- Email: [email protected]
Time from first symptom experience to help seeking for colorectal cancer patients: Associations with cognitive and emotional symptom representations
Patient Education and Counseling 99 (2016) 807–813
Contents lists available at ScienceDirect
Patient Education and Counseling journal homepage: www.elsevier.com/locate/pateducou
Time from first symptom experience to help seeking for colorectal cancer patients: Associations with cognitive and emotional symptom representations Line Flytkjær Jensena,b,1,* , Line Hvidberga,b,1, Anette Fischer Pedersena , Arja R. Aroc , Peter Vedsteda a Research Centre for Cancer Diagnosis in Primary Care (CaP), Research Unit for General Practice, Department of Public Health, Aarhus University, Bartholins Allé 2, 8000 Aarhus C, Denmark b Section for General Medical Practice, Department of Public Health, Aarhus University, Bartholins Allé 2, 8000 Aarhus C, Denmark c Unit for Health Promotion Research, University of Southern Denmark, Niels Bohrs Vej 9–10, 6700 Esbjerg, Denmark
A R T I C L E I N F O
A B S T R A C T
Article history: Received 29 June 2015 Received in revised form 12 November 2015 Accepted 14 November 2015
Objectives: The aim was to assess the association between cognitive and emotional symptom representations prior to diagnosis and the length of the patient interval (i.e. the time from the first symptom is experienced until healthcare is sought) for colorectal cancer patients. Method: The study population included 436 newly diagnosed colorectal cancer patients. Questionnaire data were collected using the Danish Revised Illness Perception Questionnaire (IPQ-R), including cognitive and emotional symptom representations and information on the patient interval. Results: High score in treatment control was associated with short patient interval (PR = 0.52, 95% CI: 0.31–0.89) and high score on the timeline cyclical dimension was associated with long patient interval (PR = 2.14, 95% CI: 1.29–3.57). Hence, patients with negative beliefs about the treatability of their symptoms and patients with strong beliefs about the cyclical nature of their symptoms were more likely to have a long patient interval. Assigning blood in stool as the most important symptom significantly interacted in the association between the patient interval and the two cognitive symptom representations consequence and personal control. Conclusion: The results indicate that aspects of symptom representations were associated with the patient’s help-seeking. Practical implications: These findings may help clinicians and public health planners shorten patient intervals. ã 2015 Elsevier Ireland Ltd. All rights reserved.
Keywords: Cancer Patient interval Illness Perception Questionnaire Symptom representations Common-sense model Background
1. Background Colorectal cancer is one of the most commonly experienced types of cancer in Denmark [1]. For Danish colorectal cancer patients diagnosed during 2000–2002, the age-standardised relative 5-year survival was estimated to be 51.7%, which was significantly lower than in comparable countries, such as Sweden, Canada and Australia [2]. In Denmark, 25% of colorectal cancer patients are treated in stage IV in the tumour, node, metastasis
* Corresponding author. Fax: +45 86124788. E-mail addresses: [email protected] (L.F. Jensen), [email protected] (L. Hvidberg), [email protected] (A.F. Pedersen), [email protected] (A.R. Aro), [email protected] (P. Vedsted). 1 These authors contributed equally to this work. http://dx.doi.org/10.1016/j.pec.2015.11.013 0738-3991/ ã 2015 Elsevier Ireland Ltd. All rights reserved.
(TNM) system, while only 16% are treated in TNM stage I [3]. In addition, Denmark has a less favourable stage distribution than other countries [4]. Later diagnosis is thus believed to be a potential explanation for the poorer prognosis among cancer patients in Denmark compared with other countries [2,5]. This may occur as a result of longer patient intervals sometimes referred to as patient delay (i.e. the time from the first symptom is experienced until healthcare is sought) [6]. In Denmark, population screening for colorectal cancer using immunochemical faecal occult blood test (iFOBT) has been introduced in 2014 to reduce colorectal mortality through early detection [7]. Nevertheless, it is estimated that approximately 75% of colorectal cancer patients will still be detected through symptomatic presentation [8], making the patient interval essential. If we are able to identify factors that influence the patient interval, this knowledge may inform health workers and policy
808
L.F. Jensen et al. / Patient Education and Counseling 99 (2016) 807–813
makers in their efforts to reduce the length of the interval and as Torring et al. [9,10] have made clear, time matters for the cancer outcome. Leventhal’s common-sense model (CSM) is a theoretical model that has been suggested as a useful tool to identify factors related to the patient interval after self-discovery of a cancer symptom [11]. The CSM provides a framework to explain how people interpret and cope with health problems [12], and it builds on the proposition that individuals create their own common sense representations of a health problem to guide their coping efforts [13]. These representations involve the creation of cognitive perceptions of identity, timeline, consequences, control/cure and causes of the health problem and also the creation of an emotional representation of the health problem [14]. In relation to the components identity and consequences, several studies have found that patients who do not attribute their symptom to cancer or do not perceive their symptom as serious have a longer patient interval than patients who perceive their symptoms as serious or specifically as cancer [15–19]. Conversely, Pedersen et al. [20] found that experiencing the alarm symptom, rectal bleeding, was associated with long patient intervals in colorectal cancer patients although many of these patients reported to have wondered if their symptom could be attributable to cancer. The timeline component of the CSM [14] is also relevant as patients who perceive their symptoms as temporary rather than permanent have been found to have a longer patient interval [15]. Thus, previous studies have used components comparable to those from the CSM, but have not explicitly been based on Leventhal’s theoretical framework. However, several researchers have called for theoretically based studies on the patient interval [21–23] and Leventhal’s CSM has been suggested as a useful tool in this connection [23]. In this study we have used the revised Illness Perception Questionnaire (IPQ-R) [24] to assess the representations of the CSM, which is also considered the measure of choice in other studies using Leventhal’s theoretical framework [25,26]. The aim of this study was to investigate the association between the patient interval and cognitive and emotional symptom representations prior to colorectal cancer diagnosis. A further aim was to examine whether experiencing blood in stool (ranked as the most important symptom) was a potential effect modifier for the association between symptom representation and the length of the patient interval. 2. Methods 2.1. Study population The study population consisted of patients who had been registered with histologically confirmed colorectal cancer in the Danish Pathology Data Bank (DPDB) between 1 January and 1 May 2010. The DPDB is a nationwide online database containing detailed information on all pathology specimen analysed in Denmark [27]. The database is updated automatically when a pathology analysis is completed at one of the Danish hospitals [28] and for this study data were retrieved using the Danish version of the Systemized Nomenclature of Medicine (SNOMED) codes for colorectal cancer: T67*3, T68*3 and T73970. The data collection took place in the period from 12 July to 26 August 2010. In total, 1105 patients were identified in the DPDB (study base). In this present study, 206 (18.6%) were excluded because of death, unknown address or research protection (i.e. residents holding publicly recorded protection from research participation). The remaining 899 patients received a questionnaire; non-respondents received a reminder, including a new copy of the questionnaire, three weeks later. A total of 577 completed questionnaires were received during the data collection period (response rate: 64.2%).
The following groups of respondents were excluded from the analyses: 47 respondents (8.1%) were excluded because they indicated that they did not have any preceding symptoms of cancer before contacting a general practitioner (GP), 52 respondents (9.0%) because they had not stated which of their symptoms they perceived as their most important and 42 respondents (7.3%) were excluded as more than 50% of their responses were missing. Hence, a total of 436 respondents were included in the statistical analyses. 2.2. The IPQ-R The revised Illness Perception Questionnaire (IPQ-R) [24], a quantitative measure derived from the CSM, was used to measure cancer patients’ representations of their symptoms prior to diagnosis. The IPQ-R consists of nine subscales, including identity (perceptions of the symptoms associated with the health problem), timeline acute/chronic dimension (perception of the chronicity of the health problem), consequences (anticipated and experienced consequences of the health problem), personal control (perceptions of own ability to control the health problem), treatment control (perceptions regarding treatment for controlling the health problem), coherence (understanding of the health problem), timeline cyclical dimension (perceptions about the stability or changeability of the health problem), emotional representation (emotional responses to the health problem) and cause (perceived cause of the health problem) [24]. We adapted the IPQ-R to measure symptom representations among patients with colorectal cancer symptoms and validated the instrument in this setting [29]. In this modified IPQ-R, the identity subscale comprised 13 commonly experienced symptoms among colon and rectal cancer patients. Before answering the modified IPQ-R, patients were asked to think back on the time before contacting their GP. First patients were asked to rate whether they had experienced each symptom and then whether they believed that the symptom was related to their cancer. The patients were then asked to state which symptom they perceived as the most important and to think of this symptom when assessing the remaining questionnaire. Following the identity subscale, the cognitive representations were listed (timeline acute/ chronic dimension, consequences, personal control, treatment control, coherence, timeline cyclical dimension and emotional representations) and included 32 statements. All statements were rated on a 5-point Likert scale: strongly disagree, disagree, neither agree nor disagree, agree and strongly agree. The scores of each subscale were calculated as stated by Moss-Morris [30]. 2.3. Dependent variables The patient interval was calculated from the dates entered in the questionnaire by the patients. Thus, the number of days between the date of the first symptom experience and the date of the first symptom-related contact to the GP was computed for each patient and subsequently dichotomised using the 75th percentile as the cut-off value, which was 88 days on average for all colorectal cancer patients. 2.4. The IPQ-R subscales Based on whether the patients believed that any of their symptoms were related to cancer before their GP contact, the identity subscale was divided into patients who ‘considered cancer’ and patients who ‘did not consider cancer’. The subscale ratings on the timeline acute/chronic dimension, consequences, personal control, treatment control, coherence, timeline cyclical dimension and emotional representation were divided into three groups, i.e. low, middle and high scores, using the 25th and 75th
L.F. Jensen et al. / Patient Education and Counseling 99 (2016) 807–813
percentile. Higher scores indicate strong beliefs about the chronicity of the symptom, the potential negative consequences of the symptom, the personal controllability of the symptom, the perceived treatability of the symptom, the personal understanding of the symptom, the cyclical nature of the symptom and a high degree of emotional distress. From the causal subscale, only one attribution was chosen to analyse i.e. cancer. This was dichotomised for each patient into ‘agree’, ‘neither nor’ and ‘disagree’ in accordance with their belief in whether their most important symptom was related to cancer. Lastly, the variable blood in stool (yes–no) was generated if the patients had assessed this symptom to be the most important of the experienced symptoms in the questionnaire. 2.5. Socio-demographic variables Due to the Danish Civil Registration System (CRS) [31], it was possible to link survey data to socio-demographic variables at the individual level through Statistics Denmark [32]. The following variables were included: age (64 and >64 years), gender (male and female), marital status (married/cohabiting and living alone) and educational level (low: 10 years, middle: > 10 15 years and high: >15 years) according to UNESCO’s classification [33]. 2.6. Statistical analysis Data were analysed using Stata version 13.1. First, the median patient interval was computed with interquartile interval (IQI) for each of the independent variables. The Wilcoxon and the Kruskal– Wallis rank-sum tests were used to compare median patient interval for the different groups, and a p-value of .05 was used to signify statistical significance. Prevalence ratios (PRs) with 95% confidence intervals (CIs) were used to determine the association between independent variables and patient interval of 88 days. Unadjusted analyses were carried out with each of the independent variables, and an adjusted model was used to control for possible confounding by age, gender, marital status, educational level and blood in stool. Based on previous findings [20], we hypothesised that the symptom blood in stool could be a potential effect modifier for the association between symptom representation and patient interval. Consequently, we performed interaction analyses for each subscale.
