Introduction

Currently, approximately 20 million people worldwide are newly diagnosed with cancer each year [], and this number is expected to reach 35 million by 2050 []. In Japan, the number of new patients with cancer is projected to reach approximately 4 million between 2035 and 2039 []. Owing to advances in treatment methods and early detection, many individuals diagnosed with cancer are now long-term survivors, and the number of people classified as survivors with cancer who live with and beyond cancer is increasing []. Regardless of treatment status, survivors with cancer tend to face various difficulties, including physical, mental, economic, social, and emotional challenges []. In particular, the prevalence of psychological stress, such as anxiety and depression, is high, making it extremely important to address their emotional needs [,]. Against this background, health care professionals are expected to provide appropriate information and high-quality communication so that patients can confidently make decisions about their diagnosis, prognosis, and treatment options [].

The most frequently reported unmet nonmedical need among patients with cancer is information []. Accurate knowledge of the disease and its treatment affects psychological well-being, quality of life, and satisfaction with treatment []. However, the quality of information, access to it, and the individual’s level of understanding pose challenges. Improving the quality of the information provided and accurately identifying patient needs also leads to better communication between patients and physicians [,]. Nonetheless, many survivors with cancer report that the medical terms and explanations used by health care professionals are difficult to understand, resulting in insufficient information and communication [,]. Therefore, while patient-centered communication [,]—which enhances the reliability of information for survivors with cancer—is important, dissatisfaction may not stem solely from a simple lack of communication but may instead be intricately related to individual cognitive and psychological factors, which need to be elucidated.

Given this background, we conducted a pilot study using content analysis of free-text responses regarding the use of psychosocial support services among survivors with cancer aged 40 years and older []. The results suggested that factors such as trust in support providers, expectations for emotional regulation, difficulty in verbalizing emotions, a desire for independence, the personality of support providers, clarity of consultation procedures, and awareness and dissemination of services are related to the use of support services []. These findings suggest that the usage of psychosocial support services may be influenced not only by access and information deficiencies but also by individual cognitive and behavioral characteristics, including personal values. Thus, it is necessary to introduce a theoretical framework to comprehensively and quantitatively capture these complex factors.

The theory of planned behavior (TPB) has been widely used to explain behavior change among patients with chronic illnesses [,]. According to the TPB, behavioral intention is formed by 3 factors: “attitude,” “subjective norms,” and “perceived behavioral control” []. However, because the behavior of patients with chronic illnesses is influenced not only by internal cognitive factors but also by external factors such as the living environment, support networks, and health care systems, in this study we integrated external and relational factors, such as the living environment, support networks, and health care systems of survivors with cancer, into the TPB framework to deepen our understanding of the formation of behavioral intentions.

When addressing communication between patients and health care professionals, which is a central focus of this study, the patient-provider relationship model [] is important. This model explains the impact of the relationship between health care professionals and patients on health outcomes, such as survival, treatment and recovery, relief of symptoms and distress, emotional well-being, and pain control. Communication functions are classified into relationship-oriented functions, which emphasize strengthening the therapeutic relationship and responding to emotions, and task-oriented functions, which involve exchanging information, decision-making, supporting self-management, and managing uncertainty [].

Previous research has often examined information-seeking behaviors and the use of support services among survivors with cancer on the basis of single attributes, such as treatment phase or sex [-]. There have been limited attempts to comprehensively typify diverse individual behavioral characteristics and psychological factors and theoretically clarify their backgrounds. Therefore, in this study, we aimed to fill this gap by using latent class analysis (LCA) to identify behavioral patterns and by examining the characteristics of these patterns from the perspectives of the TPB and the relationship-oriented and task-oriented functions of the patient-provider relationship model.

