INTRODUCTION
Cancer is a disruptive, chronic, life-threatening illness associated
with tremendous stress. Not surprisingly then, adjustment disorder (AD)
is the most prevalent psychiatric diagnosis in people with cancer,
affecting 25–30% of all patients (Derogatis
et al., 1983; Dugan et al., 1998) and
should be equal if not higher in prevalence in those undergoing bone
marrow transplantation (BMT). Despite this high prevalence, most efforts
at screening for mental problems in cancer patients have focused on major
depression rather than so-called minor depression, or adjustment disorder.
The underdiagnosis and undertreatment of psychological problems in
those with cancer, and the subsequent negative impact on quality of
life, remains a significant problem (Katon &
Sullivan, 1990; Razavi et al., 1990).
It is clear that patients with psychological distress in general and
adjustment disorder in particular are not being diagnosed or recognized
by oncology professionals (Zabora, 1998).
Unfortunately, very little research has attempted to examine adjustment
disorder in this population. Moreover, to our knowledge, no screens
exist currently for the identification of AD for use with any
population. Our research that specifically attempted to identify AD
within the cancer population (see Part I, this issue), was unsuccessful
in using existing scales measuring related disorders (anxiety and
depression). Clearly, there is a need to develop a screen specific to
AD for use with persons with cancer.
Hypothesized Component of Adjustment Disorder
It is hypothesized that one of the personality factors that may lead
to problems of adjustment is a lack of coping flexibility on the part
of the patient regarding illness and illness-related treatment (Rogers & LeUnes, 1979; Carson et al., 1989). Those who develop adjustment
disorder may be more likely to be overly rigid in their choices of
coping strategies, attempting to address new problems in the manner
they used prior to the onset of illness, without regard for necessary
illness-based adaptations. If the prior coping and problem-solving
strategies fail, the person may begin to exhibit the emotional
reactions (depressed mood and anxiety) associated with adjustment
disorder. In contrast, those who succeed in making the adjustments to
living fully with cancer typically have an adaptive style that focuses
on coping flexibility (Lee, 1983). Moreover,
there is increasing recognition that a lack of effective, flexible
coping is prevalent in cancer patients in distress and that this lack
of flexible coping is not being recognized (American
Society of Psychosocial and Behavioral Oncology/AIDS, 1999).
However, the concept of coping flexibility has never been formally
operationalized for use in the assessment and detection of adjustment
disorder in cancer patients.
Utility of Adjustment Disorder as a Diagnosis
The purpose of screening for patients with adjustment disorder is to
identify those who may be in need of intervention but who do not meet
full criteria for DSM-IV diagnoses such as major depression or
generalized anxiety disorder. Persons with adjustment disorder have
been shown to have positive outcomes when they are treated with brief
psychotherapy (Sifneos, 1989), the usual form
of psychotherapy employed by psycho-oncologists. Thus, screening for AD
allows for the possibility of early treatment using counselors, nurses,
and other staff before a problem worsens to the point of requiring more
intensive care. The rapid identification of adjustment disorder can
prompt early psychological intervention that can help to promote the
patient's quality of life or, at the very least, may prevent the
further erosion of the patient's ability to function (Strain, 1998).
Quality of Life and Adjustment Disorder
A higher level of quality of life (QOL) is related to better
adjustment to illness. In addition to our interest in identifying a
screen for adjustment disorder, we also were interested in exploring
the relationship between quality of life, adjustment disorder, and
coping flexibility. We hypothesized that those exhibiting coping
flexibility would be less likely to have AD and more likely to report
higher QOL. Moreover, we hypothesized that poor coping flexibility
would predict lower self-reported QOL over and above the negative
effects of having cancer and a comorbid psychiatric diagnosis.
The current study was an attempt to develop and pilot test a novel
screen for the identification of adjustment disorder in BMT patients, a
group we suspected would be highly symptomatic in overall distress
levels. A pilot tool measuring coping flexibility was employed for the
purpose of detecting patients in psychological distress as defined by a
diagnosis of AD. Two critical issues were examined: whether or not the
measure could identify AD (sensitivity of measurement) and secondarily
differentiate among the DSM-IV diagnoses of major depression,
generalized anxiety disorder, and adjustment disorder (specificity of
measurement). Although there is inherent overlap of symptoms across
these diagnoses, an attempt was made to differentiate AD from major
depression and generalized anxiety disorder. Specifically, it was
postulated that adjustment disorder would differ due to the necessity
for the presence of a stressor and maladaptive reaction. Similarly, as
outlined above, a hallmark of adjustment disorder may be a lack of
coping flexibility in adapting to illness and treatment-related change.
