‘A health data delivery a day’ keeps the doctor away
A case study to structure self-monitored blood pressure measurements by adopting the
sensemaking approach - a physician's perspective to manage hypertension on a
distance
Anke Vermeulen 12-07-2018
University of Amsterdam - Faculty of Science Information Studies - Business Information Systems
1st Examiner: mw. dr. V.M. Dirksen 2nd Examinar: dhr. ir. A.M. Stolwijk Case study organization: Cardiologie Centra Nederland
Abstract: Healthcare providers and patients become increasingly engaged in telemonitoring due to the rising number of chronic diseases. For most chronic diseases, monitoring is already seen as an essential method for disease control, as it corresponds with enhanced health outcomes. However,it is not clear yet how care providers make sense of self-monitored data in order to improve clinical-decision making. The objective of this study is to provide insight into the impact and role of self-monitored health data in the process of clinical-decision making by performing a case study at Cardiologie Centra Nederland. For this case study, articles were analyzed, semi-structured interviews were conducted and patient data was organized regarding the chronic disease hypertension (high blood pressure). Findings expand knowledge on how the sensemaking approach can be used to structure digitally derived data to support healthcare providers in clinical decision-making. Hence, four focus areas were defined: 1) collecting numerous and trustworthy data from patients, 2) extracting meaningful data in order to filter out irrelevant data, 3) discovering patterns of data to give meaning to the information by visualizing data and 4) perform statistical analysis on the data to allow further action to form evidence-based decisions.
Keywords: Telemonitoring; Chronic diseases; Sensemaking; Clinical decision-making; Health data; Hypertension; Self-monitored health data, Blood pressure measurements
1.
Introduction
Healthcare providers and patients become increasingly engaged in telemonitoring due to the rising number of chronic diseases (Bodemheimer et al, 2002). According to the World Health Organization,“chronic diseases are not passed from person to person. They are of long duration and generally of slow progression. Examples are cancer, diabetes, chronic obstructive pulmonary disease (COPD) and cardiovascular diseases, like strokes and hypertension” (WHO, 2016). For most chronic diseases, monitoring is already seen as an essential method of condition control, as it corresponds with enhanced health outcomes (Paré et al, 2007). Additionally, “telemonitoring is defined as the use of information technology to monitor patients at a distance” (Meystre, 2005, p. 63).
Monitoring health through technology and the use of consumer wearables are popular tools in the fitness industry to collect physiological information, like body temperature and heart rate (Piwek et al, 2016). The devices are used to generate large amounts of data, also called big data, in order for the user to utilize this information for health related purposes. However, according to Boyd & Crawford, “big data is not notable for its size, but because of its relationality to other data” (2011, p. 4-5). The researchers emphasize that “its value comes from the patterns that can be derived by making connections between pieces of data, individuals in relation to others, about groups of people or about the structure of the information itself” (Boyd & Crawford, 2011, p. 6).
The serious use of consumer wearables transferred to healthcare, in which scientists aim to discover associations from health related data to enhance treatment plans. Despite the eagerness in discovering associations from health related data, distrust and hesitation remains with regards to the ability of healthcare providers to make sense of the data collected through monitoring and to translate it into actionable insight (Mamykina et al, 2015). This concern relates to the fact that self-monitored health data is considered unorganized, and therefore time-consuming, to filter out valuable information for clinical decision-making.
“Clinical decision making is defined as a continuous process, where data are gathered, interpreted, and evaluated in order to select an evidence-based choice of action” (Tiffen, 2014, p. 399). For example in diabetes management, the question was raised whether data of blood glucose measures could be interpreted and converted into actionable insight (Peel, 2007). “Actionable insight is a cohesive set of understandings about the problem situation based on prognostic insights derived from analytical results which enables the decision-maker to make an informed decision to solve the problem” (Tan & Chan, 2016, p. 3).
In many cases, the decision makers regarding disease management are the care providers who have to interpret the delivered self-monitored health data from which insights could be gathered to make decisions about the patient’s health. While several studies show the effectiveness of monitoring and the experiences of chronically ill patients (Chen et al, 2011; Hassan & Madani 2017; Huygens et al, 2017), it is not clear yet how healthcare providers as decision makers make sense of self-monitored data. As this is seen as the future of medicine (Paré et al, 2007), in which care providers increasingly have to interpret self-measured data of chronically ill patients, understanding and structuring this challenge is needed in order for self-monitored data to be considered in the clinical decision-making process.
Therefore, the objective of this study is to provide insight into the impact and role of self-monitored health data in the process of clinical-decision making. Hence, the following guiding research question is addressed: To what extent can self-monitored health data enhance clinical-decision making by healthcare providers?
2.
Literature review
To provide a better understanding of the context of this research, existing literature is analyzed on telemonitoring. A search was performed to systematic reviews, meta-analyses and scientific articles from journals related to health data and technology.
This knowledge will be connected to literature on sensemaking, showing that the theoretically grounded approach of sensemaking could provide guidance in structuring self-monitored health data.
2.1 - Transmission of health data
Few articles are available in the literature that display how information technology is able to advocate an amount of different applications that are mainly associated with telemedicine and disease monitoring (Aceto et al, 2018). According to a published article in the journal of Telemedicine and e-Health, “telemonitoring is defined as the use of information technology to monitor patients at a distance” (Meystre, 2005, p. 63), as mentioned earlier. Another article, published in the International Journal of Medical Informatics, defines telemonitoring as “an automated process for the transmission of data on a patient’s health status from home to the respective health care setting” (Hardisty et al, 2011, p. 735). Traditionally, monitoring took place at the clinic or in a hospital for which the patient had to travel to the the healthcare setting in order to investigate the result of the data on that moment (Hardisty et al, 2011). Nowadays, “telemonitoring transformed clinical practice in a way which allows the patient to be located globally while being monitored locally” (Andersen et al, 2011, p. e113). The collection of physiological data is a joint activity between the physician and patient, in which data can advocate decision-making regarding the initiation of distinct interventions, like the adjustment of medication (Andersen et al, 2011).
