Psychosomatic Medicine Tips for Better Browsing
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by de Jonge, P.
Right arrow Articles by Latour, C. H. M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by de Jonge, P.
Right arrow Articles by Latour, C. H. M.
Related Collections
Right arrow Consultation Psychiatry
Right arrow Health Psychology
Right arrow Quality of Life
Psychosomatic Medicine 65:534-541 (2003)
© 2003 American Psychosomatic Society

ORIGINAL ARTICLES

Medical Inpatients at Risk of Extended Hospital Stay and Poor Discharge Health Status: Detection With COMPRI and INTERMED

Peter de Jonge, PhD, Iris Bauer, MSc, Frits J. Huyse, MD, PhD and Corine H. M. Latour, CNS, RN

From Department of Psychiatry (P.D.J., I.B., F.J.H., C.H.M.L.), Vrije Universiteit Medical Center, Amsterdam; and Department of Social Psychiatry (P.D.J.), University of Groningen, Groningen, The Netherlands.

Peter de Jonge, PhD, Department of Social Psychiatry, Hanzeplein 1, Gebouw 32, PO Box 30.001, 9700RB Groningen, The Netherlands. E-mail: p.de.Jonge{at}med.rug.nl

Received for publication July 1, 2002; revision received December 9, 2002.


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 COMPRI-INTERMED
 METHODS
 RESULTS
 DISCUSSION
 ACKNOWLEDGMENTS
 REFERENCES
 
OBJECTIVE: To detect the patients in medical wards at risk of extended LOS and poor discharge health status with the use of complexity prediction instrument (COMPRI) and interdisciplinary medicine (INTERMED) instruments.

METHODS: Study 1: In a sample of 275 consecutively admitted medical inpatients, a hierarchical cluster analysis on INTERMED variables was performed. The clusters were compared on length of hospital stay (LOS) and Short Form 36 (SF-36) at discharge. Study 2: Receiver operating characteristic (ROC) analysis was used to optimal cut-off points for the COMPRI and INTERMED. Patients detected with COMPRI and INTERMED were then compared with undetected patients on LOS and SF-36.

RESULTS: Study 1: In concordance with previous findings, a cluster of patients with high biopsychosocial vulnerability was identified with significantly higher scores on LOS (p < .05) and lower scores on SF-36 (p < .001) than patients in other clusters. Study 2: A cut-off point for the COMPRI of 5/6 was found to detect patients at risk of long LOS. A cut off score for the INTERMED of 20/21 was found to detect patients at risk of poor discharge health status. Patients detected with COMPRI and INTERMED had a significantly longer LOS (p < .001) and a poorer discharge health status (SF-36 MCS: p < .001; SF-36 PCS: p = .05) than nondetected patients. Of the detected patients, 37% had an extended hospital stay and poor discharge health status; of the nondetected patients, this was only 7%.

CONCLUSIONS: The COMPRI-INTERMED can help to detect complex patients admitted to medical wards within the first days of admission, and rule out those with a small chance of poor outcomes.

Key Words: internal medicine, • psychiatric, • COMPRI, • INTERMED, • SF-36, • length of stay, • health status, • quality of life.

Abbreviations: COMPRI = Complexity prediction instrument;; INTERMED = interdisciplinary medicine;; ROC = receiver operating characteristic;; AUC = area under the curve;; LOS = length of hospital stay;; SF-36 = Short Form 36;; SPSS = Statistical Package for the Social Sciences;; PCS = SF-36 Physical Health Component Score;; MCS = SF-36 Mental Health Component Score


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
COMPRI-INTERMED
 METHODS
 RESULTS
 DISCUSSION
 ACKNOWLEDGMENTS
 REFERENCES
 
Increasingly, patients with interacting chronic somatic diseases, psychiatric co-morbidity, and limitations in physical functioning complicate general hospital care. Several intervention studies have extended the attention beyond the medical problem for which the patient was admitted, and have assessed the effects of discharge planning and follow-up (1), offering counseling and psychoeducation to the patient (2, 3), or treating psychiatric co-morbidity during hospital admission (4–7).

The integration of several of these interventions by a specially trained nurse has been referred to as case management (8–12). The major components of case management are: assessment of care needs, development of integral treatment plans, improving access to (psychosocial) care and monitoring quality of health care delivery (13). Intervention studies aimed at the effects of case management in patients with somatic diseases have been conducted in specific populations of chronic ill patients with high health care utilization, such as patients with chronic heart failure and diabetes.

In studies with chronic congestive heart failure patients (14–16), significant effects of case management were found, such as reducing mortality and the chance of readmissions. Improved outcomes have also been reported in diabetes care (19). For general medical inpatients, however, the effects of case management are unknown. Small effects of standard interdisciplinary rounds on LOS have been reported (20), but standard application to all medical inpatients seems inefficient and might prevent a wide acceptance. Instead, a screening procedure is needed to first detect patients who might benefit from such an intensive intervention; preferably, a combination of a quick screener for the detection of complex patients and an assessment instrument for the content of case complexity. Important in the evaluation of the screening procedure is its negative predictive power–the proportion of undetected patients that indeed do not develop complex care needs–as especially the proportion of "missed" patients should be as low as possible.


