Breast cancer has become the most prevalent type of cancer affecting women in Indonesia and worldwide, and the number of deaths caused by breast cancer is growing every year [1,2]. In 2018, there were an estimated 2,088,849 new cases and 626,679 mortalities related to breast cancer [1]. Breast cancer occurs with an increase in the number of malignant cells originating from the inside layer of the mammary glands [3]. In the United States, breast cancer is detected by mammography (43%, 156/361), breast self-examination (25%, 90/361), clinical screening with breast self-examination (14%, 47/361), and accidents (18%, 64/361) [4]. Breast cancer prevention requires every woman to perform a breast self-examination as an early diagnosis mechanism for all ages after the first menstruation, which is expected to help reduce breast cancer mortality [5-8].

Moreover, curiosity about seeking health care–related information through the internet and social media has been gradually on the rise. Social media users prefer to seek information from social media [9,10], as other users share a substantial amount of information pertaining to breast cancer. Almost 87% of the total posts on Facebook related to breast cancer consist of support groups [11]. Other platforms such as Twitter include surveys on breast cancer education, shared stories about breast cancer survival, treatment plans, and images showing the progress of certain treatments [12]. Consequently, patients prefer to use social media to talk about sensitive body issues (such as breast cancer) as a more convenient venue than face-to-face interaction with a physician [13-16]. At the same time, physicians are actively participating in social media and health care systems related to breast cancer [9,14,17,18], and they tend to use social media for assisting, treating, and consulting on cancer [9,18,19]. Although the physician-patient interaction on social media platforms offers many conveniences, it also has important downsides.

There are reported cases of physicians behaving inappropriately on social media, such as posting incorrect information, misrepresenting their credentials, posting improper content, and false advertising [20-22]. The impact of such unethical behavior [19,23-26] can result in embarrassing patients and losing their trust. Since trustworthiness is an essential factor in any physician-patient relationship, the decrement of trust not only affects the health care business but also causes shame and depression for the patient [20,21,23]. Thus, in this study, we explored a trust model that can support and eliminate this issue. Toward this end, we focused on enhancing the multifaceted trustworthiness model proposed by Quinn et al [27] to be adopted for health care treatment on social media. The current multifaceted trust model calculates a trust score based on social interaction on social media platforms. Thus, the two-way trust evaluation adopted in this trust model is suitable for considering the patient-physician interaction in the health care domain. However, the multifaceted model has its limitations, since there is no credential representation mechanism in building trust context, no informed consent contract between parties trying to build trust, no mechanism to protect confidential data [19], and no preservation of user privacy [14]. Thus, there is a gap to be filled regarding how to best protect confidential information and ensure that each interaction and communication on social media is based on ethics.

The trust in social media is not personalized, specific, and single-faceted, but is rather generalized in a group context [27-29], and trust level cannot be annotated [27,28]. By contrast, the existing multifaceted trust model is personalizable, specializable, and capable of measuring the accuracy of trust recommendation [28]. As a result, Chieng et al [30] presented personalized comments or photos on social media as a user-centric model.

The objectives of this study were to: (1) identify the trustworthiness indicators that are most suitable for a breast self-examination system, (2) design and propose a breast self-examination system, and (3) evaluate the multifaceted trustworthiness interaction between patients and physicians using the breast self-examination system. Implementation of the multifaceted trustworthiness model into the breast self-examination system can identify the most preferred indicator of trustworthiness, offering relationship feedback between the patient and physician based on trust value and trust level. This could ultimately provide a more trustworthy and ethical patient-physician interaction on social media platforms such as Facebook.

The trust theory introduced by Rotter [31] in 1967 is known as interpersonal trust, defined as “an expectancy held by an individual or a group that the word, promise, verbal or written statement of another individual or group can be relied upon.” The trust principle was introduced by Mayer and Davis in 1995 [32], which posits that factors related to the trustor and trustee will lead to trust. The characteristic of the trustor is trust propensity, which is a general willingness to trust others. In other words, trust propensity is a cause of risk behavior. People with different backgrounds, personality types, cultures, and experiences will differ in their propensity to trust.

On the other side, the main characteristic of the trustee is trustworthiness, which is measured as the motivation to lie. For example, if a trustee will gain something through lying, they will be seen as less trustworthy [32].