While several studies focused on the collaborative event of telemonitoring (Andersen et al, 2011; Houben et al, 2015; Dedding et al, 2011; McWilliam et al, 2008), other perspectives of telemonitoring were analyzed as well. For example, researchers focused on the cost-effectiveness of telemonitoring (Ahern et al, 2006; Whitten et al, 2002), as fewer patients visited the hospital for monitoring. Other researchers focused on the effectiveness of telemonitoring for chronic diseases, for which 65 empirical studies published between 1990 and 2006 were analyzed in a systematic review regarding different conditions, “like pulmonary conditions (18), diabetes (17), cardiac diseases (16) and hypertension (14)” (Paré et al, 2007, p. 274). Accordingly, the results of the systemic review suggest that patients comply with telemonitoring programs and the use of technologies, especially for pulmonary studies rather than diabetes and hypertension. However, another study indicated that patients who suffer from diabetes and hypertension are most eager to self-monitor (Huygens et al, 2017).
Overall, researchers argue that “chronic disease telemonitoring seems to be a promising patient management approach that empowers patients, influences their attitudes and behaviors and produces accurate and reliable data” (Paré et al, 2007, p. 274).
However, no evidence is found regarding the clinical effects of telemonitoring, nor with regards to the acceptance and interpretation of telemonitoring data by healthcare providers (Paré et al, 2007). Moreover, research lacks accurate representations of the data retrieved from telemonitoring and how it was analyzed in order to perform well-informed decisions (Hardisty et al, 2011; Clark et al, 2007; Maric et al, 2009; Jerant & Nesbitt, 2005; Louis et al, 2003). Additionally, concerns with regards to correctly structure and interpreting telemonitoring data occurs, except when collected data is connected with tools and guidance to foster valuable data discovery.
2.2 - Sensemaking theory
In an article published in the Journal of Biomedical Informatics, it is argued that in a recent systematic review the following theories are considered the most popular and influential for valuable clinical data discovery: “Self-Determination Theory, Social Cognitive Theory, Theory of Planned Behavior and Transtheoretical Model of behavior change” (Mamykina, 2015, p. 407). However, according to Mamykina (2015, p. 407), “these theories focus on psychosocial aspects and on behaviors of individuals, while few consider how to make sense of data and diseases, how new information is interpreted and there is no attention for models to direct future action.”
To discover how individuals make sense of collected data from telemonitoring (Mamykina, 2015), an approach to how clinicians make sense of self-monitored data is discussed through the theoretical perspective of sensemaking. The concept of sensemaking is defined as “the process of creating an understanding of a concept, knowledge, situation, problem or work task, often to inform an action. It is a prerequisite for problem solving and decision making and executing a plan” (Zhang & Soergel, 2014, p. 2).
2.3 - Sensemaking theory perspectives
A few theories exist on sensemaking which are explained from different perspectives in fields like organizational/social psychology and information sciences (Sharma, 2006).
2.3.1 Sensemaking in organizational/social psychology
The concept of sensemaking was introduced in organization/social sciences by Karl Weick (Sharma, 2006). The concept is explained as “the process of social construction that occurs when discrepant cues interrupt individuals’ ongoing activity, and involves the retrospective development of plausible meanings that rationalize what people are doing” (Weick, 1995, p. 2; Weick et al, 2005, p. 409). The researcher emphasizes his view on sensemaking by arguing “that the concept is tested to its extreme when people encounter an event whose occurrence is so implausible that they hesitate to report it for fear they will not be believed.” (Weick, 1995, p. 1). Therefore, Weick (1995) argued that “sensemaking can be set apart from other explanatory processes such as understanding and attribution by seven distinctive characteristics of sensemaking, which are described and understood as a process: 1) Grounded in identity construction, 2) Retrospective, 3) Enactive on sensible environments, 4) Social, 5) Ongoing, 6) Focused 7) Driven by plausibility rather than accuracy.” (Weick, 1995, p. 17). The proposed seven characteristics of organizational sensemaking each “incorporates action and context, which are key elements of sensemaking” (Weick, 1995, p. 18).
However, Weick (1995) argues that other researchers (Snowden & Boone, 2007; Klein et al, 2006; Russell et al, 1993; Dervin, 1983) defined sensemaking in different ways as ‘a method’, as a ‘frame of reference, in a ‘mental model’, in a ‘context for environmental analysis’ or as ‘a process to organize information to serve a task’ (Weick, 1995, p. 8-16). The latter three definitions are examined in ‘sensemaking in information science’. These perspectives are relevant to discuss in this study, as the three researchers (Russell et al, 1993; Klein et al, 2006; Snowden & Boone, 2007) emphasize the retrieval and discovery of information in their theories to structure data.
2.3.2 Sensemaking in information science
As described in 2.3.1, sensemaking gained different definitions, in which Russell et al (1993) defined the concept as “a process of searching for a representation and encoding data in that
(Russell et al, 1993, p. 269). In order to search for representations by encoding data to answer task-specific questions, Russell et al (1993) (Appendix A) describes “the process of information sensemaking as a learning loop: 1) Search for representations by sampling information, 2) Instantiate representations by identifying/sampling information of interest, 3) Shifting representations and redefine goals 4) Use the task specific information and repeat the process. Russell et al (1993) argues that sensemaking is concerned with retrieval of information, but most importantly, also with reorganizing data in a way that it can be used” (Russell et al, 1993, p. 411).
Klein et al (2006) describe the concept of sensemaking also as a process, but in a way to develop an individual mental model to represent external data. Klein et al (2006) propose a theory concerned with including data within frames, which are representations of somebody’s perspective on an event. According to the researchers, “frames shape and define valuable data” (Klein et al, 2006, p. 88). Moreover, the authors discuss that sensemaking “can elaborate the frame by adding details, questioning the frame and doubting the explanations. Therefore, a frame functions as a hypothesis with regards to the connections among data in which a frame can be reframed: questioning the frame leads to reconsidering, to reject the initial frame and to seek for better replacement. That way, a frame can change in the process of acquiring data” (Klein et al, 2006, p. 88) (Appendix B).
Snowden (2007) also emphasizes information sensemaking as a process for environmental analysis (Appendix C), “as it advocates the use of narrative for understanding complexity and emphasizes the social aspects of sensemaking while taking into account various environmental circumstances” (Gorze-Mitka & Okrglicka, 2014, p. 405). The authors argue that it is “an appropriate and effective framework for use in qualitative research to structure challenging information management problems, as the domains of the model (complex, complicated, chaotic and obvious) force them to continuously search for new solutions for shaping decision making processes” (Gorze-Mitka & Okrglicka, 2014, p. 402).