   COMPRI-INTERMED
TOP
 ABSTRACT
 INTRODUCTION
 COMPRI-INTERMED
METHODS
 RESULTS
 DISCUSSION
 ACKNOWLEDGMENTS
 REFERENCES
 
The COMPRI-INTERMED strategy consists of a screening instrument to rule out the group of patients not at risk for increased care utilization or longer LOS and a method to assess care needs and plan case management (21). The COMPRI is an instrument to predict LOS and care complexity of internal medical inpatients and was developed on data from 11 European medical wards (22–24). The INTERMED is a method to arrange information derived from a structured patient interview leading to a prognosis of health care needs. Information is scored with respect to biological, psychological, social, and health care aspects of disease (25) (see METHODS).

Among general medical inpatients, the INTERMED distinguished between three groups of patients with different care needs: standard, chronic, and complex patients (26). This last group, consisting of 21% of the patients, had a high biopsychosocial risk profile, as indicated by high scores on all four INTERMED domains. In addition, these patients had a longer LOS and received more medications, nurse interventions, and specialist consultations. Within the population of medical inpatients, this cluster seems relevant to detect early at admission to formulate tailored interventions.

In the present study we first tried to replicate these findings, and to demonstrate that a subset of patients with high biopsychosocial vulnerability can be identified with the INTERMED. Because cluster analysis enables only a post hoc detection of patients, ie, clusters can only be identified after completing data collection of the whole ward, the second goal was to formulate a prospective strategy to detect patients with a long LOS and a poor health status by means of COMPRI and INTERMED through the development of optimal cut-off points. For this, we related COMPRI to LOS in order to be able to rule out patients with a short LOS. We related INTERMED to health status at discharge in order to detect patients at risk of poor outcomes. After specifying the cut-off scores of both instruments, we assessed the predictive value of the COMPRI-INTERMED strategy to detect patients at risk of extended hospital stay and poor discharge health status.


   METHODS
TOP
 ABSTRACT
 INTRODUCTION
 COMPRI-INTERMED
 METHODS
RESULTS
 DISCUSSION
 ACKNOWLEDGMENTS
 REFERENCES
 
Design
A patient cohort was screened at admission and followed until discharge from the hospital. The study was part of an intervention study assessing the effects of integral medicine on length of stay and patient satisfaction, which was a joint project of the two departments of general internal medicine (42 beds) and the psychiatric consultation-liaison service of the Vrije Universiteit (VU) Medical Center of Amsterdam. The present study is based on data from the baseline period of the study.

Procedure
Patients were asked to participate in the study by a research nurse in the first 3 days of admission by means of a written informed consent procedure, as approved by the medical ethics committee of the VU Medical Center. After inclusion, the research nurse reviewed the medical and nurse chart, completed the COMPRI and conducted the patient interview (15–20 minutes) to score the INTERMED. If the patient could not be interviewed, the patient’s visiting family was interviewed. Two medical students daily checked the planned patient discharges from the ward, and handed over the SF-36 to the patient. If a patient was missed, the questionnaire was sent to the patient’s home, followed by a telephone call within 1 week after discharge. After a patient’s discharge, the medical status was reviewed.

Patients
From February through September 2000, all patients admitted at the two general internal medicine wards (general internal medicine, gastroenterology and nephrology) were considered for inclusion in the study. Patients admitted for one day, patients readmitted within the study period, and patients treated by the specialty dermatology or rheumatology (with beds on the same ward) were excluded. Sixteen patients who died during admission were retrospectively removed from the sample. Figure 1 shows the inclusion of the 275 patients originating from the two wards.



View larger version (19K):
[in this window]
[in a new window]
  		Fig. 1. Patient flow chart.

 
Variables
COMPRI
The COMPRI consists of 13 items (yes or no) of which four items to be rated by the doctor and three by the nurse. For this study, we therefore added those items to the medical and nurse chart, respectively. The research nurse, employed by the department of psychiatry, rated the six remaining items. Predictions made by the doctor: 1) Do you expect this patient to have a hospital stay of at least two weeks? 2) Do you think the organization of care during hospital stay will be complex? 3) Do you expect that this patient’s mental health will be disturbed during hospital stay? 4) Is the patient known to have a currently active malignancy? Predictions made by the nurse: 5) Do you expect this patient to have a hospital stay of two weeks or more? 6) Do you think the organization of care during hospital stay will be complex? 7) Do you think this patient will be limited in activities of daily living after discharge? Additional questions scored by the research nurse: 8) Did the patient have a negative health perception during the last week? 9) Did the patient have walking difficulties during the last three months? 10) Did the patient have more than six doctor visits during the past three months? 11) Did the patient take more than three different kinds of medication the day before admission? 12) Is this an unplanned admission? 13) Is the patient retired?