Table 1 Socio-demographic characteristics of the respondents included in this study (n = 436) and the study base (n = 1105). Numbers vary due to missing data. Socio-demographic characteristics
Respondents
Study base
%
(n)
%
(n)
Age group 64 years >64 years Age, mean (SD)
35.1 64.9 67.7
(153) (283) (11.0)
27.4 72.6 71.0
(303) (802) (11.3)
Gender Male Female
55.7 44.3
(243) (193)
53.1 46.9
(572) (506)
Marital status Married/cohabiting Living alone
68.8 31.2
(300) (136)
60.7 39.3
(654) (424)
Educational level Low Middle High
35.1 46.8 18.1
(149) (199) (77)
43.9 42.5 13.6
(453) (439) (140)
809
3. Results 3.1. Socio-demographic characteristics of respondents and study base The characteristics of the respondents (n = 436) and the study base (n = 1105) are shown in Table 1. Respondents were generally younger than the study base. Respondents also included more males, more married/cohabiting residents and more people with medium or high-level education. 3.2. Symptoms Patients reported on average 3.5 symptoms (median 3, interquartile range 2–5). Blood in stools was experienced by 47% of the patients and rated as the most important by 32.8% of the patients. Changes in bowel habits was rated as the most important by 28.9% of the patients, 22.0% rated “other symptoms” such as weight loss and fatigue as the most important, and finally 16.3% rated abdominal pain as their most important symptom (data not shown). 3.3. Median patient interval The median patient interval was 30 days for the entire population (data not shown). Table 2 shows the median patient interval with interquartile interval for each of the IPQ-R subscales, for having blood in stool and for included socio-demographic variables. Statistically significant differences in patient interval were found for three cognitive subscales: consequences, treatment control and timeline cyclical dimension. Respondents with the lowest treatment control scores had a longer patient interval (35.5 days) than respondents with the highest treatment control scores (17 days). In addition, the median patient interval was 18 (IQI 2-62) and 61 days (IQI 14-148) for respondents with lowest and highest scores for the timeline cyclical dimension, respectively. 3.4. Association between symptom representations and long patient interval Table 3 presents unadjusted and adjusted associations between symptom representations and long patient interval (88 days, the 75th percentile). High score on treatment control was statistically significantly associated with lower likelihood of long patient interval (PRadj = 0.52, 95% CI: 0.31–0.89). High score on the timeline cyclical dimension had a PR of 2.14 (95% CI: 1.29–3.57) for a long patient interval compared to respondents with low scores. 3.5. Interaction between blood in stool, symptom representation and patient interval Blood in stool (as the most important symptom) interacted statistically significantly in the association between consequence and patient interval (p = 0.036) and between personal control and patient interval (p = 0.017). For patients scoring in the middle range of the consequence scale, the stratified analyses (Table 4) showed that the likelihood for a long patient interval decreased if the most important symptom was blood in stool, which was not seen for patients without blood in the stool. 4. Discussion 4.1. Main findings In this study on symptom representations and patient interval among colorectal cancer patients, we found that the two cognitive symptom representations treatment control and timeline cyclical
810
L.F. Jensen et al. / Patient Education and Counseling 99 (2016) 807–813
Table 2 Median patient interval (in days) and interquartile interval (IQI) for IPQ-R subscales, blood in stool and socio-demographic variables. Colorectal cancer patients (n = 488) Scale (possible score ranges)
Median patient interval IQI
Identity Considered cancer Not considered cancer
30 31
4–78 4–93
Timeline acute/chronic (7–35) <25% (7–13) 25-75% (14–20) >75% (21–35)
28 31.5 21
3–94.5 7–92 3–66
Consequences (4–20) <25 % (4–6) 25-75% (7–11) >75% (12–20)
39 27 31
7–138 2–65 7–87
Personal control (4–20) <25% (4–7) 25-75% (8–12) >75% (13–20)
30 31 28
3–100 4.5–88 7–88
Treatment control (3–15) <25 % (3–9) 25-75% (10–12) >75% (13–15)
35.5 31 17
7–160.5 4.5–90.5 2–61
Coherence (5–25) <25% (5–10) 25-75% (11–16) >75% (17–25)
35 30 22
7–105 5–87 3–94
Timeline cyclical (4–20) <25% (4–8) 25-75% (9–14) >75% (15–20)
18 29 61
2–62 5–81 14–148
P valuea 0.70
0.379
0.045
0.976
0.008
0.480
0.003
Emotional representation (5–25) <25% (5–10) 25-75% (11–17) >75% (18–25)
0.061 13.5 31 34
2–72.5 5–97 8–88
Attribution to cancer Agree Neither nor Disagree
22 35 31
4–66 10–77.5 4–106.5
Blood in stool No Yes
31 30
4–82 6–100
Age 18–64 years >64 years
32 30
5–108 4–76
Gender Male Female
31 30
4–91 4–90
Marital status Married/cohabiting Living alone
31 30
5–92 3–85
Educational level Low Middle High
25 30.5 36
5–70 3.5–85.5 9–139
0.224
0.348
0.145
0.407
0.469
0.089
a Wilcoxon rank-sum test and the Kruskal Wallis rank-sum test (for variables with three categories).
dimension were statistically significantly associated with a long patient interval (88 days or more). Thus, patients reporting that treatment would not be effective and that nothing could be done to
ease their symptom were more likely to have a long patient interval than patients reporting high treatment control. Further, patients for whom symptoms came and went or changed from day to day were also more likely to have a long patient interval than patients who did not report cyclical changes in symptoms. Assigning blood in stool as the most important symptom significantly interacted with the two cognitive symptom representations consequence and personal control. A stratified analysis showed that for patients scoring in the middle range of the consequence scale the likelihood for a long patient interval decreased if the patient had stated blood in stool as the most important symptom. 4.2. Comparison with existing literature In a Danish study of newly diagnosed colorectal cancer patients [34], it was found that 25% had a patient interval of more than 56 days, i.e. about one month shorter than the patient interval found in the present study. This discrepancy can be explained by several factors. First of all, there is no standardised and validated way to measure the patient interval, and the two mentioned studies used different approaches. In the present study, the patient interval was based on dates reported by the patients, whereas, in the study by Hansen et al. [34], the GP provided the date of the first symptom presentation based on the medical record. However, a study analysing the level of agreement between patient-reported and GP-reported patient intervals [35] found that the GPs systematically reported a longer patient interval than the patient. Hence, other conditions such as different sampling procedures may explain this finding. Both patient interval and symptom representations were reported by patients who had recently received their cancer diagnosis. Consequently, the dates and the symptom representations may have been influenced by patient’s emotional turmoil. For instance, some patients might have tried to legitimise or rationalise their – in retrospect – late healthcare-seeking [36] and have underestimated the length of their patient interval. This could underestimate the associations with some of the symptom representations. Furthermore, non-response might have been more widespread among patients than among GPs. Similar to a population-based study among breast cancer patients in Germany [15], which found that the second most common rationale for a long patient interval was considering symptoms as temporary, we found that patients were more likely to report a long patient interval if they experienced cyclical symptoms (i.e. symptoms came and went). This may be a natural consequence of the quest for common sense, i.e. most people use certain coping efforts when dealing with a health problem (e.g. wait and see), however it is important that the public are aware that symptoms of cancer may have a cyclical nature. We found that patients who believed that their symptom was untreatable were more likely to report a long patient interval than patients reporting that something could be done to ease their symptom. As patients with a colorectal cancer diagnosis are often elderly, they may perceive some of the symptoms (such as fatigue and loss of appetite or weight) as normal signs of the ageing process. Thus, other data from our study show that about 20% of the respondents attributed their most important symptom to ageing. Therefore, these patients may believe that nothing can be done to ease their symptom. However, healthcare seeking can possibly be promoted if this age group becomes more aware of the increased risk of cancer with age. Thus, a Danish population-based study found that only one in four is aware of the growing risk of cancer with age [37]. The emotional representations related to the most important symptom were not associated with the patient interval. The CSM suggests that emotional representations may be an independent
L.F. Jensen et al. / Patient Education and Counseling 99 (2016) 807–813 Table 3 Patient interval of 88 days unadjusted and adjusted associations with IPQ-R subscales, blood in stool and socio-demographic variables.
