Creation of a Short Form Boston Naming Test for Individuals with Aphasia

1
Creation of a Short Form Boston Naming Test for
Individuals with Aphasia Christina M. del
Toro1,3, Diane L. Kendall1,3, Craig Velozo2,4
Malcom Randall VA RRD Brain Rehabilitation
Research Center1, Malcom Randall VA HSRD/RRD
Rehabilitation Outcomes Research Center2,
University of Florida, Department of
Communication Sciences Disorders3, University
of Florida, Department of Occupational Therapy4
Figure 1. Graves et. al,(2004) BNT Short Form
Item Map ltmoregtltraregt 3
.
T
58PALETTE
. 2
T
.

55SPHYNX . 1
SS
. 54TONGS
. 57TRELLIS
.
42STETHOSCOPE 45UNICORN 0 M
44MUZZLE 53SCROLL .
41PELICAN 24SEAHORSE
31RHINOCEROS 48NOOSE
. M
35DOMINOES .
-1 S .
19PRETZEL

.
T 32ACORN -2
. S


-3 .
ltlessgtltfrequentgt
Figure 3. New BNT Short Form Item Map
ltmoregtltraregt 4 .

T
59PROTRACTOR 3 .
58PALETTE

. S 56YOKE 2
.
55SPHYNX
. S
1 54TONGS
. 57TRELLIS
42STETHOSCOPE
. 41PELICAN
24SEAHORSE 0 . M
35DOMINOES
. M
19PRETZEL -1
.
32ACORN
.
10TOOTHBRUSH -2
S .

S -3
. 5WHISTLE

-4
T
.
-5 1BED


-6 .
lessgtltfrequentgt
Phase 3 Results Discussion
Background
Phase 1 Results Discussion

• Aim 3 Rasch Analysis of a new short form
• Total N analyzed 85 out of 100 participants. 15
  participants scored 0 correct providing
  insufficient data and were dropped from analysis
• Range of person ability item difficulty -6 to
  4 logits, an increase from the previous forms.
  The larger range at the low end of the hierarchy
  (-6) indicates the form captures lower levels of
  ability not previously measured in phase 1 and 2
  forms.
• Mean ability level of the participants is higher
  than the difficulty level of the test.
• No items misfit based on Infit Mean Square
  (1.4)
• 1 item misfit on Outfit Mean Square ( 2.0)
  ranges.
• Person separation 1.97
• Number of Strata 2.96, indicating the short
  form can divide people into approximately 3
  ability levels, similar to the other forms.
• Person reliability 0.80
• Item reliability 0.97
• No items were redundant based on the unit of
  measure (item difficulty figure 3)

• Aim 1 Rasch Analysis of Graves et. al (2004)
  short form
• Total N analyzed 74 out of 100 participants. 26
  participants scored 0 correct providing
  insufficient data and were dropped from analysis
• Range of person ability item difficulty -3 to
  3 logits
• Mean ability level of the participants is lower
  than the mean difficulty level of the test., (M
  in Figure 1), indicating the test should include
  easier items to measure individuals of lower
  ability.
• No items misfit based on Infit Mean Square
  (1.4), and Outfit Mean Square ( 2.0) ranges,
  an indication of the unidimensionality of the
  short form.
• Person separation 1.83
• Number of Strata 2.7, indicating the short form
  can divide people into approximately 3 ability
  levels.
• Person reliability 0.77
• Item reliability 0.89
• 3 of the 15 (20) items were redundant (items
  that occur at the same unit of measure or
  difficulty level Figure 1), suggesting a
  rationale for item reduction.

The Boston Naming Test (BNT Kaplan, Goodglass,
Weintraub, 1983) as well as the short form
included in the second edition by Mack, Freed,
Williams, Henderson (1992), were developed
using traditional standardization methods of
classical test theory. Standardization is
performed on the test as a whole rather than on
the individual items, without attention to the
difficulty level of items. Although item
statistics can be generated post hoc, they apply
only to that group of subjects on those
particular items. Rasch analysis however takes
advantage of recent advances in psychometrics,
allowing for analysis at the item level. Item
response theory claims that the probability of a
persons response to an item is the combined
function of that persons ability and the
difficulty level of the item (Bond Fox,
2001). One of the short forms of the BNT (Graves
et al, 2004 Bond Fox, 2001) employed Rasch
analysis to responses from neurologically-healthy
individuals and individuals diagnosed with
various forms of dementia. However, considering
the differences in underlying mechanisms of
anomia for dementia compared with aphasia, it is
not known if this particular short form is valid
for use with individuals with aphasia.
Additionally, the Mack et. al (1992) short form
currently published with the BNT was developed
using data from individuals with Alzheimers
disease and age-matched healthy controls. Thus,
there is a need for empirical evidence that these
forms are appropriate for assessing naming
function in aphasia.
Purpose
The primary aim of this study was to investigate
the validity of items from the Boston Naming Test
for use in a 15 item short form for an aphasia
population. We measured item difficulty, item
fit, item redundancy, person separation,
reliability, and the original proposed hierarchy
of lexical frequency. To address this aim we
analyzed two previously published forms (Graves,
Bezeau, Fogarty, Blair, 2004 Mack, Freed,
Williams, Henderson, 1992) and one developed
for this study.
Figure 2. Mack et. al (1992) BNT Short Form Item
Map ltmoregtltraregt 4 .
.
T T
58PALETTE
3
.

