Determinants of adherence to physical cancer rehabilitation guidelines among cancer patients and cancer centers: a cross-sectional observational study

An observational study was conducted to assess adherence to and determinants of PCR guidelines in nine cancer centers. This was done at cancer centers and on patients who have been treated in these cancer centers, while taking the clustering of data into account. The existing registration systems and patient and HCP questionnaires were utilized.

The patient cohort was recruited from the nine participating cancer centers located in categorical, university, teaching, and non-teaching hospitals in the Netherlands. The cancer registry was used for the selection of eligible patients: all patients with a history of breast, female organ, urogenital organ, gastrointestinal, and hematological malignancies diagnosed between January 2014 and June 2015 who had successfully completed their primary treatment without signs of recurrence or metastases. The treating physicians asked them whether they wanted to participate and give informed consent. One HCP at each center was asked to collect data on the characteristics of their cancer center.

Indicator scores for processes and patient outcomes of care as well as patient and center characteristics were measured to assess adherence to PCR guidelines and the determinants of guideline adherence. The indicators were based on (inter)national, evidence-based PCR guidelines [37, 87, 88]. A national panel of 10–12 professional experts and patients used the RAND-modified Delphi method to develop the indicators [89, 90]. We developed indicators that measure PCR guideline adherence that have the potential to be valuable, reliable, measurable, applicable, have improvement potential, have preferably minimum amount of missing data, and contain discriminatory capacity. The main indicator was distress screening with the DT [91, 92]. The other indicators were (1) information provision concerning PA and PCRPs, (2) advice to take part in PA and PCRPs, (3) referral to PCRPs, (4) participation in PCRPs, and (5) PA uptake (PAU). In supplement 1, the definitions of the psychometric characteristics used to develop and measure the quality of the indicators used is comprehensively explicated. All developed indicators showed to be valuable, reliable, measurable, applicable, have improvement potential, have minimum amount of missing data, and four indicators contain sufficient discriminatory capacity. Supplement 2 provides an overview of the range of potential values of each developed and measured indicator regarding the psychometric characteristics.

The indicators and patient characteristics were measured among patients by means of questionnaires. To assess cancer center characteristics, existing registry systems and questionnaires distributed among HCPs involved in cancer care in the nine cancer centers were used.

We used the Statistical Package for the Social Sciences (IBM SPSS Statistics version 22 for Windows; SPSS, Chicago, IL, USA) to enter the collected data in a database. We used descriptive analyses (frequencies, percentages, means, and SD or median and interquartile ranges) to describe patient characteristics and the scores for adherence to the indicators.

To determine whether our data were normally distributed, we examined the distribution of our continuous outcome measures and carried out quantile–quantile (Q-Q) plots. We also calculated the skewness and kurtosis of these variables. For all variables, both values of the skewness and kurtosis were between − 1 and + 1, and Q-Q plots showed a straight line; therefore, they met normality requirements.

We wanted to assess the extent to which the indicator scores could be explained by characteristics of (1) cancer centers, but also (2) the patients who were treated in these cancer centers, while taking the clustering of data into account. Our dataset contains information at the patient level from nine different cancer centers for each indicator.

For this purpose, we used SAS software (SAS 9.2 for Windows; SAS Institute, Cary, North Carolina, USA) for our multilevel multivariable regression analyses. We used the Glimmix procedure for dichotomous data and the MIXED procedure for continuous data to determine (1) the association between the scores of the indicators and patient and cancer center characteristics, and (2) the association between proper screening and the other indicators. The other indicators were information provision concerning PA and PCRPs, advice to take part in PA and PCRPs, referral to PCRPs, participation in PCRPs and PAU. The patient characteristics were included in the model as confounders in the analysis.

The intra-class coefficient (ICC) was calculated for each outcome to obtain insight into the clustering effect of the cancer centers.

The mean age of the participants was 66.3 years; 60.7% were female and 94% had a Dutch background. The participants had a history of cancer of the breast (31.1%), female organs (16.0%), urogenital organs (23.6%), gastrointestinal (27.7%), and hematological malignancies (1.5%). Eighty-five percent had undergone surgery, 39.8% received chemotherapy, 42.1% radiotherapy, and 21.7% hormonal therapy. Fifty-nine percent had received two or more cancer treatments and 31.7% had two or more comorbidities. Post-cancer treatment, 36.4% gained and 13.2% lost weight, while 50.4% kept a stable weight. Of the participants, 80.5% were cohabiting. As level of education, 38.3% of patients have finished low, 40.4% middle, and 21.2% a higher level of education and 25% were employed. The mean Global Health Status score was 77.5 (SD 18.0), the mean physical function score was 82.5 (SD 18.3), and the mean EORTC-QLQ-C30 summary score was 40.8 (SD 5.4). The Mean General Fatigue score was 10 (SD 4.6) and the Mean Physical Fatigue score was 9.6 (SD 4.4). The mean PAM-13 Total Score was 55.9 (SD 13.1).

