient magnitude to warrant separate scaling for men and women on the RBANS.

Duff et al., (2003): Data from 718 community-dwelling older adults from Oklahoma were examined for the effects of age and education on RBANS scores. The authors provided age- and education-corrected scaled scores to refine the clinical utility of the RBANS for use with patients who match this demographic cohort.

Wilk et al., (2004): This paper reported “normative” data on the RBANS from a sample of 575 patients meeting diagnostic criteria for schizophrenia or schizoaffective disorder. The RBANS was recommended as an efficient screening tool for assessing cognitive impairment in patients with schizophrenia, and data were presented on performance by level of education, for referential purposes.

Duff et al., (2008): This paper reported on discrepancy scores between demographically-predicted premorbid intellect and current RBANS scores in a sample of 719 community-dwelling older normal individuals compared to a nursing home sample (N=139) and a neuropsychology clinic sample (N=117)


Information on Test-Retest Interpretation

One of the most unique features of the RBANS is that it has four equivalent alternate forms, which allows for retesting patients without the confound of significant content-related practice effects. There are a variety of ways of calculating and interpreting neurocognitive change scores, and a complete discussion of this topic is beyond the bounds of this report. On a practical basis, it seems unlikely that most clinicians will be interested in plugging test scores into regression equations in order to compute the statistical probability of various score changes. It is often more useful to have a good understanding of the distribution of change scores for a particular test, and to use that information in clinical decision-making regarding the etiology of the observed change. It is always best, of course, to avoid relying upon a single source of information to conclude that there has been a significant change in a patient’s neurocognitive status, and the prudent clinician will consider multiple sources of information in reaching such a conclusion.


Data are provided for the interpretation of change when comparing a patient’s performance on Form A to Form B (regardless of order). Data are derived from N=280 (99 normal controls and 181 patients with schizophrenia). See Wilk et al., 2002 for more details. Change distributions for the two separate samples were comparable, and therefore the samples were combined for this purpose. Test-retest intervals ranged from 1 to 134 days, and there was no apparent effect of time on retest performance over this interval range.

The average Total Scale change score was less than 1 point.

The table below indicates the percentage of the combined sample that obtained a change score within each interval. For example, 4.6% of the sample had a increase in their Total Scale score on the second testing between 16 and 20 points (inclusive). Less than 7% of the sample declined by more than 10 points on the second testing, and less than 21% of the sample declined by more than 5 points- two bits of data that are clinically relevant.




Additional Test Retest Data

In addition to the test-retest data presented above, there have been some additional studies on test-retest stability with the RBANS. These include:

Duff et al., (2005): This study involved 223 community-dwelling older adults who were retested with the same RBANS form (A) after one year. Regression equations for predicting retest performance across subtests were derived and then cross-validated on a separate elderly sample of 222 subjects. Of particular note was the fact that the total scale score did not change (mean scores were within 1 point across test sessions, for both samples). This suggests that the same form of the RBANS can be used for one-year follow-up testing in older adults, without contamination from practice effects.

Duff et al. (2004): In a separate publication using the same database, these authors presented regression equations for predicting the six RBANS index scores, again using a separate sample for cross-validation. They reported, for example, that 87% of their validation sample had an observed Total Scale re-test score that fell within +/- 5 points of the predicted retest score; 99.5% were within +/- 10 points of their predicted retest score.

Patton et al. (2005): This study reported base-rate data for change scores in 283 community-dwelling older adults, tested at 1- and 2-year intervals, with Form A of the RBANS only. Again, mean Total Scale scores remained stable in these normal older adults over a two-year period, with no evidence of practice effects.