Yumin Lin, 11354135
Supporting Older Cancer
Patients in Treatment and
Hospital Decisions
Effects of Framing and Visual Illustrations on Older
Patients’ Comprehension of Core Information and
Informed Decision Making
Author: Yumin LinStudent Number: 11354135
School: Graduate School of Communication Science Thesis Type: Master’s Thesis
Program: Research Master’s
Supervisor: Prof. Dr. Julia van Weert Date of Completion: 30.01.2020
Yumin Lin, 11354135
Table of Content
Abstract ... 4
Introduction ... 7
Theoretical Framework ... 15
Cancer Patients’ Medical Decisions and Necessary Information ... 15
Ageing, Information Processing and Decision Making ... 18
Strategies in Decision Support Tools for Older Cancer Patients ... 21
Framing Effects in Decision Support Tools... 22
Visual Illustrations in Decision Support Tools ... 26
Interaction Effects of Framing, Visual Illustration, and Age ... 31
Study 1 (Treatment Decisions) ... 33
Method ... 33
Design and Participants ... 33
Procedure ... 35 Stimuli Development ... 37 Measures ... 38 Statistical Analysis ... 47 Results Study 1 ... 48 Randomization Check ... 48 Manipulation Check ... 48 Hypothesis Testing ... 49 Conclusion Study 1 ... 52
Study 2 (Hospital Decision) ... 53
Method ... 53
Design and Participants ... 53
Procedure ... 54
Stimuli Development ... 55
Measures ... 57
Statistical Analysis ... 60
Yumin Lin, 11354135 Randomization Check ... 60 Manipulation Check ... 61 Hypothesis Testing ... 61 Conclusion Study 2 ... 67 General Discussion ... 68
Limitations and Future Research ... 76
Theoretical and Practical Implications ... 78
Reference ... 81
APPENDIX A. Stimulus Materials Study 1 ... 119
APPENDIX B. Stimulus Materials Study 2 ... 123
APPENDIX C. Questionnaire Study 1 ... 127
APPENDIX D. Questionnaire Study 2 ... 143
APPENDIX E. Results Pilot Test ... 160
APPENDIX F. Descriptive Statistics and Correlation Matrix of Study Variables ... 167
Yumin Lin, 11354135
Abstract
Background: Making decisions on treatment and hospital can be challenging for
older cancer patients. The way how decision-related information is framed and presented in decision support tools is expected to affect patients’ decision making.
Objective: The aim of this paper is to investigate the effects of framing and visual
illustrations on comprehension of core information and informed decision making, and whether the effects are different between people at different ages and levels of health literature and numeracy. This paper also aims to discuss different decisions (treatment vs. hospital) and different types of informed decision making in comparison..
Methods: Two online experiments (NStudy 1= 275; NStudy 2= 281) were conducted
among analogue cancer patients. A 2 (framing: positive vs. negative) x 2 (visual illustrations: present vs. absent) x 2 (age: younger vs. older) design was used in both studies. Four types of informed decision making were constructed based on the definition of attitude-choice consistency (Informed Decision Making Type I: based on comparative positive attitude; Informed Decision Making Type II: based on absolute
Yumin Lin, 11354135
positive attitude) and whether deliberation was considered (Deliberated Informed Decision Making Type I and Deliberated Informed Decision Making Type II).
Results: Results showed that positive framing significantly increased Informed
Decision Making Type II (p= .047) in hospital decisions, while negative framing marginally increased Deliberated Informed Decision Making Type I (p= .093) in treatment decisions. Although no moderating effect of age was found, results revealed that people with low levels of health literacy and numeracy benefitted from adding visual illustrations in terms of higher Deliberated Informed Decision Making Type II (p= .019) in hospital decisions as compared to text-only information. However, people with moderate health literacy and numeracy showed the opposite: The group was negatively affected by visuals in the hospital decision support tool and showed a significant lower Deliberated Informed Decision Making Type II (p= .043) as compared to text-only information. Similar results were found for Deliberated Informed Decision Making Type I, in which decreases were shown among people with high levels of health literacy and numeracy (p= .041) when visual illustrations were added.
Yumin Lin, 11354135
Conclusion: It is concluded that framing effects can differ between treatment and
hospital decision making. Although adding visual illustrations showed no effects in the general population, people with low levels of health literacy and numeracy can potentially benefit from visual illustrations in informed decision making, especially when the decisions are made deliberatively. Finally, the phenomena that visual illustrations can negatively influence informed decision making among people with high health literacy and numeracy need to be noticed so that providing information in multiple formats does not necessarily have advantages among this group.
Yumin Lin, 11354135
Introduction
The population in the world is rapidly ageing. By 2050, an estimation of 16% of the people in the world will be over 65 years of age, which almost doubles the share of 9% in 2019 (United Nations, 2019). Western countries are even more rapidly ageing. One in four people living in Europe and North America is expected to age over 65 by 2050 (United Nations, 2019). As a result of this growth, an increase in cancer incidences among older adults is reported, especially lung cancer, which mainly occurs in older people above 65 years old (American Cancer Society, 2019). Lung cancer causes the highest mortality from cancer (American Cancer Society, 2019; Cancer Research UK, 2016).
In recent decades, patients report increasing preference for playing an active role in medical decision making (Winkler, Hiddemann, & Marckmann, 2011; Hawley, & Jagsi, 2015). Accordingly, shared decision making (SDM), in which health professionals and patients cooperate in making the choice of tests and treatments (Coulter, 2012; Levit, Balogh, Nass, & Ganz, 2013), is identified as a significant component of high quality care, not only to provide patient-centered care but also to
Yumin Lin, 11354135
reduce overtreatment and unwarranted practice variation (Hoffmann, Montori, & Del Mar, 2014; Mulley, Trimble, & Elwyn, 2012; Stiggelbout, Pieterse, & De Haes, 2015). Several studies have presented the benefits of SDM in terms of increasing patients’ comprehension of medical options, confidence in decisions made, and satisfaction with both the information and clinician, as well as reduced levels of decisional conflict and the levels of anxiety and depression (Edwards, & Elwyn, 2006; Fallowfield, Hall, Maguire, & Baum, 1990; Gattelari, Butow, & Tattersall, 2001; LeBlanc, Kenny, O'Connor, & Légaré, 2009; Mosen et al., 2007; Stacey et al., 2017), which in turn can result in higher quality of life (Kashaf, & McGill, 2015; Kim et al., 2001).