Symptom representations
Prevalence ratio for patient interval 88 days PRunadj. (95% CI) PRa adj. (95% CI)
Identity Considered cancer Not considered cancer
1.00 1.11 (0.77–1.60)
1.00 1.13 (0.77–1.66)
1.00 0.94 (0.64–1.39)
Consequences (4–20) <25% (4–6) 25–75% (7–11) >75% (12–19)
1.00 0.67 (0.45–0.99) 0.73 (0.47–1.13)
1.00 0.72 (0.49–1.07) 0.84 (0.53–1.35)
Personal control (4-20) <25% (4–7) 25–75% (8–12) >75% (13–20)
1.00 0.86 (0.57–1.29) 0.89 (0.54–1.47)
1.00 0.92 (0.61–1.38) 0.97 (0.59–1.59)
Treatment control (3–15) <25% (3–9) 25–75% (10–12) >75% (13–15)
1.00 0.79 (0.54–1.16) 0.62 (0.37–1.04)
1.00 0.76 (0.53–1.10) 0.52 (0.31–0.89)
Coherence (5–25) <25% (5–10) 25–75% (11–16) >75% (17–25)
1.00 0.84 (0.56–1.25) 0.90 (0.57–1.43)
1.00 0.76 (0.51–1.13) 0.69 (0.42–1.13)
Timeline cyclical (4–20) <25% (4–8) 25–75% (9–14) >75% (15–20)
1.00 1.23 (0.73–2.05) 2.01 (1.18–3.42)
1.00 1.23 (0.73–2.05) 2.14 (1.29–3.57)
Emotional representation (5–25) <25% (5–10) 1.00 25–75% (11–17) 1.25 (0.77–2.03) >75% (18–25) 1.15 (0.66–2.02)
1.00 1.36 (0.83–2.21) 1.32 (0.75–2.32)
Attribution to cancer Agree Neither Disagree
1.00 1.11 (0.61–1.99) 1.38 (0.92–2.08)
1.00 1.14 (0.63–2.06) 1.46 (0.96–2.22)
Blood in stool No Yes
1.00 1.19 (0.83–1.68)
1.00 1.23 (0.87–1.74)
Adjusted for age, gender, marital status, educational level and blood in stool.
Table 4 Analyses for statistically significantly interacting symptom representations stratified on blood in stool. Prevalence ratio for long patient interval Symptom representations
Blood in stool PRa adj. (95% CI)
No blood in stool PRa adj. (95% CI)
Consequences (4–20) <25% (4–7) 25–75% (8–12) >75% (13–20)
1.00 0.51 (0.28–0.92) 0.46 (0.17–1.27)
1.00 0.92 (0.52–1.63) 1.09 (0.60–1.99)
Personal control (4–20) <25% (4–7) 25–75% (8–12) >75% (13–20)
1.00 0.61 (0.32–1.16) 0.85 (0.42–1.72)
1.00 1.38 (0.72–2.65) 1.26 (0.78–2.74)
a
patients included in this study may have reacted differently to the emotional distress caused by their symptom because of differences in coping strategies and personal resources. This, in turn, may have led some patients to seek medical attention quickly and some patients to wait and see, which may explain why we found no associations between emotional representations and a long patient interval. 4.3. Strengths and limitations
Timeline acute/chronic (7–35) 1.00 <25% (7–13) 25–75% (14–20) 0.99 (0.66–1.49)
a
811
Adjusted for age, gender, marital status and educational level.
motivator of healthcare seeking [38], but no association with the patient interval was found. This may be explained by the fact that emotions such as fear, worry and anxiety can be motivating as well as hampering factors for healthcare seeking [18,39,40]. Therefore,
This study is unique in using an established theoretical framework to study the relationship between cognitive and emotional symptom representations and patient interval among colorectal cancer patients. Such studies have formerly been called for [21,22]. The use of the validated IPQ-R instrument, which is specifically constructed to assess representations of the health problems described in Leventhal’s CSM, is seen as a main strength of the study and has previously been suggested as a theoretical framework when studying patient interval [23]. The DPDB was used to identify incident cancer patients. The DPDB contains data of high quality on pathology investigations and diagnoses [27]. Data for approximately one third of all incident cancer patients were excluded as only one biopsy was recorded, and the cancer was hence not histologically confirmed. It is unclear how this may have affected the results as the patient interval of these patients cannot be predicted. Patient intervals were dichotomized at the 75th interval as this study focused on factors associated with long patient intervals. Dichotomizing data introduces the risk of losing information, however, as data on the patient intervals was not normally distributed statistical analysis performed on continuous interval data were not advisable. The central limitation of this study was the retrospective nature of the assessment of patient interval, symptoms and symptom representations. Thus, the respondents were already patients when answering the questionnaire and therefore this study rely on their ability to distinguish between their experiences before and after their diagnosis. Thus, recall bias cannot be excluded [41,42], as the dates used to calculate the patient interval solely relied upon the information given by the respondents. Thus, it cannot be ruled out that some patients underestimated the interval and this may infer that true associations between symptom representations and long patient interval have not been detected or have been underestimated. However, other study designs also introduce methodological challenges. A prospective design would be challenging because when following a large population only very few will develop cancer and thus be eligible for study. Therefore the chosen method may be the most feasible and fortunately validation of the psychometric properties of the Danish modified IPQ-R [29], showed for instance that experiencing the symptom blood in stools was correlated with attributing this to hemorrhoids, scratch or chink and cancer indicating that patients were actually thinking of their symptom instead of their disease when answering the questionnaire. A response rate of 64% was achieved, which is higher than reported in other studies in this area [20,43]. Nevertheless, a third of the eligible patients did not participate and differed from the study base. In addition, potential for selection bias exists as patients who died shortly after receiving the diagnosis were not included in the study and possibly people who were very sick were not able to fill in the questionnaire/participate either. These patients are likely to have had more aggressive disease. It could be hypothesized that these patients e.g. perceived their symptom as more serious (as they were more ill than those who survived) or they experienced more distressing emotional representations. However, their estimated patient interval could be either longer or
812
L.F. Jensen et al. / Patient Education and Counseling 99 (2016) 807–813
shorter than the respondents. Thus, it is not possible to say in what direction the estimates may have been affected. Lastly, due to exclusion of several patient’s the final sample size is relatively small, which has led to decreased precision of the sample estimates. Thus, the relatively few statistically significant associations between the patient interval and the IPQ-R subscales may be explained by a type II error and future studies in this area should include a larger sample. 5. Conclusion This study on patients with colorectal cancer showed that the patient interval tended to be longer for patients who believed that nothing could be done to ease their symptoms and patients who stated that their symptoms came and went. Implications The findings from this study may feed directly into the clinical and public health settings, where information about the nature and normalisation of symptoms can be addressed. However, more research is needed to answer what these associations between long patient interval and the two cognitive symptom representations treatment control and timeline cyclical dimension are functions of. If the association between high scores on the timeline cyclical dimension and long patient interval is a function of lack of knowledge that symptoms of cancer may have a cyclical nature, it may be worthwhile to educate the public about this. What is certain is that the patient interval in colorectal cancer can be improved and must be addressed if outcomes of colorectal cancer are to be improved. Conflicts of interest None. Author contribution L.F.J. and L.H. contributed equally to this work. L.F.J., L.H., A.F.P., A.R.A. and P.V. all conceived the study, participated in its design and helped draft the manuscript. L.F.J. and L.H. conducted the survey. All authors read and approved the final manuscript. Funding and ethical approvals This research was financially supported by the Danish Cancer Society, the Novo Nordisk Foundation and the Research Centre for Cancer Diagnosis in Primary Care (CaP). The study was approved by the Danish Data Protection Agency (J. no. 2010-41-4664). Informed consent I confirm all patient/personal identifiers have been removed or disguised so the patient/person(s) described are not identifiable and cannot be identified through the details of the story. Acknowledgements We would like to thank the patients who participated in the study. Also, a special thank goes to Henry Jensen for assisting in the data collection. Finally, we thank the Danish Cancer Society and the Novo Nordic Foundation for the financial support that made this project possible.
