The attitudes and safeguards expressed by participants are summarised thematically below. With regards to differences in attitudes according to the proposed type of health condition for which the data would be used, the change in research topic discussed in the interview from COVID-19 to ovarian cancer and/or bowel cancer, made no difference to the rate of participants happy to donate loyalty card to a data donation bank; an equal number of participants (15 out of 20 in each condition) would be happy to give their data. However, there were differences depending on diseases that were discussed in terms of participants’ understanding, and willingness to donate different data types. These are reported below. There was a small difference in preference for sharing health status alongside loyalty card data, where rates were higher for ovarian and/or bowel cancer compared to COVID-19 condition: 17 out of 20; 15 out of 20, respectively.
Two sub-themes were identified in relation to the a priori theme “Attitudes” and were labelled as (1) rational response, and (2) emotional response. We define rational response as based on the assessment of available information and understanding, while emotional response reflects ‘gut’ feelings and instinctive thoughts. In terms of the a priori theme “Safeguards”, we defined three sub-themes, which were labelled as (1) anonymisation/the level of data detail; (2) control, convenience and choice in sharing data; (3) transparency and data security. The frequency of participant responses can be found in Appendix 3, Table A2. We discuss each theme in turn below.
Attitudes
We asked participants in two subtly different ways whether they would donate their data, and their responses were consistent. First, participants were asked if they would donate their shopping data to health research in general, and second if they would donate their shopping data to a databank where researchers would use the data for various specific health research projects. In both cases, a majority of participants (34 out of 40) were happy to donate data. For most participants there were no health research types they would not want their loyalty card data donated to. Participants were informed about loyalty card data types which could be shared: spending category, purchase amount, timestamp, and location. The majority were willing to share all these data types (28 out of 40). The discussions that these responses evoked were both rational and emotive.
Rational response. There were three main reasons why participants were willing to donate their loyalty card data. First, participants would donate because they conferred importance to health research, because donating their data was both beneficial and helpful to society. Participants generally wanted to help by donating their data and felt if it was helping society or would “make people well”, they would be happy to do so (P123, similar in P116).
A second cluster of reasoning behind donating shopping data was whether they considered shopping data private. Those who did not consider the data private did not mind sharing it. The majority of participants were less concerned about donating loyalty card data than other forms of personal data, often stating they were “not bothered” about loyalty card data being shared and displaying attitudes of indifference: “I’m happy to share everything and once it’s out there, it’s out there.” (P102). This lack of concern was attributed by some participants to the fact that this information was public already as they were visibly purchasing in shops, and the retailer already had their loyalty card data. For example, P104 stated that: “It doesn’t really bother me if people see that, especially if it’s used for research purposes [...] I wouldn’t consider it personal. I mean, you’re in a public place anyway.”
However, there was a small number of participants (N=2) who both considered shopping data private and gave negative responses to donating the data to medical research. They were sceptical about whether the data could be kept secure if it's being used for research (P112) as well as not understanding how the data could be helpful for research (P103).
Several participants (13 out of 40, 10 of those in the COVID-19 condition) were initially confused over how donating loyalty card data could be useful in medical research: e.g.: “I think like if someone gets in contact with someone that had COVID maybe[…] oh I’m not sure how they would track it if they don’t have the person’s address or their personal information.” (P110). In the bowel cancer condition initial understanding was much higher (17 out of 20).
After a short explanation, or participants themselves reflecting on the topic, most individuals developed a sound understanding of the concept of using loyalty cards data for medical research, for example to understand “the relation between the products being bought, [...] and the factors that are observed in COVID-19 patients.” (P112). Further, the majority of participants in both conditions (24 out of 40) recognised the potential of loyalty card data to investigate disease causation through diet. For example, P137 said that: “[...]they [researchers] would be categorising a variety of products that would be, I don’t know, a, b, c, d in terms of contributing towards bowel cancer. Um, being able to match those kind of things it would give them trends […] are people [...] more likely to be, you know, susceptible to bowel cancer because they’re buying products x, y and z?”
This understanding increased throughout the duration of the interviews, and participants began to envisage the concept: “it’s a great idea [...] everybody should be doing it.” (P113). They talked about methods of collecting these data that can work best for the researchers, and interventions the research could lead to. Gaining an understanding of the potential of shopping data in health research throughout the interview helped most participants to form a positive attitude towards donating these data to medical research. Once the uncertainty of how their loyalty card data would be used in health research had been addressed, participants actively wanted loyalty card data to be used for medical research with the concept seen as beneficial:“[…]these big datasets which our society has created could – can be a real force of good..” (P140).
Emotional response. Even though the majority of participants would be willing to donate their loyalty card data, fears about data sharing were still disclosed. Some participants would donate but were cautious or reserved about donating, dependent on the circumstances: “[…]it’s good, […] I’m okay for donating it but I’m not sure how it may be used, […] it depends on who, who the data is passed on, so which organisation is using my data and how it is using it.” (P134).
Over a third of participants (14 out of 40) were worried about donated loyalty card data being obtained by a third party. Most of these participants feared that it would be used by private companies to market and advertise products to them, even though such a possibility was never mentioned or implied. There was a distrust of the motives of private companies, and a belief that profitability would take precedence over the welfare of participants: “So, fears I guess would just be that data kind of being breached and that privacy being breached and being used for things that aren’t beneficial […] just for kind of like advertising or – like algorithms that kind of promote certain products to you that are actually detrimental […] people’s data just being like sold – er – to loads of different companies.”(P138).