The list of items was derived from an extensive list of 117 potential risk factors for hospital-based health care utilization originally covering: admission status, subjective clinical predictions by doctor and by nurse, case complexity/severity of illness, living/working situation, stress/social support, activities of daily living functioning, health perception/worrying, relation with doctors, past health care utilization, drug abuse, compliance, and emotional state. In a prospective study of 2,158 patients from 10 hospitals in 7 European countries, the items that were most predictive of LOS and a series of other indicators for hospital-based care utilization were selected (22, 23). A simple scoring procedure was developed to enhance clinical use, including a reduction of answering categories and weighing of individual items (24). Items 1 to 3 and 5 to 7 are given a weight of 2 for every positively rating; the remaining items are given a weight of 1 for every positive rating. The scores are summed up, resulting in a potential score range of 0 to 19. We found that the admission COMPRI score was correlated to a series of outcomes at discharge (eg, LOS: 0.47; number of medications during hospital stay: 0.49; complexity rating by doctor: 0.46; complexity rating by nurse: 0.49).

INTERMED
The INTERMED consists of a grid with four domains: biological, psychological, social, and health care. Of each of the four domains five variables are rated 0 to 3 according to a manual with clinical anchor points, resulting in a potential score range of 0 to 60. Scoring is based on a patient interview and a review of the medical chart. The following variables are scored: (1) chronicity, (2) past diagnostic uncertainty, (3) severity of illness, (4) current diagnostic uncertainty, (5) complications and life threat, (6) restrictions in coping, (7) past psychiatric dysfunction, (8) resistance to treatment, (9) severity of psychiatric symptoms, (10) mental health threat, (11) restrictions in social intergration, (12) social dysfunction, (13) residential instability, (14) restrictions in social network, (15) social vulnerability, (16) intensity of prior treatment, (17) prior treatment experience, (18) organizational complexity, (19) appropriateness of admission or referral, and (20) biopsychosocial care needs.

Elsewhere, we reported on the development, reliability, validity, and applications of the INTERMED (25–35). The INTERMED was developed in a group of general hospital psychiatrists (25) and first tested on face validity. We then assessed its content and criterion validity in patients admitted to internal medicine (26), patients with low back pain (27, 28), terminal cancer (29), diabetes (30), rheumatoid arthritis (31), and multiple sclerosis (32). On the basis of a cut-off score of 20/21, we found good interrater reliability between two raters, as indicated by a {kappa} of 0.85 (33). In patients with multiple sclerosis, we found a good test-retest reliability (r = 0.75; {kappa} = 0.60) with a period of one year between two ratings (34). Internal consistency estimates (Cronbach’s {alpha}) range between 0.78 and 0.94 in several samples of patients with somatic illness (35).

SF-36
Patients rated their health status at discharge by means of SF-36. The SF-36 consists of 36 items, organized into eight scales (physical functioning, social functioning, role limitations–physical, pain, mental health, role limitations–emotional, vitality, and general health) (36, 37). The number of response choices per item ranges from two to six. When scores on one or two of the scales were missing, the median score on that scale was used for extrapolation. The scales were recoded into standardized scores, subsequently used to construct two summary scores: a PCS and a MCS, based on the factors found by Hays and Stewart (38) with a scoring range between 0 and 100 (100 = optimal functioning). We used the Dutch version of the SF-36, which has been developed and validated in the International Quality of Life Assessment Project (39).

Additional Variables
Socio-demographic (age, sex, marital status, and living situation) and medical data (past hospital admissions, medications, and admission diagnosis) were scored from the medical chart. In order to ensure reliability of these data, two medical students scored the variables independent of each other. In case of discrepancy, data were scored after a discussion with the first author.

Statistical Analyses
Study 1
Hierarchical cluster analysis was performed to identify similar groups with relatively similar INTERMED scores. Hierarchical cluster analysis is a nonparametric method to identify similar groups of patients, based on their interindividual Euclidean distances on all specified variables (40). We used Ward’s method to form clusters, as included in SPSS 7.5. Clusters were compared on INTERMED scores for interpretation of the resulting clusters, and subsequently on COMPRI scores, LOS, PCS, and MCS. Differences were tested by means of t tests for normally distributed data and Mann-Whitney U-tests for nonnormal data ({alpha} = 0.05).

Study 2
Two groups of patients were formed based on median LOS, median MCS, and median PCS. Sensitivity and specificity were calculated for the COMPRI to detect long LOS, and for the INTERMED to detect poor MCS and PCS. Optimal cut-off points for the COMPRI and the INTERMED in relation to outcomes were based on a ROC analysis (41).

ROC analysis is based on a graph, a ROC curve, in which the sensitivity of a screener in relation to an outcome is plotted against 1 specificity for all potential cut-off points of the screener, normally resulting in a curved line. One can test whether the AUC significantly differs from a straight line (diagonal with AUC = 0.5), which describes the case when the screener is unrelated to the outcome. Also, in the curved line, one can find suggestions for cut-off scores of the screener, ie, the points that are far away from the diagonal line. In our analyses, we optimized sensitivity and specificity simultaneously, ie, the highest product of sensitivity and specificity was considered the optimal cut-off point for the screener.

Patients with positive scores on both COMPRI and INTERMED were compared with the remaining sample on admission data, past admissions, diagnosis, medication use, LOS, PCS, and MCS. To test statistical significance, {chi}2 tests for categorical data, t tests for normal distributed data, and Mann-Whitney U tests for nonnormal data were applied ({alpha} = 0.05). We calculated sensitivity, specificity, and positive and negative predictive power to assess to proportions or correctly and incorrectly detected patients based on their outcomes (LOS >= 8 days and MCS >= 62 and PCS <= 41).