Trust can therefore be defined as the confidence in somebody or a belief that somebody is good and honest [33]. Trustworthiness has been mentioned in the context of character honesty and integrity in health care [34], and it is a context-dependent and personalized characteristic [30]. According to Quinn [28], the multifaceted trustworthiness model is personalizable and specializable [28]. The trust characteristic in social media has been defined according to the following four traits [30,35-37]. The first is asymmetry, as trust between two users is not identical. That is, individual A could trust individual B, whereas individual B might not essentially trust individual A and vice versa. Second, trust is transitive: longer links create less trust. For example, consider that A and B are friends, and they trust each other. B is friends with C, but A does not know C. Since A knows B and trusts B’s friends, A might trust C to a certain extent. However, C is friends with D, whom neither A nor B knows, and thus A finds it difficult to trust D because of the distant connection and the fact that they do not know each other. Third, trust is context-dependent, according to time, situation, and experience. People tend to exhibit differences in trust based on the context. Fourth, trust is personalized as a subjective view. That is, the trustworthiness of a particular person might be viewed differently by two different people.

The key indicators of the multifaceted trust model (Figure 1) are honesty, reputation, competency, reliability, credibility, belief, confidence, and faith [27]. Honesty means the person makes good-faith agreements, tells the truth, and fulfills any promises made. Competency is the ability of one person to fulfill another person’s needs. Confidence is “a feeling of certainty or easiness regarding a belief one holds” [38]. Reputation is part of the social notion of trust [39]. Belief is justified and should be accepted (ie, acceptable without argumentative support) [34].

Figure 1 shows how the concept of personalization allows the user to declare the idea that competency is influenced by reputation (ie, competency derived from reputation), credibility is influenced by belief (ie, credibility informed by belief), and so on. This concept can be repeated to construct a trust model to suit the user’s needs and to reflect the user’s subjective view of trust.

The trust measurement model in an open social network can be classified into five main models [36]: online reputation models, Marsh trust management, multicontext trust, trust inference for social networks (TISoN), and action-based trust.

Breast self-examination is a method for early detection by which women examine their breasts to facilitate detection and alleviate any fear of cancer [45]. Breast self-examination is a regular monthly breast check using a mirror to observe any abnormal changes on the breasts [3], and is also considered to be the best tool for early breast cancer detection [46].

As patients prefer using social media to make appointments, receive reminders, diagnostic test results, provide information about their health, and as a forum for asking general questions related to health care [14], some specific features have been requested by patients as a reference to develop a breast self-examination system. Based on existing breast self-examination systems (Table 1), nine standard features should be embedded in any online breast self-examination system, including user account management, calendar, self-exam wizard, history, chat room, location, knowledge, video tutorial, and forum.

The physician-patient interaction is a communication process that describes the shared nature of the problem, treatment aims, and psychosocial care [54]. The physician-patient interaction emphasizes the behaviors of physicians toward patients. Behavior consent is the content given by the physician on health solutions (instrumental behavior) and the capability of physicians to treat patients (affective behavior) [55].

The research flow is outlined in Figure 2, following ethical phenomena in social media. The multifaceted trust model was selected for this study, as it is able to provide a subjective view of individual trust for each user. The prototype of the breast self-examination system was based on existing breast self-examination systems (Table 1). The architecture of the breast self-examination system with trustworthiness indicators was designed as a rating for the trust value for each patient consultation with a physician. Based on a survey with patients and physicians, the relationship between a particular physician trust value and trust level was identified. Thus, any patient who wants to choose a physician for consultation could refer to the physician’s trust level.

The comparison among the existing breast self-examination systems in Table 1 highlights the user account as an essential feature, which is based on the Beyond the Shock app [49] user account for securing patient and physician data. The remaining six apps do not have user accounts because they do not secure patient data, especially with respect to female breasts. The calendar is the feature that establishes the menstrual cycle period for performing a monthly self-exam. Of the seven available apps, four (BSE, Keep A Breast, Breast Control, and Makna) provide a calendar for setting a monthly self-exam reminder. The self-exam wizard feature explains how to perform a self-exam systematically, which should exist in any breast self-examination system. Only Beyond the Shock does not provide this feature. The history feature helps to record the patient’s activity and the interaction process with their physician, which is present in several apps (BSE, Dr. K’s Breast Checker, and Breast Control). The history feature records data by keeping a medical history for each patient, which will help physicians to trace each patient’s performance. It is vital that this feature is secure due to the private nature of the information. The chat room allows for direct interaction between the patient and physician, which is essential for allowing patients to have direct interaction and communication with a physician without a face-to-face meeting. None of the apps currently supports the chat room feature, as they focus only on self-exam without connecting to a physician.