55SPHYNX
2 . S
S .

42STETHOSCOPE
45UNICORN 1 .
39HAMMOCK 52TRIPOD
. 29BEAVER
.
0 M
13OCTOPUS . M 23VOLCANO
36CACTUS
.
26CANOE -1
.
10TOOTHBRUSH
.
S 20BENCH -2
S .


. -3 4HOUSE
7COMB

T .
-4 . T
lessgtltfrequentgt
Phase 2 Results Discussion
Methods

• Aim 2 Rasch Analysis of Mack et. al (2004) short
  form
• Total N analyzed 83 out of 100 participants. 17
  participants scored 0 correct providing
  insufficient data and were dropped from analysis
• Range of person ability item difficulty -4 to
  4 logits, indicating this form captures a wider
  range of ability than Graves et. al (2004) form.
• Mean ability level of the participants is nearly
  equivalent to the mean difficulty level of the
  test (M on Figure 2). This form includes several
  easy items thus, measuring individuals of lower
  ability not measured by the Graves et. al (2004)
  form.
• No items misfit based on Infit Mean Square
  (1.4) 3 items misfit based Outfit Mean Square
  ( 2.0) ranges, suggesting unexpected responses
  away from a persons ability (naming a harder
  item or not naming an easier item).
• Person separation 2.09
• Number of Strata 3.12, indicating the short
  form can divide people into approximately 3
  ability levels similar to the Graves et. al
  (2004) form.
• Person reliability 0.81
• Item reliability 0.96
• 3 of the 15 (20) items were redundant (items
  that occur at the same unit of measure (item
  difficulty Figure 2), suggesting a rationale for
  item reduction.

• Participants
• 100 individuals with aphasia
• a single left hemisphere stroke at least 6 months
  prior to enrollment
• right-handed
• monolingual English speaking
• no prior CVA or pre-existing neurological illness
• No severe impairment in vision or hearing

• Overall, a wide range of difficulty level (easy
  and hard items) is needed to capture the large
  range of ability levels seen in aphasia.
• The redundancy observed in the 60 item corpus,
  suggests the need for a shortened version. Thus,
  for individuals who have aphasia as a result of
  stroke, it is recommended to use 15 items which
  span a large difficulty range without redundancy
  in the measure of difficulty.
• The original hypothesis for the creation of the
  BNT items was based on lexical frequency however,
  discrepancies between item difficulty and word
  frequency suggests other factors may be affecting
  word production (e.g. syllable length,
  phonotactic probability, etc).
• Thus, our lab will conduct two further studies

• Data Collection
• Archival data from the patients screened at the
  VA Brain Rehabilitation Research Center
• All participants were given the full 60 item BNT
  by a licensed speech-pathologist following
  original directions for administration and
  scoring
• Only spontaneous responses without cueing were
  included

• Data Analysis
• Rasch analysis was computed using the responses
  of 100 individuals
• with aphasia in 3 phases
• 1) 15 items from the Graves et. al (2004) short
  form
• 2) 15 items from the Mack et. al (1992) short
  form
• 3) 15 items newly chosen for this study based on
  the results of phases 1 2
• Items from the Graves et. al (2004) and the Mack
  et. al (1992) forms were kept based on Infit and
  Outfit Mean Squares
• Overlapping items were excluded
• 4 items were added to fill in gaps in the
  hierarchy of difficulty level
• and expand the range to include easier items
• Participant responses were analyzed with the
  WINSTEPS Rasch analysis computer program (Bond
  Fox, 2001 Linacre, 2004, 1994).
• Infit mean squares greater than or equal to 1.4
  were considered to misfit
• Outfit mean squares greater than or equal to 2.0
  were considered to misfit

• The original 60-items of the BNT will be
  analyzed using responses from the rating scale
  scoring system which includes the semantic and
  phonologic cues. Inclusion of participants
  responses with cues will provide additional
  information of each participants ability
  allowing a more fine-grained comparison to the
  tests difficulty.
• The original 60-items of the BNT will also be
  analyzed using a hierarchy of psycholinguistic
  variables such as phonotactic probability,
  syllable length, cluster effect, and neighborhood
  density. This analysis will determine if the BNT
  is strictly a test of lexical/semantic frequency.
  

This presentation is based upon work supported
with resources and the use of facilities from the
Office of Research and Development,
Rehabilitation RD Service, and the Brain
Rehabilitation Research Center, of the Department
of Veterans Affairs.