Psychological predictors of admission and discharge Global ...

Aging & Mental Health, November 2004; 8(6): 505-513

ORIGINAL ARTICLE

Psychological predictors of admission and discharge Global Assessment of Functioning Scale scores for geropsychiatric inpatients J. A. WHITNEY', M. E. KUNIK^ V. MOUNARI^ F. G. LOPEZ' & T. KARNER^ University of Houstoyi, Houston, Texas, Baylor College of Medicine, Houston, Texas & University of South Florida, Tampa, Florida, USA

Abstract This study examined psychological predictors of Global Assessment of Functioning Scale scores (GAF) at intake and at discharge for geropsychiatric inpatients. A total of 301 predominantly male geropsychiatric inpatients were recruited from the Houston Veterans Affairs Medical Center. Path analysis was used ro test a model proposing causal paths of psychological predictors (cognitive status, depression, agitation, general psychiatric status) to GAF scores on admission and discharge. At admission, all four psychological predictors had positive paths to the admission GAF scale scores but at discharge, only two (i.e., cognidve status and general psychiatric status) of the four psychological predictors had positive paths to the discharge GAF scale scores. The admission GAF scale scores also had a positive path to the discharge GAF scale scores. The overall functioning level of geropsychiatric inpatients may be predicted by measures which assess overall cognitive status and general psychiatric functioning during hospitalizadon. Results prompt consideration of a modified global scale specifically reflecting an older adult's activities during this period of life.

Introduction An evaluation of psychological symptoms and rating of the overall functioning level of the older adult is commonly conducted upon psychiatric admission and at discharge. One of the most widely used scales to measure overall functioning level by mental health professionals is the Global Assessment of Functioning Scale (GAF) of the Diagnostic and Statistical Manual of Mental Disorders (fourth edition). The GAF is the fifth axis (AXIS V) of this multi-axial diagnostic scheme, which provides a single index to represent a patient's highest level of psychological, social, and occupational functioning. Widely used by mental health professionals in both inpatient and outpatient settings, psychiatric facilities, and managed care organizations, the total score derived from the GAF is seen to measure improvement of symptoms, to guide intervention and treatment protocols, and to direct managed (behavioral health) care allocation of funding (Piersma & Boes, 1997; U.S. Veterans Health Administration, 1997). Older psychiatric patients generally present vt'ith an array of psychological disorders (e.g., cognitive, affective, and behavioral symptoms) that

compromise their overall daily functioning level and adversely impact specific aspects of their mental and physical health. Declines in cognitive functioning are associated with poorer performance on tasks of daily living and with tasks involving short-term memory (Bakey et al., 1997). Moreover, depression has been frequently noted to have a significant impact on the older adult's level of social, physical, and psychological functioning (Reynolds ei al., 2001). Psychotic disorders may cause marked dysfunction in social and daily activities. Behavioral changes such as agitation, aggressive impulses, and social withdrawal frequently accompany cognitive and/or psychiatric disorders in older adults. Hence, the escalating demand for efifective geriatric mental health assessment and treatment supports the need for greater knowledge about the contribution of each domain (i.e., psychological, social, occupational) to the overall functioning level of older adults. Although several investigators have found that GAF ratings are closely associated with patient's symptoms (Brekke, 1992; Skodol et al, 1988), there have been a limited number of scientific investigations exploring the relation of psychological predictors, independent of social and occupational

Correspondence to: Professor Victor Molinari, PhD, ABPP, Department of Aging and Mental Health, Louis de la Parte Florida Mental Health Institute, University of South Florida, 13301 Bruce B. Downs Blvd., Tampa, Florida 33612-3899, USA.Tel: +1 (813) 974 1960. Fax: +1 (813) 974 1968. E-mail: vmolinari(^frnhi.usf.edu Received for publication 16th October 2003. Accepted 25th January 2004. ISSN 1360-7863 print/ISSN 1364-6915 online/04/06000505-0513 © Taylor & Francis Ltd DOI: 10.1080/13607860412331303784