A multivariate analysis was conducted to study the independent effect of a positive COMPRI-INTERMED score on LOS, with sex, age (>=65), independent living, marital status, being admitted from home, and the number of hospital admissions in the past 5 years (>=4) as confounders. As a dependent variable, a binary variable based on the 90th percentile of LOS (>31 days) was used.

To study the concordance between the post hoc detection of complex patients based on the cluster analysis and the prospective detection based on the cut-off scores of COMPRI and INTERMED, we calculated the {kappa}-statistic (42, 43).


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 COMPRI-INTERMED
 METHODS
 RESULTS
DISCUSSION
 ACKNOWLEDGMENTS
 REFERENCES
 
Sample Characteristics
The study sample consisted of 275 patients with a mean age of 59 years and an equal sex distribution. Most of the included patients were living independently; they had a median LOS of 8 days (mean: 14.5).

We were able to score the INTERMED based on a patient interview within the first three days of admission in 84% of the sample. Of the remaining patients, scoring was based on a review of the medical chart and an interview with a visiting family member. Complete SF-36 data at discharge were obtained from 208 patients (76%). Patients with missing SF-36 data had significantly higher COMPRI (p = .03) and INTERMED (p = .03) scores than patients with SF-36 data, but did not differ on age, sex, LOS, marital status, and living situation.Go


View this table:
[in this window]
[in a new window]
  		TABLE 1. Sample characteristics
 
Study 1
Hierarchical cluster analysis of the INTERMED scores resulted in three clusters: 1) 42 (15%) patients with relatively low scores on all domains, 2) 128 (47%) patients with low scores on the psychological and social domain and high scores on the biological and health care domain, and 3) 104 (38%) patients with high scores on all domains. The clusters were interpreted as 1) low biopsychosocial complexity (standard), 2) uncomplicated chronic somatic illness (chronic), and 3) high biopsychosocial complexity (complex), in correspondence with earlier results (29). Patients in the complex cluster had higher INTERMED (p < .001) and COMPRI (p < .001) scores, longer LOS (p < .05) and poorer MCS (p < .001) compared with the two other clusters (see Table 2). No significant difference between the complex cluster and the other two clusters was found with respect to PCS.


View this table:
[in this window]
[in a new window]
  		TABLE 2. Comparison between INTERMED clusters
 
Study 2
A significantly positive correlation was found between COMPRI and the natural logarithmic transformation of LOS (r = .50; p < .001). The ROC curve of the COMPRI to predict LOS >=8 days are shown in Figure 2.



View larger version (17K):
[in this window]
[in a new window]
  		Fig. 2. ROC curve of COMPRI to predict LOS (>=8 days). *AUC = .73 (95% CI: 0.67–0.79;p < .01).

 
Three COMPRI cut-off points are suggested in the ROC: 3/4, 5/6, and 9/10. An optimal combination of sensitivity and specificity was found at the cut-off point of 5/6, resulting in 104 true positives, 45 false positives, 43 false negatives and 76 true negatives (sensitivity 0.71 and specificity 0.63, positive predictive value = 0.70, negative predictive value = 0.64).

Significantly negative correlations were found between INTERMED and MCS (r = -.36; p < .001) and PCS (r = -.19; p < .01). The correlations were comparable for patients of whom scoring was based on the patient interview (MCS: -.34; PCS: -.19) and those of whom scoring was based on the medical chart and a family member interview (MCS: -.53; PCS: -.18). The optimal cut-off point of the INTERMED to detect patients with MCS <=62 was found to be at 20/21, resulting in 62 true positives, 29 false positives, 42 false negatives, and 75 true negatives (sensitivity = .60, specificity = .72, positive predictive value = .68, and negative predictive value = .64) (Figure 3).



View larger version (16K):
[in this window]
[in a new window]
  		Fig. 3. ROC curve of INTERMED to predict SF-36 MCS (<=62). *AUC = .68 (95% CI: 0.60–0.75; p < .01).

 
Two cut-off points of the INTERMED to detect patients with PCS <=41 were suggested by the ROC: 13/14 and 20/21 (Figure 4). The optimal combination of sensitivity and specificity was found at 20/21, resulting in 55 true positives, 36 false positives, 50 false negatives, and 67 true negatives (sensitivity = 0.52, specificity = 0.65, positive predictive value = 0.60, negative predictive value = 0.58).



View larger version (16K):
[in this window]
[in a new window]
  		Fig. 4. ROC curve of INTERMED to predict SF-36 PCS (<=41). *AUC = .60 (95% CI = 0.52–0.67;p = .016)

 
With the use of COMPRI 5/6 and INTERMED 20/21 as cut-off points, a distinction can be made between "detected" patients with both high COMPRI and INTERMED scores (N = 99; 36%), and the remaining "undetected" patients (N = 176; 64%). Comparing the patient groups, significant differences were found on LOS, MCS, and PCS. The positively screened patients were also older, more often dependent on others, used more medications, and had more previous admissions. Significant differences occurred on six of the 8 SF-36 subscales; no differences were found on "role physical" and "pain" (data not shown).