Despite the different perspective on information sensemaking, all mentioned viewpoints complement each other, as these perspectives are able to create a better understanding of human sensemaking processes. According to Malhotra (2001, p. 12), “it is critical to understand how information processed through information systems is appropriated by human users and converted into knowledge and resulting action and performance.”Figure 1 summarizes the different perspectives on sensemaking processes in information sciences in a theoretical model, as it contributes to the understanding of different processes that can be encountered for structuring relevant data in the process of clinical decision-making.
2.4 Sensemaking in health data
As illustrated, sensemaking is considered by researchers as a helpful approach to structure information. As described in section 2.1, making sense of health data is an emerging field in which corresponding theories from the perspective of information sciences could assist in structuring self-monitored data to utilize it for treatment planning. However, few studies regarding sensemaking in health data exist (Aselmaa et al, 2017; Martin et al, 2015; Mamykina, 2015; Senier et al, 2018). Moreover, corresponding studies were not applied to the understanding and interpretation of health data by care providers and no evidence is found regarding clinical effects of telemonitoring, nor with regards to the acceptance and interpretation of health data through telemonitoring by healthcare providers (Paré et al, 2007, p. 274). Moreover, it was discussed in section 2.1 that different studies lack accurate representations of the data retrieved from telemonitoring and how it was evaluated in order to perform well-informed decisions (Hardisty et al, 2011; Clark et al, 2007; Maric et al, 2009; Jerant & Nesbitt, 2005; Louis et al, 2003).
As demonstrated, there has been a large request in the emerging field of health data analytics to enhance clinical decision-making. Pressure is rising due to the increasing use of digital technologies that facilitate patients in self-monitoring. Therefore, data discovery tools are required in healthcare settings that influence treatment plans in a desired direction. Accordingly, clinical decisions are traditionally based on derived information from the patient in the doctor’s office from which a treatment plan is created by the healthcare provider. Since more and more data from the patient is digitally gathered outside of the doctor’s office by patients themselves via telemonitoring, more research is needed on this topic.
Therefore, the guiding research question in this study is: To what extent can self-monitored health data enhance clinical-decision making by healthcare providers?
In this study, the following sub questions will be answered to answer the research question: 1. What is the influence of self-monitored data on the daily practice of a healthcare
provider? (section 4.1)
2. What are the implications of self-monitored health data for clinical-decision making? (section 4.2)
3. What insight and knowledge can be derived from self-monitored health data to improve clinical-decision making? (section 4.3)
This study will focus on a healthcare setting in which self-monitored data is already considered as an important asset to improve patient care. Specifically, a case study on Cardiologie Centra Nederland is performed in which a chronic disease is examined. The chronic disease which is examined is a cardiovascular disease called hypertension (high blood pressure).
An unhealthy lifestyle is the main contributor of hypertension which corresponds with rising treatment costs, as the condition concerns around one billion people worldwide (Lawes, 2004; Sayarlioglu, 2013). For this, new techniques are considered in Europe that are able to continuously monitor blood pressure levels in order to manage the condition better and to reduce treatment costs (EHealth Action Plan 2012-2020). One of these techniques is a monitor device which is connected to a mobile application and a health information system. This technique is used by Cardiologie Centra Nederland in a telemonitoring program for patients, called ‘HartWacht’, in order to monitor blood pressure measurements from a distance. In this program, blood pressure measurements are retrieved through the device. Yet, how to utilize rising amounts of measurements to enhance clinical-decision making remains unclear.
3.
Method
This study applies a qualitative focus and is considered an empirical study due to its exploratory and inductive approach to collect and analyse data of only primary data. Derived knowledge emerged from observations and experiences rather than theory to grasp reality aspects. Hence, this study focuses on describing the current situation of the studied phenomenon and on describing studied characteristics for a new context of research. For this, hypothesis are suggested which can be tested in alternative studies. In the following paragraph the research process for this study is outlined in order to discuss the three quality aspects of validity, reliability and generalizability in one of the last sections of this study.
To provide a focus to this research, a literature review was performed. Through the literature review, information was gathered about telemonitoring for chronic disease management and the theoretical grounded approach of sensemaking. This information is used to investigate how healthcare providers deal with self-monitored data in a case study at Cardiologie Centra Nederland. For the case study, semi-structured interviews were held to investigate the influence of self-monitored health data on daily practice and how clinical decision-making can be enhanced by it. Healthcare professionals (cardiologists) are interviewed about their work, the increasing data sets of chronically ill patients and how they deal with this information.
Bryman (2016) identifies semi-structured interviews as one of the most applicable ways to analyze the knowledge of individuals on a certain subject. This method is the best fit for this study because it lets insiders disclose what is the most important information according to themselves. All interviews were conducted in 2018 and were analyzed using the grounded theory approach to derive research themes, by coding the interviews with open, axial and selective codes. The coding table can be found in Appendix D.
For this case study, the chronic condition hypertension (high blood pressure) is analyzed. To explain: “Blood pressure is a measure of the force that the circulating blood exerts on the walls of the main arteries. The pressure wave is easily felt as the pulse - the highest (systolic) pressure (SBP) is created by the heart contracting and the lowest (diastolic) pressure (DBP) is measured as the heart fills. Blood pressure is considered as a continuous variable with mean and standard deviation values” (Lawes et al, 2004, p. 284). Moreover, blood pressure should preferably not rise above SBP 140 / DBP 90 mmHg, as it increases the risk of cardiovascular complications.
The participant selection for the interviews was based on the knowledge and influence of individuals in understanding and interpreting blood pressure measurements. Through this selection, ten healthcare providers with experience in guiding patients through telemonitoring programs, information systems and health data were interviewed. The participant list can be found in Appendix E.
Moreover, a database of 78 patients (figure 2) including 458 blood pressure measurements over six weeks were derived from the health information system of Cardiologie Centra Nederland. The patients were selected based on telemonitoring participation. From this, data was organized and analyzed by the researcher to provide insight of hypertension measurements to illustrate the importance of structuring information for data interpretation.
4.
