Understanding middle‐aged and older adults' first associations with the word “cancer”: A mixed methods study in England

Abstract Objective Cancer is still widely feared and often associated with death. Fatalistic beliefs adversely affect help‐seeking for cancer symptoms and engagement in cancer prevention. This study aims to understand middle‐aged and older adults' first association with the word “cancer” and their relationship with sociodemographic factors, cancer fear, and cancer information avoidance. Methods We conducted a cross‐sectional survey of 1464 community‐based adults aged 50 to 70 living in England in April 2015. First associations with cancer were measured qualitatively and analysed using content analysis. We used binary logistic regression to analyse associations between the most common first association of cancer and sociodemographic characteristics, cancer fear, and cancer information avoidance. Results Cancer was most commonly associated with “death” (26%). Respondents with lower levels of education, living in the Midlands or North of England where cancer mortality is higher, or with close friends or family members with a cancer history, were more likely to associate cancer with death. Cancer fear was significantly associated with death associations, but cancer information avoidance was not. Conclusions Despite improved cancer outcomes, middle‐aged and older adults often associate cancer with death. Further efforts to decrease fatalistic associations in this age group may be needed.

more deprived backgrounds. 3 In addition, although older age is a major uncontrollable risk factor for cancer, it is estimated that 43% of cancers could be prevented through healthier lifestyle choices, 13 and avoidance of cancer information due to fatalistic beliefs may thus represent a missed opportunity for cancer prevention. It is therefore important to monitor negative beliefs about cancer in the population and identify subgroups that are more likely to hold these beliefs and may be deterred from cancer prevention and early diagnosis.
However, many of the studies that have quantitatively monitored these negative beliefs have also measured positive beliefs about cancer, and endorsement rates of these beliefs tend to be high. For example, 80% to 90% of US and UK adults believe that cancer can often be cured, and there is near-universal agreement that presenting promptly with symptoms, or getting checked regularly for cancer, could improve survival chances. [3][4][5] This suggests that a substantial proportion of the general population may concurrently hold positive and negative beliefs about cancer.
A disadvantage of these quantitative population-based studies is that they use prompted methodology to measure agreement with predefined statements. Previous research suggests using prompted (recognition) versus unprompted (recall) measures of cancer-related knowledge yields very different results, 14 which may also be the case with attitudes. In addition, such quantitative studies cannot determine whether it is the positive or the negative beliefs about cancer that come to mind more readily. A qualitative study of 30 UK adults provides a more in-depth analysis of the balance of negative and positive beliefs. 15 The study found the majority of interviewees had initial reactions to cancer that were fearful and fatalistic, although these "gut responses" were often followed by more optimistic and hopeful ones. These unprompted responses to cancer may affect our basic tendency to approach or avoid.
For example, associations of cancer with death may elicit fear and avoidant behaviours, while associations of survival and behavioural control may elicit an openness to learn more about cancer prevention and early detection. The content of unprompted beliefs about cancer may thus be important to inform cancer control messages and strategies.
The aim of the current study was to extend previous qualitative findings, 15 using a mixed-methods design to examine people's first association with the word "cancer" in a large, population-based study of middle-aged and older adults, for whom the threat of cancer is relevant due to their age. We also examined the sociodemographic distribution of the most common first association (which we found to be "death"), and its relationship with cancer fear and cancer information avoidance. On the basis of previous studies, 6,9,[16][17][18][19] we hypothesised that those from lower socioeconomic status and ethnic minority backgrounds would be more likely to have fatalistic cancer beliefs. We predicted more negative beliefs in the Midlands and the North of England, where cancer mortality is higher than in the South. 20 Expectations of cancer outcomes may also vary by experience of cancer in others, although previous findings in relation to this have been mixed. 3,15,16,18,21 2 | METHODS

| Design
Data were obtained from the second Attitudes, Behaviour, and Cancer-UK Survey (ABACUS), a series of 4 population-based surveys designed to assess attitudes to cancer and cancer screening in England in 2014 to 2016. 12 The fieldwork was conducted by TNS Research International using home-based computer-assisted personal interviewing in April 2015. Residences were selected based on random location sampling using the 2011 census small-area statistics and postcode address file (stratified by Government Office Region and social grade), with quotas set for age, gender, children in the home, and working status at each location.