We defined a group of patients with poor outcomes on LOS (>=8 days), MCS (>=62), and PCS (>=41), which comprised of 35 patients (17% of the sample). With the detection procedure of COMPRI 5/6 and INTERMED 20/21, we correctly detected 23 patients (true positives), incorrectly detected 39 patients (false positives), correctly not detected 130 patients (true negatives), and missed 11 patients (false negatives), equal to sensitivity of 0.68, specificity of 0.77, positive predictive value of 0.37, and negative predictive value of 0.93.

The multivariate prediction of LOS (<=31 days vs. >31 days) resulted in a significant independent effect of a positive COMPRI-INTERMED score (OR = 4.18; 95% CI: 1.54–11.30; p = .005), after controlling for sex, age (>=65), independent living, marital status, being admitted from home, and number of hospital admissions in the past 5 years (>=4). In addition, independent living (OR = 0.07; 95% CI: 0.008–5.26; p = .010) and age (>=65) (OR = 0.29; 95% CI: 0.11–0.75; p = .037) were independently related to long LOS.

Finally, relating the prospective screening to the INTERMED patient clusters, virtually all (40/42) "standard" patients and most "chronic" patients (112/127) would have remained undetected. In contrast, most of the "complex" patients (80/104) would have been identified. The correspondence between prospectively detecting the "complex" patients and not detecting the "standard" and "chronic" patients can also be expressed in the {kappa} statistic ({kappa} = 0.66), indicating good agreement (38).


   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 COMPRI-INTERMED
 METHODS
 RESULTS
 DISCUSSION
ACKNOWLEDGMENTS
 REFERENCES
 
The goal of this study was to formulate a standardized detection strategy for patients with a high LOS and poor health status based on a two-step procedure with COMPRI and INTERMED. In this strategy, the COMPRI is used as a case-finder: patients scoring above the screening threshold are considered at risk of an extended hospital stay. Consequently, they are interviewed with the INTERMED, a method to integrate biological, psychological, social and health care aspects of a patient, based on which tailored case-management can be initiated.

To determine the optimal cut-off points for the COMPRI and the INTERMED to standardize treatment allocation, we related the COMPRI to LOS and the INTERMED to discharge health status (SF-36). First, we replicated a previous finding of three INTERMED clusters of medical inpatients. A subgroup of 38% of patients was identified with a high biopsychosocial vulnerability, which was also reflected in a long hospital stay and a poor discharge health status. This group of patients is highly comparable to the cluster of complex patients we reported on before (26) both in terms of INTERMED scores and in outcomes. In the current study we found that these patients also had high COMPRI scores, suggesting that some patients may already be detected by the less time-consuming screening by the COMPRI.

The COMPRI was developed as a quick screening instrument to detect medical patients at risk of considerable diagnostic and treatment efforts and subsequently an extended hospital stay (22–24). In the current prospective study, the COMPRI correlated strongly with LOS, indicating that 25% of the variance of LOS can be explained by a simple questionnaire that can be scored within the first days of a patient’s admission. This finding provides prospective support for its validity. Similarly, the INTERMED was predictive of discharge health status, which is also comparable to previous reports in other patient populations: low back pain patients (27) and end stage renal disease patients (44).

To date, no information was available on how COMPRI and INTERMED could be used in a single model to detect patients at risk of needing additional care during hospital stay. The data presented in this study suggest that the COMPRI cut-off of 5/6 can identify a subgroup of about one-half of the patients admitted to a general medical ward. After this initial screening, the more time-consuming screening with the INTERMED cut-off of 20/21 can rule out another one third of the sample, of which most patients can be considered as low complex or standard chronic patients. We found that this way the amount of patients can be reduced to about one-third to whom additional care should be considered. This group of patients differed clearly from the remaining patients as they had a longer hospital stay and a poorer discharge health status. These detected patients were also more frequently dependent on external help at home, had more frequently past hospital admissions, and would receive more medications during the current admission episode. The findings of this study therefore provide evidence of the utility of the COMPRI and INTERMED for detecting internal medical patients in need of case-management.

Most clearly, this is seen when comparing patients who would have been detected to those that remained undetected. Of the undetected patients, only 7% had a poor outcome, defined as a hospital stay of >1 week and a discharge status lower than the median of the sample. In contrast, 37% of the detected patients had poor outcomes, a >5 times increased risk. It is difficult to say whether these findings are convincing enough for implementing the COMPRI-INTERMED procedure into clinical practice. We believe, however, that it is. With this procedure, which may take 1 hour in total, a trained nurse can rule out those patients who have only a small chance of developing poor outcomes during hospital stay. This may save a lot of energy, which may be dedicated to those at risk: of the detected patients, one-third to one-half will have poor outcomes, which may prove to be preventable in some.