References [1] Nordcan Cancer stat fact sheets. Denmark—colorectal, 2014. [2] M.P. Coleman, D. Forman, H. Bryant, J. Butler, B. Rachet, C. Maringe, U. Nur, E. Tracey, M. Coory, J. Hatcher, C.E. McGahan, D. Turner, L. Marrett, M.L. Gjerstorff, T.B. Johannesen, J. Adolfsson, M. Lambe, G. Lawrence, D. Meechan, E.J. Morris, R. Middleton, J. Steward, M.A. Richards, ICBP Module 1 Working Group, Cancer survival in Australia, Canada, Denmark, Norway, Sweden, and the UK, 1995–2007 (the International Cancer Benchmarking Partnership): an analysis of population-based cancer registry data, Lancet 377 (2011) 127–138. [3] Danish Colorectal Cancer Group (DCCG), National årsrapport 2013: Landsdækkende database for kræft i tyk-og endetarm [In Danish]. Herlev: Dansk Kolorektal Cancer Database, 2014. [4] C. Maringe, S. Walters, B. Rachet, J. Butler, T. Fields, P. Finan, R. Maxwell, B. Nedrebo, L. Pahlman, A. Sjovall, A. Spigelman, G. Engholm, A. Gavin, M.L. Gjerstorff, J. Hatcher, T.B. Johannesen, E. Morris, C.E. McGahan, E. Tracey, D. Turner, M.A. Richards, M.P. Coleman, ICBP Module 1 Working Group, Stage at diagnosis and colorectal cancer survival in six high-income countries: a population-based study of patients diagnosed during 2000–2007, Acta Oncol. 52 (2013) 919–932. [5] F. Berrino, R. De Angelis, M. Sant, S. Rosso, M. Bielska-Lasota, J.W. Coebergh, M. Santaquilani, EUROCARE Working group, Survival for eight major cancers and all cancers combined for European adults diagnosed in 1995–99: results of the EUROCARE-4 study, Lancet Oncol. 8 (2007) 773–783. [6] D. Weller, P. Vedsted, G. Rubin, F.M. Walter, J. Emery, S. Scott, C. Campbell, R.S. Andersen, W. Hamilton, F. Olesen, P. Rose, S. Nafees, E. van Rijswijk, S. Hiom, C. Muth, M. Beyer, R.D. Neal, The Aarhus statement: improving design and reporting of studies on early cancer diagnosis, Br. J. Cancer 106 (2012) 1262–1267. [7] The Danish Cancer Society. Screening for colon cancer, 2015. [8] J.S. Juul, P. Vedsted, Uncharacteristic colorectal cancer symptoms and their value in general practice, Ugeskr. Laeger 174 (2012) 710–713. [9] M.L. Torring, M. Frydenberg, W. Hamilton, R.P. Hansen, M.D. Lautrup, P. Vedsted, Diagnostic interval and mortality in colorectal cancer: U-shaped association demonstrated for three different datasets, J. Clin. Epidemiol. 65 (2012) 669–678. [10] M.L. Torring, M. Frydenberg, R.P. Hansen, F. Olesen, P. Vedsted, Evidence of increasing mortality with longer diagnostic intervals for five common cancers: a cohort study in primary care, Eur. J. Cancer 49 (2013) 2187–2198. [11] S.E. Scott, E.A. Grunfeld, M. McGurk, Patient’s delay in oral cancer: a systematic review, Commun. Dent. Oral Epidemiol. 34 (2006) 337–343. [12] H. Leventhal, D. Nerenz, D. Steele, Illness representation and coping with health threats, Handbook of Psychology and Health, Lawerence Eribaum, Hillsdale, 1984, pp. 219. [13] A.M. Barsevick, K. Whitmer, L. Walker, In their own words: using the common sense model to analyze patient descriptions of cancer-related fatigue, Oncol. Nurs. Forum 28 (2001) 1363–1369. [14] E. Leventhal, H. Leventhal, L. Cameron, Representations, procedures, and affect in illness self-regulation: a perceptual-cognitive model, Handbook of Health Psychology, Lawrence Erlbaum, Hillsdale, 2001, pp. 19. [15] V. Arndt, T. Sturmer, C. Stegmaier, H. Ziegler, G. Dhom, H. Brenner, Patient delay and stage of diagnosis among breast cancer patients in Germany—a population based study, Br. J. Cancer 86 (2002) 1034–1040. [16] J. Corner, J. Hopkinson, L. Roffe, Experience of health changes and reasons for delay in seeking care: a UK study of the months prior to the diagnosis of lung cancer, Soc. Sci. Med. 62 (2006) 1381–1391. [17] G. Meechan, J. Collins, K.J. Petrie, The relationship of symptoms and psychological factors to delay in seeking medical care for breast symptoms, Prev. Med. 36 (2003) 374–378. [18] M. O'Mahony, J. Hegarty, Factors influencing women in seeking help from a health care professional on self discovery of a breast symptom, in an Irish context, J. Clin. Nurs. 18 (2009) 2020–2029. [19] M. Pineros, R. Sanchez, R. Cendales, F. Perry, R. Ocampo, Patient delay among Colombian women with breast cancer, Salud Publica Mex. 51 (2009) 372–380. [20] A.F. Pedersen, R.P. Hansen, P. Vedsted, Patient delay in colorectal cancer patients: associations with rectal bleeding and thoughts about cancer, PLoS One 8 (2013) e69700. [21] A. Bish, A. Ramirez, C. Burgess, M. Hunter, Understanding why women delay in seeking help for breast cancer symptoms, J. Psychosom. Res. 58 (2005) 321– 326. [22] S. Scott, M. McGurk, E. Grunfeld, Patient delay for potentially malignant oral symptoms, Eur. J. Oral Sci. 116 (2008) 141–147. [23] S.E. Scott, E.A. Grunfeld, M. McGurk, Patient’s delay in oral cancer: a systematic review, Commun. Dent. Oral Epidemiol. 34 (2006) 337–343. [24] R. Moss-Morris, J. Weinman, K.J. Petrie, R. Horne, L.D. Cameron, D. Buick, The revised illness perception questionnaire (IPQ-R), Psychol. Health 17 (2002) 1. [25] M. Fischer, M. Scharloo, J. Abbink, A. van ‘t Hul, D. van Ranst, A. Rudolphus, J. Weinman, K. Rabe, A.A. Kaptein, The dynamics of illness perceptions: testing assumptions of Leventhal’s common-sense model in a pulmonary rehabilitation setting, Br. J. Health. Psychol. 15 (2010) 887–903. [26] M.S. Hagger, S. Orbell, Illness representations and emotion in people with abnormal screening results, Psychol. Health 21 (2006) 183–209. [27] R. Erichsen, T.L. Lash, S.J. Hamilton-Dutoit, B. Bjerregaard, M. Vyberg, L. Pedersen, Existing data sources for clinical epidemiology: the Danish National Pathology Registry and Data Bank, Clin. Epidemiol. 2 (2010) 51–56.
L.F. Jensen et al. / Patient Education and Counseling 99 (2016) 807–813 [28] R. Erichsen, T.L. Lash, S.J. Hamilton-Dutoit, B. Bjerregaard, M. Vyberg, L. Pedersen, Existing data sources for clinical epidemiology: the Danish National Pathology Registry and Data Bank, Clin. Epidemiol. 2 (2010) 51–56. [29] L. Hvidberg, L.F. Jensen, A.F. Pedersen, A.R. Aro, P. Vedsted, Measurement properties of the Danish version of the Illness Perception QuestionnaireRevised for patients with colorectal cancer symptoms, J. Health. Psychol. 19 (2014) 1279–1290. [30] R. Moss-Morris, The Illness Perception Questionnaire 2011, 2014. [31] C.B. Pedersen, The Danish Civil Registration System, Scand. J. Public Health 39 (2011) 22–25. [32] Statistics Denmark. http://www.dst.dk/en/OmDS (assessed November 09.11.15.). [33] Unesco Institute for Statistics. ISCED: International Standard Classification of Education. http://www.uis.unesco.org/Education/Pages/internationalstandard-classification-of-education.aspx (assessed 09.11.15.). [34] R.P. Hansen, P. Vedsted, I. Sokolowski, J. Sondergaard, F. Olesen, Time intervals from first symptom to treatment of cancer: a cohort study of 2212 newly diagnosed cancer patients, BMC Health Serv. Res. 11 (2011) 284. [35] M.B. Larsen, R.P. Hansen, I. Sokolowski, P. Vedsted, Agreement between patient-reported and doctor-reported patient intervals and date of first symptom presentation in cancer diagnosis—a population-based questionnaire study, Cancer Epidemiol. 38 (2014) 100–105.
813
[36] R.S. Andersen, B. Paarup, P. Vedsted, F. Bro, J. Soendergaard, ‘Containment' as an analytical framework for understanding patient delay: a qualitative study of cancer patients' symptom interpretation processes, Soc. Sci Med. 71 (2010) 378–385. [37] L. Hvidberg, A.F. Pedersen, C.N. Wulff, P. Vedsted, Cancer awareness and socioeconomic position: results from a population-based study in Denmark, BMC Cancer 14 (2014) 581. [38] L. Cameron, E.A. Leventhal, H. Leventhal, Symptom representations and affect as determinants of care seeking in a community-dwelling, adult sample population, Health Psychol. 12 (1993) 171–179. [39] N.C. Facione, P.A. Facione, The cognitive structuring of patient delay in breast cancer, Soc. Sci. Med. 63 (2006) 3137–3149. [40] L.K. Smith, C. Pope, J.L. Botha, Patients’ help-seeking experiences and delay in cancer presentation: a qualitative synthesis, Lancet 366 (2005) 825–831. [41] M. Pineros, R. Sanchez, R. Cendales, F. Perry, R. Ocampo, Patient delay among Colombian women with breast cancer, Salud Publica Mex. 51 (2009) 372–380. [42] S.E. Scott, E.A. Grunfeld, V. Auyeung, M. McGurk, Barriers and triggers to seeking help for potentially malignant oral symptoms: implications for interventions, J. Public Health Dent. 69 (2009) 34–40. [43] R.P. Hansen, F. Olesen, H.T. Sorensen, I. Sokolowski, J. Sondergaard, Socioeconomic patient characteristics predict delay in cancer diagnosis: a Danish cohort study, BMC Health Serv. Res. 8 (2008) 49.