In the study, one categorical, two university, two teaching, and four non-teaching hospitals participated related to respectively 5.5, 20.5, 25.2, and 48.7% of the accrued patients. One center had a MORB available. Five hospitals performed standardized screening with the DT. Eight hospitals delivered a PCRP, of which four hospitals delivered internally and four hospitals externally.

Table 1 outlines the characteristics of the patients treated for the various types of cancer. Table 2 outlines the characteristics of the nine cancer centers.

Table 4 shows multilevel associations of patient characteristics with the measured indicators.

Univariate and multivariable multilevel regression analyses were performed but showed less variation in scores of the indicators between the different cancer centers. The ICCs of the outcomes varied between 0 and 0.085. This means that maximum 8.5% of the variation in an indicator could be explained by differences between cancer centers. This variation between centers was too low to detect any association between center characteristics (such as type of hospital, availability of a MORB, existence of standardized DT screening, and existence of PCRPs) with the indicators.

Screening is a key aspect in the delivery of healthcare. Patients who will accept and benefit from rehabilitation programs can be identified and encouraged to participate in PCRPs by using the DT [59, 60]. Encouraging screening with the DT in daily cancer care will help promote the implementation of PCR guidelines. In our study, 47% of the patients were screened with the DT. Other studies found comparable percentages of screening with the DT of 40–50% [99]. The score of screening with the DT shows room for improvement, especially because the screening was significantly positively associated with higher scores of the other indicators, with ORs between 1.69 and 2.04. Therefore, encouraging screening with the DT is a good first step toward improving adherence to the current PCR guidelines.

This is one of the first studies to investigate determinants at the patient and cancer center levels for PCR guideline-based indicators. Other studies examining determinants of guideline adherence have been carried out in several other areas of cancer care, including treatment guidelines for lung, prostate, and gastrointestinal cancers. They found low guideline adherence rates and differences in delivered care associated with hospital type and patient age, gender, and disease stage [100‐103] as well as educational level [104, 105] and employment status [106]. In addition, implementation strategies developed with knowledge of determinants in these other areas of cancer care did achieve improvement of guideline adherence. Therefore, knowledge of determinants is useful in creating tailored strategies for implementing PCR guidelines.

We detected a higher screening score for women, but higher scores for information provision, referral, and participation in PCRPs for men. Gender disparity in the use of cancer rehabilitation care and other healthcare services has been noted before in the literature [83, 107‐109].

Traditional notions of masculinity that emphasize the values of being autonomous and less emotional may lead men to be reluctant to express emotion or distress [110‐112]. In addition to the higher levels of media attention being paid to the post-cancer physical and psychosocial symptoms experienced by women, HCPs might be influenced by gender bias and less aware of screening men for distress. Men also seem to be more eager for sufficient explanations concerning screening in order to make a decision to participate [56, 57].

Besides screening, gender bias has been reported to affect HCP referral and treatment decisions and may also influence decisions on advising women for increasing PA or referral to PCRPs [113‐115]. Women’s gender-specific roles and PA preferences may also contribute to women not participating in PCRPs. Contemporary the burden of cancer is evenly distributed between the different sexes. Currently, one in five men and one in six women will be diagnosed with cancer [116]; therefore, attention should be paid to improve screening of males, and improving information provision and referrals to PCRPs for female survivors of cancer. We did not distinguish our strategies on sex in our research. Future research can be used to differentiate which strategies are more effective for men and women.

We also found tumor type to be a determinant. Patients with breast cancer receive more screening, information, and referral to PCRPs, and they participate more in PCRPs. Indicator scores were lower for patients with a history of female organ, urogenital organ, and gastrointestinal malignancies. One reason for this is that most initiatives for improving PCR guideline adherence are designed for and focused on breast cancer. Screening of patients with breast cancer with the DT was relatively well adhered to, as not screening means no accreditation for breast cancer care as required by the patient organization for patients with breast cancer. Patients with gastrointestinal and female organ malignancies judge their cancer care to be of lower quality than that of patients with other tumor types [117]. After completing their primary treatment, patients with gastrointestinal malignancies also rated the information provided as significantly lower in quality than that of patients with breast cancer [118]. Worldwide, malignancies of the gastrointestinal, reproductive, and urogenital systems account for approximately 35% of all malignancies, which is three times the incidence of breast cancer [119, 120]. Therefore, it might be beneficial to preferably focus the strategy on cancer patients and their HCPs in the care pathways for gastrointestinal, female organ, and urogenital organ oncology.

