Professional Documents
Culture Documents
1177/0093854805278805
CRIMINAL JUSTICE AND BEHAVIOR
Miller / GENERALIZABILITY & UTILITY OF THE M-FAST
THE MILLER-FORENSIC
ASSESSMENT OF SYMPTOMS
TEST (M-FAST)
Test Generalizability and Utility Across
Race, Literacy, and Clinical Opinion
HOLLY A. MILLER
Sam Houston State University
The Miller-Forensic Assessment of Symptoms Test (M-FAST) was developed to provide clini-
cians with a reliable and valid screen for malingered mental illness. Previous studies have dem-
onstrated the development and initial validation of the M-FAST. The current analyses extend the
M-FAST generalizability across literacy status and race and compare predictive utility with clin-
ical opinion. Study 1 includes a sample of 280 forensic male psychiatric patients, 5 psychiatrists,
and 8 psychologists. The psychiatric participants were administered the SIRS, M-FAST item
pool, M Test, and MMPI-2. Study 2 includes an independent sample of 50 male forensic psychi-
atric participants and the 13 mental health professionals. Results provide evidence of
generalizability of the M-FAST across literacy and racial groups. Results also indicate that the
M-FAST produced higher classification accuracy than the M Test and clinical opinion, signifi-
cantly reducing the number of false negatives at the screening stage of malingering assessment.
I t has been estimated that base rates of malingering vary across set-
tings. Survey data have indicated malingered mental illness rates
from 7% to 21% in general and forensic settings, respectively (Rogers,
1986; Rogers, Salekin, Sewell, & Goldstein, 1996; Rogers, Sewell, &
AUTHOR NOTE: Partial funding for this dissertation work was provided by the
Kellogg Dissertation Research Grant Fund. Please address correspondence to Holly
A. Miller, College of Criminal Justice, Sam Houston State University, Huntsville, TX
77341-2296. E-mail: hmiller@shsu.edu
CRIMINAL JUSTICE AND BEHAVIOR, Vol. 32 No. 6, December 2005 591-611
DOI: 10.1177/0093854805278805
© 2005 American Association for Correctional and Forensic Psychology
591
Generalizability
Utility
Current Analyses
METHOD
Participants
Instruments
nine items were rationally derived from the following seven strategies
that previously have been demonstrated to distinguish individuals
who were malingering from honest responders: (a) unusual hallucina-
tions, (b) reported versus observed, (c) extreme symptomatology, (d)
rare combinations, (e) negative image, (f) unusual symptom course,
and (g) suggestibility (see Miller, 2001, for a more detailed descrip-
tion of the strategies).
M Test. The M Test (Beaber, Marston, Michelli, & Mills, 1985) was
among the first attempts to develop a screening measure to detect
malingering, specifically malingered schizophrenia. The M Test was
used as an additional validity measure because of previous use as a
concurrent validity measure and its focus on malingered schizophre-
nia. It was expected that the current psychiatric population would
encompass patients diagnosed with schizophrenia as well as patients
who were malingering schizophrenic symptoms.
Procedure
RESULTS
Generalizability
pants and indicated that the AUC was .89 (SE = .04), with a 95% confi-
dence interval of .80 to .97 (see Figure 2). An ANOVA was performed
to compare the M-FAST total score across groups and indicated a
nonsignificant difference between the literate and illiterate groups,
F(1, 280) = .15, p = .70. These results provide preliminary evidence
that there is little difference on M-FAST results between illiterate and
literate individuals.
Clinical Opinion
Clinical opinion data were collected on 136 of the 280 forensic psy-
chiatric participants. Although the final 25 items had not been chosen,
overall rates of accuracy were analyzed to examine the difference
between clinical opinion and use of an actuarial measure to screen for
malingered mental illness. General comparisons are provided through
ROC analyses (see Figure 3). For the psychiatric group, the AUC was
.72 (SE = .05), with a confidence interval of .62 to .81. The ROC for
the psychologist group produced an AUC of .80 (SE = .04), with a
confidence interval of .72 to .88. The M-FAST total score (79 items)
ROC Curve
1.00
.75
.50
Sensitivity
.25
0.00
0.00 .25 .50 .75 1.00
1 - Specificity
Diagonal segments produced by ties.
Figure 1: Receiving Operating Characteristic (ROC) Curve for the Literate Par-
ticipants (n = 222) in Study 1
ROC Curve
1.00
.75
.50
Sensitivity
.25
0.00
0.00 .25 .50 .75 1.00
1 - Specificity
Diagonal segments produced by ties.
Figure 2: Receiving Operating Characteristic (ROC) Curve for the Illiterate Par-
ticipants (n = 58) in Study 1
METHOD
Participants
1.00
.75
.50
Sensitivity
.25
0.00
0.00 .25 .50 .75 1.00
1 - Specificity
Diagonal segments produced by ties.
1.00
.75
.50
Sensitivity
.25
0.00
0.00 .25 .50 .75 1.00
1 - Specificity
1.00
.75
.50
.25
Sensitivity
0.00
0.00 .25 .50 .75 1.00
1 - Specificity
Age (years)
Mean (SD) 33 (8.7) 39 (11.9)
Range 21 to 51 18 to 70
Education (years)
Mean (SD) 10 (2.2) 12 (2.5)
Range 5 to 14 8 to 16
Marital status (%)
Single 93 58
Married 7 8
Divorced 0 31
Separated 0 3
Race (%)
Caucasian 21 56
African American 57 39
Hispanic 14 3
Other 8 2
Literacy (%)
Literate 86 83
Illiterate 14 17
RESULTS
Generalizability
1.00
.75
.50
Sensitivity
.25
0.00
0.00 .25 .50 .75 1.00
1 - Specificity
Diagonal segments produced by ties.