The next feature is location, which identifies available physicians nearest to the patient. This feature helps patients find a physician for further consultation based on their self-exam results. This feature is found in Dr. K’s Breast Checker and Makna, which provide an address of a hospital or clinic for consultation with a physician. The knowledge feature provides scientific information related to breast cancer prevention, which is also an essential feature for obtaining breast cancer–related information for patient education on breast cancer. This feature is offered by several apps (Beyond the Shock, Dr. K’s Breast Checker, Breast Control, and Makna). The video tutorial feature refers to any related breast cancer information provided via video as guidance. Video tutorials are provided in Beyond the Shock, Daisy Wheel, and Makna as an essential feature to help patients view information related to living with breast cancer. The forum is an open space to find current news or cases from physicians and patients, which offers a space where physicians provide general information to all patients on breast cancer prevention. Beyond the Shock uses a forum as an essential space for discussion between all physicians and patients in the same area. We included all criteria shown in Table 1 in our proposed breast self-examination system.

We refer to Quinn et al’s [27] multifaceted trust model, which uses the idea of implementing a trust management model to act as the subjective view on trust. The breast self-examination system involves the eight indicators of trustworthiness as a rating system in a chat room setting. These indicators will determine the value of trust based on the user’s interaction experience that is entirely personalized, transitive, and context-dependent. The personalized view will allow users to choose their trust value (ie, patients will give a value to each indicator in reference to their physician, and vice versa). The transitive view will offer each physician trust value as a reference when a patient recommends a physician to another patient. The context-dependent view will give patients flexibility in rating a physician; they can edit their trust level regarding physicians from time to time based on several consultations.

The trust level task is designed through the average rating value (ARV), which is used to calculate the average trust value given by the patient to their physician. The ARVs were generated based on the idea of Marsh [41] and Daskivich [56] to identify the trustworthiness level of a physician or patient with independent values. The trustworthiness level and independent value of a physician denoted by ARV are shown in Table 2.

If ARVs are between 9 and 10, the trustworthiness level is considered to be high, and the independent value is 5, whereas lower values indicate higher levels of distrust. Therefore, the correlation analysis will depend on the independent value as the critical element.

The prototype architecture of our breast self-examination system was based on Quinn et al’s [27] trust model, miniOSN [30], and miniOSN2.1 [30]. Patients can personalize accessibility to posted information, comments, and shared history data using the rating feature, as well as limit the physician to view the content of the patient data. The patient allows viewing the trust value and trust level of physicians. A physician’s trust value is based on the average of the trustworthiness indicators [28] and the ARV as the rating trust level [41,56]. The trust level of a physician is then identified by the trust value. To identify the ranking among multifaceted trustworthiness indicators, we evaluated the relationships between each indicator and the trust value. The higher ranking of a trustworthiness indicator is determined by a stronger correlation between the trust value and each indicator. Therefore, the ranking could evaluate the importance of the trustworthiness indicator in the multifaceted model. In the breast self-examination system, patients can personalize accessibility to set the value of posting and comments according to the trustworthiness indicators or trust level.

The trustworthiness mechanism on the breast self-examination system is measured through chat rooms and forums. When a patient request is accepted by a particular physician for consultation, the patient will encourage the trust level option of the physician, which is only seen by the patient. The patient can edit the trust level of a physician by choosing values ranging from 1 to 10 from the trustworthiness indicator’s ARV. The patient can also update the level of trust from time to time. For example, Figure 3 shows that the request of Reka (patient) for a consultation with physicians Sandana and Lukman was accepted by both physicians. After the meeting, she gave the experience a rating of 8 as the value for the physicians.

An action-based trust algorithm was further implemented for the computational mechanism on the proposed breast self-examination system. This algorithm can measure the credibility of users on social media. The capabilities for evaluating and calculating the trust factors on user content disclosure include sharing personal records, sharing a post, comment, photo, and posting a message [44]. The computation of trust value is for a physician that acts on user content disclosure, namely as a trust factor. A physician trust factor may decrease or increase based on whether the patient selects a sensitive or not sensitive option. For example, the patient will likely select a sensitive option for a self-exam photo after completing the monthly self-examination. The sensitive option must be accompanied by informed consent from the patient before it is shared with the physician.

The action-based trust algorithm divides the measurement of each user action into weights, including the weight of action (Wa), weight of post (Wp), and weight of category (Wc). At the same time, Wc is a function of the weight for category. These weights, Wa, Wp, and Wc, are identified as the parameters of the trust factor. Table 3 shows the cluster of weights for test cases simulating the algorithm [44].