Results
4.1 What is the influence of self-monitored data on the daily practice of a healthcare provider?
Nowadays, the healthcare providers from Cardiologie Centra Nederland face a patient increase who participate in the telemonitoring program. From this increase, the amount of health data from cardiovascular patients, in this case self-monitored blood pressure measurements, is rising. This increase of measurements influences the daily practice of the physician in the healthcare setting they operate from, as providing care moves from a physical towards a digital environment. However, interview results indicate that it is hard to base a judgement on only digitally derived information (respondent 5, 6, 7, 9). Moreover, healthcare providers are used to acquire information from the patient with physiological senses when providing care in the traditional setting in which the patient sits in front of the care provider. In order to obtain a ‘full picture’ of the patient in a digital setting, a call or email is send to obtain additional information about the disease. Also, results show that digital contact via realtime video contact would be an easy way to obtain additional information, which replaces the feeling of making a judgement of the patient’s disease via physiological senses (respondent 5, 6).
In addition, in the digital environment, action can only be taken on health data. To base any judgement on health data, a protocol is developed for the enactment on blood pressure measurements when the measurements would show deviant values. When this happens, the care provider takes action on the deviant values and tries to find out the reason why the deviant values occurred. The action involves a conversation with the patient by phone about the patient’s lifestyle, as different aspects of someone’s lifestyle influence blood pressure values, like the level of physical activity, the intake of specific nutrients (like salt), (over)weight, the amount of alcohol intake and the level of stress (interviewee 5, 6, 7, 9). However, it is often not clear which lifestyle variables, also called risk factors, influence blood pressure values negatively, which causes confusion during daily practice (respondent 1, 2, 3, 4, 5, 6, 7, 8, 9, 10). Despite the confusion, results show that healthcare providers try to move towards a digital clinical setting in which personalized care can be provided, with attention for self-monitoring including risk factor reduction and shared decision-making (respondent 1, 2, 5).
An interesting note is to mention is that interview results additionally indicate that job satisfaction decreases when highly educated healthcare providers would work from behind the computer reading deviant values and try to discover the corresponding risk factors, as this task is considered less challenging (respondent 6, 8, 10). A respondent emphasized the concern: ‘I think my work is going to be very boring when I have to analyze information of self-monitored data all the time and that’s not the reason why I became a healthcare provider.’
However, lower educated healthcare providers seem to enjoy reading digitally derived information and to act on the data, although it is preferred to combine digital and physical contact with patients to stabilize job satisfaction during daily practice (respondent 3, 6, 7). Accordingly, results show that high educated healthcare providers (cardiologists) argue to move the discovery and meaning of data towards the gatekeeper of the healthcare system: the general physician, as this professional also treats hypertension on a general level. Only when a serious additional complication is discovered from the usual self-monitored data, the general physician should transfer the treating task again to the cardiologist (respondent 4, 9).
4.2 What are the implications of self-monitored health data for clinical-decision making? To make sense of self-monitored data, interview results (1, 2, 3, 4, 5, 6 and 7) state that blood pressure measurements from patients need to contain two main aspects: it should be of a high-volume and the data should be trustworthy. High-volume data means that a patient is expected to measure his/her blood pressure three times a day for a long period of time to influence the care plan in a positive way: when the blood pressure measurements of a patient seem to be alarming, an intervention can be performed based on the data in which unnecessary routine interventions are filtered out.
According to the above mentioned respondents, trustworthiness of measured data means that the patient delivers his/her data through the validated device that is given at the start of the telemonitoring program, the measurements should correspond with the according therapy, the instructions for the patient should be clear and the data should be numerous. This way, it would be possible to extract detailed patient profiles and treatment strategies (respondent 4, 6 and 7). With this respect, one main concern was mentioned; the opportunity to utilize data succeeds or fails with the amount of data that is delivered by the patient; physicians fear that when patients are not seeing a doctor anymore physically, the amount of taken measurements would likely drop. The reason for this is argued by the questioned healthcare providers (respondent 2, 5, 6, 8 and 9) who consider physician-patient contact still as the main activity to direct clinical decisions and a patient’s treatment plan, as they consider the amount of data gathered as a tool to trigger personal contact. However, this argument is counter-reasoned, in which it is argued that physical contact will be much less due to the amount of blood pressure measurements, even when important clinical decisions should be made regarding the patient’s therapy (respondent 1, 3, 7 and 10). Also, results show that the doctor should be available in a ‘command center’ with other care providers like nurses and assistants who use digital media to stay in touch with the patient. This way, important clinical decisions, like a change in medication for hypertensive patients, would still be possible as digital contact is still considered personal.
4.3 What insight and knowledge can be derived from self-monitored health data to improve clinical-decision making?
As stated in 4.2, interview results indicate (respondent 1, 2, 3, 4, 5, 6 and 7) that blood pressure measurements from patients need to contain two main aspects: it should be of a high-volume and the data should be trustworthy. High-volume data means that a patient should at least measure his/her blood pressure three times a day for a long period of time to influence the care plan in a positive way: when the blood pressure measurements of a patient seem to be alarming, an intervention can be performed based on the data in which unnecessary routine interventions are filtered out. An intervention is seen as the most important part of a treatment plan by all interviewees (1 - 10), this entails a conversation with the patient with a care provider who reads and interprets the measured data from a distance. This intervention-conversation consists of six questions and are asked as a result from deviant values that have been measured by the patient, as discussed in section 4.1. Deviant values are values that rise above 140/90 mmHg. When the alarming values, meaning a blood pressure above 140/90 mmHg, is analyzed by the care provider, a protocol is followed: after the deviant values are discussed with the patient, the patient needs to measure his/her blood pressure every week for four weeks. After the four weeks, an evaluation conversation takes place in order to find out whether the measured values have dropped over four weeks time. However, it is not clear which risk factors influence the drop in blood pressure as described in 4.1.
An important note is that the mentioned protocol is seen as a pilot, which can be adjusted by leading healthcare providers and managers when the taken intervention seems to be inefficient. To examine whether this intervention is a promising part of a treatment plan and to
make sense of the blood pressure measurements, patient data of 78 hypertensive patients who are taking part in the telemonitoring program for blood pressure measurements were organized for this study after an intervention took place (figure 3).
Figure 3. Average of 458 blood pressure measurements of 78 hypertensive patients after the intervention In figure 3, the average of 458 blood pressure measurements are shown over four weeks after the intervention took place which involved discussing the deviant values with the patient. By reading this line diagram, one could state at first sight that blood pressure dropped over four weeks from an average of 145/91 mmHg to 138/87 mmHg.