On the basis of the above theoretical framework, we focused on the following two research objectives in this study:

To identify latent groups that can be typified from the viewpoint of the TPB regarding information-seeking behavior of survivors with cancer, trust in information sources, difficulties faced while seeking information, and use of psychosocial support services.To examine how the identified latent groups are characterized from the perspectives of the relationship-oriented and task-oriented communication functions of the patient-provider relationship model and to clarify the contextual factors behind information-seeking behaviors and the use of support services.
MethodsResearch Design

This cross-sectional survey was conducted on the basis of the CHERRIES (Checklist for Reporting Results of Internet E-Surveys) [] ().

The questionnaire, created in Japanese, consisted of 21 screens and 52 items on the website and was designed to be answerable within 25 minutes (). To prevent missing data, all questions were set as mandatory in the web-based questionnaire. To eliminate duplicate responses, any responses that matched IP address, sex, age, and place of residence were considered to be from the same respondent, and only the first response was included in the analysis.

Questions regarding patients with cancer information-seeking behavior were developed with reference to previous studies [] that aimed to clarify the information needs and difficulties encountered when searching for information and using psychosocial support services among patients with ovarian cancer, as well as the patient experience survey [] conducted by the National Cancer Center. During development, 3 researchers with experience in medical research and survey studies involving individuals with diseases examined whether each question appropriately reflected the research objectives and theoretical framework from the perspective of content validity and revised the questions for clarity of wording and to minimize respondent burden.

First, 12 items on sociodemographic information (including sex, age, place of residence, and occupation) and clinical information (including type of cancer, cancer stage, treatment status, and recurrence or metastasis) were collected using a single-choice format during the screening phase. This allowed for the collection of basic attribute information to uniquely classify each respondent. In addition, as a basic indicator related to information seeking, “device used to gather information” (1 item) was assessed using a single-choice format.

For information-seeking behavior regarding cancer, we started with the question “Have you ever searched for information about cancer?” Four multiple-choice checkbox items were used: “sources of cancer-related information,” “social networking service (SNS) or apps used for collecting cancer-related information,” and “advantages of psychosocial support and information-gathering services recommended by medical institutions.” For these questions, mutually exclusive options such as “did not search,” “do not wish to obtain information,” “do not use SNS or apps,” and “not applicable” were provided to ensure consistency in responses.

Furthermore, skip logic was introduced: if a respondent answered “did not search” to the question “Have you ever searched for information about cancer?” they were not required to answer questions about “sources of cancer-related information,” “devices used to collect information,” or “SNS or apps used for collecting cancer-related information,” and were instead directed straight to the question about “difficulties in information seeking.” Additionally, even if respondents had searched for information, if they selected either “do not wish to obtain information” or “do not use SNS or apps,” they were not asked to answer the question about the “device used to gather information.”

For the intention to use psychosocial support and information provision services recommended by medical institutions (corresponding to “behavioral intention” in the TPB), a yes or no question—“use of hospital-recommended apps and services for counseling support and information gathering”—was used (1 item). Only participants who answered “Yes” were able to proceed to the subsequent question on “advantages of psychosocial support and information-gathering services recommended by medical institutions.”

“Difficulties in information seeking” consisted of 7 items, “trusted information sources” consisted of 8 items, and “evaluation of interpersonal relationships between patients and health care providers” consisted of 9 items. To reduce the respondent burden, these were presented in a matrix format and measured using a 5-point Likert scale.

To assess personality traits, the Japanese version of the 10-Item Personality Inventory (TIPI-J) [], which uses 10 items and a 7-point Likert scale, was used. The TIPI-J measures the Big Five personality dimensions—extraversion, agreeableness, conscientiousness, neuroticism, and openness—with 2 items each and has shown high internal consistency: extraversion (α=.92), agreeableness (α=.85), conscientiousness (α=.82), neuroticism (α=.91), and openness (α=.86) []. The TIPI-J has also been shown to be associated with the emotional and cognitive health of patients with brain tumors [], making it a valid indicator for this study.

To ensure accessibility and usability, a test page simulating the survey environment was created before the actual survey to confirm that the question display and skip logic functioned accurately.