The central purpose of the study was to gather initial data and
suggestions for items to be included on potential screening tools to
undergo future full psychometric exploration.
RESULTS
The average age of the sample was 45.76 years (SD = 11.72)
and was comprised of 41 women (43.2%) and 54 men (56.8%). About a third
of the participants had a high school education (33.7%, n =
32), 27.4% (n = 26) had some college course work, and 13.7%
(n = 13) had a college degree. Most were currently married
(68.4%, n = 65), 14.7% (n = 14) were single, and
10.5% (n = 10) were divorced. The vast majority (92.6%,
n = 88) were Caucasian, whereas the remainder (7.4%,
n = 7) were African-American. In addition, the majority
(52.6%, n = 50) were disabled, although some were currently
employed full-time (23.2%, n = 22) or were homemakers (8.4%,
n = 8). The type of tumor varied across the sample, with acute
myeloid leukemia (24.2%, n = 23), multiple myeloma (23.2%,
n = 22), and chronic myeloid leukemia (15.8%, n = 15)
being most prevalent. At the time of the interview, most of the
participants were either preautologous transplant (32.6%, n =
31) or postallogeneic transplant (28.4%, n = 27).
Analyses were conducted to explore the relationship of the
demographics/medical information variables to the study measures.
None of the demographic variables were significantly related to SCID
diagnosis or to the C-Flex.
Based on the SCID results, slightly more than half of the
participants were suffering from a currently diagnosable mental
illness, 34.7% (n = 33) received a diagnosis of adjustment
disorder, 11.6% (n = 11) received a diagnosis of major
depression, and an additional 5.3% (n = 5) met the criteria
for the diagnosis of generalized anxiety disorder. The remaining 46
participants (48.4%) did not meet criteria for any of the disorders
examined.
It was also of interest to explore the potential impact of stage and
type of transplant on psychiatric diagnosis. A total of 31 patients
(32.6%) were preautologous, 16 were postautologous (16.8%), 21 were
preallogeneic (22.1%), and 27 were postallogeneic (28.4%) transplant.
There were no significant differences in number or type of psychiatric
diagnoses as a function of BMT type (χ1,32 =
0.64, n.s.), specific transplant status
(χ1,32 = 2.02, n.s.), or a combination of
both factors (χ1,92 = 3.05, n.s.).
Participants displayed considerable variation in their scores on the
measures. Scores on the C-Flex ranged from 40 to 95 (mean = 69.33,
SD = 11.12), accounting for nearly 70% of the possible range
(20 to 100) on the instrument. In addition, the C-Flex exhibited
adequate internal consistency as measured by a Cronbach's alpha of
0.88.
Internal Consistency, Item Analysis, and Factor
Analysis of the C-Flex
The 20 items on the C-Flex were examined from three different
approaches. First, item descriptive statistics were calculated, with
attention paid to the item distributions, means, variance, skewness,
range, and item-total correlations. The item distributions were
examined to identify items with ceiling or floor effects, low
variability, or small range, which would make the items suspect. All
items were deemed acceptable for further analysis.
Second, a principal components factor analysis was completed. An
orthogonal varimax rotation was used to help identify simple factor
structure. The results of the factor analysis (see Table 1) revealed a 5-factor solution (the number
of factors to retain was based upon factors with eigenvalues over 1.0
and a factor scree plot). Upon examination, the eight items from the
first, and strongest, factor (items 8, 9, 11, 14, 16, 17, 19, and 20)
seemed to best tap the construct that could be deemed as coping
flexibility to illness and illness-related change. The second factor,
consisting of items 1, 2, 3, 7, 10, and 15, seemed to be tapping
interest in activities such as work, hobbies, and sex. The third factor
(items 12 and 13) clearly measured anger. The fourth factor (items 5
and 18) assessed attitudes toward medications, and the final factor
(items 4 and 6) focused on family roles. Because some authorities
suggest using an oblique rotation for exploratory factor analysis
(Pedhazur & Schmelkin, 1991), the factor
analysis also was performed using an oblique rotation using the same
criteria for factor retention as above (eigenvalues over 1.0 and
analysis of a scree plot). The results of this second factor analysis
identified the same factors although individual item factor weightings
shifted somewhat.
Results of a principal components factor analysis on the C-Flex using
an orthogonal varimax rotation
The two items loading on Factor III were tentatively deleted because
they appeared to measure anger rather than coping flexibility. Although
the deletion of Factor III can be justified from the standpoint of
classical test construction, which focuses on scale development based
on construct identification of latent variables, it is also possible to
utilize an approach based on criterion-related validity. Even from this
empirically keyed approach, however, Factor III would be marked for
exclusion by its lack of significant correlation to SCID-adjustment
disorder (r = −0.094, n.s.). Thus, it was deemed that
Factor III offered no incremental value in predicting adjustment
disorder and it was dropped from further consideration.