As a limitation of the study, the considerable proportion of patients with missing SF-36 questionnaires at discharge should be mentioned. Due to organizational problems, mainly a shortage of manpower on the one side and the short notice of patient discharges on the other sides, many patients left the hospital before the questionnaire could be handed over. Because the drop out is selected as patients who were missed, had a longer LOS and higher COMPRI and INTERMED scores, it might have affected the current results. We expect however that this may be in the direction of an under-estimation of the relations between the admission screening and the outcomes.

In the study, we related INTERMED scores to health status at discharge for the total group of patients. Because our goal was to develop a stepped detection in which the COMPRI precedes the INTERMED, it would also have been an option to relate INTERMED to SF-36 only in the COMPRI-detected patients. The disadvantage of this procedure was that due to the loss of patients, these analyses would concern only a minority of patients included in the sample, ie, about one-half of the patients that we did now, which may not have been sufficient to achieve a stable ROC curve. Preliminary results of such an analysis indicate a somewhat less strong relation between INTERMED and SF-36 scores, possibly due to restriction in range, but because the optimal cut-off point in this sub sample was also found to be 20/21, our conclusions did not alter.

Apart from detecting patients at risk, the INTERMED can support clinical decision making with respect to multidisciplinary treatment. The variables included in the instrument can help the clinician, such as a clinical nurse specialist, who is added to the medical ward for this goal, to think of a series of specific risk factors that may complicate treatment, some of which necessitate a consult from medical or paramedical specialist. Perhaps, the operationalization of case management may further develop this way from a "black box" definition.

Timely detection of patients at risk of a long hospital stay and a poor discharge health status is important for improving care directed at patient needs. This has been an issue in consultation-liaison (C-L) psychiatric research for many years. The development of psychiatric screening instruments adapted to the medical setting is needed to enhance the quality of C-L psychiatric health service research, because the lack of an empirical-based valid and reliable case finder prevents research demonstrating the effectiveness of liaison interventions (45). Our study presents such a case finding procedure. The effectiveness of timely detection and treatment of patients by a nurse specialist, who uses the INTERMED to direct interdisciplinary care, is evaluated elsewhere (46).Go


View this table:
[in this window]
[in a new window]
  		TABLE 3. Comparison of positive and negative screened patients*
 

   ACKNOWLEDGMENTS
TOP
 ABSTRACT
 INTRODUCTION
 COMPRI-INTERMED
 METHODS
 RESULTS
 DISCUSSION
 ACKNOWLEDGMENTS
REFERENCES
 
This study was supported by Dutch Association of Academic Hospitals/Dutch College of Health Insurers Grant 99038: the effect of integral medicine on length of stay and patient satisfaction (F.J. Huyse). The authors also wish to express their gratitude to Liesbeth van Gemert, clinical nurse specialist, for data collection, and the staff of the internal medicine ward—in particular Diny Lanser and Theo Brouwer, head nurses, and Coen van Guldener and Roos Peerenboom, internists—for help in implementing the study.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 COMPRI-INTERMED
 METHODS
 RESULTS
 DISCUSSION
 ACKNOWLEDGMENTS
 REFERENCES
 

