Pilot study: Is the cost of acute treatment increased in cancer patients with depression?

In Germany, around 430,000 people develop cancer every year (RKI 2010), and the incidence of cancer is constantly increasing. Improved treatment options lead to a reduction in mortality. At the same time, chronic courses of the disease are increasing. For many people, it is now possible to live longer with their cancer. Psycho-oncology as a clinical and scientific discipline has been concerned with the psychological reaction patterns to cancer and psychological comorbidities of tumor patients for many years. The prevalence of psychological disorders and stress is estimated at around 25 to 30 % in these patients (Mehnert et al. 2006, Zabora et al. 2001, Härter et al. 2001, Reich 2008, Massie 2004, CMP 2007). In the studies, anxiety disorders were found in 1 to 49 %, depressive disorders in 0 to 58%, adjustment disorders in 2 to 52 % and post-traumatic stress disorders in 0 to 35 % of cases (Mehnert et al. 2006). Pain, distress, emotional strain of grief, helplessness and hopelessness as well as impairments in health-related and general quality of life are other common side effects of cancer (Mehnert et al. 2009, Weidner et al. 2006, Stock et al. 2006, Härtl et al . 2011, Stremmel et al. 2010).

Serious and especially long-lasting psychological problems and disorders have a multifaceted negative impact on cancer patients. Patients with anxiety or depression not only suffer from a significantly reduced quality of life (Ballenger et al., 2001, Keller 2004, Somerset et al. 2004), but also find it more difficult to actively participate in the examination and treatment measures of oncological therapy throughout the course of their treatment (Golden 2004, Patridge et al. 2002, Somerset et al. 2004). Anxiety and depression are also associated with more frequent complications and side effects during cancer patients' therapy, more frequent hospital admissions, and reduced effectiveness of cancer therapy (Ballenger et al. 2001, Haghighat et al. 2003, Molassiotis et al. 2002, Simpson et al. 2001, Watson et al. 1999) and increased mortality (Pinquart & Duberstein 2010).
In addition to the consequences that mental health problems can have on the patient and their recovery, an increasing number of scientific studies are investigating whether such problems also have consequences for hospitals. Studies in general medical-surgical hospital settings demonstrate longer lengths of stay per visit for somatically ill patients when they have a depressive disorder (Hosaka et al. 1999), cognitive impairment (Fulop et al. 1998), pain and a high degree of psychopathological abnormalities (Levenson et al. 1990), psychiatric comorbidity (Fulop et al. 1987), or dementia (Wancata et al. 2001). More recent studies, in which hospital care is provided under a managed care approach, support these findings. Bourgeois et al. (2005) showed in more than 31,000 cases over a period of 3 years that of the 33 to 35 % of somatically ill patients with psychiatric abnormalities, increased lengths of stay were observed for all comorbidities except substance abuse. Lobo et al. (2007) also confirmed this association for comorbidity with anxiety and depression in pulmonary patients.
Regarding the treatment costs of patients with somatic illnesses and mental health comorbidities, large-scale studies have demonstrated a correlation between depression and increased outpatient care costs (Na et al. 2007) and higher overall annual healthcare expenditures (Welch et al. 2009). Significant impacts on healthcare costs were also found by Richardson et al. (2008) in asthma patients with depressive disorders and by Rutledge et al. (2009) in women with myocardial infarction. In cancer patients, Butler et al. (2006) found higher annual healthcare expenditures among those who exhibited poor adjustment to their illness. This poor adjustment correlated with depression.
The link between additional mental health problems in patients with somatic illnesses and increased healthcare costs can now be considered relatively well-established. Studies indicate that depression plays a central role, particularly in healthcare costs. Less is known, however, about whether and which symptoms of mental distress specifically affect the costs of acute hospital care for cancer patients with life-threatening illnesses, and whether these potential correlations also hold true under the conditions of a diagnosis-related group (DRG) reimbursement system.
Pilot project on the effort involved in
Acute care for cancer patients
The pilot project at Herford Hospital conducted a retrospective analysis of routine clinical data on the psychological distress of cancer patients and related data on the care requirements during acute hospital treatment. The study investigated whether significant psychological distress increases the workload in the acute care of cancer patients.
The following questions were investigated: Do patients with high anxiety and/or depressive symptoms show [something] in the first year after admission to the hospital?
1. a higher number of hospital stays,
2. a higher number of hospital stays
as well as in the key figures of the DRG system, one per patient and treatment stay on average.
3. longer length of stay,
4. higher number of coded medical diagnoses,
5. Higher number of coded operations and procedures (OPS),
6. higher medical case severity (PCCL),
7. higher valuation ratios ([nominal CW] and effective relative weight [eff. CW])
compared to patients with no or mild anxiety and/or depression symptoms.