Figure 4: Receiving Operating Characteristic (ROC) Curve Indicating Diagnos-
tic Efficiency of the Miller-Forensic Assessment of Symptoms Test
(M-FAST) With the Literate Group in Study 2
ROC Curve
1.00
.75
.50
Sensitivity
.25
0.00
0.00 .25 .50 .75 1.00
1- Specificity
Diagonal segments produced by ties.
Figure 5: Receiving Operating Characteristic (ROC) Curve Indicating Diagnos-
tic Efficiency of the Miller-Forensic Assessment of Symptoms Test
(M-FAST) With the Illiterate Group in Study 2
Literate Illiterate
M-FAST Scale M (SD) M (SD) df F p
Total score
Malingerers 13.2 (6.2) 10.5 (5.0) 1 0.33 .58
Honest 2.6 (3.1) 1.7 (3.2) 1 0.46 .50
Rare combinations
Malingerers 3.8 (1.8) 4.2 (2.0) 1 0.79 .39
Honest 0.5 (0.8) 0.1 (0.4) 1 1.13 .30
Unusual hallucinations
Malingerers 2.5 (1.6) 1.5 (2.1) 1 0.61 .45
Honest 0.5 (0.9) 0.3 (0.8) 1 0.35 .56
Extreme symptomatology
Malingerers 3.7 (2.0) 3.5 (1.9) 1 0.62 .45
Honest 0.9 (1.7) 0.7 (1.2) 1 0.36 .63
Reported versus observed
Malingerers 1.4 (1.2) 0.5 (0.7) 1 1.12 .31
Honest 0.2 (0.4) 0.3 (0.5) 1 0.25 .62
Utility
Group NPP PPP Specificity Sensitivity FP FN
Note. NPP = negative predictive power; PPP = positive predictive power; FP = false
positives; FN = false negatives.
Utility
Total score
Malingerers 14.3 (4.1) 12.9 (4.9) 1 0.72 .43
Honest 2.9 (3.7) 2.2 (2.8) 1 0.50 .49
Rare combinations
Malingerers 3.6 (1.4) 2.9 (2.3) 1 1.90 .20
Honest 0.6 (0.8) 0.4 (0.8) 1 0.78 .39
Unusual hallucinations
Malingerers 3.1 (1.6) 1.7 (1.5) 1 1.94 .20
Honest 0.9 (1.2) 0.4 (0.6) 1 2.80 .11
Extreme symptomatology
Malingerers 4.0 (1.2) 2.9 (1.7) 1 1.49 .32
Honest 0.9 (1.2) 0.9 (1.3) 1 0.05 .83
Reported versus observed
Malingerers 2.0 (0.9) 1.8 (0.7) 1 0.52 .61
Honest 0.3 (0.5) 0.2 (0.4) 1 0.32 .58
Utility
Group NPP PPP Specificity Sensitivity FP FN
Note. NPP = negative predictive power; PPP = positive predictive power; FP = false
positives; FN = false negatives.
GENERAL DISCUSSION
.75 .75
.50
.50
.25
Sensitivity
.25
Sensitivity
0.00
0.0
0.00 .25 .50 .75 1.00
0.00 .25 .50 .75 1.00
1- Specificity
1- Specificity
Diagonal segments produced by ties.
Diagonal segments produced by ties.
.75
.75
.50
.50
.25
Sensitivity
.25
Sensitivity
0.00
0.00 .25 .50 .75 1.00 0.00
0.00 .25 .50 .75 1.00
1- Specificity
1- Specificity
Diagonal segments produced by ties.
Diagonal segments produced by ties.
tion, similar to the SIRS research results across race (Rogers, 1997b),
the M-FAST appears to generalize to both African American and Cau-
casian individuals. These studies indicate almost identical M-FAST
results for African Americans and Caucasians.
These studies also provide the first evidence comparing actuarial
prediction with the clinical opinion of malingering. Specifically, these
studies compared the M-FAST with the screening procedure gener-
ally used by forensic psychiatrists and psychologists (clinical inter-
views and file review). Although both the psychiatrists and psycholo-
gists in these studies produced accuracy above chance, their rate of
false-negative prediction was most problematic. The psychiatrists in
the sample produced a sensitivity rating of only .50, which translates
into their opinion of malingering missing 50% of the actual malinger-
ers in the sample. The psychologists’ opinions produced a sensitivity
rating of .71, which indicates that they missed 29% of the malingerers
in the sample. In contrast, the M-FAST produced a sensitivity score of
.93, indicating that the screening measure only missed 7% (n = 1) of
the malingerers in the sample.
False-negative prediction in a screening procedure should be mini-
mized, unlike the false-positive prediction where the examiner will
conduct further assessment to determine the malingering diagnosis. It
may be that those individuals who successfully fake mental illness at
the screening stage will never be assessed more specifically for malin-
gering and be considered legitimate patients. This is problematic
because these individuals will likely receive treatment or monetary
gain that should have been provided to a bona fide mentally ill person.
These analyses indicate that using an actuarial malingering screening
measure produces higher classification rates than clinical opinion.
In fact, the M-FAST significantly reduced the rates of false-negative
prediction.
As with any screening measure, the M-FAST should be used as an
initial level of assessment. To make a final diagnosis of malingering, a
thorough evaluation would likely include the SIRS, MMPI-2, or Per-
sonality Assessment Inventory (PAI; Morey, 1991) indicators, clinical
interviews, and collecting collateral information. The final determina-
tion of malingering should not rely on a screening measure but be
deferred until a complete assessment is conducted.
NOTE
1. The M Test and Minnesota Multiphasic Personality Inventory-2 (MMPI-2) were given as
part of the initial studies for item selection and validation purposes. These results were reported
in the M-FAST Professional Manual (Miller, 2001).
REFERENCES