| Participants
Overall, 1464 participants aged 50 to 70 completed the interview.
Participants with missing sociodemographic information were excluded from the present analyses, as well as those with a history of cancer, because they were not asked questions about cancer fear to avoid distress. This study was exempt from ethics approval under the UCL Research Ethics Committee guidelines, because it involved "the use of non-sensitive, completely anonymous […] survey and interview procedures" and "the participants are not defined as 'vulnerable' and participation will not induce undue psychological stress or anxiety" (https://ethics.grad.ucl.ac.uk/exemptions.php). All participants provided informed consent.

| Materials and procedure 2.3.1 | First association with cancer
Respondents' first associations with cancer were assessed qualitatively by asking: "What is the first thing that comes to mind when you think about cancer?" (adapted from 22 ). No time or word limit was set for responses. The interviewer recorded the verbatim response in a freetext box using a stylus pen.

| Sociodemographic factors
All other responses were measured quantitatively using simple questions. Sociodemographic variables included age, gender, marital status, ethnicity, education level, and region of residence. Marital status was recorded as married/living as married, single, or separated/widowed/ divorced. Due to small numbers in each of the ethnic minority groups, the ethnicity variable was dichotomised as "White" vs "Black, Asian, and Minority Ethnic (BAME)" (including mixed ethnic background).
Education level was collapsed into 5 categories (see Table 1). Experience of cancer in close family or friends was assessed by asking, "Have any friends or family members that are close to you ever been diagnosed with cancer?" and scored as "yes" or "no." Don't know and refused responses were coded as missing throughout.

| Cancer fear
Cancer fear was assessed using 2 items relating to intensity ("How anxious do you feel when you think about cancer?" scored as "not at all," "slightly," "quite," and "extremely") and frequency ("In general, how often do you worry about getting cancer yourself?" scored as "never," "occasionally," "sometimes," "often," and "very often"), adapted from a previous survey. 23 Three levels of cancer fear were created as in a previous study 12: "no cancer fear" ("not at all" anxious and "never" worried about cancer), "moderate cancer fear" ("slightly" anxious, or "occasionally" or "sometimes" worried), and "high cancer fear" ("quite" or "extremely anxious," or "often" or "very often" worried; see Data S1).

| Cancer information avoidance
Cancer information avoidance was assessed using 3 "yes"/"no" items (adapted from 24 ) asking about avoidance of 3 cancer information sources (stories in newspapers, magazines, or online; other people; TV programmes; see Data S2). Overall cancer avoidance was dichotomised into "no avoidance" ("no" on all 3 items) or "avoids at least one information source" ("yes" to at least one item).

| Qualitative analysis
Open-ended responses were transcribed and analysed using content analysis. JW, CV, and EA inductively coded the data, with any ambiguity resolved by discussion. When the initial codes were examined, they broadly fitted with the 5 dimensions of Leventhal's common sense model of illness representations (CSM), 25 so this theory was used to refine the coding frame.
The CSM states that the illness representations of a health threat determine an individual's coping response to the threat. 25 Illness representations have 5 attributes: "identity" refers to ideas about the (somatic) representation of an illness; "timeline" refers to beliefs about the timeframe of that illness and development over time; "causes" refers to beliefs about an illness' causes; "controllability" refers to the perceived controllability of the illness, and "consequences' refers to representations of the anticipated consequences of the illness. According to the CSM, these attributes together determine the cognitive appraisal of a health threat, which-along with the emotional appraisal of the threat-determine an individual's behavioural response. 25 We added 4 categories ("death (incurability)," a specific anticipated outcome of cancer that also has aspects of its expected development over time, "emotional response," "social networks," and "miscella-

| Sample characteristics
After excluding those with a cancer diagnosis (n = 127), missing sociodemographic characteristics (n = 73) or missing data on the first association of cancer question (n = 159), 1105 participants were included for analysis. Mean age was 60 years (SD = 6.27) and 51% were male ( Table 1). The majority were married (62%) and of White ethnic origin (94%), comparable to the general population in this age group in England (65% and 92%, respectively). 28 About a quarter (26%) had no formal qualifications, compared with 25% in this age group in the general population. 28 The regional distribution was similar to the general population. 29 Most respondents (72%) had experience of cancer in close friends or family members. There were no significant differences in sociodemographic characteristics between those who were included in the analyses and those excluded due to missing data (results not shown).