Targeting and Survey Distribution

The survey was conducted privately by Asmarc Inc (a Japanese marketing research company) from December 10 to December 12, 2024. The company operates an online research panel with over 16 million registered members. After agreeing to the membership terms, registered members completed preliminary and final registration, during which they provided basic demographic information, such as age, sex, place of residence, employment status, and medical history.

The survey was conducted using the D-style web and Monitas platforms (Asmarc Inc) and was conducted using a list of available questionnaires. The eligibility criteria for this study were based on patient-reported outcome criteria for survivors with cancer 1‐5 years after diagnosis []. Participants were required to be between 20 and 89 years of age at the time of the survey, to have been diagnosed with cancer at least 1 year prior, and to be either currently undergoing treatment or within 5 years of completing treatment. The following exclusion criteria were applied: participants who reported their treatment status as “before cancer treatment” or “other,” those who had been diagnosed with cancer less than 1 year prior, and those for whom more than 10 years had passed since the end of treatment.

The sample size was calculated using cancer incidence statistics in Japan [] as the reference population. Assuming a 90% CI and a 5% margin of error, the required sample size was estimated to be approximately 270 participants. This calculation was performed using the pwr package (version 1.3-0) in R. The pwr package was originally developed by Stéphane Champely and is currently maintained by Helios De Rosario. Considering a previous recommendation that at least 300 participants are desirable for conducting LCA [], the minimum target sample size for this study was set at 300 participants.

Statistical Analysis

In this study, only data for participants with responses to all questions were included, and statistical analyses were performed using R (version 4.3.2; R Foundation for Statistical Computing). Sociodemographic information was stratified on the basis of the use of psychosocial support services, specifically by responses of “Yes” or “No” to “use of hospital-recommended apps and services for counseling support and information gathering.” Descriptive statistics were calculated for each group. The sociodemographic information included sex, age, place of residence, occupation, and “device used to collect cancer information.” Cancer-related information consisted of “cancer stage,” “cancer type,” “presence or absence of recurrence,” “presence or absence of metastasis,” “current treatment status,” “time of cancer diagnosis,” and “time of cancer treatment completion.” Place of residence was recategorized into 6 categories: Hokkaido and Tohoku, Kanto, Chubu, Kinki, Chugoku and Shikoku, and Kyushu and Okinawa. To simplify the classification of cancer types, lung cancer and small cell lung cancer were grouped as lung cancer; esophageal, gastric, small intestine, and colorectal cancers were grouped as gastrointestinal cancer; liver, biliary tract, and pancreatic cancers were grouped as liver, biliary tract, and pancreatic cancer; kidney cancer and cancers of the renal pelvis and ureter and urethra, and bladder cancer were grouped as kidney, ureter, and bladder cancer; ovarian, endometrial, and cervical cancers were grouped as gynecologic cancer; leukemia, malignant lymphoma, and multiple myeloma were grouped as hematologic cancer; and brain tumors, head and neck cancer, thyroid cancer, cancer of unknown primary, skin cancer, and malignant melanoma were grouped as other. For information-seeking behaviors of patients with cancer, descriptive statistics were calculated regarding “having searched for information about cancer,” “sources of cancer information,” “SNS and apps used to gather cancer information,” and “benefits of using psychosocial support and information-gathering services recommended by medical institutions.” To understand the relational structure among items, analyses were conducted using tetrachoric correlations. Exclusive options such as “did not search” for “having searched for information about cancer,” “did not want to obtain information” for “sources of cancer information,” “do not use SNS or apps” for “SNS and apps used to gather cancer information,” and “not applicable” for “benefits of using psychosocial support and information-gathering services recommended by medical institutions” were excluded from the statistical analyses. For the TIPI-J, descriptive statistics were calculated, and the relational structure among the items was examined using polychoric correlations.