There was further sensitivity amongst participants about sharing location data. A number of participants indicated they would not donate data with associated location information (10 out 40). Participants frequently mentioned their “address” when asked which personal data was the most sensitive. Participants were informed that loyalty card location data would not necessarily correspond to their address but rather a store location or geographical region. Nonetheless, location data was often negatively associated with surveillance, with concerns being more identifiable in those participants answering questions about COVID-19. Participants questioned why this data type was needed: “[...]except location, because, that’s just getting too much […] you can determine exactly my pattern of movement around London […] a little bit much for just health research.” (P117). Those whose interview included questions about COVID-19, frequently proposed loyalty card data could be used to track individuals’ movements when asked how it could be useful for health research. In response to the same question, in interviews where the specific research topic had been cancer, no participants mentioned using location data in this manner. However, the fact that a participant voiced concern that location data could be used to track movements did not translate to a decreased willingness to share such data; the majority interviewed in the COVID-19 scenario responded positively to donating location data (19 out of 20). In comparison, nearly half of participants (9 out of 20) interviewed about cancer were not prepared to share location.
Despite expressing potential fears, most participants indicated a high level of trust in researchers, although some stated that the level of trust they could would confer would be dependent on who the researcher was. Only one participant answered they would not trust researchers at all. Participants often contextualized the higher trust they had in researchers by comparing them to private companies in whom they had less faith: “I would trust university researchers more than private researchers” (P109). In particular, they referred to the high standards universities maintained across either ethics, working towards good causes, or data policies.
Safeguards
While the majority were positive about data donation, even those who were positive mentioned some caveats in the sharing of their data, with the importance of the following items being common: data anonymisation; data security; and a transparent choice as to what their data was used for. We discuss these three issues related to safeguarding the use of loyalty cards data in turn below.
Anonymisation and the level of data detail. Participants' statements on their happiness to share loyalty card data often came with a stipulation that their information was kept safe and secure. However, the main caveat given was that the information could not be used to identify them. The main concern of participants was their name being shared, and a higher level of detail of information was associated with a higher chance of being identified. These concerns fit with participant fears expressed across the theme of emotional response, and apprehension of third parties obtaining their data and/or being located and tracked: “[…]happy to share all of it, I don’t think that’s really that sensitive, […] as long as it wasn’t linked to my name.” (P124).
A majority of participants were willing to share their health status, alongside donating their loyalty card data, in order to help researchers investigate specific diseases. Participants appeared to be more comfortable sharing their health status compared to their medical records as a lower level of detail was needed, providing the participant with a higher level of control over what was shared: “If I could just tick a box, for example, I had a certain condition I’d be okay with that.” (p136).
A few participants (17 out of 40) named medical data as more sensitive than other types of personal data, but they still stated they were willing to share this data for research. In contrast to medical data, sharing loyalty card data was frequently viewed as sharing data already in the public domain. Subject to appropriate anonymisation, a large proportion of participants (27 out of 40) were willing to share any data from their shopping regardless of the types of products they bought: “[…] because if it's anonymised, gathering data about what people are purchasing from a supermarket, I personally wouldn’t find that intrusive.” (P128).
Control and convenience in sharing data. There was no consensus on the control mechanisms and processes for sharing loyalty card data and a few issues were discussed. A majority of participants (24 out of 40) wanted the retailer to take care of donating their data rather than having to donate it directly to the researcher themselves. However, exactly half of participants wanted to be able to have control over which information they wanted to remove from the shopping data before donating it for research. Over half of participants preferred consenting to sharing already collected data, yet almost half would consent to share the data that would be collected about their shopping in the future, too. Despite no consensus on these controls a distinctive theme on convenience emerged from the answers. Participants stated they wanted to donate but were unlikely to do so if the process was not quick and easy. Participants often labelled themselves as lazy and forgetful, or weighed up convenience versus control: “If it was just all done for me that would be my preference.” (P133).
Transparency, choice and data security. The minority of participants (9 out of 40) did not want to donate to all medical research in general and they preferred to have a choice as to specific application domains. However, there was no consistency as to the types of health research that were inadmissible; only mental health was mentioned by two participants as a particularly sensitive domain. Participants were almost divided equally over whether they attached import to “choosing” the types of health research their data would be used for: 21 participants did not require a choice at all if it was for the public good, while 18 participants stated that making a choice as to application area was important to their data donation decision. For such participants, transparency about what research they are donating their data for was of value.
Interestingly, even though participants stated they sought an explicit choice, almost half of participants that attached an importance to having such a choice also said there were no health research types they would not donate to. These participants wanted the choice to choose from any domain: “I’m not overly concerned about where it ends up, or rather, which health sector specifically have it. I would like to maintain the choice [as an option].” (P118). Corresponding with the theme of transparency, even when participants did not attach an importance to having a choice, it still remained crucial for them to be informed of what the data would be used for.
Finally, transparency was the top requirement given by participants if the public was to be encouraged to trust researchers with their loyalty card data, and was referred to throughout the interview process by participants. Participants wanted it to be clear not only what their loyalty card data were going to be used for, but who would have access to it. Again, participants who desired safeguards were motivated by fears aligned to the emotional response theme. One participant suggested direct interactions would encourage data donation: “[...]build trust by having conversations with – with people […] make it clear that the university has expectations of its staff and its researchers”. (P140). While participants referred positively to the high standards universities kept for ethics, working for a good cause or data policies, a quarter of participants still wanted additional reassurances that data would be kept safe and secure.