Online surveys facilitate access to many potential respondents, especially those from hidden populations , and as the internet population-penetration rate in Japan was 83.5% in 2016 , we conducted an online questionnaire survey from September 10 to 13, 2019 to collect responses in this cross-sectional study of patients with primary cancer (gastric, colorectal, lung, breast, and prostate cancer) through an online-research company (Rakuten Insight, Inc. ). We included only those respondents whose family dental clinic was the same as that before they were diagnosed with cancer. This study targeted the enrolment of patients with the five commonest cancers that have a high morbidity, as reported by the Cancer Information Service of the National Cancer Center Japan . The exclusion criteria were: cancer recurrence, cancer types not covered by this study, multiple cancers, and "current family dental clinic is not the dental clinic the respondents visited before the onset of cancer." The questionnaires in which all items were answered were counted as valid questionnaires. The questionnaires were accepted until there were 100 respondents for each cancer.
The screening and study surveys were conducted by an online-research company (Rakuten Insight, Inc. ), who provided only anonymized data. We did not communicate directly with the respondents and did not obtain any personal identifiable information. The survey followed an opt-in method to ensure tacit consent for participation. This study was approved by the ethics committee of Tokyo Medical and Dental University in February 2019 (M2018-236).
We generated the original questionnaire used in this study based on previous reports [17, 18] and clinical experience. The questionnaire (available as supplementary table 1) consisted of 34 items from the following five sections: (1) attribute of responders such as the year of diagnosis of cancer and the cancer treatment [attribute section]; (2) knowledge about general and oral health [knowledge section]; (3) opinion about the handling of health information of cancer patients by medical/dental/healthcare professionals [opinion section]; (4) experience after being diagnosed with cancer [experience section]; and (5) thoughts on health and oral health [characteristic section]. We used the following response options for all items except for the attribute items Q4a (physician’s advice) and Q4b–d (around the seat) in this survey: 4 (strongly know/strongly agree), 3 (know slightly/agree slightly), 2 (don’t know slightly/disagree slightly), and 1 (strongly don't know/strongly disagree). Response options for Q4a-c were yes, no. Response options for Q4d were private room, wall, curtain, partition, no machine, no partition, and other (free comments).
The dependent variable in this study was the presence or absence of reporting; therefore, respondents who answered “yes” to either of the following two questions and those who answered “no” to both questions were assigned to the “reporting group” and the “non-reporting group,” respectively: Q4b: “Did you tell your family dentist that you had cancer?” and Q4c: “Did you write in the medical questionnaire at your family dental clinic that you had cancer?”
To investigate the effect of the dental clinic structure, we created "private group" and "non-private group" as a dummy variable based on whether the patient's line of sight was blocked, by using question number Q4d: “Is there anything between the treatment seats in your family dental clinic?” and patients who answered “private room, wall, curtain, or partition” and “no machine or no partition” were assigned to the private group and non-private group, respectively. Moreover, as the answer to this question included “Others (free comments),” we decided that a patient who answered “reservation for only one person” and “some private rooms” should be assigned to the private group and non-private group, respectively.
To investigate the effects of residence, we decided that “designated cities and special wards” were “urban” and “others” were “rural.” A previous study  classified their study population into four groups based on the type of residence; however, we assigned the respondents to two groups to facilitate the analysis. In Japan, the designated city is the official term for large cities, defined by the Local Autonomy Law as a city with a population of 500,000 or more, as designated by cabinet order. However, actually, cities with a population of more than one million, or more than 700,000 that are expected to reach one million soon, receive this designation ; thus, there are 20 cities, excluding Tokyo, in Japan. There are 23 special wards in Tokyo that have a similar scope as ordinary cities  and, therefore, these two regions were designated as “urban” areas because of their size and nature.
Reverse item processing was performed to facilitate the interpretation of the results of questions Q3d, Q3e, Q4e, Q4j, and Q5f. For the response items using sores 4 to 1, the analysis was conducted with scores 4 or 3 considered as “know/agree” and scores 2 or 1 as “do not know/disagree.”
All statistical analyses were conducted using the BellCurve for Excel version 3.20 (Social Survey Research Information Co., Ltd, Tokyo, Japan). We conducted a factor analysis on 23 items related to awareness to confirm the questionnaire's supposed substructure (principal factor method with varimax rotation). We calculated Cronbach's alpha to investigate the reliability of the extracted factors (six factors as patient awareness factors). We performed a univariate analysis of 36 variables, including the abovementioned 23 items plus sex, age, residence, the environment of the dental treatment seat, etc., to identify items related to the dependent variable of this study (p < 0.1). The multivariate logistic regression analysis included the variables that were identified as significant on univariate logistic analysis (p < 0.05).
Respondents who answered yes or no to Q4a were assigned to the group with physician's advice and the group without physician's advice, respectively, and a chi-square test was used to compare both groups for the presence or absence of reporting (p < 0.05). Next, we conducted multivariate logistic regression analysis (p < 0.05) against the group without physician's advice by using each responder's factor scores calculated for the six factors in the factor analysis as the independent variables.