Contents lists available at ScienceDirect
Patient Education and Counseling journal homepage: www.elsevier.com/locate/pateducou
Time from first symptom experience to help seeking for colorectal cancer patients: Associations with cognitive and emotional symptom representations Line Flytkjær Jensena,b,1,* , Line Hvidberga,b,1, Anette Fischer Pedersena , Arja R. Aroc , Peter Vedsteda a Research Centre for Cancer Diagnosis in Primary Care (CaP), Research Unit for General Practice, Department of Public Health, Aarhus University, Bartholins Allé 2, 8000 Aarhus C, Denmark b Section for General Medical Practice, Department of Public Health, Aarhus University, Bartholins Allé 2, 8000 Aarhus C, Denmark c Unit for Health Promotion Research, University of Southern Denmark, Niels Bohrs Vej 9–10, 6700 Esbjerg, Denmark
A R T I C L E I N F O
A B S T R A C T
Article history: Received 29 June 2015 Received in revised form 12 November 2015 Accepted 14 November 2015
Objectives: The aim was to assess the association between cognitive and emotional symptom representations prior to diagnosis and the length of the patient interval (i.e. the time from the first symptom is experienced until healthcare is sought) for colorectal cancer patients. Method: The study population included 436 newly diagnosed colorectal cancer patients. Questionnaire data were collected using the Danish Revised Illness Perception Questionnaire (IPQ-R), including cognitive and emotional symptom representations and information on the patient interval. Results: High score in treatment control was associated with short patient interval (PR = 0.52, 95% CI: 0.31–0.89) and high score on the timeline cyclical dimension was associated with long patient interval (PR = 2.14, 95% CI: 1.29–3.57). Hence, patients with negative beliefs about the treatability of their symptoms and patients with strong beliefs about the cyclical nature of their symptoms were more likely to have a long patient interval. Assigning blood in stool as the most important symptom significantly interacted in the association between the patient interval and the two cognitive symptom representations consequence and personal control. Conclusion: The results indicate that aspects of symptom representations were associated with the patient’s help-seeking. Practical implications: These findings may help clinicians and public health planners shorten patient intervals. ã 2015 Elsevier Ireland Ltd. All rights reserved.
Keywords: Cancer Patient interval Illness Perception Questionnaire Symptom representations Common-sense model Background
1. Background Colorectal cancer is one of the most commonly experienced types of cancer in Denmark [1]. For Danish colorectal cancer patients diagnosed during 2000–2002, the age-standardised relative 5-year survival was estimated to be 51.7%, which was significantly lower than in comparable countries, such as Sweden, Canada and Australia [2]. In Denmark, 25% of colorectal cancer patients are treated in stage IV in the tumour, node, metastasis
* Corresponding author. Fax: +45 86124788. E-mail addresses: [email protected] (L.F. Jensen), [email protected] (L. Hvidberg), [email protected] (A.F. Pedersen), [email protected] (A.R. Aro), [email protected] (P. Vedsted). 1 These authors contributed equally to this work. http://dx.doi.org/10.1016/j.pec.2015.11.013 0738-3991/ ã 2015 Elsevier Ireland Ltd. All rights reserved.
(TNM) system, while only 16% are treated in TNM stage I [3]. In addition, Denmark has a less favourable stage distribution than other countries [4]. Later diagnosis is thus believed to be a potential explanation for the poorer prognosis among cancer patients in Denmark compared with other countries [2,5]. This may occur as a result of longer patient intervals sometimes referred to as patient delay (i.e. the time from the first symptom is experienced until healthcare is sought) [6]. In Denmark, population screening for colorectal cancer using immunochemical faecal occult blood test (iFOBT) has been introduced in 2014 to reduce colorectal mortality through early detection [7]. Nevertheless, it is estimated that approximately 75% of colorectal cancer patients will still be detected through symptomatic presentation [8], making the patient interval essential. If we are able to identify factors that influence the patient interval, this knowledge may inform health workers and policy
808
L.F. Jensen et al. / Patient Education and Counseling 99 (2016) 807–813
makers in their efforts to reduce the length of the interval and as Torring et al. [9,10] have made clear, time matters for the cancer outcome. Leventhal’s common-sense model (CSM) is a theoretical model that has been suggested as a useful tool to identify factors related to the patient interval after self-discovery of a cancer symptom [11]. The CSM provides a framework to explain how people interpret and cope with health problems [12], and it builds on the proposition that individuals create their own common sense representations of a health problem to guide their coping efforts [13]. These representations involve the creation of cognitive perceptions of identity, timeline, consequences, control/cure and causes of the health problem and also the creation of an emotional representation of the health problem [14]. In relation to the components identity and consequences, several studies have found that patients who do not attribute their symptom to cancer or do not perceive their symptom as serious have a longer patient interval than patients who perceive their symptoms as serious or specifically as cancer [15–19]. Conversely, Pedersen et al. [20] found that experiencing the alarm symptom, rectal bleeding, was associated with long patient intervals in colorectal cancer patients although many of these patients reported to have wondered if their symptom could be attributable to cancer. The timeline component of the CSM [14] is also relevant as patients who perceive their symptoms as temporary rather than permanent have been found to have a longer patient interval [15]. Thus, previous studies have used components comparable to those from the CSM, but have not explicitly been based on Leventhal’s theoretical framework. However, several researchers have called for theoretically based studies on the patient interval [21–23] and Leventhal’s CSM has been suggested as a useful tool in this connection [23]. In this study we have used the revised Illness Perception Questionnaire (IPQ-R) [24] to assess the representations of the CSM, which is also considered the measure of choice in other studies using Leventhal’s theoretical framework [25,26]. The aim of this study was to investigate the association between the patient interval and cognitive and emotional symptom representations prior to colorectal cancer diagnosis. A further aim was to examine whether experiencing blood in stool (ranked as the most important symptom) was a potential effect modifier for the association between symptom representation and the length of the patient interval. 2. Methods 2.1. Study population The study population consisted of patients who had been registered with histologically confirmed colorectal cancer in the Danish Pathology Data Bank (DPDB) between 1 January and 1 May 2010. The DPDB is a nationwide online database containing detailed information on all pathology specimen analysed in Denmark [27]. The database is updated automatically when a pathology analysis is completed at one of the Danish hospitals [28] and for this study data were retrieved using the Danish version of the Systemized Nomenclature of Medicine (SNOMED) codes for colorectal cancer: T67*3, T68*3 and T73970. The data collection took place in the period from 12 July to 26 August 2010. In total, 1105 patients were identified in the DPDB (study base). In this present study, 206 (18.6%) were excluded because of death, unknown address or research protection (i.e. residents holding publicly recorded protection from research participation). The remaining 899 patients received a questionnaire; non-respondents received a reminder, including a new copy of the questionnaire, three weeks later. A total of 577 completed questionnaires were received during the data collection period (response rate: 64.2%).
The following groups of respondents were excluded from the analyses: 47 respondents (8.1%) were excluded because they indicated that they did not have any preceding symptoms of cancer before contacting a general practitioner (GP), 52 respondents (9.0%) because they had not stated which of their symptoms they perceived as their most important and 42 respondents (7.3%) were excluded as more than 50% of their responses were missing. Hence, a total of 436 respondents were included in the statistical analyses. 2.2. The IPQ-R The revised Illness Perception Questionnaire (IPQ-R) [24], a quantitative measure derived from the CSM, was used to measure cancer patients’ representations of their symptoms prior to diagnosis. The IPQ-R consists of nine subscales, including identity (perceptions of the symptoms associated with the health problem), timeline acute/chronic dimension (perception of the chronicity of the health problem), consequences (anticipated and experienced consequences of the health problem), personal control (perceptions of own ability to control the health problem), treatment control (perceptions regarding treatment for controlling the health problem), coherence (understanding of the health problem), timeline cyclical dimension (perceptions about the stability or changeability of the health problem), emotional representation (emotional responses to the health problem) and cause (perceived cause of the health problem) [24]. We adapted the IPQ-R to measure symptom representations among patients with colorectal cancer symptoms and validated the instrument in this setting [29]. In this modified IPQ-R, the identity subscale comprised 13 commonly experienced symptoms among colon and rectal cancer patients. Before answering the modified IPQ-R, patients were asked to think back on the time before contacting their GP. First patients were asked to rate whether they had experienced each symptom and then whether they believed that the symptom was related to their cancer. The patients were then asked to state which symptom they perceived as the most important and to think of this symptom when assessing the remaining questionnaire. Following the identity subscale, the cognitive representations were listed (timeline acute/ chronic dimension, consequences, personal control, treatment control, coherence, timeline cyclical dimension and emotional representations) and included 32 statements. All statements were rated on a 5-point Likert scale: strongly disagree, disagree, neither agree nor disagree, agree and strongly agree. The scores of each subscale were calculated as stated by Moss-Morris [30]. 2.3. Dependent variables The patient interval was calculated from the dates entered in the questionnaire by the patients. Thus, the number of days between the date of the first symptom experience and the date of the first symptom-related contact to the GP was computed for each patient and subsequently dichotomised using the 75th percentile as the cut-off value, which was 88 days on average for all colorectal cancer patients. 2.4. The IPQ-R subscales Based on whether the patients believed that any of their symptoms were related to cancer before their GP contact, the identity subscale was divided into patients who ‘considered cancer’ and patients who ‘did not consider cancer’. The subscale ratings on the timeline acute/chronic dimension, consequences, personal control, treatment control, coherence, timeline cyclical dimension and emotional representation were divided into three groups, i.e. low, middle and high scores, using the 25th and 75th
L.F. Jensen et al. / Patient Education and Counseling 99 (2016) 807–813
percentile. Higher scores indicate strong beliefs about the chronicity of the symptom, the potential negative consequences of the symptom, the personal controllability of the symptom, the perceived treatability of the symptom, the personal understanding of the symptom, the cyclical nature of the symptom and a high degree of emotional distress. From the causal subscale, only one attribution was chosen to analyse i.e. cancer. This was dichotomised for each patient into ‘agree’, ‘neither nor’ and ‘disagree’ in accordance with their belief in whether their most important symptom was related to cancer. Lastly, the variable blood in stool (yes–no) was generated if the patients had assessed this symptom to be the most important of the experienced symptoms in the questionnaire. 2.5. Socio-demographic variables Due to the Danish Civil Registration System (CRS) [31], it was possible to link survey data to socio-demographic variables at the individual level through Statistics Denmark [32]. The following variables were included: age (64 and >64 years), gender (male and female), marital status (married/cohabiting and living alone) and educational level (low: 10 years, middle: > 10 15 years and high: >15 years) according to UNESCO’s classification [33]. 2.6. Statistical analysis Data were analysed using Stata version 13.1. First, the median patient interval was computed with interquartile interval (IQI) for each of the independent variables. The Wilcoxon and the Kruskal– Wallis rank-sum tests were used to compare median patient interval for the different groups, and a p-value of .05 was used to signify statistical significance. Prevalence ratios (PRs) with 95% confidence intervals (CIs) were used to determine the association between independent variables and patient interval of 88 days. Unadjusted analyses were carried out with each of the independent variables, and an adjusted model was used to control for possible confounding by age, gender, marital status, educational level and blood in stool. Based on previous findings [20], we hypothesised that the symptom blood in stool could be a potential effect modifier for the association between symptom representation and patient interval. Consequently, we performed interaction analyses for each subscale.