However, the willingness and ability to participate in SDM depends on various aspects, including disease severity and age (Coulter, 2003). In case of a
life-threatening disease like cancer, patients sometimes encounter difficulties in participating in decision making (Sinding et al., 2010), or refuse to take
Yumin Lin, 11354135
An essential reason for the hesitation to participate in medical decision making is that patients often have difficulties to appropriately process and understand the provided information needed to select tests and treatment together with clinicians (Gaissmaier, & Gigerenzer, 2008; Peters, Hibbard, Slovic, & Dieckmann, 2007; Reyna, 2008). Patients are usually expected to process, understand and recall a relatively high amount of complex information about potential treatments in order to make an informed decision (Bekker et al., 1999), including numerically estimating chances of potential outcomes, both short- and- long-term (Trevena et al., 2013). This type of information is even more challenging for older patients to process. When people age, their working memory capacity declines (Bopp, & Verhaeghen, 2005), which creates difficulties for the older patients to extract relevant information from memory for processing, and reduces their abilities in deliberate reasoning and decision making (Malenka, Nestler, & Hyman, 2009; Miyake, & Shah, 1999). Older adults become more selective in the use of the deliberative mode of thinking, and invest more capacities in emotionally meaningful information, resulting in a decline in deliberative information processing (Peters, Hess, Västfjäll, & Auman, 2007;
Yumin Lin, 11354135
Santrock, 2008). Due to the changes as mentioned above in cognitive capacity and emotional focus among older adults, age could reduce their ability to make an informed decision. An informed decision is considered a vital element of quality health care in clinical field (Rimer, Briss, Zeller, Chan, & Woolf, 2004), and is defined by Bekker and colleagues (1999, p.1) as a decision “where a reasoned choice is made by a reasonable individual, using relevant information about the advantages and disadvantages of all the possible courses of action, in accord with the individual’s beliefs.” There are currently debates about the definition of informed decision,
indicating that apart from reason-based strategies, non-rational and emotional
approaches of information processing can also result in high quality decisions (Peters et al., 2008), especially among older adults (Eliott, & Olver, 2005).
In order to assist informed decision making for both older and younger cancer patients, patient decision support tools have been developed. Decision support tools aim to provide information about the advantages and disadvantages of treatment options, as well as to help the patients to clarify their values for the potential
Yumin Lin, 11354135
people in hospital decision making are developed recently. Although decision support tools are considered effective approaches to enhance informed decision making, many of them are not adapted to the specific needs and abilities of older people (van Weert et al., 2016). Most of the decision support tools show evidence-based information on the benefits and harms of different options, often accompanied by numerical
estimations (Trevena et al., 2013), which are especially difficult for older adults to process, and might lead to miscomprehension (Mata, Schooler, & Rieskamp, 2007). It has thus become vital to take older adults’ abilities and needs into account in the question of how to appropriately present decision-relevant information to them. Therefore, it is necessary to develop and test decision support tools which are tailored to older adults’ skills and needs (van Weert et al., 2016). The present study aims to test the effectiveness of two strategies to present information in decision support tools, which might support the comprehension of core information and informed decision making, especially among older patients.
Firstly, the framing of the risk information in decision support tools could strongly influence how older individuals perceive and react to the message, and thus affect
Yumin Lin, 11354135
decision making (Rothman, Bartels, Wlaschin, & Salovey, 2006). Negatively-framed information encourages people to accept potential risks (Kahneman, & Tversky, 2013), while positively-framed information makes people become risk-averse (Kahneman, & Tversky, 2013), and provides emotional support, especially to older patients (Gurm & Litaker, 2000; McNeil, Pauker, & Tversky, 1988). In studies concerning decision support tools, information can be either presented as the chance of a risk occurring, or the chance of a risk not occurring. This is a way of how information can be framed
differently. A number of studies indicate advantages of the positive frame in medical
decision making by positively influencing emotions (Gurm & Litaker, 2000; McNeil,
Pauker, & Tversky, 1988), which might lead to higher willingness to in-depth process the provided risk information, and in turns result in better informed decision making (Veldwijk et al., 2016). These findings further support the advantage of providing positive-framed risk information to older adults, based on the phenomenon that emotionally meaningful information has more effects on them (Santrock, 2008). However, other studies suggest that framing has no consistent influence on decision
Yumin Lin, 11354135
making in clinical settings (Edwards, Elwyn, Covey, Matthews, & Pill, 2001; Levin, Schneider, & Gaeth, 1998).
Additionally, decision support tool developers often include visual illustrations to enhance the comprehension of the numerical information (Elwyn et al. 2006). While there is evidence that adding visual assistance of numeric information is beneficial for understanding medical information among young adults, the effects for older adults remain inconsistent (Levie, & Lentz, 1982; Liu, Kemper, & McDowd, 2009; Morrow, Hier, Menard, & Leirer, 1998; Sojourner, & Wogalter, 1998). Several studies in decision psychology and risk communication have demonstrated the effectiveness of combining written texts with visual illustrations in meaningful learning (Mayer, 2002; Mayer, & Moreno, 2003). In particular, so-called “cognitive” visual illustrations illustrating the content of the text (e.g. graphs such as icon arrays, bar and line charts. From now on summarized as visual illustration) increase the understanding of the risk information (Brotherstone, Miles, Robb, Atkin, & Wardle, 2006) by embedding meaning to abstract numerical information (Visschers et al., 2012), which is positively related to information processing and comprehension, which in turns increase
Yumin Lin, 11354135
informed decision making. Also, visual illustrations in risk communication can potentially reduce undesired framing effects (Garcia-Retamero, & Galesic, 2010), for example framing bias (Hodgkinson, Maule, Bown, Pearman, & Glaister, 2002). Until now, to the best of our knowledge, no studies have investigated how framing and visual illustrations in decision support tools synchronously affect comprehension of core information and informed decision making among older cancer patients. Since older patients are experiencing changes in information processing and abilities, it is important to test strategies that can potentially support their decision making, and develop decision support tools tailored for them. Also, the main effects of framing and visual illustrations remain both theoretically and empirically inconsistent. The
research question in the present study is: What are the effects of using positive
framing (compared to negative framing) and visual illustrations (compared to text-only) in decision support tools on comprehension of core information, and informed decision making? Do these effects differ between older (>=65) and younger (<65) adults? The age cut-off of 65 years was decided because it is commonly used in
Yumin Lin, 11354135
clinical trials, as well as the people’s self-identification of older age (Staudinger, & Bluck, 2001).
Since older adults’ information processing and decision making in patient decision support tools are barely investigated in previous studies, the outcomes of the present study will make contributions to the relevant literature. In terms of practical
instructions, health professionals and developers of patient decision support tools can use the knowledge gained from the present study as guidelines to provide high-quality information, especially for older patients.
Theoretical Framework
Cancer Patients’ Medical Decisions and Necessary Information
Patients have to make various types of decisions in a clinical setting. In case of a cancer diagnosis, patients need to choose between several treatment options, but also between hospital where they can receive this treatment. In the present study, I have decided to focus on both the decision for the preferred treatment and for the preferred hospital.