506

J. Whitney et al.

functioning, to the GAF score within a geropsychiatric population. Mean GAF scores of very low functioning demented patients increased significantly between admission and discharge; concurrently, general psychiatric symptoms, depression, and agitation decreased significantly during the same time period (Bakey et a/., 1997). Hence, the GAF scores from admission to discharge reflected improvement in this particular group of patients' global functioning. A similar study vtath a geriatric inpatient population examined the treatment outcomes of a group of nursing residents and found that GAF scores significantly increased from admission to discharge, thus reflecting the patients' improvement in their global functioning. Decreased levels of general psychiatric symptoms, depression, and agitation were noted from admission to discharge (Kunik et al., 1996). Other studies have also noted predictable concurrent associations between the severity of patients' symptoms and clinician's ratings of global impairment (Roy-Byrne et al., 1996). Since the GAF was developed to provide a score which represented the 'psychological, social, and occupational functioning of individuals on a hypothetical continuum of mental health-tnental illness' (American Psychiatric Association, 1994, p. 32), it has been difficult to differentiate the strongest predictor of the GAF scale. With the GAF scale composed of three diverse domains (i.e., psychological, social, and occupational functioning), it is unlikely that each domain varies mutually in each geropsychiatric patient. Each domain must be independently evaluated to gain accurate information about the overall functioning level of each patient. TTiis study examined: (1) the potential contributions of several psychological predictors (i.e., cognitive status, depression, agitation, general psychiatric status) to the GAF (DSM-IV) scores among geriatric patients undergoing inpadent treatment; (2) the direct contributions of the exogenous variables (i.e., cognitive status, depression, agitation, general psychiatric status) to the endogenous variable (i.e., the GAF scale score) at admission and at discharge; and (3) the ability of intake GAF ratings to predict discharge GAF ratings independent of the psychological predictors. Clarifying what accounts for the variance of the GAF score is an imponant step toward understanding what the GAF measures and for providing a more accurate measurement of the overall functioning level of geropsychiatric inpatients. Methods This study used a secondary analysis of a pre-existing database to examine the contributions of cognitive status, agitation, depression, and general psychiatric

status to admission and discharge GAF scores obtained from an inpatient geropsychiatric sample. IRB approval was received by Baylor College of Medicine, the Houston Veterans Affairs Medical Center (HVAMC), and the University of Houston before data analysis proceeded on this pre-exisdng data set.

Sample The population included all admissions (using computerized medical records and tiie unit's database) to tlie HVAMC Geropsychiatry Unit between October 1993 and May 1995. A final sample from the above population, about 80% (H —301), was chosen based on the criteria that admission and discharge total scores for each predictor and criterion measure were available in the computerized database. The sample was predotninantly retired (9Q%), Caucasian (78%), males (98%) with a mean age of about 72 years. At least 68% had a high school education or less, and 26% attended or graduated from college. Five percent continued their education after the undergraduate level and 26 subjects had missing education data. Forty-three percent were married and 44% were divorced, separated or widowed. About 13% were single, less than 1% fell within the 'other' category, and seven subjects had missing marital data. Many of the inpatients either resided at home alone (28%) or needed parttime or fuU-dme supervision (43%). About 5% resided in a personal care home, 16% within an intermediate care facility, 1% within an skilled care facility, 5% were listed in the *other' category, and four subjects had missing data for living arrangements.

Procedure

Patients received a comprehensive evaluation by a multidisciplinary team that included two geriatric psychiatrists, a geropsychologist, a psychiatric nurse, a social worker, and a physician assistant. On admission and at discharge, cognitive status, general psychiatric status, and behavioral symptoms of each patient were rated using a standardized battery consisting of the Mini-Mental State Examination (MMSE) (Folstein et ai, 1975), the Hamilton Rating Scale for Depression (HDRS) (Hamilton, 1960), the Cohen-Mansfield Agitation Inventory (CMAI) (Cohen-Mansfield et al., 1989), and the Brief Psychiatric Rating Scale (BPRS) (Overall & Gorham, 1962) by the same attending physician. Admission and discharge GAF scores were also determined by the same team of mental health

P^chological predictors of GAF scores at admission and discharge professionals from psychiatry, psychology, and social work who had access to all prior test scores.

