Introduction

The COVID-19 crisis highlighted the importance of using the potential of patient generated health data (PGHD) not only to track personal encounters, but also for big data research and to understand various under-researched conditions that lead to acute or chronic diseases. For example, the Robert Koch Institute (RKI) has launched a COVID-19 data donation app in Germany [1]. Users may allow the app access to their wearables and donate any data recorded from health apps or tracking devices. The use of health apps and wearables enjoys ever-increasing popularity within society [2]. However, the willingness to donate self-tracked data might be of short duration, considering the special circumstances of the COVID-19 pandemic and the therefore current high level of trust the RKI holds within German society. A more sustainable and lasting solution to gain access to self-tracked data for research and development (R&D) purposes might be the payment of monetary compensation—a digital dividend.

The lack of transparency concerning data processing among some health app providers triggers trust issues and data privacy activism in most countries that are members of the Organization for Economic Co-operation and Development (OECD). Calls for compensation strategies such as a digital dividend are increasing steadily, as evidenced in German political debates about regulatory possibilities of such a dividend [3]. In most OECD countries, personal data is understood to be personal property. Regulations, such as the European General Data Protection Regulation (GDPR) and its translation into German national law, DSGVO, have evolved to protect the individual’s data property, which means that a digital dividend becomes an ever more realistic future scenario.

Health data is understood to be among the most valuable personal data. The pharmaceutical industry in particular could experience significant benefits from engagement in big data research of realistic daily health data and increased efficiency over long and costly clinical trials as a result [4]. German government aims to become a data hub and buy large amounts of personal data as part of their digital strategy [5]. In December 2019, Germany passed the digital care act (Digitale Versorgung Gesetz, DVG), enabling physicians to prescribe health apps. Statutory health insurance companies will reimburse the costs for health apps to the individuals they insure, which is an internationally unique system of health app usage encouragement [6].

First, to research users’ estimated willingness to accept a certain monetary compensation payment—a digital dividend—for sharing self-tracked health data and second, to determine the main factors for such a trade, we conducted a discrete choice experiment (DCE) in Germany. DCE is a choice-based survey method used to research people’s preferences for different options in realistic choice scenarios, contrary to asking them directly about their preferences. We chose this method because people tend to have strong opinions about selling personalized data when asked directly [3]. Previous experiments in lab settings led to demands for high monetary compensation for personal data [7–9]. We used a specific type of DCE, separated adaptive dual response (SADR), to get a more realistic picture of preferences. The experiment was launched before the outbreak of COVID-19 in Germany.

Therefore, we are going to answer the research question under which circumstances wearable and health app users would accept a digital dividend for their self-tracked health data. To answer the research question, we conducted two identical studies with different prices to test the general robustness of price estimates, as we expected people to overestimate the monetary value of their self-tracked data.

Background

Research about price expectations for self-tracked health data is still sparse. As the RKI COVID-19 data donation app was found to fight the severe social and physical effects of the pandemic, the data donation model enjoyed a significant trust advantage. Hence, there is already severe criticism of the RKI data donation model because the donated data is not pseudonymized directly. This leaves a security loophole and facilitates data misuse as well as identification of personalized fitness histories [10]. As soon as the RKI loses people’s trust in its approach to data processing and the threat of the COVID-19 pandemic has eased, the donation model might become obsolete. Even though Skatova and Goulding argue that people donate their data mainly for altruistic reasons [11], the donation model does not reach a large-scale and diverse group of people [10], which would be needed for big data research with self-tracked data.

New technologies such as artificial intelligence and big data offer unprecedented possibilities for use of self-tracked health data in research and development (R&D), which would therefore decrease the costs and duration of expensive control group studies. This might be the next milestone in medical research. Self-tracked data is going to provide new insights using prescriptive, descriptive, predictive analysis and simulations [12–14], potentially improving treatment quality, preventive care, and diagnostics [13]. Hence, research about incentive systems to trigger a large and diverse group of users to share their data for R&D purposes is essential. Bataineh et al. criticized the market for personal data because it shows a significant imbalance [15]. App users are currently not compensated for the provision of their data, and the authors further criticized the lack of an adequate platform to monetize and trade self-tracked data [15]. The question about the compensation that people are willing to accept remains unsolved. This is despite the fact that there have already been some research attempts at answering this question, as described in the following section. However, an SADR study, i.e. a well-aligned, trade-off-based approach, has not yet been performed, even though the settings in this type of study allow for more realistic results.