In addition, interview results indicate that diagrams such as shown above would increase the understanding of digitally derived data, specifically for patient profiles (respondent 1). Patient profiles are groups of patients that share the same risk factors for high blood pressure, like overweight and stress. This way, healthcare providers can start with working towards personalized treatment plans to optimize patient care (respondent 1). Moreover, other conditions, like heart failure and arrhythmia could also benefit from clear representations of self-monitored health data (respondent 1, 2, 3, 7, 9, 10). Although these conditions are also considered in the telemonitoring program, results show that data results from heart failure and arrhythmia requires a different response from the healthcare provider.
Accordingly, the two mentioned conditions have a higher risk of complications for the patient on the short term, so in order to keep track of a patient’s health, a clinical decision should be initiated immediately when these conditions would show deviant values (respondents 3, 7, 9).
For hypertension, initiating action is necessary on the long term: when a patient would show deviant values for over a long period of time, the chance of developing cardiovascular complications would likely increase (respondents 4, 5). Therefore, healthcare providers should increasingly watch for ‘trends’ instead of ‘snapshots’ of high blood pressure (respondent 2). An important note to mention is that results indicate that discovering trends from data should be connected to predictive software, also called an ‘assisting decision system’ or an ‘automated protocol system’, from which a ‘risk score’ could be developed. This way, healthcare providers would be able to recognize profiles and deviant trends at a glance in which additional diseases and complications also could be discovered, like sepsis and pulmonary embolism (respondent 1 and 2). Hence, a concern from a respondent with regards to the development of an ‘automated protocol system’ occurred, which involved the following question: ‘Who should take responsibility when the wrong decision is made about the patient’s health: the physician or the software developer who created the system?’This concern is relevant to mention, as it provides room for discussion in future studies.
5.
Discussion and Conclusion
This study aimed to provide insight into the impact and role of self-monitored health data in the process of clinical-decision making by adopting the sensemaking perspective. Hence, the following research question was raised: ‘To what extent can self-monitored health data enhance clinical-decision making by healthcare providers?’
The results from the case study illustrate the importance of structuring information for data interpretation, as it is considered a big challenge to extract valuable data from all measurements to initiate action. Therefore, the research question can only be answered through an explanation of clinical-decision making and in what way self-monitored data could enhance this process.
The current overall model of clinical-decision making (figure 4) is based on clinical decisions that are traditionally taken. Due to the increase of self-monitored data, this model needs to be reconsidered.
According to the results, self-monitored data needs to be structured in order to derive relevant data. As structuring information is not considered as a task in the current process of clinical decision-making, four focus areas are defined for healthcare providers to consider in order to structure relevant data: 1) collecting numerous and trustworthy data from patients, 2) extracting meaningful data in order to filter out irrelevant data, 3) discovering patterns of data to give meaning to the information by visualizing it and 4) perform statistical analysis on the data to allow further action to make evidence-based decisions. These four areas are discussed from the perspectives of the mentioned sensemaking theories in which theoretical and practical implications are reflected.
From section 4.1, it became clear that the delivery of data from blood pressure measurements rises due to the increase of patients who participate in the telemonitoring program of Cardiologie Centra Nederland. Due to the amount of self-monitored data, complexity and uncertainty remains regarding shaping decision-making when the datasets become too large for direct interpretation. For this, the environmental model of Snowden & Boone (2007), called the Cynefin framework (Appendix C), can be used to understand externally derived data. This model could guide this challenge, for which the four main categories could be applied in the process of information sensemaking to aid decision-making: complex, complicated, chaotic and obvious (Snowden & Boone, 2007, p. 4) (Appendix C). The model enables healthcare providers to discover unfamiliar situations. The increase of health data could be considered in the ‘complex’ domain (Appendix C), as it is concerned with an emerging situation to discover patterns from the data that is not already known. “This domain involves the ‘probe-sense-respond’ reasoning, as informative patterns could emerge” (Snowden & Boone, 2007, p. 4). This can be the case when additional complications are noticed from the actual delivered data. However, due to the amount of data, new discoveries can be made which not only concern the intended analyzed disease, as additional complications could be found from the data as well.
Another aspect from the results in section 4.1 which is worth to discuss is the division of tasks from high educated medical professionals to lower educated healthcare providers and to the general physician. Since the model of Snowden & Boone (2007) emphasizes the collective context, this could be an effective approach to structure data, as it “advocates the use for understanding complexity and emphasizes the social aspects of sensemaking while taking into account various environmental circumstances” (Gorze-Mitka & Okrglicka, 2014, p. 402). Results indicated that the digital environment in which healthcare providers have to operate from also have implications on job satisfaction. The social aspects of sensemaking regarding ‘making sense’ of the task that is being transferred should be considered in this context, as new tasks, like analyzing large amounts of self-monitored data, are changing the profession of the healthcare provider. To illustrate, a concern from the results is highlighted: high educated professionals are aware of the advantages of digitization, however, concerns regarding their profession exists in terms of analyzing external data instead of physical contact with a patient. A quote from one the interviewees emphasizes the concern: ‘ I think my work is going to be very boring when I have to analyze information of self-monitored data all the time and that’s not the reason why I became a healthcare provider’.
When taking the task division and social aspects into account in the model of Snowden & Boone (2007), the domain ‘complicated’ is best represented, as it indicates that a setting is familiar, although the outcome of it is unfamiliar. This reasoning process is called ‘sense-analyze-respond’ according to Snowden & Boone (2007, p. 4), in which the healthcare provider is able to assess information, analyze it and chooses the best practice to make a decision based on his/her expertise. For example, when a task is transferred from a cardiologist to the general physician, the latter professional should apply the ‘sense-analyze-respond’ reasoning process (Snowden & Boone, 2007, p. 4) to make sense of the newly transferred task, in which data needs to be structured to reason to the right clinical-decision, for example in the occasion when medication should be adjusted.
In section 4.2, it became apparent that data should contain two main aspects to extract valuable knowledge from the information: self-monitored data should be of high-volume and it should be trustworthy. To reach the level of high-volume and trustworthiness, results indicated that healthcare providers expect from patients that they should at least measure their blood pressure three times a day and that patients use the validated device. This process to motivate patients to reach high-volume and trustworthy data corresponds with the concept of individual information sensemaking of Klein et al (2006). According to the researchers, “this process to structure information requires an individual mental model, or a frame, to represent external data. A frame shapes and defines relevant data and sensemaking could elaborate the frame by adding details, questioning the frame and doubting explanations” (Klein et al, 2006, p. 88).