Yumin Lin, 11354135
Cancer patients often find making treatment decisions challenging, because they are unsure about the effectiveness and the balance of benefits and risks of the
treatment options, and their potential consequences (Reyna, Nelson, Han, & Pignone, 2015). Patient Decision Aids (PDAs) systematically present the benefits and harms of the treatment options in a balanced manner, so that the patients can make informed and evidence-based decisions that are in line with their personal values and
preferences (Stacey et al., 2017). PDAs are designed as evidence-based tools to support patients in making medical decisions among available options, as defined by the International Patient Decision Aids Standards (IPDAS) Collaboration (Elwyn et al. 2006).
The understanding of patients’ hospital choice has for long been limited to
traditional economic theories and methodology (Fischer, Pelka, & Riedl, 2015), where the location of the hospital plays a significant role (Maynard, & Beckingham, 2016; Trinh et al., 2013; Wouters et al., 2009). However, more and more patients decide to bypass their nearest cancer care centers and search for a cancer care provider based on indicators such as waiting time, advancement of technology, and specific advantages
Yumin Lin, 11354135
compared to other providers (Aggarwal et al., 2017; Aggarwal et al., 2018). Since many patients are willing to travel beyond their nearest cancer care provider, it has become increasingly important for the patients, especially in western countries, to seek for comparative performance information (CPI) in decision support tools, which offers the patients opportunities to make explicit comparisons between the services of health care providers, in order to make an informed decision about which provider best suits them (Damman, van den Hengel, van Loon, & Rademakers, 2010; Hussey, Luft, & McNamara, 2014). CPI in decision support tools (from now on summarized as CPI) is commonly applied in western countries to encourage patients to participate in decision making, as well as to enhance competition-based reforms among health care providers (Ketelaar, 2015).
The present study focuses on two types of decision support tools for cancer patients: PDAs for making a treatment decision and CPI for making a hospital decision. A number of studies have demonstrated that using decision support tools positively influences patients’ knowledge, feeling informed, the accuracy of risk perception, clarity of individual values, and willingness to participate in SDM, compared to usual
Yumin Lin, 11354135
care (Stacey et al., 2017; van Weert et al., 2016). Consequently, there is evidence to support the advantage of decision support tools to enhance informed decision making among cancer patients (Aning, Wassersug, & Goldenberg, 2012; Hersch et al., 2015; O’Connor et al., 2007; O’Connor, & Edwards, 2009).
Ageing, Information Processing and Decision Making
In the past decades, it has become generally accepted in the field of human judgement and decision making that people use two modes of thinking while processing information: the intuitive mode and the deliberative mode (Kahneman, 2003). The intuitive mode works when individuals process the information implicitly, associatively, automatically and relatively fast, while the deliberative mode does the opposite when individuals apply a more systematic approach, and think analytically, reason-based, consciously and relatively slow (Kahneman, 2003). The intuitive mode of thinking demands relatively low levels of cognition, and remains constantly active; but if sufficient cognitive capacity is available or accessible, the deliberative mode of thinking is activated, and operates in parallel (Smith, & DeCoster, 2000; Strack, &
Yumin Lin, 11354135
Deutsch, 2004). The parallel operation of both the intuitive and deliberative modes of thinking is essential in decision making, because high quality decisions are expected when individual needs are meet both emotionally and cognitively (Peters, Diefenbach, Hess, & Västfjäll, 2008). This indicates that separate use of intuitive or deliberative mode of thinking can result in poor decision making. For example, there are situations in which the use of the intuitive mode alone can result in poor decisions, for example, when the deliberative mode fails to detect or correct the errors caused by the intuitive mode of thinking (Gilbert, 1989; Kahneman, & Frederick, 2002; Stanovich & West, 2000). Also, there is evidence that the intuitive mode of thinking can outperform the deliberative mode of thinking (Kruglanski, & Gigerenzer, 2011; Usher, Russo, Weyers, Brauner, & Zakay, 2011). The operation of both modes of thinking requires the ability and motivation to process information (Petty & Cacioppo, 1986), which change significantly with ageing.
Previous studies demonstrate that age is negatively associated with the ability to apply the deliberative mode of thinking, especially in unfamiliar situations. This is supported by the fact that older adults possess poorer working memory capacity
Yumin Lin, 11354135
(Bopp, & Verhaeghen, 2005), and encounter difficulties to quickly process the information (Salthouse, 1992), exclude irrelevant information (Hasher, & Zacks, 1988), recall the information (Kausler, 1994), and control the regulation of cognition (Chen, 2002). However, previous studies suggest that the intuitive mode of thinking where affect plays a significant role in many cases, seem to remain largely active with ageing (Peters, Diefenbach, Hess, & Västfjäll, 2008; Strough, Bruin, & Peters, 2015). This could be potentially explained by accumulated knowledge and experience related to the aspects to be weighed, so that older adults rely more on the intuitive mode of thinking (Reyna, 2004).
As a result of the more frequent use of the intuitive mode of thinking compared to the deliberative mode among older adults, affect plays a more vital role as the source of information (Peters et al., 2007; Peters et al., 2008). Additionally, with the end of life approaching, older adults focus more on emotional meaningful goals because of the changes in time perspective (Carstensen, 2006), and the importance of short-term emotional well-being increases (Charles, Mather, & Carstensen, 2003). Therefore, older patients tend to pay more attention to, and invest their cognitive resources more
Yumin Lin, 11354135
in information relevant to their emotional goals (Carstensen, 2006; Peters et al., 2008). However, encountering life-threatening diseases like cancer creates strong negative emotions, and thus reduces the positivity focus (Peters et al., 2008). Evidence is provided that even emotions unrelated to the decision problem can affect information processing (Isen, 1997), and consequently influence informed decision making. Generally speaking, it is more challenging for older cancer patients to deliberatively process and comprehend information, especially factual and statistical information, compared to younger patients. This could result in miscomprehensions and the application of more intuitive, heuristic and less reason-based strategies (Mata et al., 2007). An important challenge in patient decision support tools is how to reduce the cognitive effort necessary to process and understand information for older patients, as well as to encourage them elaborate on the information.
Strategies in Decision Support Tools for Older Cancer Patients
Older cancer patients could potentially benefit from several strategies to present information in decision support tools. The present study investigates two possible
Yumin Lin, 11354135
strategies which might positively affect information processing and decision making among older patients: The use of 1) framing and 2) visual illustrations.
Framing Effects in Decision Support Tools
Presenting probabilities of potential consequences of health care options is an essential element in decision support tools (Elwyn et al. 2006), the same consequence could be framed in different manners by presenting the probability of positive outcomes, or presenting the probability of negative outcomes. In case of cancer incidences, for example, the potential consequences could be framed positively as the chance of survival or being free of side-effects over a certain period of time, or negatively the chance of death or experiencing side-effects over a certain period of time. The way in which potential consequence are framed, especially risky outcomes, can affect various aspects in information processing and decision making. According to prospect theory, when thinking about potential gains, individuals tend to become risk-averse and prefer certainty; but when thinking about potential losses, individuals tend to become risk seeking and prefer uncertainty in order to avoid losses
Yumin Lin, 11354135
(Kahneman, & Tversky, 2013).
