Attendance to Follow-Up Care in Survivors of Adolescent and Young Adult Cancer: Application of the Theory of Planned Behavior.

PURPOSE
The aim was to study follow-up care attendance in adolescent and young adult (AYA) cancer survivors to investigate: (1) correlates of the intention to attend follow-up care and (2) whether the intention is associated with the actual attendance, applying the theory of planned behavior (TPB).


METHODS
We conducted a questionnaire survey in AYA cancer survivors diagnosed 1990-2005 at age 16-25 years, registered in the Cancer Registry Zurich and Zug, Switzerland, who had survived at least 5 years. Structural equation modeling was applied to investigate TPB-related correlates (attitudes, subjective norms, and perceived behavioral control) of intention to attend follow-up care. Logistic regression analysis was used to study the association between intention and actual attendance.


RESULTS
We included 160 AYA cancer survivors in the study (mean age at study: 34.0 years, mean age at diagnosis: 21.6 years, 98 [61.3%] male). Positive attitudes toward follow-up care (coefficient = 0.32, 95% confidence interval [CI]: 0.05 to 0.60) and supportive subjective norms (coefficient = 0.59, 95% CI: 0.41 to 0.78) were associated with higher intention to attend follow-up care. Perceived behavioral control was not associated with intention to attend (coefficient = -0.13, 95% CI: -0.36 to 0.10), but with actual attendance (odds ratio [OR] = 4.55, 95% CI: 1.83 to 11.31). Higher intention was associated with actual follow-up care attendance (OR = 14.29, 95% CI: 5.80 to 35.21).


CONCLUSION
Positive attitudes and supportive social norms were associated with higher intention to attend follow-up care, and higher intention was associated with actual follow-up care attendance. Increasing awareness of the importance and benefits of follow-up care not only among survivors but also family, friends and healthcare professionals may help increase follow-up care attendance among AYA cancer survivors.


Introduction
Five-year survival of European adolescent and young adult (AYA) cancer patients has increased remarkably and reached 87% for patients diagnosed between 1995 and 2002. 1 However, late effects of cancer and its treatment are common and include various medical conditions, 2, 3 decreased mental health 4 and health-related quality of life, 3, 5, 6 and psychosocial problems. [6][7][8][9] Attendance to follow-up care is important to detect late effects early and to treat and support survivors accordingly. Consequently, it is important to investigate correlates of follow-up care attendance among AYA cancer survivors.
Research showed that female and older survivors were more likely to attend recommended follow-up care, 4, 10 and attendance rates decreased with more time after diagnosis. [11][12][13][14] More than five years after diagnosis, 40% of survivors reported not having attended a routine medical visit during the last year. 14 Furthermore, lower quality of life and more health problems were associated with more medical out-patient visits in long-term survivors 10 indicating that there might also be a link to follow-up care attendance.
Perceived barriers to attend medical care included lack of insurance, high costs, 13, 14 no guidance from the oncologist, no perceived need for follow-up care 13,15 , not enough information regarding follow-up care, conflict with other responsibilities such as family and work, fear of new cancer, and avoidance of care related to the former cancer disease. 15 Knowledge on cognitive correlates of follow-up attendance is still lacking. To address this gap, we applied the theory of planned behaviour (TPB 16 , Figure S1 in Supplementary Material) to investigate cognitive correlates of follow-up care attendance in a populationbased sample of AYA cancer survivors in Switzerland. The TPB suggests that the intention to perform a behaviour predicts the actual behaviour. The intention itself is predicted by three constructs: attitudes towards the behaviour (referred to as attitude), subjective norm (norm) and perceived behavioural control (control). Additionally to the intention, the actual behaviour is predicted by control. Attitude refers to a person's favourable or unfavourable opinions about the behaviour of interest, norm refers to perceived expectations among the social environment to perform the behaviour, and control refers to the perceived easiness or difficulty to perform the behaviour. Generally, more positive attitudes and norms, and greater control are expected to be associated with a higher intention to perform the behaviour, and a higher likelihood that the behaviour is actually performed. 16 Survivors' positive attitudes towards follow-up care, the perception that family, friends and health professionals support or expect their attendance, and the perceived easiness to attend care are thus expected to increase the intention of survivors to attend follow-up care.
The aim of our study was to investigate whether the TPB helps predicting follow-up care attendance in Swiss AYA cancer survivors. Specifically, we aimed at investigating: i) TPBrelated correlates of the intention to attend follow-up care, and ii) whether the intention is associated with the actual attendance.

Study participants
Eligible survivors had been diagnosed with cancer in the Canton of Zurich, Switzerland, between 1990 and 2005, aged 16-25 years at diagnosis, registered in the Cancer Registry Information on diagnosis (classified according to the International Classification of Childhood  Cancer, third edition (ICCC-3) 21 ), age at diagnosis, treatment, and time since diagnosis were extracted from the registry. Treatment was hierarchically coded as surgery only, chemotherapy (may have had surgery), and radiotherapy (may have had surgery and/or chemotherapy).