Structural Validity of the Exploratory Factor Analysis

Exploratory factor analysis (EFA) was conducted according to the reporting standards proposed by Watkins []. EFA was chosen to identify the underlying factor structure. The measurement scales used were “behavioral barriers experienced when collecting information about cancer,” “reliable information sources,” and “relationships with doctors, medical professionals, and society.” Each item was rated on a 5-point Likert scale ranging from 1 to 5. The appropriateness of conducting EFA was confirmed using the Bartlett test of sphericity []. The criteria for determining the number of factors included the Kaiser criterion of eigenvalues greater than 1 [] and the minimum average partial (MAP) method []. Eigenvalues exceeded 1 up to the fifth factor (fifth factor=1.26), and the MAP exhibited its lowest value (0.03) at the fifth factor; therefore, a 5-factor solution was adopted. To choose the factor extraction method, multivariate normality was examined using the Mardia test []. Because skewness (skewness=62.14; P<.01) and kurtosis (kurtosis=613.96; P<.01) were significant, multivariate normality was rejected, and a method that did not assume normal distribution was required. Moreover, because the items used in this study were ordinal scales based on a 5-point Likert scale, factor analysis using the minimum residual method with a polychoric correlation matrix, which is suitable for ordinal data, was adopted [,]. Oblimin rotation [] was used for simplification and theoretical convergence. Because oblique rotation was used, pattern coefficients of ≥0.37 were judged to be theoretically meaningful loadings, and coefficients of ≥0.40 were considered the primary criterion [] for substantial interpretation. Cross-loadings of ≥0.3 were examined for interpretative caution. Internal consistency was determined using Cronbach α and McDonald ω, and factors with a reliability of ≥0.7 [,] and theoretical significance were judged as appropriate. The analysis was conducted using the psych package (version 2.6.2) in R. The psych package was developed and is maintained by William Revelle [].

The LCA

In this study, we hypothesized that information-seeking behaviors and the use of psychosocial support services among survivors with cancer are not uniform and that multiple latent profiles exist. To identify these subgroups on the basis of statistical modeling, we used LCA []. Unlike traditional distance-based cluster analysis, LCA estimates latent structures using probabilistic models, allowing for objective model comparisons using information criteria, making it well suited for identifying latent profiles []. The analysis followed the reporting standards of Weller et al [], and the optimal number of latent classes was determined on the basis of statistical fit indices and theoretical interpretability []. For the measurement variables, we recoded variables to ensure estimation stability, categorizing age, occupation, treatment status, and information-gathering devices. To examine the indicators included in the LCA, we conducted a polychoric correlation analysis and included 10 variables—sex, age, employment status, cancer type, stage, recurrence, metastasis, treatment status, time since completing treatment, and information-gathering devices—which were confirmed to have intervariable correlations of ±0.3 or higher. Model fit was assessed primarily using Bayesian information criterion (BIC), adjusted Bayesian information criterion (aBIC), and consistent Akaike information criterion (cAIC), and the model with the smallest value was selected as a candidate. Entropy was checked, with ≥0.8 considered acceptable and ≥0.9 considered ideal []. For the final model selection, we carefully evaluated whether each class had a sample size of n>50 or >5% of the total sample and whether each class profile was clinically meaningful. The R poLCA package (version 1.6.0.1), developed by Drew A Linzer and Jeffrey B Lewis, was used for the analysis [].

Multiple Comparison Tests Between Identified Latent Classes

We examined the differences in each survey item among the latent classes. The variables compared included “counseling support and information gathering through apps and services recommended by the hospital,” “survey items related to cancer,” “sources of information about cancer,” “SNS and apps used to gather cancer information,” and “advantages of using psychosocial support and information-gathering services recommended by medical institutions,” as well as the factors identified in the EFA and the TIPI-J. To assess statistical significance, we adopted nonparametric testing methods with a significance level of P<.05. For binary items, we used the chi-square test for independence, and for ordinal items, we used the Kruskal-Wallis test. For items where significant differences were observed, we conducted residual analysis for binary items and the Wilcoxon rank-sum test (P=.02) for ordinal items as post hoc analyses using the Bonferroni correction and evaluated the association between the latent classes and each survey item.