In order to create a frame for a patient to realize high-volume and trustworthy data, healthcare providers should elaborate more on the purpose of measuring when explaining the concept of telemonitoring to the patient. Then it would be more likely that the expectations of the healthcare provider are met. Yet, a discrepancy can occur when the expectations of healthcare providers aren’t met when the patient would deviate from measuring three times a day. From this, when data isn’t numerous, no clinical-decisions can be performed as the data would also not be considered trustworthy. In this case, the frame is questioned in which another important concern can be added from the results: healthcare providers fear that when patients do not see any doctor anymore physically, the amount of measurements would likely drop. This concern is strengthened by the results from the mentioned systematic review in the literature review regarding telemonitoring compliance of patients, which suggests that patients with hypertension are part of the group conditions that comply the least to telemonitoring programs (Pare et al, 2007), while another study indicates that hypertensive patients are most eager to self-monitor (Huygens et al, 2017). Nevertheless, management and healthcare providers should reconsider their expectations towards hypertensive patients and their measurement behavior and try to find a solution in order to motivate patients to measure their blood pressure in in a consistent way. This way, the individual mental model or frame is improved and structured by details and explanations which is relevant for continuous patient care.
The most important results from section 4.3 which involved data from 78 patients are best discussed by the theoretical lense of Russell et al (1993), who defined the concept as “a process of searching for a representation and encoding data in that representation to answer task-specific questions, like decision-making and problem-solving” (Russell et al, 1993, p. 269). Therefore, the theories of Snowden & Boone (2007) and Klein et al (2006) do not apply fully in this situation, as Russell et al (1993), focus on the meaning of data for collective understanding, and does not emphasize individual sensemaking. In a knowledge-intensive data setting like described in 4.3, it is necessary to utilize the perspective of Russell et al (1993) for information sensemaking to guide all healthcare providers involved in clinical decision-making. From the results in 4.3, it became clear that an intervention is seen as the most important part and task of a treatment plan, as it can steer the intended therapy for the patient. However, in order to know when to perform any action to benefit the patient based on digitally derived data, encoding data for clear representations is needed.
This process is described by Russell et al (1993) “as a learning loop, which is demonstrated as follows: 1) Search for representations by sampling information, 2) Instantiate representations by identifying/sampling information of interest, 3) Shifting representations and redefine goals 4) Use the task specific information and repeat the process” (Russell et al, 1993, p. 411). The first phase (sampling information) can be seen to assemble blood pressure measurements from the patients who are involved in the telemonitoring program ‘HartWacht’. In this case, 458 blood pressure measurements from 78 patients were organized and analyzed. The diagram which followed (figure 2) assisted in understanding the information, for which Russell et al (1993) argue that sensemaking is most concerned with structuring data in a way it can be utilized. When data cannot be used, the corresponding research goals need to be refined.
The diagram (figure 2) indicates that the average blood pressure dropped over four weeks from 145/91 to 138/87, from which one can argue that the intervention is a positive action to execute for hypertensive patients. Yet, to prove whether the intervention is effective, a statistical analysis on the derived sample should be performed. Hence, “misinterpretations may emerge once inappropriate research strategies are applied and it's thus necessary to contemplate the study objectives and therefore the variety of information obtained in clinical investigations before any measure is applied” ( Prashanth, 2011, p. 548). From this, Russell’s learning loop can be explained: information is sampled, the type of data is understood and goals are refined when misinterpretations occur when applying inappropriate measures.
An appropriate measure that should be considered is the repeated measures analysis-of-variance (ANOVA), which is a method that can be used for analysing treatment effects in cardiovascular research (Cleophas et al, 2009). To clarify, “a repeated measures design
is one in which multiple, or repeated, measures are made on the same experimental unit serially in time, such as weekly blood pressure measurements” (Sullivan, 2008, p. 1238).
An experimental unit is in this case the group of 78 patients that participate in the telemonitoring program for which two variables are considered: the independent variable is time and the dependent variable is blood pressure. The test ‘one-way repeated measures ANOVA’ would be applicable, as it shows developments in mean averages over time. The test involves two study designs: 1) examining developments in averages over three or more stages in time or 2) examining developments in averages under three or more different circumstances. The first one can be used to examine the outcome of the 4-week intervention on blood pressure on three different stages (1st week, 2,5 weeks, 4 weeks) (Lund Research, 2018).
In this respect, a one-way repeated measures ANOVA hypothesis can be defined regarding this study:
The null hypothesis (Ho) = average blood pressure is equal at the three stages in time (1st week, 2,5 weeks, 4 weeks).
The alternative hypothesis (Ha) = average blood pressure is non-identical at one or more stages in time (1st week, 2,5 weeks, 4 weeks).
The second study design of one-way repeated measures ANOVA examines developments in averages under three or more different circumstances (Lund Research, 2018). Accordingly, this design could be used to discover patient profiles. For example, ‘type of treatment’ could be the independent variable with ‘risk factors 1 (like smoking)’, ‘risk factors 2 (inactivity)’ and ‘risk factors 3 (weight)’ as the three levels of the independent variable. As mentioned in section 4.3, when such patient profiles can be discovered and tested, interview results indicate that this data should be connected to predictive software, functioning as an ‘assisting decision-system’, in order for healthcare professionals to recognize deviant values at a glance and to move towards ‘personalized medicine’ due to the representation of risk factors per patient group. This way, clinical-decision making could be supported by technology.
To conclude, clinical-decision making can be enhanced by self-monitored data by structuring the data in different ways. It became clear from the reflection that the ‘complex’ and ‘complicated’ domains of the Cynefin framework of Snowden & Boone (2007) (Appendix C) are relevant to structure external data from patients. Moreover, to make sense of self-monitored health data, information from patients should be numerous and trustworthy. In order to structure data in this context, the individual information sensemaking model of Klein et al (2006) could provide guidance by creating a mental model / frame for the patient. This frame can be used as a reminder to deliver numerous and trustworthy data, as it forms the basis of clinical decision-making in the future. Moreover, the theoretical lense of Russell (1993) on information sensemaking is the most suitable to structure information in order to serve a task. In this case, it could be best applied to extracting meaningful knowledge from numerous blood pressure measurements. In order to obtain meaningful knowledge to serve the task ‘clinical-decision making’, statistical analysis has to be performed in which the ‘one-way measures ANOVA’ is a possible candidate to prove whether an intervention is effective by hypothesis testing. When an intervention is proven effective, further development towards a ‘decision system’ can be realized in which clinical-decisions can be made through automated protocols. Hence, the responsibility towards the patient’s health should be discussed regarding decisions from an automated protocol system or a healthcare provider.
