Predictors for use of psychosocial services in patients with metastatic colorectal cancer receiving first line systemic treatment

Demographic and clinical characteristics were obtained from patients and medical charts at baseline, i.e. start of first line systemic treatment. In addition, during the 48 weeks of the TES study, information on clinical events was collected. At baseline, patient-reported psychological characteristics were measured with the following tools. The HADS (Hospital Anxiety and Depression Scale) and DT/PL (Distress Thermometer and corresponding Problem List) were used to assess distress and the need for referral with a recall period of two weeks, respectively. The HADS consist of 14 questions with scores ranging from 0 to 42, the recommended cut off score for distress in cancer patients is ≥13 [20]. The DT/PL comprises of the Distress Thermometer ranging from 0 to 10, followed by a list of problems evaluating a wide range of problems and a single question evaluating the need to talk to a professional [2]. A score of 5 or higher on the DT has been shown to indicate presence of distress [2]. QOL was measured by the EORTC-QLQ-C30 version 3.0 [21], consisting of five functional scales (physical, role, cognitive, emotional, and social), symptom scales and individual symptom items, and perceived financial impact of the disease.

During treatment, the following clinical events were assessed by clinicians: treatment modification or discontinuation, clinical benefit during the first 10 weeks of treatment (progression vs stable disease or partial response), occurrence of severe (i.e. grade III/IV) adverse events (AEs, these were graded and assigned by clinicians using the National Cancer Institute’s Common Terminology Criteria for Adverse Events (NCI-CTCAE) as mild (Grade 1), moderate (Grade 2), severe (Grade 3), or life-threatening (Grade 4) [22]) and total number of unscheduled admissions. Psychosocial service utilization in- and outside the hospital was assessed using an adapted version of the TiC-P questionnaire [23] after 10, 24 and 48 weeks of treatment with a recall period of three months. The TiC-P is a feasible and reliable instrument for collecting data on medical consumption and productivity losses [23, 24]. The first part of the TiC-P includes 14 structured yes/no questions on relevant medical resource items each followed by a question on the volume of medical consumption. The questions include contacts within the mental healthcare sector (regional mental healthcare organization, psychiatrist/psychologist or psychotherapist in private practices or outpatient hospital, institutional day-care treatment, Consultation Agency for Alcohol and Drug addiction (CAD), self-help group) and contacts with general healthcare providers (general practitioner, allied health professional, social worker, providers of alternative medicine, outpatients visits to medical specialists, hospital admission and contacts with a company doctor) and the use of medication. For the current analyses we evaluated contacts with social workers, contacts with services in regional mental healthcare organizations, contacts with psychiatrists/psychologists in private practice, and contacts with psychiatrists/psychologist in hospital settings.

Descriptive statistics were calculated to summarize all variables and to outline the pattern of use of psychosocial services in patients from the TES trial. Differences between patients who had used psychosocial services versus those who had not used psychosocial services during follow-up were computed using the chi-square test for categorical data and t-test for metric data. Two separate multivariable analyses were undertaken to identify predictors for the use of psychosocial support. In the first analysis, baseline variables potentially related to the use of psychosocial support were evaluated. Explanatory variables included gender (male vs female), age at randomization, married/living together vs living alone, education level (low: primary education, middle: lower general secondary education, intermediate vocational education or high school, high: higher vocational education or university), employment (working yes vs no), functional status (defined by the Eastern Cooperative Oncology Group Performance Status (ECOG PS), which is a score ranging from zero (“fully active”) through three (“capable of only limited self-care”) to five (“dead”) [25]), right or left sided tumor, number of prior cancer-related treatments, type of chemotherapy (capecitabine vs CAPOX vs FOLFOX or other), total number of (organs with) metastases, time since initial diagnosis in months (< 1.5 months vs 1.5–10 months vs > 10 months), time since diagnosis of metastatic disease in months (< 1.0 months vs 1.0–2.5 months vs > 2.5 months), comorbidity (number of comorbid conditions as reported by patients at baseline), allocation to treatment arm of the parent study (intervention vs control) and participation in other clinical studies (yes vs no), distress on HADS (score of < 13 or ≥ 13), distress score on DT (< 5 or ≥ 5), number of physical problems reported on PL, number of emotional problems reported on PL, need to talk to a professional (yes vs no), scores on functional QOL scales, global QOL, fatigue and pain score on EORTC-QLQ-C30. Variables were included in the multivariable logistic regression equation if they had an univariable association (p ≤ .25) with the criterion [26]. A backward stepwise regression model was computed to select the final model that included predictive factors which were statistically significant at the 0.10 level. Variables not significantly contributing to the multivariable logistic regression equation were removed. In the second analysis it was evaluated whether clinical factors during treatment predict use of psychosocial support. Explanatory variables included clinical benefit during the first 10 weeks of chemotherapy treatment, number of severe AEs, any severe AE during treatment, any admission during treatment, and number of admissions. Using the final model on baseline predictors as a base, variables were included in this model if they had an univariable association (p ≤ .25) with the criterion. A backward stepwise regression model was computed to select the final model that included predictive factors which were statistically significant at the 0.10 level. Variables not significantly contributing to the multivariable logistic regression equation were removed. The explained variance of the model was assessed by use of Nagelkerke’s R². Since multicollinearity can influence the estimated beta parameter, the variable inflation factor (VIF) for each variable included in the regression equations was calculated (i.e. a VIF greater than 2.5 indicates multicollinearity) [27, 28]. Data were analyzed using IBM SPSS statistics version 22.0 (IBM Corp., Armonk, NY).

Limitations of the present study include the self-reporting of use of psychosocial services, that may have under- or overestimated the actual use of these care services. Self-reported data are argued to be less reliable and threatened by bias arising from social desirability and recall period [44]. Second, this study did not make a clear distinction between type of psychosocial needs and problems. For some needs, especially needs resulting from practical or logistical issues, psychosocial services may not be appropriate. In addition, the analyses were conducted on a homogenous sample of particularly middle-aged, male, patients with advanced CRC participating in a CRT. Further research is needed to determine whether these findings apply to community-based settings as well and whether these findings are generalizable to younger or older patients. Furthermore, the need for psychosocial support was assessed with one single item question only. Another limitation was the sample size. Even though a total of 349 patients participated in the original trial, only 70 made use of psychosocial support. As a consequence some of the subgroups only contained a small number of patients, which could limit the generalizability of the results, warranting external validation.

Our results can be used to inform clinicians in intervention practices as these provide insight into characteristics of patients receptive to psychosocial care. Identification of predictors for use of psychosocial services in patients with mCRC may allow resources to be directed more efficiently to patients who need and want help. In addition, our study findings underscore that relying on screening for distress to identify patients that use psychosocial services is less successful than has been assumed [33]: the need to talk to a professional, instead of the level of distress, predicted use of psychosocial services. Instead of relying on screening for distress, psychosocial services could be targeted to patients expressing the need to talk to a professional. This could be supplemented with more targeted assessments of patients at risk for psychosocial problems [45]. In addition, it should be emphasized that there may be a group of individuals at risk of suffering from negative psychosocial outcomes that is unlikely to use psychosocial services (e.g. less educated, older man), future research should focus on potential barriers to this use and on novel targeting tools to include these patients that my benefit from support.