Ethical Considerations

This study was approved by the Research Safety and Ethics Committee of Tokyo Metropolitan University of Industrial Technology (approval no 23020). The survey was conducted in compliance with the Declaration of Helsinki and relevant ethical guidelines, and only those who provided informed consent participated in the study. The collected data were securely stored in a password-protected database that was accessible solely to the research team. Participants were compensated with reward points according to the regulations of each platform. The allocation and exchange of points were managed exclusively by the survey company, and the research team was not informed about individual point allocations. Participants who responded via the D-style web platform received 2 points for the preliminary survey and 15 points for the main survey, totaling 17 points upon completing both surveys. Participants who responded via the Monitas platform received 2 points for the preliminary survey and 9 points for the main survey, totaling 11 points upon completing both surveys.

ResultsParticipant Characteristics

Of the 4183 members registered on the D-style web and Monitas platforms, 1095 individuals who indicated in their preregistration information that they “currently have cancer or have had cancer in the past” were selected for the screening survey. Of these, 627 individuals responded (participation rate: 57.3%). Among the respondents, 350 met the eligibility criteria for the main survey and proceeded to participate (transition rate: 55.8%). Accordingly, the final analysis included 350 survivors with cancer (350/1095, 31.9%; ).

The survey participants comprised 175 men and 175 women (175/350, 50%). The average age was 57.9 (SD 10.8) years. The largest age group was 60‐69 years (138/350, 39.4%), followed by 70‐79 years (89/350, 25.4%). Regarding occupation, the most common response was unemployed (108/350, 30.9%), followed by company employees (84/350, 24%). Most participants resided in the Kanto region (156/350, 44.6%). Regarding clinical characteristics, the most common types of cancer were breast cancer (93/350, 26.6%) and gastrointestinal cancer (91/350, 26%). Many patients were at an early stage, with stage 1 accounting for 156 of 350 (44.6%) participants. Regarding treatment status, more than half had completed treatment and were undergoing regular follow-up visits at the hospital (202/350, 57.7%). Smartphones (193/350, 55.1%) were the most commonly used devices to gather information about cancer. In contrast, 105 of 350 (30%) participants reported that they did not search for information. Regarding the use of hospital-recommended apps and services for counseling support and information gathering, 282 of 350 (80.6%) participants answered “yes,” and 68 of 350 (19.4%) participants answered “no.” Among participants who reported that they did not search for information in response to the question about devices used to gather information about cancer (105/350, 30%), most also reported “yes” regarding the use of hospital-recommended apps and services for counseling support and information gathering (91/105, 86.7%), with only 14 of 105 (13.3%) reporting “no.” Of those who answered “yes,” 161 of 282 (57.1%) had completed treatment and were under regular follow-up, and 72 of 282 (25.5%) were receiving outpatient or home medical care. Smartphones were the most commonly used devices to gather information about cancer (161/282, 46%; ).