The effect of framing in PDAs for treatment decisions can be influenced by a number of factors, including the risk of the treatment and the data format (Gong et al., 2013). Waters and colleagues (2009) suggest positive framing as a more popular strategy in presenting online cancer-related information. Several studies have also demonstrated the general advantage of a positive frame over a negative frame in PDAs aim to assist treatment decisions. Armstrong and colleagues (2002) found that in a hypothetical treatment choice, being exposed to positive frames increases patients’ comprehension of risks and key information. Veldwijk et al. (2016) concluded that individuals make strategical and informed treatment decisions more often based on positive frames, compared to negative frames. A study by Carling and colleagues (2010) demonstrated that positive frames led to higher consistency between a fully informed decision and the decisions participants made. These benefits of positive frames can potentially be explained by the fact that positive frames prime the patients for positive feelings for the information. This might increase the willingness to process the provided information, while negative frames lead the feelings to an
Yumin Lin, 11354135
opposite direction (Gurm & Litaker, 2000; McNeil, Pauker, & Tversky, 1988). To
illustrate, Um and colleagues (2007) suggest that positive emotions facilitate learning, and thus support knowledge construction. Emotion, especially emotions that are felt at the time of decision making, plays a significant role in decision making (Lerner, Li, Valdesolo, & Kassam, 2015). According to socio-emotional selectivity theory, emotional goals become increasingly important among older patients (Carstensen, 2006), and they gradually focus more on feeling good at the moment (Charles et al., 2003). Therefore, it is assumed that positive frames in PDAs that create more positive feelings, and thus have stronger effects on increasing older patients’ comprehension of core information and informed decision making than negative frames. Therefore, the following hypothesis is formulated:
H1: Being exposed to a positive framed message in PDAs for treatment decisions, compared to a negative framed message, lead to more comprehension of core information (H1a) and informed decision making (H1b), and these effects are stronger for older cancer patients (>=65), compared to younger cancer patients (<65) (H1c).
Yumin Lin, 11354135
However, a number of studies also suggested the benefits of using a negative frame, especially in CPI aiming to assist care provider choice. Hibbard and colleagues (2000) investigated how framing affected the decisions on health care providers. The results indicate that negative framing concerning potential risks or losses significantly outperformed the positive framing featuring potential benefits or gains, for increasing the comprehension of core information, and the weight of the information in CPI in decision making. The advantage of negative framing might be explained by the fact that preferential reactions to positive information might result in positivity bias, which negatively affects informed decision making. And negative frames are expected to correct these positivity biases. Since older people are more sensitive to positive information, the challenge to correct positivity biases might increase with ageing (Carstensen, 2006). Additionally, the proper application of weighing comparisons in decision making requires a data-driven strategy, which is less likely to be used by older patients (Woodhead, Lynch, & Edelstein, 2011). This can be explained by the fact that their accumulated knowledge and experience lead to a higher likelihood to apply experience-driven strategy (Woodhead et al., 2011), which in terns result in
Yumin Lin, 11354135
more frequently use of intuitive mode of thinking when people are ageing (Reyna, 2004). Therefore, the hypothesis below is formulated to make a clear distinction between the assumptions for CPI concerning hospital decisions:
H2: Being exposed to a negative framed message in CPI for hospital choice, compared to a positive framed message, lead to more comprehension of core information (H2a) and informed decision making (H2b), and these effects are weaker for older cancer patients (>=65), compared to younger cancer patients (>65) (H2c).
Visual Illustrations in Decision Support Tools
Another strategy which can potentially assist information processing and decision making among older cancer patients is to add visual illustrations in patient decision support tools. Visual illustrations are “simple graphical representations of numerical expressions of probability and include icon arrays, bar and line charts, and others” (Garcia-Retamero, & Cokely, 2017). According to the multimedia principle of the cognitive theory of multimedia learning (CTML), the combination of written text and visual illustrations is more effective in meaningful learning, compared to information
Yumin Lin, 11354135
in single format (Mayer, 2002). The assumption is based on the dual coding approach, which indicate that individuals possess different systems to process verbal and visual information (Paivio, 2013). Because verbal and visual information are coded separately in memory, the combination of both lead to a better retrieval of information from memory (Paivio, 2013). Sprague and colleagues (2012) indicate positive effects of text-plus-graph information on correct comprehension of risks, compared to text-only information. This is especially relevant in patient decision support tools in cancer incidences, where patients have to process a lot of numerical information (Trevena et al., 2013), and their evaluations of risk information might be biased, because they find it hard to understand numerical information. Additionally, visual illustrations provide a more concrete representation of the information, which corrects the biases caused by only verbal presentation of the numerical information (Gigerenzer, Gaissmaier, Kurz-Milcke, Schwartz, & Woloshin, 2007). Also, a number of studies suggest that visual illustrations facilitate information processing (Garcia-Retamero, & Galesic, 2010; Garcia-Retamero, & Cokely, 2011; Kurz-Milcke, Gigerenzer, & Martignon, 2008).
Yumin Lin, 11354135
Making an informed decision in the context of risk requires a high level of accuracy in the comprehension and evaluation of risk information, which may result in accurate decisions (Garcia-Retamero, & Cokely, 2017). In the field of patient decision support tools, previous studies indicate that adding visual illustrations to textual information lead to increased knowledge and comprehension for the relevant information (Gaissmaier et al., 2012; Kakkilaya et al., 2011; Trevena et al., 2013), and assist with the evaluations of complex concepts (Zikmund-Fisher et al., 2008). Due to the fact that ageing results in limited working memory capacity (Bopp, & Verhaeghen, 2005), it is expected that older patients will benefit more from visual illustrations. An additional argument to support this assumption is that visual illustrations of numerical information increase patients’ satisfaction with the information provided by making the information comprehension and processing easier, and thus provide emotional support especially for older patients (Bol et al., 2014; Bol et al., 2015). Therefore, along with the previous findings, the following hypothesis is proposed:
Yumin Lin, 11354135
tools (both PDAs and CPI), compared to text-only information, lead to more comprehension of core information (H3a) and informed decision making (H3b), and these effects are stronger for older cancer patients (>=65), compared to younger cancer patients (>65) (H3c).
Another factor that might influence the effectiveness of visual illustration in patient decision support tools is health literacy, with health numeracy being an important element indicating the relevant numerical skills (Garcia-Retamero, & Cokely, 2017). Health literacy refers to “the degree to which individuals can obtain, process, and understand basic information and services they need to make appropriate health decisions,” in which health numeracy demonstrates the skills to manipulate and interpret quantitative information (Hernandez, 2013; Paulos, 1988). A number of studies have presented that individuals with lower health literacy encounter difficulties in processing and interpreting the information in decision support tools (Damman, Hendriks, Rademakers, Spreeuwenberg, Delnoij, & Groenewegen, 2012; Hibbard, Peters, Dixon, & Tusler, 2007; Zwijnenberg et al., 2012). However, when individuals with lower health literacy are provided with the option to take part in
Yumin Lin, 11354135
decision making, they will not refuse to participate (Mancini, Jansen, Julian‐Reynier, Bechlian, Vey, & Chabannon, 2014).