Table 1 Socio-demographic characteristics of the respondents included in this study (n = 436) and the study base (n = 1105). Numbers vary due to missing data. Socio-demographic characteristics
Respondents
Study base
%
(n)
%
(n)
Age group 64 years >64 years Age, mean (SD)
35.1 64.9 67.7
(153) (283) (11.0)
27.4 72.6 71.0
(303) (802) (11.3)
Gender Male Female
55.7 44.3
(243) (193)
53.1 46.9
(572) (506)
Marital status Married/cohabiting Living alone
68.8 31.2
(300) (136)
60.7 39.3
(654) (424)
Educational level Low Middle High
35.1 46.8 18.1
(149) (199) (77)
43.9 42.5 13.6
(453) (439) (140)
809
3. Results 3.1. Socio-demographic characteristics of respondents and study base The characteristics of the respondents (n = 436) and the study base (n = 1105) are shown in Table 1. Respondents were generally younger than the study base. Respondents also included more males, more married/cohabiting residents and more people with medium or high-level education. 3.2. Symptoms Patients reported on average 3.5 symptoms (median 3, interquartile range 2–5). Blood in stools was experienced by 47% of the patients and rated as the most important by 32.8% of the patients. Changes in bowel habits was rated as the most important by 28.9% of the patients, 22.0% rated “other symptoms” such as weight loss and fatigue as the most important, and finally 16.3% rated abdominal pain as their most important symptom (data not shown). 3.3. Median patient interval The median patient interval was 30 days for the entire population (data not shown). Table 2 shows the median patient interval with interquartile interval for each of the IPQ-R subscales, for having blood in stool and for included socio-demographic variables. Statistically significant differences in patient interval were found for three cognitive subscales: consequences, treatment control and timeline cyclical dimension. Respondents with the lowest treatment control scores had a longer patient interval (35.5 days) than respondents with the highest treatment control scores (17 days). In addition, the median patient interval was 18 (IQI 2-62) and 61 days (IQI 14-148) for respondents with lowest and highest scores for the timeline cyclical dimension, respectively. 3.4. Association between symptom representations and long patient interval Table 3 presents unadjusted and adjusted associations between symptom representations and long patient interval (88 days, the 75th percentile). High score on treatment control was statistically significantly associated with lower likelihood of long patient interval (PRadj = 0.52, 95% CI: 0.31–0.89). High score on the timeline cyclical dimension had a PR of 2.14 (95% CI: 1.29–3.57) for a long patient interval compared to respondents with low scores. 3.5. Interaction between blood in stool, symptom representation and patient interval Blood in stool (as the most important symptom) interacted statistically significantly in the association between consequence and patient interval (p = 0.036) and between personal control and patient interval (p = 0.017). For patients scoring in the middle range of the consequence scale, the stratified analyses (Table 4) showed that the likelihood for a long patient interval decreased if the most important symptom was blood in stool, which was not seen for patients without blood in the stool. 4. Discussion 4.1. Main findings In this study on symptom representations and patient interval among colorectal cancer patients, we found that the two cognitive symptom representations treatment control and timeline cyclical
810
L.F. Jensen et al. / Patient Education and Counseling 99 (2016) 807–813
Table 2 Median patient interval (in days) and interquartile interval (IQI) for IPQ-R subscales, blood in stool and socio-demographic variables. Colorectal cancer patients (n = 488) Scale (possible score ranges)
Median patient interval IQI
Identity Considered cancer Not considered cancer
30 31
4–78 4–93
Timeline acute/chronic (7–35) <25% (7–13) 25-75% (14–20) >75% (21–35)
28 31.5 21
3–94.5 7–92 3–66
Consequences (4–20) <25 % (4–6) 25-75% (7–11) >75% (12–20)
39 27 31
7–138 2–65 7–87
Personal control (4–20) <25% (4–7) 25-75% (8–12) >75% (13–20)
30 31 28
3–100 4.5–88 7–88
Treatment control (3–15) <25 % (3–9) 25-75% (10–12) >75% (13–15)
35.5 31 17
7–160.5 4.5–90.5 2–61
Coherence (5–25) <25% (5–10) 25-75% (11–16) >75% (17–25)
35 30 22
7–105 5–87 3–94
Timeline cyclical (4–20) <25% (4–8) 25-75% (9–14) >75% (15–20)
18 29 61
2–62 5–81 14–148
P valuea 0.70
0.379
0.045
0.976
0.008
0.480
0.003
Emotional representation (5–25) <25% (5–10) 25-75% (11–17) >75% (18–25)
0.061 13.5 31 34
2–72.5 5–97 8–88
Attribution to cancer Agree Neither nor Disagree
22 35 31
4–66 10–77.5 4–106.5
Blood in stool No Yes
31 30
4–82 6–100
Age 18–64 years >64 years
32 30
5–108 4–76
Gender Male Female
31 30
4–91 4–90
Marital status Married/cohabiting Living alone
31 30
5–92 3–85
Educational level Low Middle High
25 30.5 36
5–70 3.5–85.5 9–139
0.224
0.348
0.145
0.407
0.469
0.089
a Wilcoxon rank-sum test and the Kruskal Wallis rank-sum test (for variables with three categories).
dimension were statistically significantly associated with a long patient interval (88 days or more). Thus, patients reporting that treatment would not be effective and that nothing could be done to
ease their symptom were more likely to have a long patient interval than patients reporting high treatment control. Further, patients for whom symptoms came and went or changed from day to day were also more likely to have a long patient interval than patients who did not report cyclical changes in symptoms. Assigning blood in stool as the most important symptom significantly interacted with the two cognitive symptom representations consequence and personal control. A stratified analysis showed that for patients scoring in the middle range of the consequence scale the likelihood for a long patient interval decreased if the patient had stated blood in stool as the most important symptom. 4.2. Comparison with existing literature In a Danish study of newly diagnosed colorectal cancer patients [34], it was found that 25% had a patient interval of more than 56 days, i.e. about one month shorter than the patient interval found in the present study. This discrepancy can be explained by several factors. First of all, there is no standardised and validated way to measure the patient interval, and the two mentioned studies used different approaches. In the present study, the patient interval was based on dates reported by the patients, whereas, in the study by Hansen et al. [34], the GP provided the date of the first symptom presentation based on the medical record. However, a study analysing the level of agreement between patient-reported and GP-reported patient intervals [35] found that the GPs systematically reported a longer patient interval than the patient. Hence, other conditions such as different sampling procedures may explain this finding. Both patient interval and symptom representations were reported by patients who had recently received their cancer diagnosis. Consequently, the dates and the symptom representations may have been influenced by patient’s emotional turmoil. For instance, some patients might have tried to legitimise or rationalise their – in retrospect – late healthcare-seeking [36] and have underestimated the length of their patient interval. This could underestimate the associations with some of the symptom representations. Furthermore, non-response might have been more widespread among patients than among GPs. Similar to a population-based study among breast cancer patients in Germany [15], which found that the second most common rationale for a long patient interval was considering symptoms as temporary, we found that patients were more likely to report a long patient interval if they experienced cyclical symptoms (i.e. symptoms came and went). This may be a natural consequence of the quest for common sense, i.e. most people use certain coping efforts when dealing with a health problem (e.g. wait and see), however it is important that the public are aware that symptoms of cancer may have a cyclical nature. We found that patients who believed that their symptom was untreatable were more likely to report a long patient interval than patients reporting that something could be done to ease their symptom. As patients with a colorectal cancer diagnosis are often elderly, they may perceive some of the symptoms (such as fatigue and loss of appetite or weight) as normal signs of the ageing process. Thus, other data from our study show that about 20% of the respondents attributed their most important symptom to ageing. Therefore, these patients may believe that nothing can be done to ease their symptom. However, healthcare seeking can possibly be promoted if this age group becomes more aware of the increased risk of cancer with age. Thus, a Danish population-based study found that only one in four is aware of the growing risk of cancer with age [37]. The emotional representations related to the most important symptom were not associated with the patient interval. The CSM suggests that emotional representations may be an independent
L.F. Jensen et al. / Patient Education and Counseling 99 (2016) 807–813 Table 3 Patient interval of 88 days unadjusted and adjusted associations with IPQ-R subscales, blood in stool and socio-demographic variables.