‎Figure 1. Flowchart of participant selection process. Table 1. Patient characteristics (N=350).CategoryTotal, n (%)Yes, n (%)No, n (%)Use of hospital-recommended apps and services for counseling support and information gatheringSexMale175 (50)141 (40.3)34 (9.7)Female175 (50)141 (40.3)34 (9.7)Age (years)Mean (SD)57.9(10.8)20‐293 (0.9)3 (0.9)0 (0)30‐397 (2)6 (1.7)1 (0.3)40‐4932 (9.14)28 (8)4 (1.1)50‐5976 (21.7)56 (16)20 (5.7)60‐69138 (39.4)114 (32.6)24 (6.9)70‐7989 (25.4)71 (20.3)18 (5.1)80‐895 (1.44)4 (1.14)1 (0.3)JobCompany employee84 (24)73 (20.9)11 (3.1)Civil servant9 (2.6)8 (2.3)1 (0.3)Self-employed21 (6)18 (5.1)3 (0.9)Company director6 (1.7)3 (0.9)3 (0.9)Freelance6 (1.7)4 (1.1)2 (0.6)Housewife and househusband67 (19.1)48 (13.7)19 (5.4)Part-time job44 (12.6)38 (10.9)6 (1.7)Unemployed108 (30.9)85 (24.3)23 (6.6)Other5 (1.4)5 (1.4)0 (0)AreaHokkaido and Tohoku30 (8.6)23 (6.6)7 (2)Kanto156 (44.6)129 (36.9)27 (7.7)Chubu52 (14.9)38 (10.9)14 (4)Kinki68 (19.4)53 (15.1)15 (4.3)Chugoku and Shikoku22 (6.3)19 (5.5)2 (0.9)Kyushu and Okinawa22 (6.3)20 (5.7)2 (0.6)Cancer typeLung cancer29 (8.3)24 (6.9)5 (1.4)Breast cancer93 (26.6)74 (21.1)19 (5.5)Gastrointestinal cancer91 (26)74 (21.1)17 (4.9)Liver, biliary tract, and pancreatic cancer14 (4)9 (2.6)5 (1.4)Kidney, ureter, and bladder cancer17 (4.9)14 (4)3 (0.9)Prostate cancer39 (11.2)31 (8.9)8 (2.3)Gynecological cancer21 (6)16 (4.6)5 (1.4)Blood cancer30 (8.6)27 (7.7)3 (0.9)Other16 (4.6)13 (3.7)3 (0.9)StageStage 1156 (44.6)129 (36.9)27 (7.7)Stage 287 (24.9)67 (19.1)20 (5.7)Stage 368 (19.4)54 (15.4)14 (4)Stage 439 (11.1)32 (9.1)7 (2)Cancer recurrenceYes62 (17.7)55 (15.7)7 (2)No288 (82.3)227 (64.9)61 (17.4)Cancer metastasisYes55 (15.7)47 (13.4)8 (2.3)No295 (84.3)235 (67.1)60 (17.1)Current treatment statusInpatient treatment5 (1.4)5 (1.4)0 (0)Outpatient or home visit treatment87 (24.9)72 (20.6)15 (4.3)Completed treatment and under regular observation202 (57.7)161 (46)41 (11.7)Completed treatment, no hospital visits56 (16)44 (12.6)12 (3.4)When was the cancer diagnosed?Within 3 years127 (36.3)109 (31.1)18 (5.1)Within 5 years85 (24.3)61 (17.4)24 (6.9)Within 10 years96 (27.4)75 (21.4)21 (6)>11 years42 (12)37 (10.6)5 (1.4)Time when cancer treatment endedDuring treatment58 (16.6)49 (14.0)9 (2.6)Within 1 year114 (32.6)96 (27.4)18 (5.1)Within 3 years86 (24.6)60 (17.1)26 (7.4)Within 5 years92 (26.3)77 (22.0)15 (4.3)Devices used to gather information about cancerSmartphone193 (55.1)161 (46.0)32 (9.1)Personal computer12 (3.4)10 (2.9)2 (0.6)Tablet (eg, iPad)8 (2.3)6 (1.7)2 (0.6)Other32 (9.1)14 (4)8 (5.1)Not searching for information105 (30)91 (26)14 (4)Information-Seeking Behavior of Patients With CancerTopics Researched Regarding Cancer

The most frequently researched topic related to cancer was “my disease,” which accounted for 75.7% (265/350) of responses. This was followed by “about treatment” (68.6%, 240/350) and “progress and prognosis” (55.4%, 194/350). “Treatment costs” accounted for 38.9% (136/350) of responses, “life during treatment” for 36% (126/350), and “other patients’ treatment and daily life” for 27.1% (95/350). Regarding correlations between categories, “life during treatment” had a strong positive correlation with “about my own condition” (ρ=0.7) and “progress and prognosis” (ρ=0.64). “Other patients’ treatment and daily life” showed a moderate positive correlation with “about my own condition” (ρ=0.61).

