CC BY
29
Zheyu Li,
11th Grade, Hangzhou Foreign Languages School
Hangzhou, China E-mail:1641025778@qq.com
WILLINGNESS-TO-PAY FOR MEDICAL ARTIFICIAL INTELLIGENCE: AN EMPIRICAL ANALYSIS OF THE DIABETES PREVENTION AND TREATMENT SYSTEM
Abstract: For the past 20 years, diabetes has become one of the primary health threats in China. Patient-based self-management was recognized as a key aspect in the treatment of diabetes, but patients do need external empowerment such as self-management programs or apps. Artificial Intelligence can enhance such systems to help monitor the patients, moreover, can do diagnosis of disease progression and connect the patients to professional medical practitioners as necessary, even before the patient realizes the peril. However, as the general public has not yet fully developed an understanding of the cost of intellectual property, it is important to study patient's willingness-to-pay. Through a comprehensive survey, it was found that an average patient does have a relatively high need for intelligent treatment programs and is indeed willing to pay for such services.
Keywords: diabetes management, artificial intelligence, willingness-to-pay.
1. Introduction ity and efficiency of medical services, it failed to
Over the past few decades, the incidence of dia- track the patient self-management of their diabetes.
betes has increased dramatically in China. In 2015, the count of diabetic Chinese has reached 109 million, an increase of11.2 million compared to 2013. According to the International Diabetes Federation (IDF), Chinese diabetes patients will continue to increase by 50% by 2040 (IDF [3]). The central government has taken a series of measures to combat the disease but hasn't achieved much effect. This failure is thought to be mainly due to the prevalent incompetence to control and manage the disorder (Ye and Huang, [14]), that the general public simply lacks the knowledge to notice early symptoms of diabetes, not to mention to effectively diagnose and seek treatment and medication (Wac et al. [8]; Xu [13]). Take the city of Hangzhou as an example, the government has established many Wise Medical System (WMS), which integrates modern technology into the entire process of medical assistance (appointment, streamlined itinerary, billing, etc.). However, while the system has improved the qual-
WMS cannot offer emergency response to sudden rises or falls of the blood sugar level, nor does it record detailed diagnose reports or evaluate the result of each treatment cycle. Starting from 2013, there has been several professional support systems and phone applications that came into the market, targeting chronic disease control. However, as support systems have limited functions and needed expensive support devices, and the phone apps needed human doctors behind the application interfaces, they both failed to liberate patients from their frequent, mandatory visits to the hospital (Wang [10]).
Despite the slow evolution in purely medical trials in chronic disease management, the thriving research in artificial intelligence (AI) offers brand-new approaches. Since the 1970s, many research institutes have attempted to introduce AI into computer-assisted medical diagnosis and treatment. Early "expert systems" such has MYCIN, Internist-
I, DXPLAIN were commercialized. At the turn of the 21st century, medical AI continues to develop, and the medical robot, Watson, developed by IBM in early 2000s, has proved to be four times more precise in its diagnosis than physicians fresh out of medical school (Wang [9]; Wu [12]). In the field of diabetes prevention and treatment, AI research has successfully evolved from theoretical to practical phase, and the main mode is integrate information of different aspects into a whole diagnose advice, including monitoring of blood sugar, diet, workout, medication, and diabetes education (Li [5]). It offers real-time feedback based on its interaction with the patient (Du [1]). Through physical sensors, the AI system observes the patient's behavior and aids in the formation of appropriate habits, assisting in all phases of the disorder, including diagnosis and treatment (Luo [6]; Wei et al. [11]). The fast development of internet further fueled up these developments of AI-enhanced systems and apps, enabling the personalized services. However, will diabetes patients trust AI-assisted treatment enough? Are they willing to pay for it? These questions are not studied by existing literature and would be answered by this study.
2. Imagining the AI-Based Intelligent Medical System
In the fully developed version of Diabetes Intelligent Medical System (DIMS), there will be three parts: 1) the AI component collecting patient data and performing diagnostic modeling behind the scene, 2) hardware monitoring the patient's physical metrics (either digital or manually input by patients), 3) the service layer of phone application and healthcare practitioners, also behind the scene. In a typical scenario, periodic data collection from diabetes patients or high-risk people is collected and analyzed by the AI component periodically (can be a few minutes, or every hour, depending on the patient's preference). If any abnormality is detected, the AI component will contact the patient and his family, and suggest the best available doctor
based on the doctor's credentials. It is exactly these added services of on-going diagnosis and real-time connection with healthcare providers that set the AI-based app apart from regular self-management apps. Therefore, the pricing of such apps should also be based on the added services.