Symptom representations
Prevalence ratio for patient interval 88 days PRunadj. (95% CI) PRa adj. (95% CI)
Identity Considered cancer Not considered cancer
1.00 1.11 (0.77–1.60)
1.00 1.13 (0.77–1.66)
1.00 0.94 (0.64–1.39)
Consequences (4–20) <25% (4–6) 25–75% (7–11) >75% (12–19)
1.00 0.67 (0.45–0.99) 0.73 (0.47–1.13)
1.00 0.72 (0.49–1.07) 0.84 (0.53–1.35)
Personal control (4-20) <25% (4–7) 25–75% (8–12) >75% (13–20)
1.00 0.86 (0.57–1.29) 0.89 (0.54–1.47)
1.00 0.92 (0.61–1.38) 0.97 (0.59–1.59)
Treatment control (3–15) <25% (3–9) 25–75% (10–12) >75% (13–15)
1.00 0.79 (0.54–1.16) 0.62 (0.37–1.04)
1.00 0.76 (0.53–1.10) 0.52 (0.31–0.89)
Coherence (5–25) <25% (5–10) 25–75% (11–16) >75% (17–25)
1.00 0.84 (0.56–1.25) 0.90 (0.57–1.43)
1.00 0.76 (0.51–1.13) 0.69 (0.42–1.13)
Timeline cyclical (4–20) <25% (4–8) 25–75% (9–14) >75% (15–20)
1.00 1.23 (0.73–2.05) 2.01 (1.18–3.42)
1.00 1.23 (0.73–2.05) 2.14 (1.29–3.57)
Emotional representation (5–25) <25% (5–10) 1.00 25–75% (11–17) 1.25 (0.77–2.03) >75% (18–25) 1.15 (0.66–2.02)
1.00 1.36 (0.83–2.21) 1.32 (0.75–2.32)
Attribution to cancer Agree Neither Disagree
1.00 1.11 (0.61–1.99) 1.38 (0.92–2.08)
1.00 1.14 (0.63–2.06) 1.46 (0.96–2.22)
Blood in stool No Yes
1.00 1.19 (0.83–1.68)
1.00 1.23 (0.87–1.74)
Adjusted for age, gender, marital status, educational level and blood in stool.
Table 4 Analyses for statistically significantly interacting symptom representations stratified on blood in stool. Prevalence ratio for long patient interval Symptom representations
Blood in stool PRa adj. (95% CI)
No blood in stool PRa adj. (95% CI)
Consequences (4–20) <25% (4–7) 25–75% (8–12) >75% (13–20)
1.00 0.51 (0.28–0.92) 0.46 (0.17–1.27)
1.00 0.92 (0.52–1.63) 1.09 (0.60–1.99)
Personal control (4–20) <25% (4–7) 25–75% (8–12) >75% (13–20)
1.00 0.61 (0.32–1.16) 0.85 (0.42–1.72)
1.00 1.38 (0.72–2.65) 1.26 (0.78–2.74)
a
patients included in this study may have reacted differently to the emotional distress caused by their symptom because of differences in coping strategies and personal resources. This, in turn, may have led some patients to seek medical attention quickly and some patients to wait and see, which may explain why we found no associations between emotional representations and a long patient interval. 4.3. Strengths and limitations
Timeline acute/chronic (7–35) 1.00 <25% (7–13) 25–75% (14–20) 0.99 (0.66–1.49)
a
811
Adjusted for age, gender, marital status and educational level.
motivator of healthcare seeking [38], but no association with the patient interval was found. This may be explained by the fact that emotions such as fear, worry and anxiety can be motivating as well as hampering factors for healthcare seeking [18,39,40]. Therefore,
This study is unique in using an established theoretical framework to study the relationship between cognitive and emotional symptom representations and patient interval among colorectal cancer patients. Such studies have formerly been called for [21,22]. The use of the validated IPQ-R instrument, which is specifically constructed to assess representations of the health problems described in Leventhal’s CSM, is seen as a main strength of the study and has previously been suggested as a theoretical framework when studying patient interval [23]. The DPDB was used to identify incident cancer patients. The DPDB contains data of high quality on pathology investigations and diagnoses [27]. Data for approximately one third of all incident cancer patients were excluded as only one biopsy was recorded, and the cancer was hence not histologically confirmed. It is unclear how this may have affected the results as the patient interval of these patients cannot be predicted. Patient intervals were dichotomized at the 75th interval as this study focused on factors associated with long patient intervals. Dichotomizing data introduces the risk of losing information, however, as data on the patient intervals was not normally distributed statistical analysis performed on continuous interval data were not advisable. The central limitation of this study was the retrospective nature of the assessment of patient interval, symptoms and symptom representations. Thus, the respondents were already patients when answering the questionnaire and therefore this study rely on their ability to distinguish between their experiences before and after their diagnosis. Thus, recall bias cannot be excluded [41,42], as the dates used to calculate the patient interval solely relied upon the information given by the respondents. Thus, it cannot be ruled out that some patients underestimated the interval and this may infer that true associations between symptom representations and long patient interval have not been detected or have been underestimated. However, other study designs also introduce methodological challenges. A prospective design would be challenging because when following a large population only very few will develop cancer and thus be eligible for study. Therefore the chosen method may be the most feasible and fortunately validation of the psychometric properties of the Danish modified IPQ-R [29], showed for instance that experiencing the symptom blood in stools was correlated with attributing this to hemorrhoids, scratch or chink and cancer indicating that patients were actually thinking of their symptom instead of their disease when answering the questionnaire. A response rate of 64% was achieved, which is higher than reported in other studies in this area [20,43]. Nevertheless, a third of the eligible patients did not participate and differed from the study base. In addition, potential for selection bias exists as patients who died shortly after receiving the diagnosis were not included in the study and possibly people who were very sick were not able to fill in the questionnaire/participate either. These patients are likely to have had more aggressive disease. It could be hypothesized that these patients e.g. perceived their symptom as more serious (as they were more ill than those who survived) or they experienced more distressing emotional representations. However, their estimated patient interval could be either longer or
812
L.F. Jensen et al. / Patient Education and Counseling 99 (2016) 807–813
shorter than the respondents. Thus, it is not possible to say in what direction the estimates may have been affected. Lastly, due to exclusion of several patient’s the final sample size is relatively small, which has led to decreased precision of the sample estimates. Thus, the relatively few statistically significant associations between the patient interval and the IPQ-R subscales may be explained by a type II error and future studies in this area should include a larger sample. 5. Conclusion This study on patients with colorectal cancer showed that the patient interval tended to be longer for patients who believed that nothing could be done to ease their symptoms and patients who stated that their symptoms came and went. Implications The findings from this study may feed directly into the clinical and public health settings, where information about the nature and normalisation of symptoms can be addressed. However, more research is needed to answer what these associations between long patient interval and the two cognitive symptom representations treatment control and timeline cyclical dimension are functions of. If the association between high scores on the timeline cyclical dimension and long patient interval is a function of lack of knowledge that symptoms of cancer may have a cyclical nature, it may be worthwhile to educate the public about this. What is certain is that the patient interval in colorectal cancer can be improved and must be addressed if outcomes of colorectal cancer are to be improved. Conflicts of interest None. Author contribution L.F.J. and L.H. contributed equally to this work. L.F.J., L.H., A.F.P., A.R.A. and P.V. all conceived the study, participated in its design and helped draft the manuscript. L.F.J. and L.H. conducted the survey. All authors read and approved the final manuscript. Funding and ethical approvals This research was financially supported by the Danish Cancer Society, the Novo Nordisk Foundation and the Research Centre for Cancer Diagnosis in Primary Care (CaP). The study was approved by the Danish Data Protection Agency (J. no. 2010-41-4664). Informed consent I confirm all patient/personal identifiers have been removed or disguised so the patient/person(s) described are not identifiable and cannot be identified through the details of the story. Acknowledgements We would like to thank the patients who participated in the study. Also, a special thank goes to Henry Jensen for assisting in the data collection. Finally, we thank the Danish Cancer Society and the Novo Nordic Foundation for the financial support that made this project possible.