Sources of Cancer Information and Use of SNS and Apps

The most common source of cancer-related information was “internet/SNS” (58.9%, 206/350), followed by “doctors, nurses, and other medical staff” (50.6%, 177/350). Other sources included “newspapers, magazines, and books” (17.7%, 62/350), “family, friends, and acquaintances” (15.7%, 55/350), “National Cancer Center website” (15.4%, 54/350), and “cancer counseling support center” (7.4%, 26/350). Regarding correlations between categories, a moderately strong correlation (ρ=0.69) was observed between “newspapers, magazines, and books” and “television and radio,” Among the SNS and apps used, YouTube (Google LLC) was the most common (13.4%, 47/350), followed by blogs (10%, 35/350), LINE, provided by LY Corporation (7.4%, 26/350), and X (formerly Twitter, provided by X Corp; 6%, 21/350). These items exhibited positive correlations with each other (ρ=0.48-0.6).

Benefits of Using Psychosocial Support and Information-Gathering Services Recommended by Health Care Institutions

The most commonly identified benefits of psychosocial support services or apps were “being able to obtain reliable information” (59.7%, 209/350) and “deepening knowledge about illness and treatment” (58.6%, 205/350), with more than half of the respondents emphasizing the quality and understanding of information. “Receiving information and support tailored to one’s own symptoms or situation” was cited by 45.4% (159/350) of patients, and “facilitating communication with doctors or medical staff” was cited by 29.4% (103/350) of patients. Regarding correlations between items, a strong positive correlation (ρ=0.71) was found between “deepening knowledge about illness and treatment” and “receiving information suited to one’s own symptoms or situation.” “Easy to operate and user-friendly” showed a moderate positive correlation with both “deepening knowledge about illness and treatment” (ρ=0.61) and “facilitating communication with doctors or medical staff” (ρ=0.6; and ).

Table 2. Trends in information-seeking behavior among patients with cancer.ItemValue, n (%)Things I have researched about cancerAbout treatment240 (68.6)My disease265 (75.7)Course or prognosis194 (55.4)Information about other patients95 (27.1)Life during treatment126 (36)Cost of treatment136 (38.9)Other20 (5.7)Sources of cancer informationCancer Consultation Support Center26 (7.4)Doctors, nurses, and so on177 (50.6)Contact at public health center or health center5 (1.4)Newspapers, magazines, and books62 (17.7)Television and radio40 (11.4)National Cancer Center website54 (15.4)Internet and SNS206 (58.9)Family, friends, and acquaintances55 (15.7)Other6 (1.7)SNS and apps used to gather cancer informationFacebook (Meta Platforms, Inc)14 (4)Instagram (Meta Platforms, Inc)15 (4.3)LINE (LY Corporation)26 (7.4)X (formerly Twitter; X Corp)21 (6)YouTube (Google LLC)47 (13.4)Blog35 (10)Other46 (13.1)Benefits of using psychosocial support and information-gathering services recommended by health care institutionsAccess to reliable information209 (59.7)Easy to operate and use76 (21.7)Access to the latest medical information152 (43.4)Easy to understand correct usage and operation78 (22.3)Personal information is kept safe58 (16.6)Facilitate communication with doctors and medical staff103 (29.4)Gain knowledge about illnesses and treatments205 (58.6)Get information and support tailored to your symptoms and situation159 (45.4)

aSNS: social networking services.

TIPI-J

The mean scores reported by respondents were highest for neuroticism at 8.62 (SD 2.47), followed by openness at 8.24 (SD 2.43), extraversion at 7.65 (SD 2.6), and conscientiousness at 7.15 (SD 2.29). Agreeableness exhibited the lowest value among the 5 traits, at 5.84 (SD 2.03). A Shapiro-Wilk test [] was conducted to assess the distribution of each scale, and the results were significant for extraversion (W=0.98; P<.001), agreeableness (W=0.95; P<.001), conscientiousness (W=0.98; P<.001), neuroticism (W=0.98; P<.001), and openness (W=0.98; P<.001), indicating that the data were not normally distributed. The correlations among the subscales were uniformly low (r=−0.43 to 0.37; ).

