This study was interested in three groups of subjects. These groups were based on the two independent variables highlighted. The first independent variable was whether someone had lived in the United States. The second variable was whether someone was a layperson or had expert knowledge about the healthcare system. Originally, the plan was for four groups with health experts split into those who had studied in the US and those who had not. Ideally, it would have remained that way, allowing for me to control for the difference between expert knowledge and living abroad among policy experts. However, of the five policy experts, all of them had lived abroad, four in the United States, and one in Britain, so it was decided to go with three groups, folding all of the healthcare experts into one group. This was changed for both ease of research and to get more meaningful results. The original proposal called for twenty people, and the ultimate tally of subjects turned out to exceed that total. The final number of respondents was twenty-four with twenty-two them being fully usable. (Some respondents did not fit all of the criteria for
77 (Sofaer and Firminger 2005 p. 520)
33
inclusion but that only became apparent after the interview had been started.) More specifics on the respondents appear in the Analysis section.
The first group, henceforth known as A, was composed of Taiwanese people who had lived in Taiwan their entire lives. By “living in Taiwan their entire lives”, I mean people who had not lived outside of Taiwan for more than six months at any one time.
Group B was composed of Taiwanese people who had spent more than six months living in the United States within the past fifteen years. Group C was composed of people who either worked or studied the National Health insurance system professionally.
The subjects were contacted over the course of several months, the first in May of 2014 with the last in September that same year. Interview subjects were found using a variety of methods. For groups A and B, the snowball method was used, starting with people who were already acquainted with the researcher. A purposive sampling snowball method was then used where interview subjects recommended more interview subjects in order to expand the group. Group C was found using purposive sampling; the author personally sent emails to health policy experts and asked them if they would participate.
Interviews were conducted at restaurants, coffee shops, personal offices and private homes and recorded with the subject's consent. Interviews ranged from thirty minutes to an hour. Those recordings were then transcribed and coded. As part of the snowball method, some of the subjects referred to me were actually living in the United States at the time of my research. The survey was sent on to them to be filled out at their own convenience. In these cases, there was no in-person interview.
5.3 Framework
The original framework I proposed for this paper had eight variables. (see Figure 1).
34
Figure 1.
The design of this method was intended to address some of the biggest problems in researching perception and satisfaction. Initially, it was suspected that open-ended questions and long interviews make it easier to separate experience from expectation.78 The literature suggested that personal interviews increase the response rate and get longer and more candid answers then mail questionnaires.79 This was found to be the case.
There were certain subjects in Group B that I was unable to directly interview so I had to settle for forwarding the questions. The subjects interviewed directly produced far more material.
I chose to omit discussions of different specific providers in the US because my method was designed to look at how people view the system globally. It was permissible to omit the providers since studies have shown that they actually are not very strong
78 (Sofaer and Firminger 2005 p. 520)
79 (Crow et al. 2002 p. 31)
35
determinants of patient perception.80
The format of the original conceptual model was drawn from the tentative one that Sofaer and Firminger proposed for patient perception, and was my initial guess at a basic format that would elicit what drives perceptions of people about their healthcare system. The circles are the inputs and the outputs of how a person forms his/her healthcare perceptions. In this study, not all of the interview groups have all of those circles. Group C, the healthcare policy experts, have “expert knowledge.” This group is composed of people who analyze healthcare at the policy or system level and not just at the personal consumer level. This study was also looking for effects related to living abroad and perceptions of value so it was important to delineate those who have experience with the US system vs. those who have experience with only the Taiwan system.
The purpose of this framework was to guide the creation of the interview
questions and the subsequent analysis of the interviews. Though this framework detailed many possible influences on perception, the interviews at first focused on eight main variables. Five variables were designated input variables, ideas that influence initial opinions of healthcare before perception is formed. Three others are designated dependent variables, measurements of perception itself.
That framework was revised as the interviews went on and dropped to six
variables. The circles that were marked “possible contributors” were still measured by the survey but I wanted to make the framework show only the explicit variables measured. I also folded certain variables together. For example, there was no practical way to separate
“Direct Experience with the Healthcare System” from “Previous Experience.” To get meaningful results, it required that subjects remember a past direct experience and then remember experiences that were themselves previous to the past experience. Similarly,
“Knowledge of What to Expect” and “Expectations” were not separable. One of the problems with the initial framework was that it was unidirectional. From the surveys, it was clear that there were feedback mechanisms between these variables. “Perceptions of Direct Experience” were influenced by “Expectations” but those “Perceptions” would in turn go back and influence “Expectations.” “Perceptions of Direct Experience” could also
80 (Hekkert et al. 2009 p. 72)
36
influence someone's personal “Ideals” which in turn would influence their “Expectations”
which would then alter how they perceived subsequent experiences. Indeed, this study postulates that laypeople who live in the United States have gone through the loop once already, starting from “Direct Experience with the United States” and that this has altered their “Expectations” and “Ideals” so that they react differently when they have “Direct Experience with the Taiwan System.”