Statistical analysis
Descriptive statistics, chi-square tests and t-tests were used to compare participants and non-participants.

Aim 1: Structural equation modelling (SEM)
We used structural equation modelling (SEM) to investigate correlates of the intention to attend follow-up care. SEM allows investigating associations between latent (unmeasured) factors based on measured indicator variables. A two-step approach to SEM was used. 22 First, an adequate measurement model was built; then the TPB-based structural model was added. The final SEM consisted of two parts: a measurement model depicting the relationships between indicators (indicator variables from the questionnaire) and latent factors (correlates and intention), and a structural model representing the paths between latent factors (Figure 1).
The measurement model consisted of four parts representing the four latent factors attitude, norm, control, and intention measured by indicators. Higher scores indicate more positive attitude, more supportive norm, higher control, and higher intention. Cronbach's alpha measuring internal consistency was calculated for the four latent factors. Principal component factor analysis (PCFA) was applied for each factor separately to test whether the indicators loaded on one factor each. A confirmatory factor analysis (CFA) was conducted for the four factors together. Modification indices (MI) based on CFA and theoretical justification were used to modify the measurement model. The structural model consisted of three paths pointing from the correlates to intention ( Figure 1).
Maximum likelihood estimation taking into account all available information in the data in the presence of missing values was applied. Thus, no imputation of missing values was necessary to run the SEM. To estimate the precision of the model parameters the robust Huber/White/sandwich estimator was used.
To assess goodness of fit the following measures were used: Χ² (chi-squared) test statistic, root-mean-square error of approximation (RMSEA) with corresponding 90% confidence interval, and the standardized root-mean-square residual (SRMR) for overall model fit, and Comparative Fit Index (CFI) and Tucker-Lewis Index (TLI) for the comparison with baseline models. 23 For programming reasons the goodness of fit measures had to be obtained without applying the robust standard error estimation.

Aim 2: Logistic regression analysis
We applied multivariable logistic regression analysis to investigate the association between intention and actual attendance to follow-up care (Figure 1). Control and intention were included in the analysis as factor scores derived from the measured indicators. For the calculation of the factor scores, missing values in indicators were imputed with the mean of the available indicators of control or intention, respectively.

Covariates
Socio-demographic and clinical covariates were included in the SEM and multivariable logistic regression analyses if they were associated with intention or attendance in univariate regression analyses at p<0.05 level (

Study participants
Among 469 cancer survivors eligible for the study, 389 (82.9%) could be contacted. Of those, 160 (41.1%) returned the questionnaire and were included in the analysis.
Participants and non-participants were similar regarding sex, age at study, diagnosis, cancer treatment, age at diagnosis and time since diagnosis (Table 1). More than half of the survivors (n=92; 57.5%) reported to attend follow-up care.

Structural equation modelling
Internal consistency measured by Cronbach's alpha was α=0.80 for attitude, α=0.88 for norm, α=0.99 for intention, and α=0.43 for control. According to the PCFA, the indicators for attitude, norm and intention loaded on one factor each ( Table S2 in Supplementary Material). The indicators for control loaded on two factors, one for the indicators C1 and C2, and another one for indicator C3. Therefore, indicator C3 was excluded for further analyses. Internal consistency was α=0.71 for indicators C1 and C2 of control.
The MI of the CFA proposed to free the covariances between two pairs of attitude indicators (indicators A1 and A2: MI=35.5; indicators A3 and A4: MI=29.3). Since these two indicators were also related from a theoretical point of view, the covariances were freed. The loadings of the indicators on the factors were strong and statistically significant (Tables S3, S4 in Supplementary Material).
The results were similar with and without robust standard error estimation with slightly larger confidence intervals without the robust estimation. The CFA showed acceptable to good model fit (23) (Tables S3, S4 in Supplementary Material).

Aim 1: Correlates of intention
Positive attitude towards follow-up care (Coeff.=0.32, 95%CI: 0.05-0.60) and higher supportive norm (Coeff.=0.59, 95%CI: 0.41-0.78) were associated with higher intention to attend follow-up care (Table 2, Figure 1). Control was not associated with intention (Coeff.=-0.13, 95%CI:-0.36-0.10). Having had a relapse was associated with higher intention. The model explained 73.8% of the variance in intention ( Table 2). The results were similar with and without robust standard error estimation. The model showed acceptable to good model fit. 23

Aim 2: Intention and actual attendance
Both, higher intention to attend follow-up care (OR=13.77, 95% CI: 5.58-33.95) and higher control (OR=6.30, 95% CI: 2.26-17.53) were associated with actual attendance (Table 3, Figure 1). Having late effects, having had a relapse and higher fear of detecting late effects during follow-up care were associated with attendance to follow-up care (Table 3).