The third approach to exploring the items was to conduct a
reliability analysis. Total scale internal consistency, when all 20
items were included in the analysis, yielded a coefficient alpha of
0.88. As mentioned above, based on results of the factor analysis, the
reliability analysis also was completed after eliminating the two items
of Factor III, because they did not seem to assess coping flexibility
and were not correlated to the diagnosis of adjustment disorder on the
SCID. Other items were also examined to see if their removal would
increase overall alpha; however, none were identified. The remaining 18
items still exhibited a coefficient alpha of 0.88 for the total scale
and all were retained for hypothesis testing analyses. Reliability of
the four individual subscales also was examined. Alphas ranged from
0.43 to 0.93.
Correlations between Study Measures and SCID Diagnosis
Table 2 displays the correlations between
the C-Flex and the SCID diagnoses. It should be noted that negative
correlations do not necessarily indicate an indirect relationship
between constructs but may be artifacts of scoring. Higher scores on
the C-Flex indicate higher functioning. As shown in the table, the
C-Flex (r = −0.086, n.s.) was not a significant
predictor of either adjustment disorder or generalized anxiety
disorder. However, the C-Flex was significantly related to major
depression (C-Flex: r = −0.436, p <
0.001).
Correlations between the C-Flex and the SCID diagnoses of adjustment
disorder, major depression, and generalized anxiety disorder along with
established measures
It was also of interest to determine whether the C-Flex was
correlated with the established measures to determine an initial
picture of both convergent and divergent validity. The C-Flex was
significantly correlated to all of the other scales, indicating
substantial overlap, and was especially strongly correlated with the
ZSDS (r = 0.614, p < 0.000) and the FACT-G scale
(r = 0.740, p < 0.000). As a test of divergent
validity, the C-Flex was correlated with scores on the Physical
Subscale of the FACT-G. The C-Flex exhibited poor divergent validity as
evidenced by strong correlations with the subscale.
Sensitivity and Specificity Results
Given the nonsignificant correlations with AD, sensitivity and
specificity statistics were not investigated for the C-Flex.
Quality of Life
It was also of interest to explore whether coping flexibility
measured by the C-Flex was indicative of perception of enhanced quality
of life. From the earlier correlations, it was evident that coping
flexibility was significantly related to global quality of life as
measured by the FACT-G (r = 0.74, p < 0.001). To
further test this relationship, a multiple regression analysis was
conducted. After controlling for the influence of psychiatric diagnosis
(R2Δ = 0.37, F = 53.45, p
< 0.001) and specific type of BMT (n.s.), coping flexibility was not
found to be a significant predictor of quality of life.
Exploratory Analyses
Sensitivity and Specificity in Detecting Any SCID
Diagnosis
Analyses were repeated for the ability of the C-Flex to identify a
criterion defined as any of the SCID diagnoses (major depression,
generalized anxiety disorder, or adjustment disorder). The C-Flex
(r = −0.44, p < 0.001) was significantly
correlated to the presence of any disorder. Sensitivity and specificity
were then examined using the cutoff score that produced the greatest
overall accuracy (total number of AD negatives correctly identified
plus total number of AD positives correctly identified divided by total
N). The C-Flex exhibited poor specificity (30.43%) and good
sensitivity (91.84%) in detecting any diagnosis.
Predictive Utility of C-Flex Subscales
A further question concerned whether any of the four
factor-analysis-derived subscales of the novel C-Flex scale could stand
alone as predictors of either adjustment disorder or any diagnosis. To
examine this question, a series of correlations was conducted between
the C-Flex subscales and (1) any SCID diagnosis and (2) SCID adjustment
disorder diagnosis. None of the C-Flex factors were significantly
correlated to the SCID–adjustment disorder diagnosis. However
Factor I (r = −0.44, p < 0.001) and Factor
II (r = −0.49, p < 0.001) were significantly
related to the presence of any SCID disorder. As a follow-up,
sensitivity and specificity statistics were then calculated for Factors
I and II using a cutoff of ≤30 and ≤20, respectively. Factor I
showed the best overall predictive accuracy in predicting any diagnosis
with a specificity of 76.09% and sensitivity of 81.63%.
DISCUSSION
This study was an attempt to develop a brief, rapid screen for the
presence of adjustment disorder diagnosis in BMT patients. To this end,
a novel screening instrument was developed to attempt to screen for
adjustment disorder.