| First association with cancer
The results of the qualitative analysis of participants' first association with the word "cancer" are presented in Table 2. Overall, 86% of responses were coded under one theme only, while 14% were coded under multiple applicable themes. "Death" was the most frequent association with cancer (mentioned by 26%), followed by references to the controllability of cancer (25%), its identity (23%), emotional responses to cancer (15%), social networks (10%), and cancer consequences (8%). A small minority (4%) mentioned potential causes of cancer, such as smoking or genetics.
In terms of cancer controllability, most respondents made references to cancer treatments (eg, "chemotherapy" or "surgery"), with fewer respondents referring to cancer survival or early detection. In terms of identity, most respondents had negative beliefs about cancer (eg, "horrible disease"), while fewer people had more neutral associations of cancer as "just another health condition." Only very few people spontaneously reported cancer as something to be avoided (2%). Negative emotional responses (eg, "panic") were much more common than positive, hopeful, or empathetic ones (eg, "hope people do survive").
About a third of those who referred to their social networks mentioned someone who had died from cancer. Finally, a small proportion associated cancer with its physical (eg, "pain") or other consequences.

| Sociodemographic distribution of having "death" as first association
We examined the sociodemographic distribution of having "death" as first association of cancer (  Table 1). Almost a quarter (23%) reported avoiding at least one source of cancer information, but there was no association between "death" associations and cancer information avoidance.

| CONCLUSIONS
This was the first population-based study to qualitatively examine first associations with cancer in a large sample of middle-aged and older adults, for whom the threat of cancer is relevant due to their age.
Despite improvements in cancer outcomes over recent decades, cancer was most commonly associated with death. Positive and hopeful associations, and references to modifiable risk factors, early detection, and survival, were infrequent.
This study extends and quantifies previous findings by Robb et al 15 that initial responses to cancer tend to be negative, by indicating how common "death" and other negative responses to cancer are and by examining the sociodemographic patterning of these responses. Our study replicates the finding that almost a quarter of middle-aged and older adults in England avoid cancer information. 12 The proportion of participants who spontaneously reported death associations was also similar to the proportion agreeing to prompted statements that "cancer is a death sentence" found in previous quantitative studies in the UK. 3,30 Death associations were more common among those with lower levels of education, from regions with higher cancer mortality, and among those with previous experience of cancer in others. This is consistent with suggestions by other authors that fatalistic cancer attitudes may be the indirect consequence of under-education and poverty, leading to a focus on day-to-day survival rather than healthpromoting behaviours, and potentially causing delays in help-seeking for symptoms and poorer cancer outcomes. [31][32][33][34][35] Fatalistic beliefs in families and communities may be reinforced by witnessing this cycle of poorer cancer outcomes. 31,32,36 The "availability heuristic," or the ease with which instances of cancer deaths can be brought to mind, 37 may further reinforce fatalistic beliefs.

| Clinical implications
Our study furthers our understanding of public attitudes to cancer.
Leventhal's CSM states that the cognitive representation of a health threat (eg, "cancer means death") is processed alongside the emotional representation of the health threat (eg, "fear"), and that these parallel processes jointly determine the behavioural response to the threat. 25 Although previous studies suggest that people may be "in two minds" about cancer, 3,15 our study shows that the first response to cancer is often negative. Cancer fear and fatalism undermine cancer prevention and early detection behaviours, [7][8][9][10][11][12]

| Study limitations
This study had some limitations. First, generalisability may be affected by the sample not being a random probability sample, although quota research could ask about specific types of cancer.
In conclusion, a significant proportion of middle-aged and older adults in England associate "cancer" with "death," especially those from lower socio-economic backgrounds or regions with higher cancer mortality. These fatalistic associations may undermine engagement with cancer prevention and early detection initiatives. Efforts are needed to understand how associations with "cancer" are formed, and how unduly negative associations may be altered to improve cancer outcomes.

ACKNOWLEDGEMENTS
This work was supported by a programme grant from Cancer Research UK awarded to Professor Jane Wardle (C1418/A14134). We should like to acknowledge the intellectual contribution that Professor Wardle has made to the design of this study before her death in October 2015.
Cancer Research UK was not involved in the design of this study; the collection, analysis, or interpretation of the results; in the writing of the manuscript; or in the decision to submit for publication.

CONFLICT OF INTEREST
The authors declare that they have no competing interests.

CONTRIBUTORS
CV and JW conceived of the study and its design and were responsible for data collection. EA, JW, and CV developed the coding framework, which was refined with input from RD. EA, RD, and CV conducted the qualitative coding. EA and RD conducted the quantitative analyses, with input from CV. EA and RD prepared the final manuscript, with input from JW and CV. All authors have seen and approved the final manuscript before submission.