  1. Boone CR, Coulton CJ, Keller SM. The impact of early and comprehensive social work services on length of stay. Soc Work Health Care 1981; 7: 1–9.
  2. Mumford E, Schlesinger HJ, Glass GV. The effects of psychological intervention on recovery from surgery and heart attacks: an analysis of the literature. Am J Public Health 1982; 72: 141–51.[Abstract/Free Full Text]
  3. Devine EC, Cook TD. A meta-analytic analysis of effects of psycho-educational interventions on length of postsurgical hospital stay. Nursing Res 1983; 32: 267–74.
  4. Hengeveld MW, Ancion FAJM, Rooijmans HGM. Psychiatric consultations with depressed medical inpatients: a randomised controlled cost-effectiveness study. Int J Psych Med 1988; 18: 33–43.[Medline]
  5. Strain JJ, Lyons JS, Hammer JS, Fahs M, Levobits A, Paddison PL, Snyder S, Strauss E, Burton R, Nuber G, Abernathy T, Sacks H, Nordlei J, Sack C. Cost offset from a psychiatric consultation-liaison intervention with elderly hip fracture patients. Am J Psychiatry 1991; 148 (8): 1044–49.[Abstract/Free Full Text]
  6. Levenson JL, Hamer RM, Rossiter LF. A randomised controlled study of psychiatric consultation guided by screening in general medical inpatients. Am J Psych 1992; 149 (5): 631–37.[Abstract/Free Full Text]
  7. Gater RA, Goldberg DP, Evanson JM, Lowson K, McGrath G, Tantam D, Millom L. Detection and treatment of psychiatric illness in a general medical ward: a modified cost-benefit analysis. J Psychosom Res 1998; 45 (5): 437–48.[CrossRef][Medline]
  8. Tracey F, Taylor IC, McConnell JG. A prospective study of the process of assessment and care management in the discharge of elderly patients from hospital. Ulster Med J 1998; 67 (1): 36–40.[Medline]
  9. Gibbs B, Lonowski L, Meyer PJ, Newlin PJ. The role of the clinical nurse specialist and the nurse manager in case management. JONA 1995; 25 (5): 28–34.
  10. Schillinger D, Bibbins-Domingo K, Vranizan K, Bacchetti P, Luce JM, Bindman AB. Effects of primary care coordination on public hospital patients. J Gen Int Med Vol 2000; 15: 329–36.
  11. Edelstein EL, Cesta TG. Nursing case management: an innovative model of care for hospitalised patients with diabetes. Diabetes educator 1993; 19 (6): 517–21.
  12. Tennier LD. Discharge planning: an examination of the perceptions and recommendations for improved discharge planning at the Montreal General Hospital. Soc Work Health Care 1997; 26 (1): 41–60.[CrossRef][Medline]
  13. Holloway F, Oliver N, Collins E, Carson J. Case management: a critical review of the outcome literature. Europe Psychiatry 1995; 10: 113–28.
  14. Stewart S, Marley JE, Horowitz JD. Effects of a multidisciplinary, home-based intervention on unplanned readmissions and survival among patients with chronic congestive heart failure: a randomised controlled study. Lancet 1999; 354: 1077–83.[CrossRef][Medline]
  15. Rich MW, Beckham V, Wittenberg C, Leven CL, Freedland KE, Carney RM. A multidisciplinary intervention to prevent the readmission of elderly patients with congestive heart failure. N Engl J Med 1995; 333 (18): 1190–5.[Abstract/Free Full Text]
  16. Cline CMJ, Israelsson BYA, Willenheimer RB, Broms K, Erhardt LR. Cost effective management programme for heart failure reduces hospitalisation. Heart 1998; 80: 442–46.[Abstract/Free Full Text]
  17. Harris LE, Luft FC, Rudy DW, Kesterson JG, Tierney WM. Effects of multidisciplinary case management in patients with chronic renal insufficiency. Am J Med 1998; 105: 464–70.[CrossRef][Medline]
  18. Beder J. Evaluation research on the effectiveness of social work intervention on dialysis patients: the first three months. Soc Work Heath Care 1999; 30 (1): 15–30.
  19. Aubert RE, Herman WH, Waters J, Moore W, Sutton D, Peterson BC, Bailey CM, Kaplan JP. Nurse case management to improve glycemic control in diabetic patients in a health maintenance organisation. A randomised, controlled trial. Ann Intern Med Vol 1998; 129: 605–12.[Abstract/Free Full Text]
  20. Curley C, McEachern JE, Speroff T. A firm trial of interdisciplinary rounds on the inpatient medical wards. An intervention designed using continuous quality improvement. Med Care 1998; 36 (8): AS4–AS12.[CrossRef][Medline]
  21. Huyse FJ, Herzog T, Lobo A, Lyons JS, Slaets JPJ, Fink P, Stiefel F, De Jonge P. Detection and treatment of mental disorders in general health care. Europe Psych 1997; 12 (Suppl 2): 70S–80S.[CrossRef]
  22. de Jonge P, Huyse FJ, Slaets JPJ, Herzog T, Lobo A, Lyons JS, Opmeer BC, Stein B, Arolt V, Balogn N, Cardoso G, Fink P, Rigatelli M, van Dijck R, Mellenbergh GJ. Care complexity in the general hospital. Results from a European study. Psychosomatics 2001; 42: 204–212.[Abstract/Free Full Text]
  23. de Jonge P, Huyse FJ, Herzog T, Lobo A, Slaets JP, Lyons JS, Opmeer BC, Stein B, Arolt V, Balogn N, Cardoso G, Fink P, Rigatelli M. Risk factors for complex care needs in general medical inpatients. Results from a European study. Psychosomatics 2001; 42: 213–221.[Abstract/Free Full Text]
  24. Huyse FJ, de Jonge P, Slaets JPJ, Herzog T, Lobo A, Lyons JS. COMPRI–an instrument to detect patients with complex care needs. Results from a European study. Psychosomatics 2001; 42: 222–228.[Abstract/Free Full Text]
  25. Huyse FJ, Lyons JS, Stiefel FC, Slaets JP, de Jonge P, Fink P, Gans RO, Guex P, Herzog T, Lobo A, Smith GC, van Schijndel RS. INTERMED: a method to assess health service needs: I. Development and first results on its reliability. General Hosp Psychiatry 1999; 21: 39–48.
  26. de Jonge P, Huyse FJ, Stiefel FC, Slaets JPJ, Gans ROB. INTERMED-a clinical instrument for biopsychosocial assessment. Psychosomatics 2001; 42 (2): 106–109.[Abstract/Free Full Text]
  27. Stiefel FC, de Jonge P, Huyse FJ, Guex P, Slaets JP, Lyons JS, Spagnoli J, Vanotti M. "INTERMED": a method to assess health service needs. II. Results on its validity and clinical use. Gen Hosp Psychiatry 1999; 21: 49–56.[CrossRef][Medline]
  28. Stiefel FC, de Jonge P, Huyse FJ, Slaets PJ, Guex P. INTERMED–an assessment and classification system for case complexity. Results in patients with low back pain. Spine 1999b; 24 (4): 378–85.[CrossRef][Medline]
  29. Mazzocato C, Stiefel F, de Jonge P, Levorato A, Ducret S, Huyse FJ. Comprehensive assessment of patients in palliative care: a descriptive study utilising the INTERMED. J Pain Symptom Manage 2000; 19 (2): 83–90.[CrossRef][Medline]
  30. Fisscher CJ, Stiefel FC, de Jonge P. Case complexity and clinical outcome in diabetes mellitus: a prospective study using the INTERMED. Diabetes Metab 2000; 26 (4): 295–302.[Medline]
  31. Koch N, Stiefel F, de Jonge P, Franssen J, Chamot A, Gerster JC, Huyse F, So AK. Identification of complex patients with rheumatoid arthritis and increased health care utilisation. Arthritis Rheum 2001; 45 (3): 216–21.[CrossRef][Medline]
  32. de Jonge P, Huyse FJ, Ruinemans GMF, Stiefel FC, Lyons JS, Slaets JPJ. Timing of psychiatric consultations. The impact of social vulnerability and level of psychiatric dysfunction. Psychosomatics 2000; 41 (6): 505–11.[Abstract/Free Full Text]
  33. de Jonge P, Latour C, Huyse FJ. Interrater reliability of the INTERMED in a heterogeneous somatic population. J Psychosom Res 2002; 52: 25–7.[CrossRef][Medline]
  34. Deleted in proof.
  35. de Jonge P, and Stiefel FC. Internal consistency of the INTERMED in patients with somatic diseased. J Psychosom Res 2003; 54 (5): 497–499.[CrossRef][Medline]
  36. Brazier JE, Harper R, Jones NMB, O’Cathain A, Thomas KJ, Usherwood T, Westlake L. Validating the SF-36 health survey questionnaire: new outcome for primary care. Br Med J 1992; 305: 160–4.
  37. Sherbourne CD, Mazel M. The RAND 36-item health survey 1.0. Health Econ 1993; 2: 217–27.[Medline]
  38. Hays RD and Stewart AL. The structure of self-reported health in chronic disease patients. Psychological assessment. J Consult Clin Psych 1990; 2: 22–30.
  39. Essink-Bot ML, Krabbe PFM, Bonsel GJ, Aaronson NK. An empirical comparison of four generic health status measures. Med Care 1979; 35 (5): 522–37.
  40. Aldenderfer MS, Blashfield RK, editors. Cluster analysis. In: Quantitative applications in the social sciences. Beverly Hills CA: Sage, 1984.
  41. Swets A. Measuring the accuracy of diagnostic systems. Science 1988; 240: 1285–93.[Abstract/Free Full Text]
  42. Brennan P, Silman A. Statistical methods for assessing observer variability in clinical measures. BMJ 1992; 304: 1491–4.
  43. Cohen J. A coefficient of agreement for nominal scales. Edu Psych Meas 1960; 20: 37–46.
  44. Deleted in proof.
  45. McKegney FP, Beckhardt RM. Evaluative research in consultation-liaison psychiatry. Review of literature 1970–1981. Gen Hosp Psych 1982; 4: 197–218.
  46. De Jonge P, Latour CHM, Huyse FJ. Implementing psychiatric interventions on a medical ward: effects on patients’ quality of life and length of hospital stay. In Press. Psychosom Med.