3. Patients' Willingness-to-Pay
Obviously, today's internet enables patients
to access a lot of free medical information online, however, the vast amount of information can also be confusing and overwhelming to a regular patient. Specialized apps such as Good Doctor Online (hao yisheng zaixian, and Chu-nyu Doctor (chunyu yisheng, offer information in more organized ways, and there are also diabetic focused apps such as Diabetic Nurse (tang hushi, ffi^i). Most of their contents are free to the patients now, but they have very limited service functions.
On the other hand, Chinese are gradually getting used to pay for knowledge online, as online classes and services get popular. The willingness to pay can be measured by the amount of money that one accepts to pay for a product, which demonstrates the estimated value of the object/service to the individual (Lang and Lan [4]; Pan and Sun [7]). Research of this concept has been centered on price-setting of usage of public facilities or services (entry price for tourist sites, for example). Applying AI to chronic disease prevention and treatment is an important tendency in healthcare for chronic disease patients. There is a forming ground for willingness-to-pay for such AI-enhanced diabetic management apps.
4. Methodology
This study has selected diabetes patients in the central city of Hangzhou as the survey sample. 700 questionnaires were distributed, 680 retrieved and 662 considered valid, yielding a 95% effective response rate. The questionnaire is composed of four parts: demographic information, willingness to pay for DIMS, sources of willingness to pay, desired im-
provements of the new system. Within the section of willingness to pay, this study has largely employed the Contingent Valuation Method (CVM) (Dutta et al. [2]). Considering excessive flexibility of open-end questions and the consequent difficulty to reach a conclusion, this survey is composed primarily of referendum and payment-card questions, asking the sample members to indicate the strength of their preferences and the maximum amount they are willing to pay for certain services. Additionally, this study has interviewed 20 medical professionals to compile their opinions on the DIMS and the future of diabetes treatment. The statistics collected through the survey and the interviews are then compared against published data, so as to provide a more extensive analysis of the public's willingness to pay for DIMS.
5. Results on Willingness-to-Pay
Sample Composition
Overall, the sample selected for this study has a good representative distribution in demographic attributes. 47% of the respondents are male and 52% female, which confirms the representativeness of the sample, having a gender ratio close to 1:1, clearing up any potential bias from gender differences. 40% have high school or associate degrees, while 45% have college degrees or higher. The respondents are mostly in two age ranges: 31-44, and 45-59, representing 30% and 33% of the respondents respectively. In terms of profession, 37% are corporate employees, and the majority of the rest employed by government and public organizations. The diversity and representativeness of the sample sets a solid base for our statistical research.
Willingness-to-pay
proportion
004%
033% 033% 018% 010%
002%
0% 10% ■ under 100
20% 30% 40% 1100-499 ■ 500-1499
50% 60% 70% .1500-2999 ■ 3000-4999
80% 90% ■ above 5000
100%
Figure 1 Amount willing to pay for DIMS Currency Chinese yuan, CNY
Figure 2. Willingness to Pay by Gender, Occupation (Yuan)
As (Fig. 1) shows, the majority only accepts to pay relatively small amounts. 33% only wanted to pay less than 100 yuan, and another 33% wanted to pay between 100 and 499 Yuan. 18% were willing to pay between 500 and 1499 Yuan, and only 16% were willing to pay over 1500 for DIMS programs.
Patients' willingness to pay doesn't differ much by gender, occupation (Fig. 2), or age (Fig. 3). However, as education level gets higher, the respondents were more willing to pay a higher price 100%
for DIMS, as shown in (Fig. 4). Most noticeably, 23% post-graduate degree holders were willing to pay over 1500 Yuan, compared to average 15% of other groups. We believe this is because higher degree holders tend to have higher incomes, thus can more easily afford expensive services. This is confirmed by Fig. 5, which shows the willingness-to-pay by income level. The higher income groups, 200k-500k, and 500k+, jumped out for willingness-to-pay above 1500 Yuan.