The recruitment of patients with abdominopelvic cavity tumors to PCRPs is difficult [7, 8, 26, 121]. HCPs are more hesitant to refer patients who have undergone major abdominal surgery to PCRPs and typically advise patients to refrain from PA for a number of weeks after surgery [7]. Teaching the HCPs about the positive associations of PA with less physical and psychosocial symptoms and even improved mortality [33, 34, 122‐124] might be a good strategy. In addition, tailored PA guidelines need to be developed since these patients require different PCRPs due to a different range of morbidities and needs. The introduction of accreditation for PCR guideline-based care that has proved successful for patients with breast cancer into the pathways for gastrointestinal, female organ, and urogenital organ oncology might be another strategy.

Two or more cancer treatments showed to be a determinant. Patient with fewer treatments overall have fewer visits to the cancer center and encounter fewer HCPs who provide them PCR guideline-based care. For all treatment modalities, it should be clear when, who, and where the PCR care is delivered, preferably stated in a treatment protocol. PCRPs delivered through practical avenues such as print materials, telephone counseling, and web-based programs are an alternative [125‐129] for patients with fewer visits to the cancer center. Web-based PCRPs with online encouragement, online diaries, and online physical activity programs proved to be feasible with median vigorous PAU over time, and the burden for HCPs appeared to be limited [130‐132].

Patients with post-cancer treatment weight gain/loss had better adherence to PCR guideline-based indicators. Cancer patients experience weight changes due to the cancer itself or to the cancer treatments, such as loss of muscular mass and increased fat mass [133‐136]. The weight change might be the reason for paying more attention to PCR. A referral to a dietician might more readily lead patients to PCRPs as a means of returning to their old weight. In addition, PA is one of the main treatments for weight imbalance since it reduces fat mass and improves muscle mass and has a potential role in preventing and treating cachexia [136, 137].

The ORs of age being 0.96–0.98 and ORs of the MFI-20 mean general score being 1.07–1.10 are numerically very close to an OR of 1.00. The absolute influence of the determinants age and fatigue (the MFI-20 mean general score) on the indicator scores will therefore be negligible and not clinically relevant.

However, a higher age has previously been found to be associated with negative patterns in delivered care and lower levels of PA [138‐142]. In addition, most cancer patients have higher fatigue scores [95], and fatigue is a common and debilitating side effect of cancer and its treatment [143]. It is known that PA can reduce fatigue after the treatment of cancer [144]. More research is necessary to explore the additional effect of strategies focusing on patients with fatigue and of a higher age.

Our study has several strengths. One is that we thoroughly followed the RAND-modified Delphi method [89, 90], which led to the discovery of valid indicators which formed an important basis for measuring guideline-based PCR care. Another is the large study sample of 999 patients, which might have contributed to a reliable dataset for the investigation of the adherence and the analysis of the determinants associated with optimal PCR care.

There are also some limitations which need to be addressed; for example, possible selection bias. Only patients of cancer centers who were willing to participate in our study were included. One can assume these patients have better adherence to PCR guideline-based indicators since these centers are more dedicated to improving this aspect of cancer care. Thus, we expect lower indicator scores in centers less committed to achieving this goal. Further research should also include patients from cancer centers not motivated to implement PCRPs.

One could expect that also organizational characteristics would be associated with performance on indicators related to the provision of distress screening and rehabilitation programs to cancer survivors. Univariate and multivariable multilevel regression analyses showed less variation in scores of the indicators between the different cancer centers. The ICC is calculated as the ratio of the between variance and the total variance (between and within variance). The ICC gives information of the degree of correlation among patients within a cancer center and the proportion of total variance that is attributed to the cluster level (cancer centers). The ICCs of the outcomes varied between 0 and 0.085. This means that maximum 8.5% of the variation in an indicator could be explained by differences between cancer centers, predicting a low chance of between-cluster variability. This variation between centers was too low to detect any association between center characteristics with the indicators. This might be caused by the limited sample size of nine cancer centers and the variation in characteristics between them, indicating that participation of more centers with more variation in characteristics is needed in future research to analyze cancer centers’ characteristics associated with the indicators.

Possible determinants influencing PCR guideline implementation often arise at multiple levels in the healthcare system (patient, HCPs, cancer center, and healthcare organization levels). Currently, cancer care is frequently provided by a multidisciplinary team of HCPs situated in a cancer center. This results in interactive, coordinated care; therefore, we only explored determinants of the cancer centers. However, there may be compelling reasons for both lack of adherence and adherence due to determinants of the individual HCPs, particularly because HCPs’ limited knowledge and skill levels, negative approach, non-commitment to PCRPs, difference in attitude about timing and strategies for cancer rehabilitation, and fear of additional workload all hinder proper PCR care [77, 78, 145, 146]. On the level of the referring providers, limited knowledge levels concerning PCRPs and PCR guidelines hinder proper screening of patients. Moreover, lack of knowledge and skills among HCPs resulted in a lack of qualified information provision for the patients. It also resulted in a lack of guidance in finding the right PCRP and a successful referral for joining the PCRP, both being barriers that impede proper PCR care [77, 78, 145, 146].