Data and methods
At the Herford Hospital, between 2004 and 2008, the psychological burden of 968 oncology patients was routinely assessed upon inpatient admission (pre-treatment time) using the German version of the "Hospital Anxiety and Depression Scale (HADS)" (Herrmann et al. 1995) and the psychosocial burden was assessed using an evidence-based checklist (CMP 2007). To initiate psycho-oncological care, patients were divided into three distress groups based on the thresholds of the HADS total score (HADS-G): low distress (HADS-G < 15), moderate distress (HADS-G ≥ 15 and < 22) and high distress (HADS-G ≥ 22) or based on the HADS-Anxiety subscale (HADS-A < or ≥ 8) and the HADS-Depression subscale (HADS-D < or ≥ 8) (for clinical evidence see Bjelland et al. 2002, Ibbotson et al. 1994, for treatment procedures see CMP 2007).
For 841 of the 968 patients, the corresponding billing-relevant hospital data, such as DRG codes, hospital stay days, number of diagnoses, number of operations and procedures, medical case severity, and nominal and effective relative weight (cost weight), could be assigned. Data analysis was performed using anonymized datasets. A total of 3,726 treatment episodes from 841 patients between 2004 and 2008 were included in the data pool. To control for the catalog effect on DRG parameters, both the current DRG catalog and those for the following two years were applied to the billing-relevant data. By transferring the data to two subsequent catalog years, a direct comparison of DRG parameters from three different years was possible.
Further data selections were made from the data pool of 841 patients with 3,726 cases. For statistical analysis, only data sets from patients who met the following criteria were selected: 1. Patients with a first episode (including patients with at least one year relapse-free), 2. Patients who completed the questionnaires within the first four weeks after hospital admission, 3. Patients admitted to the hospital on or after January 1, 2006 (the data from 2006 to 2008 were used as the basis for calculation, based on the 2008 catalog year), 4. Patients admitted to the hospital no later than December 31, 2007 (the calculation period for the care expenditure was 12 months from hospital admission). 360 patients met these criteria and, with 1,219 treatment episodes, formed the sample for further statistical data analysis. Table 1 lists the sample of patients for both stress groups in HADS-Anxiety and HADS-Depression.
The initial exploratory data analysis investigated whether relevant differences in patient care needs existed based on the HADS total score (low, moderate, severe) or on the anxiety and depression subscales (low, high). The data analysis revealed relevant correlations with care needs based solely on the depression subscale scores. Therefore, only the results relating to patients with low or high levels of distress on the HADS depression scale are presented below.
Of the 360 patients, 245 (68.1 %) were classified as not depressed and 115 (31.91 TP3T) as depressed according to the HADS-D subscale threshold. There was no significant difference between the two subgroups with regard to age (T = -.946; p = 0.345), sex (x² = 0.362; p = 0.548), and partition frequencies (x² = 2.297; p = 0.317), but there was a significant difference with regard to cancer diagnosis (x² = 59.031; p = 0.021). The unequal distribution of cancer diagnoses in both groups can be explained by the high number and heterogeneity of the ICD diagnoses of the underlying disease. No specific clustering of cancer diagnoses in either of the stress groups was apparent from the distributions.
Results
An exploratory data analysis was conducted using multiple linear regression analyses with the predictors age, sex, HADS anxiety total score, and HADS depression total score. The dependent variables relevant to the effort involved (criterion) were: the total number of treatment episodes and the total number of days spent in the year since hospital admission, as well as the average number of days of stay, coded diagnoses, coded operations and procedures, the severity level, and the nominal and effective relative weight per patient and episode over the year, all derived from the DRG codes.
In the multiple regression analyses, the HADS anxiety total score contributed only a very small amount to the explained variance in one effort parameter. With one exception, age consistently emerged as the primary predictor explaining the variance. The highest coefficients of determination were achieved by the predictors age, HADS depression total score, and sex, with a 21.31 TP3T contribution to the explained variance for the mean number of diagnoses per patient and episode and a 17.31 TP3T contribution to the explained variance for the mean case severity (PCCL) per patient and episode (see Table 2). Age and the depression total score explained 14.31 TP3T of the explained variance for the mean length of stay per patient and episode. However, this regression could not be confirmed in the cross-validation.
The results from the exploratory analysis indicate that, in addition to the known correlation between age and gender and care burden, depression, rather than anxiety, may be a significant factor. Tests for group differences between patients with low (HADS-A < 8) and high anxiety scores (HADS-A ≥ 8) on the HADS anxiety scale revealed no significant differences in DRG parameters. Therefore, further analyses focused on the groups of patients with low and high levels of distress as measured by the HADS depression scale.
