A growing body of evidence suggests the effectiveness of using big data to promote various health-enhancing behaviors in patients with chronic diseases.1 For example, data-driven efforts have been made to establish prediction models for health events, assist medical decision-making, and provide intelligent care. Nurses have opportunities to proactively extract insights from these data to improve health outcomes and provide patient-centered care.2 As data are collected from multiple sources, how to make the data most effective is essential to identify possible engagement opportunities to aid patient care. Therefore, with proper theoretical guidance, big data can play a better role in generating high-quality and meaningful data or conclusions.

In the era of big data, mobile technology enables the convergence of theory- and data-driven discovery to generate new intervention patterns. Integrating both theory and big data as dyadic drivers may be a promising and iteratively updated approach to innovate the cancer care practice. Our research team has developed a series of mHealth applications for patients with breast cancer according to symptom occurrence and development rules. In 2018, based on the concept and technology of "user portrait" and the assumption of the recommended guide to the appropriate patient population, we assumed that the "recommendation system" could also be applied to the symptom management of cancer. Guided by the symptom management theory and personalized recommendation theory, we developed a personalized recommendation system for the symptom management of patients with breast cancer. The personalized recommendation system involved patient-reported information measures and a knowledge base for symptom management. The patient data and knowledge base were matched with machine learning algorithms including deep neural networks and association rules, and the accuracy of recommendations can be continuously improved according to the update of big data. The usability test results verified the availability and acceptability of the recommendation system, which provided empirical evidence on the effectiveness and feasibility of practice driven by both theory and big data. In this perspective, our research team also adopted the concept of "user portrait" to build a symptom management support system for parents of children with leukemia. Further, we construct a system with early warning and intervention response function for patients with breast cancer by a deep neural network. However, a more targeted theory was necessary to promote orderly cycles of data to continuously improve the quality-of-care system. Thus, we conducted a systematic literature review and found that the Learning Health System (LHS) can integrate the data in the individual dimension, circulate to self-learning iteration and intelligent optimization of the whole system, and is valuable to improve patient health outcomes and continuous improvement of the system. Therefore, we adopted LHS as the framework and focused on elements of symptom management including assessment and response, aiming to construct an Internet-based intelligent symptom management model for patients with breast cancer. In the system, the computer-adaptive tests for patient-reported outcome measurement information system were used for data collection, and LHS cycles contributed to the optimization of the symptom management system, showing that continuous optimization based on big data and theory can inject new momentum into the sustainable development of cancer care plans.

Contemplating the past and looking forward to the future, there are some considerations that theory and big data can play an important role in promoting cancer care in the future. First, although some studies start to demonstrate the effectiveness of big data or theory, few of them adopt the dyadic driving model combining theory and big data. The combination of big data and theory has the potential to integrate, explore, and analyze a large number of complex and heterogeneous data with certain research aims. Second, most theories being used in cancer survivors are health-related. Cooperation that integrated cross-discipline theories such as sociology, economics, management, and informatics might provide new solutions for cancer care. For example, a data-driven user portrait empowered by big data is a way forward to empower care based on patient preferences, in which artificial intelligence contributes to expanding the matching levels between patients and care services. Third, considering a lack of sustainable long-term follow-up for cancer survivors, future studies can consider including the theory that motivates patients' adherence to follow-up and make good use of big data to maintain patients' interest to participate in their health management. For example, design thinking and interaction design theory have been recently used in the nursing field, showing its potentials in driving the design of healthcare programs and products.3,4 By using the big data collected by an information system, the care needs of patients can be refined, and the precise management in different scenarios can be achieved, which can maintain patients' intention to be involved. Finally, the use of big data and data mining allows the ability of self-adaptation for the information system to support personalized care.5 Although multiple information products in the Internet sale field have realized self-adaptation, few health applications can be iteratively optimized by big data. To achieve this, relevant theories and technology that contribute to self-adaptation of the system need to be used. In addition, end-use and multidisciplinary stakeholders should be involved in the design and test process of information systems with the guidance of relevant theories. Future studies should consider the combined drive of theory and big data to address the aforementioned issues and promote the informatization and intelligent development of cancer care.

References