Wathieu and Friedman analyzed the lack of confidence in online services and app providers and the general rejection of personal data trade [16]. Spiekermann et al. researched the conditions of the international market for personal data. They concluded that people tend to overestimate the monetary value of their data because they do not feel comfortable with the possibility of their data being linked back to them [17]. Many people have little knowledge about the market for personal data and barely know that the data of a single person is rather worthless. However, statistical clustering and digital phenotyping might help to find patterns or gain new insights about the occurrence of different chronic diseases [17]. Therefore, it might be difficult to find a realistic price for health data, depending on personal attitudes and preferences. People often claim strong data privacy concerns, but over 30 million Germans use the payback system, a commercial bonus program offered by the American Express Group. This program enables the trade of consumer shopping information for bonus points. One bonus point equals one Euro Cent [18].

Cvrcek et al., Grossklags and Acquisti, and Hubermann et al. conducted experimental studies to estimate willingness to accept (WTA) and willingness to pay (WTP) for the disclosure of different personal information [9–11]. WTA demonstrates the minimum amount of money one has to offer an individual in order for them to give up a specific good or service. At the same time, WTP illustrates the maximum amount of money an individual would be willing to pay for a service or a good [19]. Many studies showed a significant disparity between WTA and WTP for the same good, which may relate to some form of loss aversion [19]. Aquisti et al. state that “what people say their data is worth depends critically on the context in which they are asked” [20]. This means that people’s perception of their data’s monetary value depends to a large extent on whether they are asked for their data to be protected or whether they are asked to sell their data.

Estimating WTA for PGHD seems difficult because personal price expectations are influenced by opinions, experience, settings, and attitudes. The amount may differ for each individual and depends on how emotionally charged the information is. For example, Cvrcek et al. discovered that participants would agree to publish their smartphone location points for 43.00€ [7]. Grossklags & Acquisti, on the other hand, conducted a quiz with participants and found that people would publish their quiz results on average for $7.06 USD and their personal information for $31.80 USD [8]. The WTP for protecting the same data was below one USD. When asked about number of sexual partners, the WTA averaged $2291.30 USD, and the WTP to protect the same data was $12.10 USD [8]. Such examples illustrate the significant gap between people’s WTP and WTA for data privacy, which results from missing transparency about data processing and a lack of data privacy education. Many app providers offer their services free of charge for the user but sell user data to third parties or use advertisements on their platforms to finance their businesses. Users are often unconscious of these financing models. Furthermore, WTA depends to a large extent on the type of information provided. Hubermann et al. asked participants to auction information about their age, weight, and height for their study. They concluded that participants with a higher body mass index demanded higher prices for the information about their weight than people with a body mass index below average [9].

These experimental studies are groundbreaking in drawing a bigger picture about the practicability of a digital dividend. Participants were not asked to share their continuous real world health data in other experimental studies. In some studies, researchers were able to link the information back to the individuals [7–9,19], whereas in reality, data should be stored and sold anonymized or at least pseudonymized. Making personalized data publicly available often leads to a feeling of embarrassment and, therefore, leads to people demanding higher prices for their information [20]. In most experiments [8,9,19], participants would agree to a one-time data trade for a single payment, whereas a continuous data exchange would be more feasible for large research projects; this is why a continuous digital dividend seems more realistic than a single reward for our study. The literature shows that transparency about data selling to third parties, as well as the nature of health data in particular, seem to be essential factors for the dividend demanded [3,9,16,21]. Research about different incentive systems to trigger people to share their data with different R&D institutions is becoming increasingly important because large-scale self-tracked health data simulations could become a game changer in the fight against the COVID-19 pandemic and beyond. This paper contributes to the current political and academic debate about a digital dividend for self-tracked health data, offering a different and more realistic methodological setting.

Data and methods

To research the digital dividend demanded for self-tracked data, we conducted a DCE. Contingent valuation was no option for the purpose of our study because users had no price demand experience. Our study settings did not allow respondents to directly state their WTA but rather asked about their preferences for different real-world scenarios [22]. If asked directly, people with strong data privacy concerns and opinions tend to block any imagination about selling their self-tracked health data in surveys or interviews, leading to observations of extreme response behavior. Previous experiments in lab settings led to high demands for monetary compensation [8]. Another method might have been a quantitative data analysis with proxy values for data prices. Databases for digital data prices barely exist. Known models to estimate the price of personal data might include payback estimates [18]. However, these are not suitable to calculate the price of health data because health and health data are special goods. We expected that patient generated health data would be more valuable for the stakeholders than shopping data. In our online experimental survey, we showed respondents various offers for their self-tracked data. They had to decide which combination of circumstances they preferred and whether they accepted the offer. Based on these decisions, we estimated the extent to which each attribute level contributed to the observed decision [23].