As can be seen in Table 3, significant differences (in bold) in the indicators for healthcare expenditure in the first year after hospital admission were found between the HADS-D subgroups. Although the number of hospital admissions in the first year of acute treatment was even slightly higher for patients with low levels of distress than for those with high levels, patients with high levels of depression stayed in the hospital for almost six days longer. The DRG indicators calculated per patient and episode consistently showed higher values for patients with high levels of depression. The differences between the subgroups were significant except for the nominal and effective relative weights.
It is noteworthy, and therefore also included in Table 3, that the number of hospital stays and the DRG codes per patient and episode already differ significantly between the HADS-D subgroups in the first six months of acute treatment. This is not surprising, as the total length of stay (days) shows that the majority of treatment episodes occurred in the first six months. This result indicates that the presence of high levels of depression is already associated with higher DRG codes in the initial phase of cancer therapy.
The study further investigated whether the observed differences in DRG indicators between the HADS-D subgroups were also evident when considering only patients with the same cancer diagnosis. In the two largest groups, colon cancer (N=63 patients) and breast cancer (N=56 patients), all DRG indicators suggested a higher healthcare workload for those with the highest burden on the HADS-D subscale. However, these group differences were not statistically significant.
Further analyses of the differences in care costs based on case-related calculations, i.e., relating to individual treatment episodes (cases) that could be assigned to the low or high burden group in terms of depression, are not presented in this paper. The difficulty here lies in the fact that the cases are considered as isolated, person-independent events, and it is not taken into account that the determined anxiety and depression scores of a patient, as person-specific characteristics, are related to a varying number of associated cases or treatment episodes for that individual patient.
discussion
This project is one of the first studies in German-speaking countries to analyze the relationship between psychological distress in cancer patients and the level of care required under DRG (Diagnosis Related Groups) conditions. The study is exploratory in nature, as it addresses the question of how appropriate it is to conduct more in-depth research in this area. Further investigation into the impact of psychological symptoms in cancer patients is worthwhile because this study found indications of a higher level of care required for patients experiencing psychological distress. Among the newly diagnosed or one-year relapse-free patients who were hospitalized, correlations between psychological distress and care were only observed in the area of depression.
Among the 115 of 360 patients who exhibited high depression scores on the HADS, higher values in the DRG indicators were already evident within the first six months after hospital admission. Significant differences between patients with low and high levels of depression were observed in total length of stay, average length of stay, number of coded diagnoses, operations and procedures, and average medical case severity. Although the differences in the relative weights were not statistically significant, the difference in the relative weights of 0.203 (nominal) and 0.235 (effective) reflects a considerable increase in the hospital's workload for the group of patients with high levels of depression.
Limitations of this study lie in the fact that the project was conducted using data from consecutively admitted patients and not from a selected sample of a patient group with a specific indication. Therefore, no conclusions can be drawn for a specific disease group. Potentially relevant factors such as differing prognoses and stages of cancer, changes in cancer therapy within the outpatient and inpatient sectors, sociodemographic and anamnestic factors, and other relevant data may have been affected.
Patient data could not be considered or controlled in the exploratory study.
Furthermore, it remains unclear whether the symptoms of severe depression in a patient are due to pre-existing depressive symptoms, a stress reaction caused by the illness, a stress reaction resulting from the particular severity of the illness, or a combination of several of these factors. Regardless of whether the depression is attributable to the illness, it can be stated that the discussion regarding the burden of care and healthcare costs contributes to further exploring whether psycho-oncological care for cancer patients with depression not only positively influences their quality of life, as scientifically proven, but also potentially has a positive impact on the burden of care.
As Kishi et al. (2004) demonstrated, the early identification and treatment of patients with psychological distress leads to a reduction in hospital stays. Similarly, in the randomized controlled trial by Simpson et al. (2001) in breast cancer patients, group therapy resulted in a reduction of depression, mood swings, and psychiatric symptoms, as well as an increase in quality of life, and was associated with a 23.5 % reduction in healthcare costs at the 2-year follow-up compared to the control group without group therapy. Di Matteo (2000, 2004) identifies patient adherence as the crucial factor that leads to reduced treatment participation when psychological distress is present. Successfully reducing psychological distress enables patients to actively participate in their treatment to a sufficient degree and has a positive impact on the course of treatment.
Consequently, efficient healthcare for specific patient groups will primarily be achieved through effective psychosocial and psychotherapeutic support. Valid identification and individualized psycho-oncological treatment play a central role in ensuring optimal treatment outcomes. This initial exploratory study has demonstrated that, especially within the DRG system, it is worthwhile to examine the healthcare burden on cancer patients more closely, taking into account their psychological stress.