From the discussion and conclusion, it became apparent that information structuring can assist in understanding and interpreting self-monitored data. To emphasize, a schematic overview is created in order to cope with self-monitored data by highlighting questions throughout the clinical decision-making process (figure 5). However, structuring and understanding such data is a challenge for all healthcare providers involved in utilizing information, as it is an emergent field in which no ‘best practice’ exists yet.
6.
Limitations and Future Research
As this research used an inductive research approach and applied qualitative research methods, theory contribution to the field of sensemaking in health data was performed by showing how healthcare providers use self-monitored data currently and how they could structure it to enhance clinical-decision making. To explain the limitations regarding this study, the three quality assessment concepts of validity, reliability and generalizability in qualitative research are addressed based on an article which was published in the Journal of Family Medicine & Primary Care by Leung (2015). In this article, the author describes that validity in qualitative studies related to health sciences, “means that the ‘appropriateness’ of the tools, processes and data are assessed” (Leung, 2015, p. 325). The ‘appropriate’ choice of methodology for this study (grounded theory approach) found several themes that comply with the context of the conducted research. Accordingly, purposeful sampling was used to illustrate the situation regarding sensemaking of health data at Cardiologie Centra Nederland: only healthcare providers who are familiar with the mentioned telemonitoring program were interviewed as well as only patients who participated in the telemonitoring program for hypertension were analyzed. One limitation occured, as only the condition of hypertension was discussed and therefore, no full picture of the whole telemonitoring program could be shown, as heart failure and arrhythmia were not analyzed. Therefore, future research should focus on studying the data from the two conditions as well to prove overall effectiveness.
For this study to assess its reliability, processes and results should be replicable. However, “this is challenging and epistemologically counter-intuitive, as data could differ in richness” (Leung, 2015, p. 326). Yet, according to Leung (2015), reliability for qualitative research should focus on consistency, in which study results are allowed to variate marginally. As data for this study was extracted from the eccentric sources, triangulation was performed by comparing and verifying the data in context with peer researchers (Leung, 2015). Triangulation with peer researchers is another limitation, as this study could only be fully understood by peer researchers working in the same context. However, reliability was also impacted on a positive note, as the interview data and quantitative data from patients created a more comprehensive picture of the studied phenomenon. Hence, for future research, statistical analysis on the derived hypothesis should be tested in order to prove whether the intervention is effective for hypertension.
As a last remark, generalizability is assessed. This study is considered as empirical research and since the study is conducted on a explicit topic in a distinct population and context, the results are considered not generalizable. However, future research should focus on the generalizability of the studied topic by proving statistical associations between the suggested variables.
7.
References
Aceto, G., Persico, V., & Pescapé, A. (2018). The role of Information and Communication
Technologies in healthcare: taxonomies, perspectives, and challenges. Journal of Network and Computer Applications.
Ahern, D. K., Kreslake, J. M., & Phalen, J. M. (2006). What is eHealth (6): perspectives on the evolution of eHealth research. Journal of medical Internet research, 8(1).
Andersen, T., Bjørn, P., Kensing, F., & Moll, J. (2011). Designing for collaborative interpretation
in telemonitoring: Re-introducing patients as diagnostic agents. International journal of
medical informatics, 80(8), e112-e126.
Aselmaa, A., van Herk, M., Laprie, A., Nestle, U., Götz, I., Wiedenmann, N., ... & Jolnerovski,
M. (2017). Using a contextualized sensemaking model for interaction design: A case
study of tumor contouring. Journal of biomedical informatics, 65, 145-158.
Bernell S and Howard SW (2016) Use Your Words Carefully: What Is a Chronic Disease? Front.
Public Health 4:159. doi: 10.3389/fpubh.2016.00159
Bodenheimer, T., Lorig, K., Holman, H., & Grumbach, K. (2002). Patient self-management of
chronic disease in primary care. Jama, 288(19), 2469-2475.
Bryman, A. (2016). Social research methods. Oxford: Oxford University Press.
Chen, S. H., Tsai, Y. F., Sun, C. Y., Wu, I. W., Lee, C. C., & Wu, M. S. (2011). The impact of
self-management support on the progression of chronic kidney disease—a prospective
randomized controlled trial. Nephrology Dialysis Transplantation, 26(11), 3560-3566. Clark, R. A., Inglis, S. C., McAlister, F. A., Cleland, J. G., & Stewart, S. (2007). Telemonitoring
or structured telephone support programmes for patients with chronic heart failure: systematic review and meta-analysis. Bmj, 334(7600), 942.
Cleophas, T. J., Zwinderman, A. H., & van Ouwerkerk, B. M. (2009). Methods for analysing
cardiovascular studies with repeated measures. Netherlands Heart Journal, 17(11), 429-433.
Dedding, C., Van Doorn, R., Winkler, L., & Reis, R. (2011). How will e-health affect patient
participation in the clinic? A review of e-health studies and the current evidence for
changes in the relationship between medical professionals and patients. Social science
& medicine, 72(1), 49-53.
European Commission. (2012). E-Health Action Plan 2012-2020. Brussels. Retrieved from: https://ec.europa.eu/digital-single-market/en/news/ehealth-
action-plan-2012-2020-innovative-healthcare-21st-century
Gorzeń-Mitka, I., & Okręglicka, M. (2014). Improving decision making in complexity environment. Procedia Economics and Finance, 16, 402-409.
Hardisty, A. R., Peirce, S. C., Preece, A., Bolton, C. E., Conley, E. C., Gray, W. A., ... & Elwyn,
G. (2011). Bridging two translation gaps: a new informatics research agenda for
telemonitoring of chronic disease. International journal of medical informatics, 80(10),
Hassan, Q. F., & Madani, S. A. (2017). Internet of Things: Challenges, Advances, and Applications. Chapman and Hall/CRC.
Houben, S., Frost, M., & Bardram, J. E. (2015, February). Collaborative affordances of hybrid
patient record technologies in medical work. In Proceedings of the 18th ACM
Conference on Computer Supported Cooperative Work & Social Computing (pp.
785-797). ACM.
Huygens, M. W., Swinkels, I. C., de Jong, J. D., Heijmans, M. J., Friele, R. D., van Schayck, O. C., & de Witte, L. P. (2017). Self-monitoring of health data by patients with a chronic disease: does disease controllability matter?. BMC family practice, 18(1), 40.
Jerant, A. F., & Nesbitt, T. S. (2005). Heart failure disease management incorporating telemedicine: a critical review. JCOM, 4(12), 4.
Klein, G., Moon, B., & Hoffman, R. R. (2006). Making sense of sensemaking 2: A macrocognitive model. IEEE Intelligent systems, 21(5), 88-92.
Lawes, C. M., Vander Hoorn, S., Law, M. R., Elliott, P. A. U. L., MacMahon, S., & Rodgers, A.
(2004). High blood pressure. Comparative quantification of health risks: global and
regional burden of disease attributable to selected major risk factors. Geneva: World
Health Organization, 281-390.
Leung, L. (2015). Validity, reliability, and generalizability in qualitative research. Journal of
family medicine and primary care, 4(3), 324-326.
Louis, A. A., Turner, T., Gretton, M., Baksh, A., & Cleland, J. G. (2003). A systematic review of
telemonitoring for the management of heart failure. European journal of heart failure,
5(5), 583-590.
Lund Research (2018). Repeated measures ANOVA. Retrieved July 12, 2018, from https://statistics.laerd.com/statistical-guides/repeated-measures-anova-statistical-guide. php
Malhotra, Y. (2001). Expert systems for knowledge management: crossing the chasm between information processing and sense making. Expert Systems with Applications, 20(1), 7-16.
Mamykina, L., Smaldone, A. M., & Bakken, S. R. (2015). Adopting the sensemaking perspective
for chronic disease self-management. Journal of biomedical informatics, 56, 406-417. Maric, B., Kaan, A., Ignaszewski, A., & Lear, S. A. (2009). A systematic review of
telemonitoring technologies in heart failure. European journal of heart failure, 11(5),
506-517.
Martin, G. P., McKee, L., & Dixon-Woods, M. (2015). Beyond metrics? Utilizing ‘soft intelligence’ for healthcare quality and safety. Social science & medicine, 142, 19-26.
McWilliam, C. L., Kothari, A., Leipert, B., Ward-Griffin, C., Forbes, D., King, M. L., ... &
Oudshoorn, A. (2008). Accelerating client-driven care: pilot study for a social
interaction approach to knowledge translation. CJNR (Canadian Journal of Nursing
Research), 40(2), 58-74.
Meystre, S. (2005). The current state of telemonitoring: a comment on the literature.
Telemedicine Journal & e-Health, 11(1), 63-69.
Paré, G., Jaana, M., & Sicotte, C. (2007). Systematic review of home telemonitoring for chronic
diseases: the evidence base. Journal of the American Medical Informatics Association,
Peel, E., Douglas, M., & Lawton, J. (2007). Self monitoring of blood glucose in type 2 diabetes: longitudinal qualitative study of patients' perspectives. Bmj, 335(7618), 493.
Piwek, L., Ellis, D. A., Andrews, S., & Joinson, A. (2016). The rise of consumer health wearables: promises and barriers. PLoS Medicine, 13(2), e1001953.
Prashanth, G. P. (2011). Comparison of oscillometric blood pressure measurement by two
clinical monitors. Indian journal of anaesthesia, 55(5), 548.
Russell, D. M., Stefik, M. J., Pirolli, P., & Card, S. K. (1993, May). The cost structure of
sensemaking. In Proceedings of the INTERACT'93 and CHI'93 conference on Human
factors in computing systems (pp. 269-276). ACM.
Sayarlioglu, H. (2013). Rational approaches to the treatment of hypertension: modification of
lifestyle measures. Kidney international supplements, 3(4), 346-348.
Senier, L., Smollin, L., Lee, R., Nicoll, L., Shields, M., & Tan, C. (2018). Navigating the
evidentiary turn in public health: Sensemaking strategies to integrate genomics into
state-level chronic disease prevention programs. Social Science & Medicine.
Sharma, N. (2006). Sensemaking: Bringing theories and tools together. Proceedings of the
American Society for Information Science and Technology, 43(1), 1-8.
Snowden, D. J., & Boone, M. E. (2007). A leader's framework for decision making. Harvard
business review, 85(11), 68.
Sullivan, L. M. (2008). Repeated measures. Circulation, 117(9), 1238-1243.
Tan, S. Y., & Chan, T. (2016). Defining and conceptualizing actionable insight: a conceptual
framework for decision-centric analytics. arXiv preprint arXiv:1606.03510.
Tiffen, J.M., Corbridge, S.J., & Slimmer, L.W. (2014). Enhancing clinical decision making:
development of a contiguous definition and conceptual framework. Journal of
professional nursing : official journal of the American Association of Colleges of
Nursing, 30 5, 399-405.
Weick, K. E. (1995). Sensemaking in organizations (Foundations for organizational science).
Thousands Oaks: Sage Publications Inc. 1-18.
Weick, K. E., Sutcliffe, K. M., & Obstfeld, D. (2005). Organizing and the process of sensemaking. Organization science, 16(4), 409-421.
Whitten, P. S., Mair, F. S., Haycox, A., May, C. R., Williams, T. L., & Hellmich, S. (2002).
Systematic review of cost effectiveness studies of telemedicine interventions. Bmj,
324(7351), 1434-1437.
WHO. Noncommunicable Diseases. (2016). Retrieved from:
http://www.who.int/topics/noncommunicable_diseases/en/
Zhang, P., & Soergel, D. (2014). Towards a comprehensive model of the cognitive process and
mechanisms of individual sensemaking. Journal of the Association for Information
8.
Appendix
A. Data sensemaking model (Russell et al, 1993)
Russell, D. M., Stefik, M. J., Pirolli, P., & Card, S. K. (1993, May). The cost structure of
sensemaking. In Proceedings of the INTERACT'93 and CHI'93 conference on Human
Appendix B. Data/frame model (Klein et al, 2006)
Klein, G., Moon, B., & Hoffman, R. R. (2006). Making sense of sensemaking 2: A macrocognitive model. IEEE Intelligent systems, 21(5), 88-92.
Appendix C. The Cynefin framework of Snowden & Boone (2007)
Snowden, D. J., & Boone, M. E. (2007). A leader's framework for decision making. Harvard business review, 85(11), 68.
