br documenting lack of symptom awareness as a common
documenting lack of symptom awareness as a common reason for delaying care among patients with cancer13,14 and low prevalence of cancer symptom awareness,10,15,16 particularly among underserved groups.16 Although dif-ferences in symptom awareness did not explain interna-tional differences in cancer survival,17 anticipated delay was associated with lack of symptom awareness across the six countries9 and within the UK.10,18 Quaife and col-
leagues18 found relationships between the lack of recogni-tion of the symptoms of lung, breast, and colorectal cancers and increased likelihood of delay by patients in seeking care; these relationships were consistent, indepen-dent of demographics and perceived healthcare access. Subsequent research has associated lower symptom awareness with regional differences in cancer survival19 and documented the preliminary impact of symptom awareness campaigns on reducing stage at diagnosis.20 In the U.S., little research has examined cancer symptom awareness and care seeking.19 To date, no population-based studies have examined U.S. cancer symptom aware-ness and care seeking across a range of cancer symptoms. Using population-based survey data modeled after the ICBP ABC instrument, this CHIR-265 study builds upon interna-tional work by examining associations between cancer symptom recognition and anticipated time to seeking care in the U.S.
Computer-assisted telephone interviews with a population-rep-resentative sample of English-speaking adults (aged ≥50 years) in the U.S. (N=1,425) were conducted using an instrument adapted from the ICBP ABC survey.11 The original ICBP ABC survey underwent substantial cognitive testing and test−retest reliability checking.11 For the U.S. version, minor changes were
made to ensure language and response codes were appropriate for the U.S. context. For example, demographic questions on educational attainment and ethnic group were adapted to match U.S. census categories, and references to the National Health Service were removed. To account for the rising number of cell phone−only households,21 landline and cell phone households were randomly sampled from regions across the U.S. using two approaches. For landline sampling, households were selected using plus-digit dialing, which systematically takes a random selection of telephone numbers from national telephone directo-ries and replaces the last two digits with randomly generated numbers. This approach increases coverage of the population by including unlisted telephone numbers, resulting in better repre-sentativeness. Households were eligible if at least one person aged ≥50 years lived there. The Rizzo method was used to ran-domly select an individual in the household when more than one person was eligible.22 For cell phones, it was not possible to use plus-digit dialing because of the restrictions on calling cellu-lar numbers in the U.S. Therefore, telephone numbers were selected at random from a database of 1,000-block records held by Survey Sampling International. Data were collected by Ipsos MORI’s Social Research Institute (a UK-based research company who administered the original ICBP ABC survey) from August to October 2014. All activities were reviewed for ethical approval by the National Cancer Institute’s Office of Human Subjects Research Protections. To equalize selection probabilities and compensate for noncoverage and nonresponse, survey design weights and nonresponse weights were developed and applied to the survey data. Design weights accounted for probability of interview selection within the household. Nonresponse weights for key demographic variables (age, sex, region, highest level of education, and race) were applied using 2012 American Com-munity Survey data to account for differences between the study sample and the U.S. population.
Anticipated time to seeking physician-based care for the follow-ing four cancer symptoms was assessed: persistent cough, rectal bleeding, breast changes (females only), and changes in mole appearance. Cancer prevention for each associated cancer is rec-ommended in the U.S. (lung cancer, colorectal cancer, breast cancer, and skin cancer), and routine screening is recommended for all except skin cancer. Respondents were instructed to indi-cate for how long, from first noticing each symptom, they would wait to go to the doctor; responses were categorized as follows: immediately, up to 1 week, 1 to 2 weeks, 2 to 3 weeks, 3 to 4 weeks, more than a month, and would not contact doctor. To compare results with previous analyses in other countries, responses indicating seeking care from nonphysicians (e.g., pharmacists) were excluded (persistent cough: n=24, rectal