This article has been cited by other articles:


Home page
 Chronic IllnessHome page
L. Montin, H. Leino-Kilpi, J. Katajisto, J. Lepisto, J. Kettunen, and T. Suominen
Anxiety and health-related quality of life of patients undergoing total hip arthroplasty for osteoarthritis
Chronic Illness, September 1, 2007; 3(3): 219 - 227.
[Abstract] [PDF]

Home page
 Psychosom. Med.Home page
E. Lobo, P. De Jonge, F. J. Huyse, J. P.J. Slaets, M.-J. Rabanaque, and A. Lobo
Early Detection of Pneumology Inpatients at Risk of Extended Hospital Stay and Need for Psychosocial Treatment
Psychosom Med, January 1, 2007; 69(1): 99 - 105.
[Abstract] [Full Text] [PDF]

Home page
 PsychosomaticsHome page
C. H.M. Latour, R. de Vos, F. J. Huyse, P. de Jonge, L. A.M. van Gemert, and W. A.B. Stalman
Effectiveness of Post-Discharge Case Management in General-Medical Outpatients: A Randomized, Controlled Trial
Psychosomatics, October 1, 2006; 47(5): 421 - 429.
[Abstract] [Full Text] [PDF]

Home page
 Evid. Based Nurs.Home page
Other articles noted: 25 Jul 03 to 7 Nov 03
Evid. Based Nurs., January 1, 2004; 7(1): e1 - 1.
[Full Text] [PDF]

Home page
 Psychosom. Med.Home page
P. de Jonge, C. H. M. Latour, and F. J. Huyse
Implementing Psychiatric Interventions on a Medical Ward: Effects on Patients' Quality of Life and Length of Hospital Stay
Psychosom Med, November 1, 2003; 65(6): 997 - 1002.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by de Jonge, P.
Right arrow Articles by Latour, C. H. M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by de Jonge, P.
Right arrow Articles by Latour, C. H. M.
Related Collections
Right arrow Consultation Psychiatry
Right arrow Health Psychology
Right arrow Quality of Life

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS