1 INTRODUCTION

Cancer patients often suffer from high distress, poor mental health, and low quality of life,1-7 and affective and anxiety disorders are highly prevalent.7 For example, a study with German cancer patients showed that the odds of being depressed were more than five times higher for cancer patients than for the general population.8 Furthermore, many patients suffer from different forms of anxiety9, 10 and even years after initial diagnosis cancer patients may have fear of recurrence or progression.11 Psychosocial interventions are effective and thus may help to improve cancer patients' mental health and quality of life.12-14 For example, a mindfulness-based intervention for younger breast cancer survivors resulted in a significant reduction of distress.15

What do we know about utilization of professional psychosocial support in cancer patients? A study in a large German sample of cancer patients showed that about one third of patients made use of psychotherapy or psychological counselling or both because of distress due to cancer, but less than half of patients currently diagnosed with a mental disorder, assessed with the Composite International Diagnostic Interview in oncology patients (CIDI-O),16 reported having made use of psychosocial support offers.17 Another evaluation of cancer patients with significant distress showed that about 70% declined help, often because patients preferred to manage distress on their own, already received help, or did not evaluate their distress as severe enough.18 Uptake of and adherence to psychosocial interventions in cancer patients is higher when therapy is offered prior to medical treatment than later.19 Furthermore, sociodemographic and clinical variables and information about support correlate with utilization of professional help.17, 20 For example, a study with long-term breast cancer survivors showed that older age, lower education, and less information about support was associated with decreased participation in past support programs.20

In the German health care system, more than half of all cancer patients attend a 3-week inpatient rehabilitation program after acute treatment.21, 22 It is financed by the German Statutory Pension Insurance to reduce long-term impairments due to cancer and to support social participation. A key element for patients to express their needs is the admission interview at rehabilitation. However, a considerable amount of cancer patients report unmet needs at both admission and discharge, whereas meeting patients' supportive care needs during inpatient rehabilitation may improve their quality of life.23

The main goal of our study (“UNSAID”) is to explore cancer survivors' unexpressed needs in the admission interview of inpatient rehabilitation. In the first study part, we explored potential barriers to patients expressing needs. We found a wide range of potential barriers, for example, limited time in the admission interview, shame, or non-empathetic behaviour of the physician.24 In the second part, we found that a quarter of patients reported unexpressed needs at admission, which was associated with lower quality of life and mental health at discharge.25 Based on these previous results, we now examine the longitudinal trajectories of cancer patients after inpatient rehabilitation. Therefore, we explore whether cancer patients with unexpressed needs show lower values in mental health and quality of life than other patients at both 3 and 9 months after discharge. Furthermore, we examine if patients with unexpressed needs at admission continue to have unexpressed needs and how they make use of professional outpatient care 3 and 9 months after discharge.

2 METHODS

The Ethics Committee of the Medical Faculty of the University of Würzburg (ref: 71/17) approved this study. It is registered on WHO International Clinical Trials (DRKS00012998) and performed in accordance with the Declaration of Helsinki. All patients provided written informed consent.

2.1 Design, patients, and recruitment

This study has an exploratory design. We conducted a monocentric questionnaire survey with cancer patients attending a 3-week inpatient rehabilitation program. Inclusion criteria comprised written informed consent, diagnosis of breast, prostate, or colon cancer, and age of 18 years or older. Exclusion criteria comprised lack of German language skills and severe, uncorrected visual impairment. We performed measurements at the end (T1) as well as 3 (T2) and 9 months (T3) after the end of inpatient rehabilitation. Routine data were available from the standard admission diagnostics at the beginning of rehabilitation (T0). During the recruitment period (08/2018 – 11/2018), a physician asked all eligible patients two weeks after admission whether they wanted to participate in the study, and provided written information. On consent, patients received the first questionnaire during the last week of their rehabilitation stay, that is, at the end of their treatment. Patients received follow-up questionnaires via mail and sent them back to the clinic.

2.2 Measures 2.2.1 Unexpressed needs

To assess the proportion of patients with unexpressed needs, we asked patients at T1 to remember their admission interview and rate the following item: “In the admission interview, I expressed all my concerns and distress”. Response options were1 absolutely not true,2 rather not true,3 rather true, and4 absolutely true. We grouped patients in unexpressed needs (absolutely not true + rather not true) versus expressed needs (rather true + absolutely true). For a complete description of the self-developed questionnaire see.25 Hence, we assessed unexpressed needs in the admission interview retrospectively.

2.2.2 Mental health

To assess symptoms of depression and generalized anxiety, we used a 2-item depression scale (PHQ-2) and a 2-item anxiety scale (GAD-2) from the Patient-Health-Questionnaire.26 Patients assess their symptoms over the last two weeks on a 4-point Likert scale, with the response options1 not at all,2 several days,3 on more than half the days, and4 nearly every day. For each subscale, a sum score ranging from 0 to six is formed. Cronbach's alpha for both PHQ-2 and GAD-2 in our data at T1 was 0.79, respectively.

Furthermore, we used the emotional functioning scale of the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire Core 30 (EORTC QLQ-C30).27 The four items contain subjective ratings about mental health during the last week on a 4-point Likert scale, ranging from 1 (not at all) to 4 (very much). Sum scores are transformed into a range from 0 to 100, with higher scores reflecting better health status. Cronbach's alpha in our data at T1 was 0.92.

We assessed fear of progression with the short form of the Fear of Progression Questionnaire (FoP-Q-SF).28 Patients rate 12 items on a five-point Likert scale ranging from 1 (never) to 5 (very often). In a sum score ranging from 0 to 60, higher scores indicate higher levels of fear of progression. Cronbach's alpha in our data at T1 was 0.89.

2.2.3 Quality of life

Quality of life was assessed with the global health and quality-of-life scale of the EORTC QLQ-C30.27 The two items contain a subjective assessment of health and quality of life during the last week on a 7-point Likert scale ranging from 1 (not at all) to 7 (very much). The sum score is transformed into a scale ranging from 0 to 100, with higher scores indicating better quality of life. Cronbach's alpha in our data at T1 was 0.90.

At T0, the routine assessment of fear of progression and quality of life showed differences to the original questionnaires in wording and order of the items as well as the formatting of scales. Therefore, we decided to present results at T0 for those outcomes separately and exclude them from the trajectory models.

2.2.4 Utilization

We asked patients at T2 and T3 if they still had a need for psychosocial support and if they talked about their needs. Furthermore, we asked them about their utilization of professional help, using self-constructed items. In particular, we wanted to know whether they currently used or had used one or more of the following services: general practitioner, specialized physician, psychotherapy or psychological counselling, cancer counselling, self-help group, physiotherapy, domestic help, or palliative medical care. If they did not report using one of the services, we asked whether they thought it was not necessary or if access was too difficult.

2.3 Statistical analysis

We analysed trajectories of patients not expressing needs compared to patients expressing needs regarding mental health and quality of life by conducting multigroup structural equation models (SEM), with expressed needs (yes/no) as grouping variable.29 Analyses were conducted separately for each outcome. The observed variables (e.g. PHQ-2 at T0, T1, T2, and T3) were modelled as 4 distinct correlated variables. A model without further restrictions is a just identified model and the estimated intercepts equal the means of each variable. Hypothesis about different trajectories were tested by including restrictions on the intercepts for each variable in each group.30 The chi2-test of this model equals a chi2-difference test with the unrestricted model. A non-significant p-value indicates that the fit of the restricted model is not worse than the fit of the unrestricted model.

We conducted the model tests as follows: First, we estimated a non-restricted model (Model 1). Second, we tested a model (Model 2) restricting all intercepts to be equal. This model tests the hypothesis of no change over time and no difference between both groups. If the chi2-test was not significant, we could not reject a no change model (i.e., constant means over time and between groups) and no further tests were carried out. A significant chi2-test indicated that the restricted model fits the data worse than the unrestricted model and further model tests were conducted. In the third step, we tested a model (Model 3) restricting intercepts within groups to be equal at all time points, but different between groups. A non-significant chi2-test was interpreted as constant means over time, but differences between groups, and no further tests were carried out. A significant chi2-test indicated that the restricted model fits the data worse than the unrestricted model and further model tests were conducted. In the last step (Model 4), we explored whether further parameter restrictions would improve model fit (e.g., restricting the group means to be equal at T0, but different at T1 to T3). This last step was purely exploratory and based on descriptive results. All models were estimated using robust maximum likelihood estimates. In all steps, model comparisons were done using Satorra-Bentler-chi2-difference test between the focal model and a less restricted model (i.e., the unrestricted model or a model with non-significant p-value). Based on the results of the final model, we computed Cohen's d as a measure of effect size for differences between groups per timepoint. Cohens' d of 0.2/0.5/0.8 was regarded as small/medium/large.31 Full information maximum likelihood (FIML) estimation with included auxiliary variables (age, sex, and diagnosis) was used to include patients with missing data in the analyses.32

We analysed associations between (not) expressing needs in the admission interview and utilization of professional help at follow-up using cross-tables, chi2-tests, and Cramer's V as a measure of effect size. Effect sizes of 0.1/0.3/0.5 were regarded as small/medium/large.31 Alpha was set to 0.05 for all analyses. We used IBM SPSS Statistics (Version 25.0.0.1) and R (Version 3.6.1). SEM were estimated using the lavaan-package.33

3 RESULTS 3.1 Sample characteristics

A total of 713 eligible patients were asked to participate in the study. Overall, 470 (65.9%) agreed to participate. Participants and non-responder showed no significant differences in gender, diagnosis, or type of rehabilitation. Non-responders were older than participants (participants: M(SD) = 58.8 (12.8), non-responders: M(SD) = 62.1 (12.9); t [704] = 3.29, p < 0.001). From 21 persons (2.9%) completed questionnaires at T1 were lacking, so that a total of 449 persons (63%) were finally included. Follow-up data were available from 418 persons (93.1%) at T2 and 401 persons (89.3%) at T3, respectively. Routine data at T0 were available from 437 persons (97.3%). Sociodemographic and clinical characteristics of the sample are given in Table 1.

TABLE 1. Sociodemographic and clinical characteristics at T1 for all patients (n = 449), patients with unexpressed needs (n = 107), and patients who expressed their needs (n = 336). Total (n = 449) Unexpressed (n = 107) Expressed (n = 336) Sex, n (%) Female 284 (63.3) 73 (68.2) 208 (61.9) Male 165 (36.7) 34 (31.8) 128 (38.1) Age (years), mean (SD) 58.8 (12.8) 59.1 (13.0) 58.5 (12.6) Marital statusa, n (%) Single 44 (9.8) 12 (11.3) 32 (9.6) Married 302 (67.6) 68 (64.2) 229 (68.4) Marriage-like relationship 18 (4.0) 5 (4.7) 13 (3.9) Divorced/separated 51 (11.4) 12 (11.3) 38 (11.3) Widowed 32 (7.2) 9 (8.5) 23 (6.9) Childrenb, n (%) Yes 350 (78.1) 80 (74.8) 264 (78.8) No 98 (21.9) 27 (25.2) 71 (21.2) Educationc, n (%) Less than junior (<10 years; basic secondary school) 121 (27.7) 23 (21.7) 98 (28.7) Junior (10 years; middle-level secondary school) 157 (35.9) 36 (34.0) 119 (33.4) Secondary specialized school (12 years) 44 (10.1) 14 (13.2) 30 (11.1) Senior (high-school graduate) 115 (26.3) 30 (28.3) 83 (24.8) Employment statusd, n (%) Fulltime 119 (26.7) 20 (18.7) 99 (29.6) Part time 74 (16.6) 21 (19.6) 52 (15.5) Less than part time 25 (5.6) 8 (7.5) 17 (5.1) Training 2 (0.4) 1 (0.9) 1 (0.3) Unemployed 18 (4.0) 7 (6.5) 11 (3.3) Disability pension 13 (2.9) 2 (1.9) 11 (3.3) Old-age pension 160 (35.9) 43 (40.2) 113 (33.8) Other 35 (7.8) 5 (4.7) 30 (9.0) Cancer type, n (%) Breast cancer 238 (53.0) 58 (54.2) 178 (53.0) Prostate cancer 134 (29.8) 30 (28.0) 102 (30.4) Colon cancer 77 (17.1) 19 (17.8) 56 (16.7) Stage of cancer, n (%) Primary cancer 407 (90.6) 98 (91.6) 304 (90.5) Recurrent/Metastatic cancer 42 (9.4) 9 (8.4) 32 (9.5) Treatment intentione, n (%) Curative (high chance of healing) 272 (76.0) 65 (78.3) 204 (75.6) Curative (high risk of recurrence) 59 (16.5) 11 (13.3) 47 (17.4) Palliative 27 (7.5) 7 (8.4) 19 (7.0) a missing: n = 2. b missing: n = 1. c missing: n = 12. d missing: n = 3. e missing: n = 91.

About a quarter of patients (24.3%, n = 107) reported unexpressed needs in the admission interview. We consider T1 as the essential time point of measurement, where we captured unexpressed needs as a baseline for all further longitudinal analyses. Therefore, we explicitly describe patients' characteristics at this time point. Bivariate correlation coefficients (Spearman's Rho) of previous analyses showed no significant correlations of not expressing needs with sociodemographic and clinical characteristics.25

3.2 Mental health

Analysing depression, a model assuming differences between groups at all time points, within-group mean differences between T0 and T1, and constant within-group means between T1 and T3 showed the best fit (chi2 = 6.12; df = 4; p = 0.19; effect sizes of differences between groups d = 0.32, T0; d = 0.39, T1 through T3, with fewer depressive symptoms in patients expressing needs; see Table 2 and Figure 1).

TABLE 2. Means and standard deviations of depression, anxiety, emotional functioning, fear of progression, and quality of life in patients with expressed versus unexpressed needs from T0 until T3 and structuring equation models Means and standard deviations T0 T1 T2 T3 M SD M SD M SD M SD Depression Expressed 1.75 1.38 1.06 1.21 1.15 1.32 1.12 1.26 Unexpressed 2.13 1.37 1.43 1.30 1.65 1.54 1.74 1.50 Anxiety Expressed 1.55 1.37 1.13 1.36 1.27 1.40 1.10 1.30 Unexpressed 1.83 1.48 1.39 1.42 1.64 1.52 1.81 1.63 Emotional functioning Expressed 59.65 27.26 66.03 27.66 65.17 26.74 66.26 27.55 Unexpressed 56.05 27.77 56.72 29.82 54.19 26.67 55.55 28.49 Fear of progressiona Expressed 27.63 9.71 29.18 9.46 27.94 9.57 27.70 10.10 Unexpressed 30.28 10.00 32.28 9.79 31.02 10.30 31.22 10.57 Quality of lifea Expressed 6.03a 1.90a 69.88 18.53 69.59 18.87 69.57 20.35 Unexpressed 5.75a 1.69a 67.06 17.96 63.45 19.00 62.50 21.72 Structuring equation models Model 2 Model 3 Model 4 Chi2 df p-value Chi2 df p-value Chi2 df p-value Depression 174.37 7 <0.001 147.31 6 0.00 6.12 4 0.19 Anxiety 77.87 7 <0.001 67.43 6 0.00 4.83 2 0.09 Emotional functioning 44.60 7 <0.001 33.55 6 0.00 2.19 5 0.82 Fear of progression 34.01 5 <0.001 23.15 4 0.00 0.75 2 0.69 Quality of life 11.38 5 0.04 4.70 4 0.32 – – – Note: T0: start of inpatient rehabilitation; T1: end of inpatient rehabilitation; T2: 3 months after inpatient rehabilitation; T3: 9 months after inpatient rehabilitation. Abbreviation: df, degrees of freedom. a Routine assessment at T0 of quality of life showed differences to the original questionnaires in the formatting of scales.

Trajectories of mental health and quality of life (means and 95% confidence interval) in cancer patients with unexpressed versus expressed needs. Notes: T0: start of inpatient rehabilitation; T1: end of inpatient rehabilitation; T2: 3 months after inpatient rehabilitation; T3: 9 months after inpatient rehabilitation. Higher scores in PHQ-2 indicate a higher risk for depression, in GAD-2 a higher risk for anxiety, in EORTC QLQ-C30 higher emotional functioning and better quality of life, and in FoP-SF a higher risk for fear of progression. We did not include values of fear of progression at T0 due to differences in in wording and order of the items at T0 compared to other measurement points. We did not include values of quality of life at T0 due to differences in our assessment in formatting of scales compared to the original

In anxiety, a model assuming differences between groups at all time points and within groups as following fitted best: In patients expressing needs, intercepts (i.e., means) differ between T0 and T1, T2 and T3, but not T1 and T2. In patients not expressing needs, intercepts differ between T0 and T1, T1 and T2, and T2 and T3, but not T0 and T3, respectively (chi2 = 4.83; df = 2; p = 0.09; effect sizes of differences between groups d = 0.32, T0, and d = 0.39, T1 through T3, with less anxiety in patients expressing needs; see Table 2 and Figure 1).

For emotional functioning, a model fitted best with differences between groups at all time points; in patients expressing needs, with differences between T0 and T1, but not among T1 through T3, and those not expressing needs with equal intercepts among T0 through T3 (chi2 = 2.19; df = 5; p = 0.82; effect sizes of differences between groups d = 0.18, T0, and d = 0.44, T1 through T3, with better emotional functioning in patients expressing needs; see Table 2 and Figure 1).

Analysing fear of progression, we only included values of T1, T2, and T3 in SEM. A model with differences between groups at all time points, within-group mean differences between T1 and T2, but not T2 and T3 fitted best (chi2 = 0.75; df = 2; p = 0.69; effect sizes of differences between groups d = 0.38, T1; d = 0.39, T2 through T3, with less fear of progression in patients expressing needs; see Table 2 and Figure 1). At T0, there was a significant difference between groups with an effect size of d = 0.27.

3.3 Quality of life

For quality of life, again we included only values of T1, T2, and T3 in SEM. A model assuming differences between groups, but, within groups, constant means over time fitted best (chi2 = 4.70; df = 4; p = 0.32; effect size of differences between groups d = 0.28, with better quality of life in patients expressing needs; see Table 2 and Figure 1). At T0, there was no significant difference between groups.

3.4 Utilization of professional help

Overall, about 20% of patients had a need for professional support and less than 18% had further unexpressed needs at T2 and T3 (see Table S1). However, patients with unexpressed needs had a higher need for professional psychosocial support at follow-up than other patients. They more often did not talk about their needs after discharge, although they would have liked to. Furthermore, they made less use of the following h