Between May and October 2019, we conducted two pre-studies with 35 and 100 respondents prior to the final data acquisition, targeting primarily students from our University’s network, who most likely used health apps or wearables to track health. Hence, we ran a power analysis using the results from the 100-respondent pre-test according to Bekker-Grob et al. [24]. The results showed that we needed a minimum of 192 respondents for each questionnaire when applying a significance level of p ≤ 0.05. In December 2019, we conducted the main study over a period of three weeks with 842 respondents, and a second, identical study with 272 respondents which contained lower prices. We launched the second study with lower prices to test the price estimates of the main study for robustness. We carried out the two studies through a professional panel provider, Norstat GmbH, and targeted German speaking respondents aged 18 years to 99 years. By signing their terms and conditions, we committed ourselves to follow their code of ethics and the ESOMAR guidelines. In line with our power analysis, more than 192 respondents participated in either of the questionnaires.

Results

Discussion

In this research, we complemented the current political discussion about a potential legal setting for digital dividends. The results of our analysis clearly showed that participants would share their health data for a digital dividend if no data sales to third parties occurred and stakeholders were transparent about data storage and processing. Hence, we derived from our results that German stakeholders in the health care sector in general and pharmaceutical companies in particular suffer from a significant lack of trust, which directly influences the price expectations for data. We agree with Spiekermann et al. and Wathieu and Friedman that people have difficulty estimating the monetary value of their data because they have little or no knowledge of its statistical use in R&D [16,17]. However, we used a different experimental setting than previous studies have because we offered a monthly bonus payment to the participants for a continuous data exchange, rather than single payment for a one-time disclosure of personal data.

Our analysis also showed that socio-economic factors are either irrelevant to or play no role in whether a person would share his or her data with a health care stakeholder, if the stakeholder earns enough trust and is transparent about data use and processing. This conclusion might be especially interesting for the enhancement of pharmaceutical research because self-tracked data from a diverse group of people might complement or replace control group studies [40–42].

The robustness test showed an anchor effect for the prices [33], i.e. that people had no expectations or knowledge about the price of digital health data. Hence, this might be a chance to introduce a digital dividend because the market may set the price at first. The estimated WTA levels resulting from the study were higher than the attribute levels, which again signals the uncertainty about prices and the participants’ diverging expectations. Future research should address the stakeholder side because we do not know if the participants’ estimated WTA matches stakeholders’ WTP. Hence, given the COVID-19 crisis and a conceivable resulting change of long-run preferences, a second study with lower prices should be repeated in the future. Price expectations might have changed because of the sensitization for the need to enhance and complement research with big data PGHD studies after the COVID-19 crisis.

Summary and conclusion

In summary, stakeholders within the health care sector should engage in more transparency about data storage and data sales to third parties. Such an effort can encourage people to share their self-tracked data with them for R&D. These companies seem to enjoy a low level of trust concerning the processing of personal data. People tend to overestimate the monetary value of their health data, which resulted in high prices demanded. Respondents were generally sensitive to further data sales to unknown third parties.

Nevertheless, the results showed that a digital dividend could be an accepted instrument to convince people to share their data for R&D purposes, having an importance weight of nearly 21%. The pharmaceutical industry in particular would benefit from the inclusion of big data PGHD research. There are high incremental costs for health insurance, pharmaceutical, and medical device companies in comparison to universities. Under certain conditions, 54.91% of the participants would sell their self-tracked data to universities, 46.65% to health insurers, and 38.87% to pharmaceutical or medical devices companies for 20€/month. When considering the scenario that all data is sold and raw data was not further sold to third parties, then universities would have to pay 145.66€/month to use the data. Health insurers would have incremental costs of 31.75€/month and pharmaceutical and medical device companies would have incremental costs of 92.15€/month compared to universities. For agents who would like to buy self-tracked data, it would be advisable to encourage more transparency and educational campaigns about their data processing and data security strategies to increase people’s trust by dissolving the fear of personal discrimination.

A continuing success of the data donation model in Germany is questionable because people were triggered by the events of the COVID-19 crisis when engaging with the RKI data donation app. The discussions about the data security of the RKI data donation app already sparked fear about a lack of transparency and possible discrimination through data misuse. Transparency, trust, and monetary compensations are effective long-term measures to convince a diverse and large group of people to share their high-quality data. Hence, this is also an encouraging result as PGHD-research may help identify the reasons for chronic diseases or severe conditions. The public discussion about patient generated health data usage in R&D might also increase general health app usage and, therefore, health education through self-monitoring.

Supporting information