Figure 2.
37
5.4 Survey Design
This is the list of the original variables, including all of the preliminary questions that touched upon these variables in some way. While they are listed by variable, it is important to emphasize that each question was not limited to just one variable. Several of these questions elicited answers that touched upon two or three variables at once. And that was invaluable for drawing connections. Another important note about these
questions is that they weren't meant to be comprehensive. The study did not actively seek to directly measure cost, access, satisfaction, doctor choice, patient knowledge, doctor knowledge, communication, time spent, facilities, etc. on a standard scale. Answers related to all of those were important to the research, and they did come up in the responses, but the interview process was designed to ask open-ended questions that allowed the patients to bring up those points which they felt were important.
The interview questions were generally in pairs. The first question requested that a response be placed on a Likert scale, while the second asked them to explain why they chose that rating in concrete terms. It should be noted that this research was more focused on the reasons behind those scores than the scores themselves. For example, the subject was asked to say how expensive the NHI is on a five point scale, then the follow-up question asked why they chose that score. The methodology required the subjects to map their perceptions to an abstract framework by asking them to think in terms of a scale and then following up to examine what this internal scale looked like. This methodology is drawn from the guidelines laid out by Kahneman et al. According to their work, a pure monetary scale would be unreliable as they found that it is more of a measure of perception or attitude. At the same time, they discovered that a categorical scale is far more reliable and consistent from person to person. The method in this study allows the use of a categorical scale which is the standard in most perception research and then mates it with a monetary scale to assess how that perception translates to monetary value according to the individual person.
Another important insight from Kahneman is that the interviewer should avoid using any real numbers when asking the questions to avoid altering the response. To quote,
“Tasks in which respondents indicate a judgment or an attitude by producing a
38
number are susceptible to an anchoring effect: the response is strongly biased toward any value, even if it is arbitrary, that the respondent is induced to consider as a candidate answer.” 81
This effect is also another reason for the use of a categorical scale followed by an elicitation of monetary or time values. The sample size is comparatively low so it is important that bias is minimized in every way possible.
5.5 Variables
These are the original eight variables and the questions that I attached to them. It needs to be stated, however, that Grounded Theory requires iterating and altering the theory as new information comes in. In this section, therefore, I will explain the original variables, the questions based on them, and how they changed as the research went on.
Variable I: Expert Training – One of the goals of this research was to examine how perceptions differ between policy experts and laypeople, which required some assessment of what separates the two groups. For the purposes of this study, people who work professionally in healthcare policy such as professors or government officials are said to possess “expert training.” This is knowledge of how their own country's system works and how healthcare systems work generally. This is the main factor delineating groups A and B from group C. Questions for this variable were open-ended as it was important to see what experts emphasized vs. what laypeople did.
How familiar are you with Taiwan's National Health insurance system on a 5-point scale with 1 being not at all familiar and 5 being very familiar?
◦ Why did you choose that rating?
This question pair was one of the most important for this thesis, and it was asked of everyone, whether or not they were a designated “healthcare expert”. It was important to identify people who might know more about the health system through their job or previous life experience. Subject B3, when asked why she gave herself a 5, responded with, “because I used to work in a (expletive deleted) hospital in Taiwan.”
81 (Kahneman et al. 2000 p. 226)
39
Can you describe how Taiwan's National Health Insurance is structured?
This question was originally supposed to elicit a large detailed response from policy experts. However, it very quickly became apparent that this question was far too open-ended. In addition, the responses elicited would be essentially reference material, and it was a waste of the experts'
valuable time to ask them a question which was better served by reading from a book. The level of detail needed and the length of responses I received the first few times I asked this question proved that it was going to be too long and useful enough to use in the study. This question was eventually removed.
How do you use cross-national comparisons in your professional work?
The goal of this question was twofold. The first purpose was to show that policy experts use cross-national comparisons in their professional work.
The second goal was to see how it was integrated with the work they were already doing.
What kind of sources do you use to get information about a health system?
This question served to illustrate the source of expert knowledge.
Variable II: Previous Experience – This variable assessed the role of previous experience with healthcare in creating expectations. Partially, that entails assessing what someone's previous experiences were but it was also trying to illuminate how those experiences generated expectations. In the end, this variable was functionally identical to Variable V: Direct Experience. So I folded them together in the final analysis. .
How often do you go to the doctor? (1-5 almost never - very often )
◦ What does that mean in times per year?
◦ This question pair was to measure both the actual experience of seeing the doctor and investigate what the subject's ideals about going to the doctor were.