EFA Results

The Bartlett test of sphericity was conducted to assess the applicability of factor analysis. On the basis of the results, the null hypothesis that the correlation matrix is an identity matrix was rejected (χ²210=5425.91; P<.001), indicating that there was a sufficiently robust correlation structure among the items. The Kaiser-Meyer-Olkin statistic was 0.81, confirming sampling adequacy and determining that factor analysis was appropriate. The number of factors was determined on the basis of eigenvalues, MAP, and theoretical consistency. Eigenvalues exceeded 1 up to the fifth factor (factor 1=6.33, factor 2=4.27, factor 3=2.56, factor 4=1.86, and factor 5=1.26), and the cumulative contribution of the 5-factor solution was 67.83%. MAP decreased with the addition of more factors and improved up to the fifth factor. In addition to these results, considering that the measurement included several conceptual domains such as “difficulties in information seeking,” “support and consultation for care,” “relationship with the physician,” “trust in information from medical institutions and specialized websites,” and “trust in information from outside medical institutions,” a 5-factor solution was adopted.

Factor 1 included items reflecting difficulties encountered in the process of seeking cancer-related information (factor loadings: 0.54-0.9) and was named “difficulties in information seeking.” Factor 2 included items related to continued employment before treatment, concerns about appearance, convalescence, precautions in daily life, and workplace considerations (factor loadings: 0.42-0.74) and was interpreted as “evaluation of support for cancer care.” Factor 3 included items on the primary physician’s knowledge and experience, ease of consultation, and trust in information provided by physicians (factor loadings: 0.67-0.81) and was named “evaluation of the relationship with the physician.” Factor 4 included trust in information from cancer-specialized or hospital or pharmaceutical company websites (factor loadings: 0.73-0.85) and was named “evaluation of the reliability of information from medical institutions and cancer-related websites.” Factor 5 included trust in sources of information outside medical institutions, such as family and acquaintances, other patients, and social networking sites (factor loadings: 0.47-0.76) and was named “assessment of the reliability of information from nonmedical institutions.”

The item “I had to visit several websites” loaded on both Factor 1 (0.60) and Factor 4 (0.33), and “the most reliable source is information from the physician” loaded on both Factor 3 (0.67) and Factor 4 (0.32). The internal consistency of the factors ranged from α=.75 to .9 and ω=0.78 to 0.92, with small interfactor correlations, the highest being between Factors 2 and 3 (r=0.47), justifying the treatment of the 5 factors as independent subconcepts ().

Table 3. Factor structure of information-seeking behavior and evaluation of information sources and psychosocial support services among survivors with cancer.ItemMean (SD)SkewnessKurtosisFactor 1Factor 2Factor 3Factor 4Factor 5H²Factor 1: difficulties in information seekingI didn’t know where to look for it2.88 (1.18)0.08−1.010.90.10−0.150.010.82I didn’t know who to ask2.81 (1.18)0.15−1.030.880.04−0.04−0.090.050.8I didn’t know what to look for2.92 (1.17)0.13−0.960.880.050.03−0.230.040.81I couldn’t find the information I wanted right away3.11 (1.14)−0.16−0.90.77−0.0100.12−0.120.61I didn’t know which information I could trust3.04 (1.08)−0.07−0.80.71−0.090.07−0.010.050.51I had to go to several websites3.31 (1.09)−0.43−0.640.60−0.040.33−0.030.47I couldn’t ask the doctor a question2.25 (1.12)0.73−0.280.54−0.1−0.39−0.010.060.57Factor 2: support and consultation for careI think a medical professional spoke to me about continuing to work before starting treatment3.15 (1.1)−0.2−0.54−0.020.74−0.060.10.020.55I felt I could talk to a medical professional about my concerns about changes in appearance3.07 (1)−0.1−0.3100.73−0.060.030.040.51I think I was able to talk to someone about my illness and medical treatment3.27 (0.97)