Measures Mini-Mental State Examination (MMSE) The MiniMental State Examination was first introduced by Folstein, Folstein, and McHugh (1975) and has become a standard tool for cognitive assessment in geriatric patients. The scores of the MMSE range from 0-30, with higher scores indicating higher cognitive functioning. Scores below 24 have been established as a cutoff point for cognitive dysfunction (Sherrell ei al., 1999). Hafnihon Rating Scale for Depression (HDRS) The Hamilton Rating Scale for Depression (Hamilton, 1960) is an observer rating scale for depression (Linden et al., 1995) that is frequendy used to assess older adults due to its brevity and psychometric qualities (Kunik et al., 1996). The Hamilton Rating Scale for Depression consists of 17 items rated on either a three- or five-point Likert-type scale with a score of 18 or higher indicating a major depression. Cohen-Mansfield Agitation Inventory (CAIAI) The Cohen-Mansfield Agitation Inventory was developed by Jiska Cohen-Mansfield (1986) to assess agitated behaviors in nursing home residents. Twenty-nine agitated behaviors are rated on a seven-point scale of frequency ('1' indicating the participant never engages in the specific agitated behavior, and '7' indicating the participant manifests the behavior on the average of several times per hour) and the scores are sumtned to yield a total score. Brief Psychiatric Rating Scale (BPRS) The Brief Psychiatric Rating Scale was developed by Overall and Gorham (1962) to assess severity of general psychiatric symptoms. Items on this 18-item measure are rated on seven-point scale anchors with ratings summed to produce a total score. A path analysis was used to test a model proposing causal paths from the manifest predictors (cognitive status, depression, agitation, and general psychiatric status) to the endogenous variable (the GAF scale scores). This model allows an examination of the contributions of the exogenous variables (cogtiitive status, depression, agitation, and general psychiatric status) to the endogenous (GAF scale score) variable at admission and at discharge.

Data analytic strategy USTREL 8.51 Qoreskog et al., 2001) was the statistical program used to systematically test the hypotheses and to insure that the model was an

507

accurate representation of the entire set of causal relations. Based on the variance/covariance matrix, this program used maximum likelihood estimation and provided multiple indices to evaluate how well the model fits the data with adequacy of fit determined by the convergence among a number of indices. Ha\Tng an acceptable fit of the model only confirmed that it is one of the several possible acceptable models that is representative of the data. Fit of the model was assessed using a variety of fit indices. The chi-square test was the index used to provide a comparison of the proposed model to the saturated model that fits the data perfectly. A nonsignificant chi-square test indicates that the model fits the data well. Because the chi-square (/") test is affected by sample size, the number of parameters estimated, non-normal distributions, and large degrees of freedom, the local fit indices such as the Root Mean Square Error of Approximation (RMSEA), the Normed Fit Index (NFI), NonNormed Fit Index (NNFI), and the Comparative Fit Index (CFI) were also examined. Results The summary statistics (i.e., means, standard deviations, and ranges) of the admission and discharge predictor measures and the criterion variable are displayed in Table 1. The means from admission to discharge indicated there was minimal change in cognitive status after treatment but significant changes in depression, agitation, psycliiatric status, and overall functioning level after treatment. Moreover, at discharge, most of the geropsychiatric patients experienced cognitive impairments with low levels of agitation and psychiatric symptoms and scored within the non-depressed range. Their overall functioning level refiected improvement from admission to discharge but most patients still scored within the 'serious impairment' range of functioning on the GAF. The bivariate correlations (of the four predictor measures and criterion variables) and the correlation matrix (Table 2) illustrated the pattern and strength of relations among the exogenous and endogenous constructs. Given the path model under study, a covariance-variance matrix was required for model estimation since it satisfied the assutnptions of die methodology and was the appropriate form of data for validating causal relations. Certain aspects of tlie correlation matrix deserve mention. Three of the four admission predictors, the MMSE, the CMAI, and the BPRS were correlated with admission GAF scores, whereas admission HDRS scores were not correlated with the admission GAF scores. All four discharge predictors, the MMSE, CMAI, HDRS, and the BPRS were correlated to the discharge GAF scores as expected.

508

J.

et al.