According to C2, Taiwanese people go to the doctor an average of fifteen times per year while Americans only go 4.1 times per year.82 It follows that what might seem “occasional” to someone in Taiwan might seem “often” to
82 (McCarthy, 2014 web)
40
someone who has been living in America.
Where do you go? (private clinics, hospitals, etc.)
◦ This was important for the study in case there was a significant difference between people who went to the clinic vs, going to the hospital. One of the things that did come up was that some subjects in Group A didn't seem to consider times they went to a clinic, as “going to a doctor.”
Have you ever been covered by the health system of another nation besides the US or Taiwan?
◦ This question was to make sure that no one had been covered in a country that wasn't the US or Taiwan.
Have you been covered under Taiwan's NHI?
What kind of healthcare coverage do (did) you have in the United States?
◦ Both of these questions were to establish what the subject was covered by in the US and Taiwan. Someone with very expansive employer-provided care might perceive the US system differently than someone who lacks insurance.
Have you ever used the health system in the United States?
Have you ever used the health system in Taiwan?
How did you choose your doctor?
Did you use it for urgent or routine care?
◦ These questions were to see the extent of their experience. There were a handful of subjects who had lived in the United States but had never used the medical system there. Their answers were different from those who had had direct experience.
When did you first get health coverage?
How did you first get health coverage?
Describe your first experience with the healthcare system as an adult.
◦ These three questions were removed from the study because the first few times I asked, subjects were unable to remember.
Variable III: Knowledge of What to Expect – This was supposed to illicit how much a person knows about healthcare, but not from direct experience. The goal behind
41
looking at this variable was examining how much people form their expectations based on external sources, schools, newspapers, media, family members etc. as opposed to direct experience.
Unfortunately, this variable was similar to “Expectations” and ran into the problem of subjects being unable to recall accurately. Most of the subjects were old enough that the first time they would have learned about the system as a responsible adult was over a decade in the past. They were not able to remember when they first heard about the system vs. experiencing it directly. It quickly became apparent that separating out the external sources of expectations as separate from direct experiences would have required an entirely new study.
How did you learn about the healthcare system?
Do you feel like you were adequately informed about the healthcare system before you used it directly? (1-5 not informed – extremely well informed)
What did you already know and what was new after your first encounter?
Where would you go to learn more about the healthcare system?
◦ None of these questions were used in the final study and this category was eliminated entirely. In retrospect, it might have been useful to ask about preparing to go to America. Alternatively, I could have checked for familiarity with certain sources.
Variable IV: Expectations – This was looking at a person's current expectations for their healthcare system. Current expectations were used because trying to assess expectations prior to direct experiences that were already in the past was likely to produce unreliable results with people unable to recall what they expected at a given time.
If you broke a bone and had to go to the hospital, describe what you think would happen.
◦ What kind of treatment would you receive?
◦ Would it be expensive to treat? (1-5 scale)
▪ How much is that in monetary terms?
◦ How long would it take to get treatment? (1-5 scale very quick – very long
42
time)
▪ How much time would that be?
▪ This set of hypothetical questions was given to subjects in Group A for the Taiwan system while subjects in Group B received questions about both the Taiwan and US system. As the study proceeded, the question was changed to “broken arm” rather than “broken bone”. The reason for using a broken arm instead was to give the subjects a common and concrete frame of reference. A broken arm is an injury that most people can think about, and it was a way to see how they would react to a similar
experience. One of the unintended side effects of this question was that it tended to elicit any similar experience that the subject might have had.
Variable V: Direct Experience Seeking Healthcare in Taiwan or the US – These questions asked people to describe the extent of their direct experience with the health system. For example, someone with a chronic illness will use the system more than someone who is healthy and their usage of the system will be different. In practice, this variable was unable to be separated from “Previous Experience”.
If you're comfortable, describe your most recent experience going to the doctor in Taiwan.
If you're comfortable, describe your most recent experience going to the doctor in the US
◦ These two questions elicited a wealth of information and experience. They also were invaluable to characterizing the depth of their experience.
Variable VI: Perceptions of the Experience with the Healthcare System – These questions asked for thoughts about the direct experience that they had. Earlier, the survey asked similar questions about a hypothetical scenario, but here, they were asking about an instance that had actually happened. Here the study was less concerned with how much someone paid objectively or how much time it took to get treatment, and was instead interested to know what the patient thought about the details of the experience,
Variable VI: Perceptions of the Experience with the Healthcare System – These questions asked for thoughts about the direct experience that they had. Earlier, the survey asked similar questions about a hypothetical scenario, but here, they were asking about an instance that had actually happened. Here the study was less concerned with how much someone paid objectively or how much time it took to get treatment, and was instead interested to know what the patient thought about the details of the experience,