References [1] Nordcan Cancer stat fact sheets. Denmark—colorectal, 2014. [2] M.P. Coleman, D. Forman, H. Bryant, J. Butler, B. Rachet, C. Maringe, U. Nur, E. Tracey, M. Coory, J. Hatcher, C.E. McGahan, D. Turner, L. Marrett, M.L. Gjerstorff, T.B. Johannesen, J. Adolfsson, M. Lambe, G. Lawrence, D. Meechan, E.J. Morris, R. Middleton, J. Steward, M.A. Richards, ICBP Module 1 Working Group, Cancer survival in Australia, Canada, Denmark, Norway, Sweden, and the UK, 1995–2007 (the International Cancer Benchmarking Partnership): an analysis of population-based cancer registry data, Lancet 377 (2011) 127–138. [3] Danish Colorectal Cancer Group (DCCG), National årsrapport 2013: Landsdækkende database for kræft i tyk-og endetarm [In Danish]. Herlev: Dansk Kolorektal Cancer Database, 2014. [4] C. Maringe, S. Walters, B. Rachet, J. Butler, T. Fields, P. Finan, R. Maxwell, B. Nedrebo, L. Pahlman, A. Sjovall, A. Spigelman, G. Engholm, A. Gavin, M.L. Gjerstorff, J. Hatcher, T.B. Johannesen, E. Morris, C.E. McGahan, E. Tracey, D. Turner, M.A. Richards, M.P. Coleman, ICBP Module 1 Working Group, Stage at diagnosis and colorectal cancer survival in six high-income countries: a population-based study of patients diagnosed during 2000–2007, Acta Oncol. 52 (2013) 919–932. [5] F. Berrino, R. De Angelis, M. Sant, S. Rosso, M. Bielska-Lasota, J.W. Coebergh, M. Santaquilani, EUROCARE Working group, Survival for eight major cancers and all cancers combined for European adults diagnosed in 1995–99: results of the EUROCARE-4 study, Lancet Oncol. 8 (2007) 773–783. [6] D. Weller, P. Vedsted, G. Rubin, F.M. Walter, J. Emery, S. Scott, C. Campbell, R.S. Andersen, W. Hamilton, F. Olesen, P. Rose, S. Nafees, E. van Rijswijk, S. Hiom, C. Muth, M. Beyer, R.D. Neal, The Aarhus statement: improving design and reporting of studies on early cancer diagnosis, Br. J. Cancer 106 (2012) 1262–1267. [7] The Danish Cancer Society. Screening for colon cancer, 2015. [8] J.S. Juul, P. Vedsted, Uncharacteristic colorectal cancer symptoms and their value in general practice, Ugeskr. Laeger 174 (2012) 710–713. [9] M.L. Torring, M. Frydenberg, W. Hamilton, R.P. Hansen, M.D. Lautrup, P. Vedsted, Diagnostic interval and mortality in colorectal cancer: U-shaped association demonstrated for three different datasets, J. Clin. Epidemiol. 65 (2012) 669–678. [10] M.L. Torring, M. Frydenberg, R.P. Hansen, F. Olesen, P. Vedsted, Evidence of increasing mortality with longer diagnostic intervals for five common cancers: a cohort study in primary care, Eur. J. Cancer 49 (2013) 2187–2198. [11] S.E. Scott, E.A. Grunfeld, M. McGurk, Patient’s delay in oral cancer: a systematic review, Commun. Dent. Oral Epidemiol. 34 (2006) 337–343. [12] H. Leventhal, D. Nerenz, D. Steele, Illness representation and coping with health threats, Handbook of Psychology and Health, Lawerence Eribaum, Hillsdale, 1984, pp. 219. [13] A.M. Barsevick, K. Whitmer, L. Walker, In their own words: using the common sense model to analyze patient descriptions of cancer-related fatigue, Oncol. Nurs. Forum 28 (2001) 1363–1369. [14] E. Leventhal, H. Leventhal, L. Cameron, Representations, procedures, and affect in illness self-regulation: a perceptual-cognitive model, Handbook of Health Psychology, Lawrence Erlbaum, Hillsdale, 2001, pp. 19. [15] V. Arndt, T. Sturmer, C. Stegmaier, H. Ziegler, G. Dhom, H. Brenner, Patient delay and stage of diagnosis among breast cancer patients in Germany—a population based study, Br. J. Cancer 86 (2002) 1034–1040. [16] J. Corner, J. Hopkinson, L. Roffe, Experience of health changes and reasons for delay in seeking care: a UK study of the months prior to the diagnosis of lung cancer, Soc. Sci. Med. 62 (2006) 1381–1391. [17] G. Meechan, J. Collins, K.J. Petrie, The relationship of symptoms and psychological factors to delay in seeking medical care for breast symptoms, Prev. Med. 36 (2003) 374–378. [18] M. O'Mahony, J. Hegarty, Factors influencing women in seeking help from a health care professional on self discovery of a breast symptom, in an Irish context, J. Clin. Nurs. 18 (2009) 2020–2029. [19] M. Pineros, R. Sanchez, R. Cendales, F. Perry, R. Ocampo, Patient delay among Colombian women with breast cancer, Salud Publica Mex. 51 (2009) 372–380. [20] A.F. Pedersen, R.P. Hansen, P. Vedsted, Patient delay in colorectal cancer patients: associations with rectal bleeding and thoughts about cancer, PLoS One 8 (2013) e69700. [21] A. Bish, A. Ramirez, C. Burgess, M. Hunter, Understanding why women delay in seeking help for breast cancer symptoms, J. Psychosom. Res. 58 (2005) 321– 326. [22] S. Scott, M. McGurk, E. Grunfeld, Patient delay for potentially malignant oral symptoms, Eur. J. Oral Sci. 116 (2008) 141–147. [23] S.E. Scott, E.A. Grunfeld, M. McGurk, Patient’s delay in oral cancer: a systematic review, Commun. Dent. Oral Epidemiol. 34 (2006) 337–343. [24] R. Moss-Morris, J. Weinman, K.J. Petrie, R. Horne, L.D. Cameron, D. Buick, The revised illness perception questionnaire (IPQ-R), Psychol. Health 17 (2002) 1. [25] M. Fischer, M. Scharloo, J. Abbink, A. van ‘t Hul, D. van Ranst, A. Rudolphus, J. Weinman, K. Rabe, A.A. Kaptein, The dynamics of illness perceptions: testing assumptions of Leventhal’s common-sense model in a pulmonary rehabilitation setting, Br. J. Health. Psychol. 15 (2010) 887–903. [26] M.S. Hagger, S. Orbell, Illness representations and emotion in people with abnormal screening results, Psychol. Health 21 (2006) 183–209. [27] R. Erichsen, T.L. Lash, S.J. Hamilton-Dutoit, B. Bjerregaard, M. Vyberg, L. Pedersen, Existing data sources for clinical epidemiology: the Danish National Pathology Registry and Data Bank, Clin. Epidemiol. 2 (2010) 51–56.
L.F. Jensen et al. / Patient Education and Counseling 99 (2016) 807–813 [28] R. Erichsen, T.L. Lash, S.J. Hamilton-Dutoit, B. Bjerregaard, M. Vyberg, L. Pedersen, Existing data sources for clinical epidemiology: the Danish National Pathology Registry and Data Bank, Clin. Epidemiol. 2 (2010) 51–56. [29] L. Hvidberg, L.F. Jensen, A.F. Pedersen, A.R. Aro, P. Vedsted, Measurement properties of the Danish version of the Illness Perception QuestionnaireRevised for patients with colorectal cancer symptoms, J. Health. Psychol. 19 (2014) 1279–1290. [30] R. Moss-Morris, The Illness Perception Questionnaire 2011, 2014. [31] C.B. Pedersen, The Danish Civil Registration System, Scand. J. Public Health 39 (2011) 22–25. [32] Statistics Denmark. http://www.dst.dk/en/OmDS (assessed November 09.11.15.). [33] Unesco Institute for Statistics. ISCED: International Standard Classification of Education. http://www.uis.unesco.org/Education/Pages/internationalstandard-classification-of-education.aspx (assessed 09.11.15.). [34] R.P. Hansen, P. Vedsted, I. Sokolowski, J. Sondergaard, F. Olesen, Time intervals from first symptom to treatment of cancer: a cohort study of 2212 newly diagnosed cancer patients, BMC Health Serv. Res. 11 (2011) 284. [35] M.B. Larsen, R.P. Hansen, I. Sokolowski, P. Vedsted, Agreement between patient-reported and doctor-reported patient intervals and date of first symptom presentation in cancer diagnosis—a population-based questionnaire study, Cancer Epidemiol. 38 (2014) 100–105.
813
[36] R.S. Andersen, B. Paarup, P. Vedsted, F. Bro, J. Soendergaard, ‘Containment' as an analytical framework for understanding patient delay: a qualitative study of cancer patients' symptom interpretation processes, Soc. Sci Med. 71 (2010) 378–385. [37] L. Hvidberg, A.F. Pedersen, C.N. Wulff, P. Vedsted, Cancer awareness and socioeconomic position: results from a population-based study in Denmark, BMC Cancer 14 (2014) 581. [38] L. Cameron, E.A. Leventhal, H. Leventhal, Symptom representations and affect as determinants of care seeking in a community-dwelling, adult sample population, Health Psychol. 12 (1993) 171–179. [39] N.C. Facione, P.A. Facione, The cognitive structuring of patient delay in breast cancer, Soc. Sci. Med. 63 (2006) 3137–3149. [40] L.K. Smith, C. Pope, J.L. Botha, Patients’ help-seeking experiences and delay in cancer presentation: a qualitative synthesis, Lancet 366 (2005) 825–831. [41] M. Pineros, R. Sanchez, R. Cendales, F. Perry, R. Ocampo, Patient delay among Colombian women with breast cancer, Salud Publica Mex. 51 (2009) 372–380. [42] S.E. Scott, E.A. Grunfeld, V. Auyeung, M. McGurk, Barriers and triggers to seeking help for potentially malignant oral symptoms: implications for interventions, J. Public Health Dent. 69 (2009) 34–40. [43] R.P. Hansen, F. Olesen, H.T. Sorensen, I. Sokolowski, J. Sondergaard, Socioeconomic patient characteristics predict delay in cancer diagnosis: a Danish cohort study, BMC Health Serv. Res. 8 (2008) 49.