TABI£ 1. Means, standard deviations and ranges of predictor variables at admission and discharge Variable MMSE.ADM MMSE_DIS HDRS^DM HDRS_DIS CMAI_ADM CMAI_DIS BPRS.ADM BPRS_DIS ADM_GAF DIS_GAF

301 301 301 301 301 301 301 301 301 301

Mean

SD

Range

20.35 21.58 15.04 7.28* 46.98 38.14* 38.08 28.53* 27.68 43.66*

8.11 8.33 8.41 5.32 17.07 9.55 11.12 7.88 10.02 15.11

0-30 0-30 0-42 0-32 31-126 30-91 18-87 18-62 5-60 10-80

Abbreviations of the following measures: MMSE_ADM, Mini Mental State Examination Admission; MMSE_DIS, Mini Mental State Examinatiiin Discharge; HDRS_ADM, Hamilton Depression Rating Scale Admission; HDRS_DIS, Hamilton Depression Rating Scale Discharge; CMAI.ADM, Cohen-Mansfield Agitation Inventory Admission; CMAJ_DIS, Cohen^Manstield Agitation Inventory Scale; BPRS^ADM, Brief Psychiatric Rating Scale Admission; BPRS_DIS, Brief Psychiatric Rating Scale Discharge; ADM_GAF, Admission Consensus Conference GAF scale; DIS^GAF, Discharge Consensus Conference GAP scale. * Significant mean differences, p < 0.05, twotailed.

With respect to the interrelations of the predictor variables at each time point, it was expected that each predictor would be correlated with all other predictors. However, the admission HDRS was not correlated to the admission MMSE or the admission CMAI. At discharge, the HDRS was correlated with CMAI and BPRS scores but not correlated with MMSE scores. An examination of the predictor correlations across time points revealed that admission MMSE and admission CMAI were not correlated with the discharge HDRS. In addition, the admission HDRS was not correlated with the discharge BPRS. Overall, the HDRS was the predictor least correlated with other predictors across the two time points.

GAF (Time 1: paths from MMSB, HDRS, CMAI and BPRS to GAF; Time 2: paths from MMSE and BPRS to GAF) were significant and in the expected directions. TTie four paths that were included to improve the fit of the hypothesized model were also significant (paths from admission MMSE to discharge BPRS and CMAI, admission BPRS to discharge GAF, and admission HDRS to discharge CMAI). The paths from discharge CMAI and HDRS to the discharge GAF were not significant, contrary to expectation. The path from the admission GAF to the discharge GAF was also significant. At Time 1, two paths, HDRS to MMSE and HDRS to CMAI were not significant, which was unexpected but, at Time 2, all paths from each predictor to all other predictors were significant, as expected.

Model building Discussion The model-building phase included testing of six models (see Table 3) beginning with model 1, the saturated model. The next model (model 2) was the hypothesized model or path model in which these researchers hypothesized causal relations among variables. Models 3, 4, 5, and 6 were the re-estimated models, each tested to improve the fit of the hypothesized model to the sample data. Because the fit of model 6 fell within an acceptable range as per the goodness-of-fit statistics, it was quite reasonable to stop adding paths at that point. All the paths added during the model-building phase were significant and overall model fit had improved. In Fig. 1, the parameters of the path model (regression coefficients and covariances) are displayed. Regression coefficients are represented by single-headed arrows that indicate a hypothesized pathway between two variables. Covariances are represented with double-headed, curved arrows between two variables and indicate no directionality. The results were relatively expected. Six of the hypothesized eight paths from the predictors to the

This study was a secondary analysis of a pre-existing database to examine the possible contributions of several psychological predictors (i.e., cognitive status, depression, agitation, and psychiatric status) to the overall functioning level of a subgroup of geropsychiatric inpatients. Results generally supponed previous research concerning the impact of the most prevalent mental health conditions (i.e., cognitive status, depression, agitation, and general psychiatric impairment) on the overall functioning level of the older adult. Across both time points, six of the eight hypothesized paths from the psychological predictors to the GAF were significant. These findings are consistent witli the results of previous studies with mixed-aged samples (Hilsenroth et al., 2000; Moos, Nichols & Moos, 2002;), which have shovm moderately robust associations between symptoms and GAF ratings. As noted, patients' diagnoses and symptoms were strongly associated with highest level of functioning. Skodol, Link, Shrout and Horwath (1988) also

Psychological predictors of GAF scores at admission and discharge

d d 1 I

I

I

M -o o M

•o o o m n rn lA ^ o --> d d d d d

* in in —o d d I

* *