Although the study results were disappointing with regard to this
primary aim, there were some potentially interesting findings. In terms
of predictive validity, the C-Flex proved to be a very poor screening
tool for adjustment disorder in this population. In contrast, the
C-Flex displayed moderately high correlations with scales measuring
depression, anxiety, and quality of life. Moreover, the scale appeared
to be useful in predicting the presence of any disorder (not just AD).
Overall, the C-Flex exhibited adequate reliability and yielded
understandable but somewhat unreliable subscales based on the factor
analysis. Also, the correlations with AD were in the expected direction
and approached significance.
Although not useful in predicting AD, the C-Flex may be a useful tool
in predicting psychological distress in general as it does not appear
to tap a construct unique to adjustment disorder. Thus, poor perceived
coping flexibility in addressing illness may be a general risk factor
for psychological distress rather than a specific marker for adjustment
disorder, and the C-Flex may be a better tool for detecting general
psychological distress than adjustment disorder specifically. Indeed,
there is some limited research to suggest that coping flexibility is
beneficial in the reduction of symptomatology for schizophrenia,
depression, and anxiety (Wheaton, 1985; Perry, 2000).
It is possible that other methods of scale development may have
improved the results. The current study emphasized construct-based
scale development. Criterion-based methods maximize predictability and
likely would have produced higher levels of detection, although
predictive correlations from tests developed in this way tend to be
difficult to replicate in new samples (Nunnally,
1978). For example, in the case of the C-Flex, the instrument
may have benefited from selecting items most highly related to AD.
Alternately, specificity may have been enhanced by selecting only those
items that were unrelated to SCID-identified major depression or
generalized anxiety disorder to assure divergent validity. Future
research may benefit from exploring these options or expanding both the
item pool and number of participants.
Portions of the C-Flex may have some utility as screens for general
psychological distress. Specifically, the eight items comprising Factor
I of the C-Flex may prove useful as a very short screen. Overall,
Factor I showed a specificity of 56.45% and sensitivity of 75.76% for
detecting adjustment disorder and a specificity of 76.09% and
sensitivity of 81.63% for detecting any psychiatric diagnosis.
Therefore, it might be a useful stand-alone tool for identifying
general malaise in this population.
The underlying rationale for the study was that a better
understanding of AD and its detection could have a helpful impact on
the delivery of services for cancer patients in the medical setting. As
stated earlier, the usefulness of adjustment disorder as a diagnostic
category for oncology is to identify those patients who may be in need
of intervention but who do not meet full criteria for DSM-IV diagnoses
such as major depression or generalized anxiety disorder. With early
identification, persons with adjustment disorder can benefit greatly
from brief psychotherapy and psychoeducation (Sifneos, 1989; Pollin &
Holland, 1992; Wise, 1994; Strain, 1998). Unfortunately, a suitable means for
screening for adjustment disorder remains elusive.
On a final note, we explored coping flexibility as measured by the
C-Flex for its ability to predict perceptions about quality of life. It
was felt that those patients with higher levels of coping flexibility
would also be likely to maintain a higher level of quality of life.
Although the two constructs were related (r = 0.74, p
< 0.001), a multiple regression analysis controlling for psychiatric
diagnosis and type of BMT failed to reach significance. Thus, although
a relationship exists between the two, it was overshadowed by other
variables in the study.
Limitations
There are several limitations to the current study. First, the sample
was almost completely Caucasian, which may limit the generalizability
of the findings. Second, the interrater reliability of the SCID
diagnoses is not known. It may be problematic that only one rater
performed the SCID diagnostic interviews. Third, the use of
face-to-face SCID interviews may have biased the results. Although
subjects seemed to be very forthcoming, the possibility exists that
there was some underreporting of symptoms. There is some evidence that
patients may answer questions on a self-report screening questionnaire
in a more forthright and honest way than they might if questioned face
to face (Zabora, 1998). Thus, future studies
may want to consider alternative procedures for diagnosis such as
computer-based diagnostic programs. Finally, the study utilized a
cross-sectional design and therefore results cannot be interpreted
causally. Subsequent research should use a longitudinal design that
would allow for the collection of data on the natural progression of AD
in patients with cancer as well as the opportunity to explore the
predictive utility of the measures.
Conclusion
A clear problem was found in trying to develop a screen for
adjustment disorder. The novel screen based on the idea of coping
inflexibility was unable to detect adjustment disorder with any degree
of reliability. Thus, the challenge is to continue work in the area to
identify the core concept that can help identify these patients and
lead to the creation of brief screening tools.