Although a negative association between age and health literacy has been found in some studies (Ashida et al., 2011; Baker et al., 2000; Von Wagner et al., 2007), several studies find no association between age and health literacy (Chesser, Keene Woods, Smothers, & Rogers, 2016; Ganzer, Insel, & Ritter, 2012; McDougall Jr, Mackert, & Becker, 2012). Therefore, it can be insightful to investigate the role of health literacy and numeracy in the effectiveness of visual illustrations separately. Patients with low health literacy or numeracy are expected to benefit more from the additional visual illustrations, compared to patients with high literacy and numeracy (Meppelink, Smit, Buurman, & van Weert, 2015). Therefore, an additional hypothesis is formulated as:
H4: The positive effects of adding visual illustrations to cancer-related risk information in decision support tools (both PDAs and CPI) concerning both treatment and hospital choices, are stronger for patients with low health literacy and numeracy.
Yumin Lin, 11354135
Interaction Effects of Framing, Visual Illustration, and Age
The main effects of framing and visual illustrations on patients’ information processing and decision making have been demonstrated in previous studies. However, the interactions effect between framing and visual illustrations remain inconsistent in previous studies: Armstrong and colleagues (2002) suggest that participants exposed to positively-framed visual illustrations have higher comprehension of provided information, compared to those exposed to negatively-framed illustration; in contrast, Zikmund-Fisher and colleagues (2007) report the advantages of negative frames in supporting participants to understand provided graphic information in details, compared to positive frames.
Additionally, a problem regarding the use of either positive or negative frame is that individuals might make decisions only based on how the information is presented, without carefully process and comprehend the information. This phenomenon refers to framing bias (Hodgkinson et al., 2002). Previous studies indicate that adding visual illustrations increases deliberative information processing, and helps to reduce framing bias, resulted from a positive or a negative frame (Gamliel, & Kreiner, 2013;
Yumin Lin, 11354135
Garcia-Retamero, & Galesic, 2010; Garcia-Retamero, & Cokely, 2011). Therefore, it is expected that a proper combination of framing and visual illustrations will be the most effective approach. As stated before, emotions play an essential role in framing effect. While older patients are more sensitive about emotions (Carstensen, 2006), because visual illustrations make patients more satisfied with the information provided by making understanding and processing information easier, and thus provide emotional support especially for older patients (Bol et al., 2014; Bol et al., 2015). It is also assumed that the combination of appropriate framing and visual illustrations is particularly effective among older patients. The following hypotheses are formulated based on the above-mentioned assumptions:
H5: Positively-framed information with visual illustrations in PDAs for treatment decisions lead to a higher comprehension of core information (H5a) and informed decision making (H5b) compared to positively-framed text-only information, negatively-framed information with visual, and negatively-framed text-only information. These effects are weaker for older cancer patients (>=65), compared to younger cancer patients (>65) (H5c).
Yumin Lin, 11354135
H6: Negatively-framed information with visual illustrations in in CPI for hospital decisions lead to a higher comprehension of core information (H6a) and informed decision making (H6b) compared to positively-framed information with visual, positively-framed text-only information, and negatively-framed text-only information., These effects are weaker for older cancer patients (>=65), compared to younger cancer patients (>65) (H6c).
Study 1 (Treatment Decisions)
Method
Design and Participants
An online experiment was conducted for Study 1 concerning treatment choices for lung cancer. A 2 (framing: positive vs. negative) x 2 (visual illustrations: present vs. absent) x 2 (age: younger cancer patients [< 65 years old] vs. older cancer patients [>= 65 years old]) between-subject factorial design was adopted to test the hypotheses.
Yumin Lin, 11354135
online panel named Dynata. Participants were recruited by Dynata by sending out invitations embedded with an URL link to participate in the study. The primary inclusion criteria of the present study were that all participants needed to be analogue patients with a personal history of being diagnosed with cancer. This is a validated method according to a meta-analysis by Van Vliet et. al. (2012). Since the questionnaire was developed in English, invitations were only sent to native English speakers, i.e. analogue patients from the database in the USA and the UK of Dynata. If an analogue patient was willing to participate, the participant was randomly assigned to one of the four conditions. Participants who had been diagnosed with lung cancer were excluded from the study, since they could potentially have better comprehension and pre-existing knowledge on the information provided in the experimental study, which might result in biases negatively affecting the internal validity of the study.
Power analysis with alpha set as .05 and probability level set as .80 indicated that 270 participants were needed in order to detect effects in medium size (.25). The calculation was based on the key dependent variable in the study, informed choice
Yumin Lin, 11354135
making.A total of 275 analogue cancer patients participated in the experiment among which 150 participants were younger than 65 years old (M= 47.71, SD= 11.16, range 18-64, 46% male), and 125 participants were older than 65 years old (M= 72.45, SD= 5.91, range 65-94, 60% male). Approximately 56% of the participants had tertiary education, and around 22% of the participants were currently receiving cancer treatments (Table 4, See APPENDIX F).
Procedure
After giving informed consent, participants were asked to answer several questions regarding demographics and cancer history. This was designed to double-check whether they met the inclusion criteria of the present study, as well as the representativeness of the sample. In case of meeting the inclusion criteria, the participants were randomly assigned to one of the following conditions: in the first condition, participants were exposed to positive framed textual information in a table about two treatment options for lung cancer (surgery and stereotactic radiotherapy), accompanied by visual illustrations; in the second condition, the same textual
Yumin Lin, 11354135
information as in the first condition was presented, but the visual illustrations were excluded; in the third condition, participants read negative textual framed information in a table about the treatment options accompanied by visual illustrations; in the fourth condition, the same textual information were provided as in the third condition, but the visual illustrations were excluded.
The participants were provided with a scenario. They were asked to imagine that they had been diagnosed with early-stage lung cancer, and their physician provided them with web-based information about two available treatment options, namely surgery, and stereotactic radiotherapy. The participants were instructed to read the provided information carefully, because they had to make a choice between the two treatment options based on the information. After reading the information, participants were asked to complete a questionnaire regarding their comprehensions of the information, attitudes towards the treatment options, their choice, their experiences reading the information, as well as personal characteristics including health literacy and numeracy.
Yumin Lin, 11354135
Stimuli Development
The stimuli applied in Study 1 (Figure 1-4, See Appendix A) were partially adapted from an existing Dutch online lung cancer PDA from a website named Keuzehulp Longkanker (http://www.keuzehulp-longkanker.nl/). The existing PDA provided comparable information about two treatment options for lung cancer (surgery and stereotactic radiotherapy), covering aspects including the chance of survival, experiencing recurrence of cancer tissue, and experiencing side-effects after receiving both treatments. Due to the fact that the numerical information regarding side-effects was not provided for stereotactic radiotherapy in the PDA by Keuzehulp Longkanker, additional numerical information was included based on two studies investigating side-effects after receiving stereotactic radiotherapy (Lo et al., 2013; Moiseenko et al., 2018).