We conducted a survey using open-ended interviews with 32 participants [57,58] and 77 interactions in the breast self-examination system. The survey was conducted from February 3, 2020 to March 30, 2020. The participants were physicians and female outpatients, all of whom had used the breast self-examination system. The 32 participants included 20 females and 12 males, comprising 16 physicians [58] and 16 female outpatients [57]. Of the 16 physicians, 12 were general practitioners and 4 were oncology specialists.

The 16 outpatients were healthy females who were aware of the health care system. Eight of these outpatients were aged 18 to 25 years old and the other eight were aged above 25 years. However, not all 32 participants ultimately completed the interaction task in the chat room and forum due to the consultation period. There were 24 active chat room participants and 22 participants interacted in the forum. The evaluation monitored the activity in the chat room (interaction between physician and patient) and the sharing of information by the physician through the forum. The interview started with an introduction to the research goal and an explanation of the flow on how to use the breast self-examination system.

This study was approved by the research ethics board of Esa Unggul University committee (No. 0155-20.133/ DPKE-KEP/ FINAL-EA/ UEU/ V/ 2020).

The prototype follows a module design (Figure 4), which is classified into four phases [59]. The first phase of the user account is registering and logging into the system. The second phase of self-exam is the phase of conducting a personal self-exam on the breast and annotation into the system. The third phase of consultation with a physician is when the user finds a doctor and has a consultation. The fourth phase of open features for the public is the opportunity for the public to access knowledge, video tutorials, and forum features without obtaining a user account. The phrases in each phase are based on user privileges.

Each of the features of the breast self-exam system shown in Figure 4 has its own function. The calendar is a reminder system for performing a breast self-exam. The self-exam wizard is a guide to performing the correct self-exam daily. History serves as a self-exam record and tracks monthly breast self-exams. The location finds the nearest physician for receiving treatment. The chat room is a space for consultations between patients and physicians. Knowledge is a collection of links to news and expert opinions on breast health. Video tutorial is a collection of videos for performing the breast self-examination correctly. The forum is where physicians can share important information for patients related to breast cancer or breast self-exam.

To select the most appropriate trust model for the breast self-examination system, we performed a literature review on papers related to the trust model. Ultimately, 11 articles were selected for comparison analysis among existing trust models. Table 4 shows the analysis of the comparison from several perspectives, including the trust model, related domains, selection of trust factors, methodology, and benefits.

The selection of trust models to suit the breast self-examination system refers to the health care, internet, and social media domains. Among the 11 articles, 7 are related to health care, 2 pertain to the internet, 1 is associated with social media, and the other is related to psychology. The trust model related to social media is the Multifaceted Trust Model for Online Social Network Environment [30]. Initially, the multifaceted trust model was introduced by Quinn [27] for the internet environment. The multifaceted trust factors support a user-centric model that requires users to personalize trust. For instance, Abbas et al [60], Montaque et al [61], and Quinn et al [27] focus on different areas with respect to reliability. Abbas et al [60] focus on the overall reliability of health care software, and Montaque et al [61] focus on the overall reliability of medical technology. In contrast, Quinn et al [27] and Chieng et al [30] introduce reliability as being personalized and specialized to a particular user or things.

From the methodology perspective, 5 of the studies were based on a qualitative approach, 5 were based on a quantitative approach, and the remaining study was based on a structured literature review. The qualitative approaches include evaluation of trust in the relationship between the patient and physician [25,63,64], whereas the quantitative approaches concentrate on a questionnaire to obtain participants’ feedback. Therefore, the qualitative approach is more effective and relevant for garnering maximum performance out of the system.

The benefit perspective brings a consistent approach to choosing the trust model for the breast self-examination system. Two of the studies explored the trust model benefit that focused on the personalized and trust recommendation measurement offered by Quinn et al [27] and Chieng et al [30]. In contrast, the remaining trust model benefits focus on the general trust model of theoretical issues such as the instrument created [63], trust theory on the patient-physician relationship [25], patient trust in technology [61], and behavior approach theory [64]. Thus, a trust model related to the user-centric model is relevant to be embedded in the breast self-examination system.

Based on analysis of the 11 filtered articles, we decided to adopt the multifaceted trust model introduced by Quinn et al [27]. The reason for selecting the multifaceted trust model (Table 4) is that this model can provide a subjective view of individual trust for each user. This model is also a user-centric model that can personalize trust features such as comments and photos on social media [30]. In particular, trust can protect the physician’s reputation before the patient makes any decision and allows the user to choose a credible physician [44]. According to Singh and Chin [36], trust is a significant factor to attract a user to use the site for recommendation to others based on a rating feature. That is, patients can consider a physician’s credibility for consultations, and physicians can consider the patient’s honesty in providing information about their health status [36].