Discussion
We found that positive attitudes towards follow-up care and the perception that the social environment expects the survivor to attend follow-up care (supportive norms) were associated with higher intention to attend follow-up care in Swiss AYA cancer survivors. Survivors with a higher intention to attend follow-up care and those who perceived it to be easier to attend (higher perceived control) were more likely to actually attend follow-up care.
Facilitating and promoting positive attitudes towards follow-up care might help to increase follow-up care attendance in AYA cancer survivors. With more time passed since diagnosis, attendance rates to follow-up care decrease, 11-14 and also attitudes towards follow-up care worsen. 4 Thus, it might be especially important to sustain and promote positive attitudes among long-term survivors. Also attitudes of health care professionals towards follow-up care might influence attitudes of survivors and thus follow-up care attendance: other studies found that reasons to not attend follow-up care were that survivors felt that health care professionals would not have enough time to provide care, 10 or that there was no guidance from their oncologist regarding follow-up care. 15 Thus, survivors but also health care professionals should be continuously informed about the importance of follow-up care. 13,15 This might be achieved by implementing survivorship care plans, 24 as there seems to be a lack of age-appropriate information for AYA cancer survivors. 4, 25-32 Supportive social norms might also increase follow-up care attendance. Although AYAs are mature, parents play an important role during but also after the end of treatment. 33 We found that expressing the importance of follow-up and the expectation that survivors attend follow-up increased the intention to attend follow-up. This is in line with a reason for not attending medical out-patient visits found in another study: non-attending survivors perceived that health care professionals felt survivors would not need follow-up care. 13 Thus, families and friends of survivors and health care professionals should be advised about their role in supporting survivors to attend follow-up care what is expressing the importance and expectation of attendance.
We did not find perceived control being associated with the intention to attend follow-up but with actual attendance. Survivors generally reported finding it easy to attend follow-up care (Table S1 in Supplementary Material). These results might mirror the fact that due to the small country size, well-maintained public transport system and mandatory health insurance in Switzerland it is straightforward to attend follow-up care if a survivor is willing to do so. Thus, logistics, health care system or insurance related barriers to attend medical care identified in other studies 13-15 might not hold for Switzerland. Perceived control might still be increased by providing survivors with guidance regarding follow-up care by for instance implementing survivorship care plans. 24 Survivorship care plans might help empowering survivors to take responsibility for their health, for instance by attending regular follow-up care. Furthermore, high perceived control is only possible if follow-up care is available and ideally tailored to the needs of AYA cancer survivors. In other studies, survivors reported the need for age-appropriate follow-up care that is accessible, affordable and flexible to be compatible with work and family. 4, 13, 25, 34 Regarding organization of follow-up, care led by an oncologist in an adult hospital was preferred, 11, 35 a multidisciplinary team involved was favoured as well. 11, 30, 33 Medical reasons for follow-up were rated higher than supportive reasons by both, survivors 11, 35 and health care professionals. 36 Thus, follow-up care should include the provision of diverse information including former cancer disease and treatment, 24, 32 and monitoring for late effects. But follow-up care should also include psycho-social support services outside the clinical setting, and cover aspects such as insurance or employment issues. 4,35 The presence of late effects, having had a relapse, and higher fear of detecting late effects during follow-up were associated with actual attendance in our study. Suffering from late effects and relapse of cancer make it necessary to seek medical care. A main aim of followup care is to detect late effects. The fear of detecting late effects might thus be due to this awareness. On the other hand, survivors attending care might be more anxious in general, and have a higher need for monitoring their health. Our findings are in line with a study in Canada reporting that most common reasons for medical care visits in AYA cancer survivors were secondary neoplasms and other health-related symptoms. 12 In contrast, a study in the USA found the fear of new cancer being a barrier to attend follow-up care. 15 This contrast might be due to a culturally influenced higher need for monitoring and higher perceived control to attend care in Switzerland compared to the USA.
A limitation of our study is its cross-sectional design. Therefore, we are not able to draw conclusions on the temporality of the investigated TPB-related correlates and intention and actual follow-up care attendance. Furthermore, follow-up attendance was self-reported and might not reflect actual attendance correctly. The response rate of 41.1% is relatively low, but in the same range as response rates in questionnaire studies among childhood cancer survivors. 37 Also, AYAs are a heterogeneous and mobile population 34 difficult to reach and include into studies. 38,39 A strength of our study is the use of the widely used TPB 40 to investigate correlates of intention and actual attendance in AYA cancer survivors. Another strength is the populationbased sampling of AYA cancer survivors. Our study participants were comparable to nonparticipants suggesting that the study sample is representative for AYA cancer survivors in a large and diverse region of Switzerland. Furthermore, we observed similar results in a cohort of Swiss childhood cancer survivors 17 , what strengthens our findings.
In conclusion, we found positive attitudes and supportive social norms being associated with higher intention to attend follow-up care, and higher intention being associated with actual follow-up care attendance. Increasing awareness of the importance and benefits of follow-up care among survivors but also family, friends and health care professionals may help to increase follow-up care attendance among AYA cancer survivors.     Abbreviations: SD, standard deviation.

Assessment of correlates, intention and actual follow-up care attendance
Correlates and intention to attend follow-up care were assessed using 7-point Likert scales.

3/8
Internal consistency and principal component factor analyses