Equivalent numerical textual information was presented in tables for all conditions, but was framed in different ways. In the positive framing conditions, the chances of experiencing positive outcomes (e.g. survival) or avoiding risks (e.g. not experiencing recurrence of cancer, not experiencing side-effects) were shown to the participants. In
Yumin Lin, 11354135
contrast, in the negative framing conditions, the chances of experiencing negative outcomes (e.g. mortality) and risks (e.g. experiencing recurrence of cancer, experiencing side-effects) were presented,.
As for the manipulation of the presence of visual illustrations, bar graphs were added in the conditions with visual illustrations for every single aspect presented in the textual information, which represented the comparisons between the two treatment options. In the conditions without visual illustrations, the bar graphs were absent. The choice of the visual presentation format was based on previous studies demonstrating the surplus in bar graphs, when the viewers were required to compare different risks (Feldman-Stewart, Kocovski, McConnell, Brundage, & Mackillop, 2000; Schapira, Nattinger, & McHorney, 2001; Schapira, Nattinger, & McAuliffe, 2006), including a study among older adults (van Weert, Alblas, Van Dijk, & Jansen, in revision).
Measures
Comprehension of Core Information. Comprehension of core information was
measured in two sub-dimensions, verbatim knowledge and gist knowledge. The participants were provided with the information they read depending on the
Yumin Lin, 11354135
experimental conditions. A pilot test among 22 participants was conducted before the study to evaluate different ways to ask questions. The aim of the pilot test was to avoid questions which were either too simple or too difficult. The questions used in the present study to measure comprehension of core information were selected and adjusted based on the results from the pilot test (Table 2, See APPENDIX E), taking the balance of questions concerning different treatment options and framed in different ways into consideration, so that the amount of positively-framed questions was the same as the amount of negatively-framed questions.
Verbatim knowledge was defined in the present study as the ability to correctly understand the numbers presented in the PDA (Hamstra et al., 2015). Four open questions were used to measure verbatim knowledge. An example of the questions was “A hundred lung cancer patients received surgery, how many of them is estimated to die within 30 days?” Participants were required to insert a number as their answer in the box below each question. Due to the fact that the participants could refer to the information while answering the question, only answers that exactly matched the provided information were defined as correct answers. At the end, one point was
Yumin Lin, 11354135
given to each correct answer, resulting in a scale ranged from 0 to 4, α= .87, M= 2.50,
SD= 1.63.
Gist knowledge was defined in the present study as the ability to understand the general meaning and purpose of the provided information, and identify the key points in the information (Hawley, Zikmund-Fisher, Ubel, Jancovic, Lucas, & Fagerlin, 2008). The question measuring gist knowledge in the present study mainly concerned participants’ ability to make comparisons on various aspects between the treatment options. Ten multiple choice questions were included in the questionnaire to measure gist knowledge. Three types of questions were formulated in the measurement, for example 1) “In case of surgery and stereotactic radiotherapy, the chance of dying within 30 days is exactly the same.” with the answer options “Correct”, “Incorrect” and “I do not know”; 2) “Dirk finds the long-term survival rate the most important element for making treatment decisions. Which treatment should he choose?” with answer options “Stereotactic radiotherapy”, “Surgery”, “No preferred option”, and “I do not know”; or 3) “Ken received surgery, and Sandy received stereotactic radiotherapy. Who had a higher chance to experience nausea?” with answer options
Yumin Lin, 11354135
“Ken”, “Sandy”, “They had the same chance”, and “I do not know”. Only one answer option was correct for each question, and one point was given to every correct answer, resulting in a scale ranged from 0 to 10, α= .89, M= 6.65, SD= 3.29.
Finally, the scores of both verbatim and gist knowledge were combined as the indicator of comprehension of core information ranging from 0 to 14, α= .92, M= 9.15,
SD= 4.63. The weight of the two sub-dimensions (4:10) reflected the main goal for
PDAs, which was to assist patients make comparisons that resulted in a decision.
Informed Decision Making Informed decision making was measured by adapting
the multi-dimensional measure of informed choice (MMIC), consisting of knowledge, attitudes, and choices (Marteau, Dormandy, & Michie, 2001; Michie, Dormandy, & Marteau, 2002). A decision is considered informed when a person has sufficient knowledge of the subject, and the attitude is consistent with the choice (Marteau et al., 2001)
Comprehension of core information was used as the indicator for knowledge. Participant were defined as having sufficient knowledge when they had both sufficient verbatim and gist knowledge. Sufficient knowledge was defined as a
Yumin Lin, 11354135
knowledge score above the median, which was 3.00 for verbatim knowledge, and 8.00 for gist knowledge.
Attitude was measured with a six-item scale for both surgery (α= .86, M= 4.33,
SD= 1.22), and stereotactic radiotherapy (α= .87, M= 4.50, SD= 1.20), adapted from
the study by Marteau et al. (2001). Participants were asked to indicated their opinions on seven-point semantic scales for each treatment option, with the response categories ranging from, for example, 1 (difficult) to 7 (simple), or from 1 (a bad idea) to 7 (a good idea). Choice was measured through a question asking which treatment option for lung cancer the participants would select: surgery or stereotactic radiotherapy. The consistency of attitude and choice was in previous research defined in two ways (Stacey et al., 2017): (1) when the participant had a more (or at least not less) favorable attitude towards the selected option, compared to the unselected one; or (2) when the participant had a positive attitude towards the selected option, where positive attitude was defined as an attitude score above the mid-point of the scale (4.00). The informed decision based on the attitude-choice congruence defined in (1) was summarized from now on as Informed Decision Making Type I (IDM-1), and
Yumin Lin, 11354135
based on the attitude-choice congruence defined in (2) as Informed Decision Making Type II (IDM-2).
Additionally, van den Berg and colleagues (2006) argued that it is necessary to add the evaluation of decision-making process to the measurement of informed decision making, especially for decisions involved high risks. They proposed measuring deliberation as a sub-dimension in informed decision making, so that a decision was informed, when the criteria of sufficient knowledge, consistent attitude and choice, as well as perceiving the decision as deliberated were fulfilled. Deliberation was measured by six items such as “I imagined how I would feel if I didn’t have to undergo either treatment.” with the answer categories ranging from 1 (Totally disagree) to 7 (Totally agree), resulting in a Cronbach’s α= .78, M= 5.43, SD= .90. A deliberated decision was defined as a deliberation score above 5.50, which was the median of the scale (van den Berg et al., 2006). Consequently, Deliberated Informed Decision Making Type I (D-IDM-1) and deliberated Informed Decision Making Type II were constructed (D-IDM-2).