041% 037% ■ above 500 030% 032%
028% 033% ■ 100-499 ■ under 100 037% 030%
031% 030% 032% 038%
80% 60% 40% 20% 0%
under 30 31—44 45—59 above 60
Figure 3. Willingness to Pay by Age (Currency: CNY)
The diabetes patients have a certain willingness to it is important to understand which functionalities pay for DIMS, however, as each individual has a differ- were deemed more important in the DIMS system. ent recognition and understanding of the new system, Three groups of six maj or functional elements were
highlighted to ask the patient to rank their perceived importance: 1) healthcare provider like functions such as AI-assisted treatment, online medical consultation, and sugar control plan design; 2) functions that patients can do themselves, such as daily blood sugar monitoring and blood sugar data management,
and; 3) additional health management. As Fig. 8 indicated, in quadrant A, patients put a high emphasis on the health provider related features, which confirms with our hypothesis that the value comes from the feature that sets the DIMS from other apps, which is the healthcare services behind the scene.
Figure 4. Willingness to pay based on education levels Yuan
Figure 5. Willingness to pay based on income (Ten Thousand Yuan; Yuan)
Sources of Willingness-to-Pay
6. Conclusions and Recommendations income levels, namely, the higher education and
This study aims to investigate the diabetes pa- income they have, the more they are willing to pay.
tients' willingness to pay for AI-assisted services. On average, most patients (about 66%) are willing
It was found that they are willing to pay, but their to pay less than 500 Yuan for such services a year,
level of price acceptance varies by education and while 34% are willing to pay over 500 Yuan. The
main motivations of such willingness to pay stem online medical consultation, and sugar control plan from the patients' hope in Al-assisted treatment, design.
Figure 8. Quadrant model of DIMS function necessity and willingness to
Additional interviews with medical profession- rupted body scan, updating algorithms, and enhanc-
als suggest that the future of DIMS lies in the con- ing machine learning, so as to ensure the sustainable
tinuous health monitoring and improving treatment growth of the DIMS in its applicability and diagno-
precision. It should fully utilize the internet spirit of sis precision. If adopted and promoted properly, this
serving the users and enhancing user experience. In system can not only be of great medical assistance to
terms of technical advancement, DIMS developers diabetes patients and high-risk population, but also
should focus on augmenting the stability of uninter- be a health management tool for the general public.
References:
1. Du J. Smart Medical Application Scenarios and Demand Analysis [J]. Telecommunication Network Technology, 2013 (8): 27-31.
2. Dutta M., Banerjee S., Husain Z. Untapped Demand for Heritage: A Contingent Valuation Study of Prinsep Ghat, Calcutta [J]. Tourism Management, 2007. 28(1): 83-95.
3. IDF (2015). Diabetes Atlas 7th Edition[EB/OL]. Retrieved on 10 April 2018 from: URL: https://www. idf.org/our-activities/advocacy-awareness/resources-and-tools/13: diabetes-atlas-seventh-edition.html.
4. Lang H., Lan Z. Hypothesis of Market Assessment Method-A Method of Exporting Willingness to Pay [J]. Public Health, 2001. 27(4): 261-275.
5. Li L. Progress and Prospects of Smart Medical [J]. China Science and Technology Industry, 2017. (1): 66-67.
6. Luo Y. Research and Application of Case-Based Reasoning Diabetes Diagnosis System [D]. Master's Thesis, Harbin Institute of Technology, 2013.
7. Pan L., Sun Y. A Study on the Willingness to Pay for Scenic Spots' Ticket Price: A Case Study of Xixi National Wetland Park [J]. Geographical Science, 2015. 35(4): 440-447.
8. Wac K., Bults R., Van B.B. et al. Mobile Patient Monitoring: The Mobi Health System [C]. Annual International Conference of the IEEE Engineering iMedicine and Biology Society(EMBC). Minneapolis: IEEE, 2009: 1238-1241.
9. Wang D. Design and Implementation of Diabetes Integrated Intelligent Medical Platform Based on Intelligent Terminal [D]. Master's Thesis, University of Electronic Science and Technology of China. 2015.
10. Wang L. ZigBee-Based Low Power Remote Medical Monitoring System [D]. Master's Thesis, Lanzhou University of Technology. 2010.
11. Wei Z., Xue X., Wang N. Application ofArtificial Intelligence in Diagnosis of Diabetes[j]. China Medical Equipment, 2017. 14(12): 142-145.
12. Wu J. Smart Medical: Can Be the New Breakthrough in Medical Institutions [j]. President of Chinese Hospital, 2017. (z1): 89-89.
13. Xu H. Design and Implementation of Telemedicine Monitoring Network Platform [D]. Master's Thesis, Beijing Jiaotong University. 2011.
14. Ye L., Huang S. Research Progress on Hypoglycemic Drugs [j]. Academic Journal of Medicine, 2008. (11): 1325-1328.