Yumin Lin, 11354135
questions from the eHealth literacy (eHEALS. Norman, & Skinner, 2006) and two questions Short Test of Functional Health Literacy in Adults (STOHFLA. Chew, Bradley, & Boyko, 2004). Since the experiment was conducted online, combing the two scales had the potential to measure participants’ knowledge and skills not only in understanding, but also in searching relevant information online. The answer categories ranged from 1 (Totally disagree) to 7 (Totally agree). A factor analysis using principal components analysis and varimax rotation was conducted with the five items. One item from STOHFLA (“I need to have someone help me read hospital materials.”) had a factor loading under .10 (.06), and was excluded from the factor. Finally, the four remaining items (e.g. “I know how to use the health information I find on the Internet to help me.”) constructed a factor which explained 54.99% of the variance (Eigenvalue= 2.75). A score of health literacy was calculated as the mean of the four items, α= .84, M= 5.16, SD= 1.12. A participant was categorized as having high health literacy when the health literacy score was above the median (Mdn= 5.25).
Yumin Lin, 11354135
Schulz, Ghazal, & Garcia-Retamero, 2012). Four questions were asked to assess relevant numerical skills. Participants were asked to insert an answer to the questions. Considering the fact that two of the four questions asked the participants to fill in the answer in percentages, an answer in decimals or without a percentage sign (e.g. .25 or 25 for 25%) was defined as correct. One point was given for each correct answer, resulting in a scale ranging from 0 to 4, α= .60, M= .71, SD= 1.01. Deleting any items from the numeracy scale would not increase the Cronbach’s alpha. This could be potentially explained by the fact that more than half (56.73%, Mdn= .00) of the participants gave a wrong answer to each of the numeracy questions. A participant was categorized as having high numeracy when at least one question was answered correctly.
The participants were categorized in three groups considering both health literacy and numeracy: having both high health literacy and numeracy (high); having either high health literacy or numeracy (moderate); or having both low health literacy and numeracy (low).
Yumin Lin, 11354135
Feeling Informed Feeling informed was measured by the informed subscale
with three items from the Decisional Conflict Scale (O'Connor, 1995), such as “I know which options are available for me.” The answer categories ranged from 1 (Totally disagree) to 7 (Totally agree). A scale of feeling informed was created as the mean of the three items, α= .80, M= 5.71 1, SD= .87.
Background Characteristics Participants were asked to indicated their age,
gender, educational level, whether they were currently receiving cancer treatment, experience with surgery and stereotactic radiotherapy.
Manipulation Check To check whether the manipulation of framing was
successful, three items were added to the questionnaire. Two question addressed how much the provided information was perceived as positive or negative (e.g. “The information in the decision aid focused on the potential positive consequences for both treatment options.”), and was measured on Likert scales ranging from 1 (Totally disagree) to 7 (Totally agree). A back-up question measured on a 7-point Likert scale how balanced the participants though the provided information was.
Yumin Lin, 11354135
Statistical Analysis
First, Chi-square tests and Analyses of Variance (ANOVA) were performed to check whether the control variables were equally distributed among the conditions. Second, independent sample t-tests were conducted to examine whether the framing of the provided information was perceived by the participants correctly. Finally, Multivariate Analysis of Variance (MANOVA) was conducted with comprehension of core information and all types of informed decision making as dependent variables, to test H1 (IV: framing; moderator: age group), H3 (IV: visual illustration; moderator: age group, H4 (IV: visual illustration; moderator: health literacy and numeracy) and H5 (IV: framing; moderator: visual illustration & age group) respectively, with experience with surgery as covariate. In case a significant interaction effect was found between the independent variable and the moderator, simple effects analyses were performed to investigate how the effects were different for different categories of the moderator.
Yumin Lin, 11354135
Results Study 1
Randomization Check
Results from Chi-square test demonstrated that participants were equally distributed among the four conditions regarding age group, χ2(3, N= 273)= 4.52,
p= .211; gender, χ2(3, N= 273)= 2.69, p= .441; educational level, χ2(9, N= 273)= 6.95, p= .642; currently receiving cancer treatment, χ2(3, N= 273)= 2.15, p= .543; and
experience with stereotactic radiotherapy, χ2(3, N= 273)= 3.34, p= .343. Additionally, results from ANOVA showed that feeling informed did not differ between conditions,
F (3, 254)= 1.10, p= .351. However, there was a significant association between the
conditions and experience with surgery, χ2(3, N= 273)= 14.24, p= .003. Therefore, experience with surgery was included as covariate in all analyses.
Manipulation Check
Results from t-test showed that the positively-framed conditions (M= 5.76, SD= .96) were perceived as more positive than the negatively-framed conditions (M= 4.78,
Yumin Lin, 11354135
4.89, SD= 1.48) were perceived as more negative than the positively-framed conditions (M= 4.35, SD= 1.63), t (258)= 2.81, p= .005. Finally, the negatively-framed conditions (M= 2.70, SD= .74) were perceived as less clearly slanted to the positive outcomes than the positively-framed conditions (M= 3.14,
SD= .92), t (241)= -4.10, p< .001. Therefore, the manipulation of framing was
successful.
Hypothesis Testing
Main Effect Framing (H1a & H1b)
Results from MANOVA showed that there was no significant main effects of framing on comprehension of core information (from now on summarized as comprehension), F (1, 268)= .01, p= .945; Informed Decision Making Type I (IDM-1),
F (1, 268)= .72, p= .397; Informed Decision Making Type II (IDM-2), F (1, 268)= .04, p= .842; and Deliberated Informed Decision Making Type II (D-IDM-2), F (1, 268)=
1.46, p= .229. However, framing had a marginal main effect on Deliberated Informed Decision Making Type I (D-IDM-1), F (1, 268)= 2.84, p= .093. There was a
Yumin Lin, 11354135
marginally higher proportion of D-IDM-1 among participants exposed to negatively-framed information (M= .16, SD= .37), compared to those exposed to positively-framed information (M= .07, SD= .26). H1a and H1b were not supported.
Interaction Effect Framing and Age (H1c)
MANOVA demonstrated that the interaction between framing and age group had no significant effects on comprehension, F (1, 268)= .84, p= .361; IDM-1, F (1, 268)= .10, p= .756; IDM-2, F (1, 268)= .28, p= .600; D-IDM-1, F (1, 268)= .06,
p= .807; and D-IDM-2, F (1, 268)= .05, p= .820. H1c was rejected.
Main Effect Visual Illustration (H3a & H3b)
MANOVA showed that adding visual illustrations had no significant main effects on comprehension, F (1, 268)= .35, p= .557; IDM-1, F (1, 268)= .01, p= .946; IDM-2,
F (1, 268)= .01, p= .918; D-IDM-1, F (1, 268)= .76, p= .384; and D-IDM-2, F (1,
Yumin Lin, 11354135
Interaction Effect Visual Illustration and Age (H3c)
Results indicated that the interaction between visual illustration and age group did not significantly affect comprehension, F (1, 268)= 2.03, p= .155; IDM-1, F (1, 268)= .12, p= .735; IDM-2, F (1, 268)= .82, p= .365; D-IDM-1, F (1, 268)= .36,
p= .547; and D-IDM-2, F (1, 268)= .63, p= .428. H3c was rejected in treatment
decision making.
Interaction Effect Visual Illustration and Health Literacy & Numeracy (H4)
MANOVA showed that the interaction between visual illustration and level of health literacy and numeracy had no significant effect on comprehension, F (1, 236)= .00, p= .999; IDM-1, F (1, 236)= .13, p= .875; IDM-2, F (1, 236)= .40, p= .673; D-IDM-1, F (1, 236)= 1.17, p= .313; and D-IDM-2, F (1, 236)= .29, p= .752. H4 was not supported in treatment decision making.
Interaction Effect Framing and Visual Illustration (H5a & H5b)
Yumin Lin, 11354135
framing and visual illustration on comprehension, F (1, 264)= .06, p= .815; IDM-1, F (1, 264)= 1.26, p= .264; IDM-2, F (1, 264)= 1.67, p= .197; D-IDM-1, F (1, 264)= .01,
p= .919; and D-IDM-2, F (1, 264)= .26, p= .613. H5a and H5b were rejected.
Three-way Interaction Effect Framing, Visual Illustration and Age (H5c)
Results indicated that there were no significant three-way interaction effects between framing, visual illustration and age group on comprehension, F (1, 264)= 1.09, p= .297; IDM-1, F (1, 264)= .00, p= .963; IDM-2, F (1, 264)= .00, p= .947; D-IDM-1, F (1, 264)= .48, p= .491; and D-IDM-2, F (1, 264)= .46, p= .499. H5c was not supported.
Conclusion Study 1
In sum, Study 1 suggested that framing marginally affected Deliberated Informed Decision Making Type I (D-IDM-1). However, the direction of the effects was in contrast to the hypothesis, so that negatively-framed message showed the potential to positively affect informed treatment decision making, when deliberations and comparative positive attitudes (not necessarily positive) was considered. No
Yumin Lin, 11354135
significant effects were found for visual illustration and the interactions between the framing, visual illustration, and age.
Study 2 (Hospital Decision)
Method
Design and Participants
Study 2 was also conducted in a hypothetical context of cancer-related decision making, but the participants were asked to make a decision between three hospitals. This study also adopted a 2 (framing: positive vs. negative) x 2 (visual illustrations: present vs. absent) x 2 (age: younger cancer patients [< 65 years old] vs. older cancer patients [> =65 years old]) between-subject factorial design.
Data was collected in parallel to Study 1 via the same panel, but with a different sample. Also, identical sampling strategies and inclusion criteria were applied as in Study 1. An identical sample size (270) was required to detect a medium-sized (.25) effect based on the power analysis in Study 1. A total of 281 participants took part in Study 2, among which 150 participants were younger than 65 years old (M= 48.48,
Yumin Lin, 11354135
SD= 10.39, range 22-64, 38% male), and 131 participants were older than 65 years
old (M= 71.82, SD= 4.84, range 65-86, 62.6% male). Approximately 54% of the participants had tertiary education, and 24.6% of the participants were currently receiving cancer treatments (Table 5, See APPENDIX F).
Procedure
The participants followed a similar procedure as in Study 1. The conditions in Study 2 were similar in terms of the use of framing and visual illustrations. The participants were also double-checked based on the inclusion criteria after giving informed consent. After this, they were randomly assigned to one of the four conditions.
The participants were asked to imagine that they were diagnosed with early-stage lung cancer, and they agreed on receiving surgery based on the proposal of their physician. The decision they had to make after reading the provided web-based PDA was in which of the three hospitals they would undergo lung cancer surgery. The participants were also instructed to read the information carefully. After reading the
Yumin Lin, 11354135
PDA, the participants filled in a questionnaire adapted from Study 1, in which identical concepts were measured, but tailored to the hospital decision context.
Stimuli Development
The stimuli created for Study 2 (Figure 5-8, See Appendix B) were adapted from an existing Dutch online hospital comparison CPI named Ziekenhuischeck (https://www.ziekenhuischeck.nl/). Three random hospitals were drawn from the list of all Dutch hospitals, and their information regarding survival and complications after lung cancer surgery were used for the development of the stimuli in Study 2. Additionally, the names of the hospitals were modified, so that bias from pre-existing experience and knowledge in regards to the hospitals were controlled. The use of modified names of existing hospitals also ensured the external validity, so that the information in the stimuli was comparable to those in real life.
Three additional aspects (recurrence, 2x waiting time) were added to the CPI retrieved from Ziekenhuischeck. The numerical information of the chance of experiencing recurrence of cancer tissue was adapted from the stimuli created for
Yumin Lin, 11354135
Study 1, so that comparable numbers were manipulated for the three hospitals based on the relevant information from Study 1. Finally, the statistics regarding waiting time were manipulated based on relevant information in other types of cancer (e.g. breast cancer).
Identical to Study 1, equivalent numerical textual information was presented in tables for all conditions, but applied different framing strategies. In the positive framing conditions, the chances of survival, not experiencing complications, not experiencing recurrence of cancer to the lung, receiving diagnosis and/ or treatment in a short time were presented, while in the negative framing conditions, the chances of mortality, experiencing complications, experiencing recurrence of cancer to the lung, not receiving diagnosis and/ or treatment in a short time were demonstrated.
To make the two studies comparable, the same approaches of the manipulation of the presence of visual illustrations were applied, so that in the conditions with visual illustrations, bar graphs representing the comparisons between the three hospitals in every aspect were added, while in the conditions without visual illustrations, only textual information was provided.
Yumin Lin, 11354135
Measures
Comprehension of Core Information As in Study 1, the measurement of
comprehension of core information was developed based on a pilot test among 21 participants (Table 3, See APPENDIX E). Similar approaches were applied as in Study 1. four questions were asked to assess participants’ verbatim knowledge, for example “In a percentage, what is the chance of not experiencing recurrence of cancer tissue to the lung after surgery in Laurentine Hospital?” Since the questions asked for a percentage, the answers in decimals or without a percentage sign (e.g. .93 or 93 for 93%) were coded as correct. The verbatim knowledge scores ranged from 0 to 4, α= .80, M= 2.30, SD= 1.54.
Gist knowledge was again measured by ten multiple choice questions in three types as in Study 1, for example “Among the three hospitals, there is the highest chance of dying within 30 days after lung cancer surgery in Albertus Schweizer Hospital.” with the answer options “Correct”, “Incorrect” and “I do not know”. The gist knowledge scores ranged from 0 to 10, α= .90, M= 6.48, SD= 3.40.
