Method

We analyzed our notes taken during fieldwork and interview transcripts using an affinity mapping exercise (Kawakita 1991) and evolving design themes clustered according to their similarity, dependence and proximity of relationship. Themes were identified from the body of evidences gathered from the field and used for the ideation of a caregiver assistive system or tool. Some of the key themes identified are motivation, persuasion strategies, monitoring and vigilance challenges, information flow, role reversals and conflicts. Patient monitoring and vigilance emerged as a key challenge influencing patient-caregiver relationship as well as caregiving dialogue exchanges, and conflicts. Success of persuasion strategies strongly depended on the effectiveness of monitoring. This motivated the design, implementation and testing of iSwear, a wearable device for patients with chronic illness, which can send messages to caregivers about patient activity related to food & medicine intake. An initial exploratory evaluation was conducted with 3 families, where iSwear was given to these patients for a week. We monitored usage patterns through patient-caregiver sms/call logs and followed up with in-person interviews in the homes of the families. We faced limitations in time to extend our probes during the pilot but the period afforded an intimate view of caregiving routines that formed around the technology probe and the challenges thwarting a more successful adoption of technologies for caregiving. We discuss the findings from our ethnography in detail in the next section.

Design of iSwear

This section will explain the design of the CDM communication and monitoring system ‘iSwear’. primarily informed by the ethnographic insights derived from caregiving situations.

iSwear is a system focalizing caregiving as a set of key practices in persuasive heath care delivery in family settings. Our aim was not to build a prototype with full capability and accuracy, but to look at usability and acceptance of such a system in a familial caregiving scenario. Our ethnography of familial caregiving revealed caregiver-devised patient monitoring and vigilance strategies focused unduly on the need for assurance and pervasive communication with the patient. With iSwear we aimed to aid some of these strategies while placing the caregiver as the central actor in caregiver-patient communication practices.

Figure 1. iSwear device & its components

We considered measuring physiological parameters such as heart rate, ECG, EEG, activity and food intake, to measure and manage multiple chronic conditions. Considering the focus of our study and caregiver intervention areas, we narrowed it down three parameters. These are 1. Measure of daily physical activity 2. Time of medicine intake 3. Time and portion size of food intake. A wearable system, iSwear, was designed to measure these physiological parameters. Figure 1 shows the complete component diagram of iSwear. iSwear consists of three main sensors, the accelerometer, tilt sensors and an RFID transceiver. These were used to measure the above-mentioned patient activity data. RFID tags placed on the medicine bottle helped to inform about medicine intake. The data collected from the sensors of the iSwear is converted into meaningful information about patient activity with the combination of different data streams.

Figure 2. Message variations received at Caregiver’s Phone

The GSM module attached to the device consisted of a specialized modem with SIM card, for sending SMS to the caregiver at a frequency of four times a day (i.e. 9am, 12pm, 3pm and 6pm). Figure 2 shows different variations of the SMS designed for actions of medicine intake, walking, and food intake respectively.

Design of Caregiving Ontology – Caregiver as Agent

To capture caregiver activity or responses on the received system messages, a caregiving ontology(CO)was designed. We use the standard OLW (Web Ontology Language) to create entities- in our case, caregiver data about existing caregiver – patient social and kinship profiles and their contextual information (such as experience of caregiving and locational availability) were created as ‘caregiver’ entity to the CO. The system would query the entities and their inter-relationship to trigger notification for any care activity related reporting. Patient entity values were analyzed in runtime, and contextual information about patient activity and medicine adherence were added to the ontology as another set of entities.

Patient-end disease information and contextual information about patient activity and medicine adherence were added to the ontology as object values. Caregiver data about existing caregiver profiles and their contextual information (such as the experience of caregiving and locational availability) are added as ‘caregiver’ entity to the CO. The CO would store all the caregiver responses as an object value for the entity caregiver.

A simple SPARQL (Simple protocol and RDF Query Language) pseudo-code run on the patient entity for the CO is used to understand the required caregiver response, and thereby send an actionable notification to the caregiver. The queried activity is then sent to the different caregivers based on their type using the messaging system.

SELECT ? cgType ? cgFeedback ? cgIntervention ? Recommendation
Where
{
cgIntervention : Activity
cgType : CGc-s
cgFeedback : hasRecieved ? hasResponded ? hasActed ?}

Once the intervention type and the nature of the caregiver have been identified, predefined message templates are used to create recommendations. These recommendations are pushed as notification to the caregivers. The caregiver messaging flow and their responses to these notifications informs us about the real-time human interaction in a caregiving ecosystem. The system places the caregiver as agent in triggering the messaging activity.

EVALUATING THE FEEDBACK DATA ECOSYSTEM

We conducted our preliminary field study with 3 families (2 conjugal, 1 filial, all in-person caregivers). All 3 patients were above 60 years of age and suffering from type-II diabetes mellitus. iSwear was given to the patients who were asked to wear it for a period of one week from 8am to 8pm. This was the time when these caregivers or patients were generally out at their workplaces. We took care to familiarize and orient users to the device and the nature of our experiment. The RFID tag was stuck to the medicine bottle/strip and dosage timing and frequency was noted. The phone number of the caregiver was noted and set as default number to send SMS through iSwear. An SMS & Call Log Backup application [33] was installed in the caregiver’s phone with their consent. We also gained consent for all phone conversations to be time logged and recorded.

We observed wearable notification response and patient action response through phone logs and wearable logs. Wearable notification response would inform us about how caregivers responded to the received notification and in how much time. The patient action response would inform us about the patient activity once the caregiver has responded to a notification in a certain way, such as making call to the patient. Additionally, at the end of every week, both the patient and the caregiver were interviewed for system feedback.

INITIAL FINDINGS

Our findings are focused on understanding care arrangements in Indian families. We probed on the nature, order and extent of various caregiving activities, health record keeping and information management involved in effective care delivery. We took specific care in understanding family dynamics that shaped and hovered around caregiving activity in order to focus on the caregiver-patient interactions and emotions surrounding the act of caregiving. We observed various forms of persuasion and motivation strategies, ranging from the subtle to the distinct, in a caregiver’s repertory of practices for wellness compliance. We probed for the overt and covert needs that caregivers expressed with the current system of access to health monitoring technologies and CDM aids. Location closeness (i.e. remote and in-person caregiving) also provided us insights on different caregiving needs and concerns.

We represent the trust-persuasion quadrants of caregiving scenarios in Figure 4. These are 1. Filial-Remote (FR), 2. Filial-In Person (FI), 3. Conjugal-Remote (CR), and 4. Conjugal-In Person (CI). This representation helps us understand the aspects of trust and persuasion among filial or conjugal caregivers with their kin. The remote and in-person caregiving setting represents the location closeness between caregiver and the patient. While experience represents the number of years spent with the patient as a caregiver, we also found the parameter of experience complimenting the trajectory work representation (Strauss et.al 1985) as caregivers gain in understanding of the patient’s explicit and implicit needs with experience. With gain of experience in longitudinal chronic caregiving, the lower trajectory work becomes part and parcel of a caregiver’s daily lifestyle. However, the same representation may not be true in case of acute episodes, which requires higher trajectory work.

We explain the aspects of trust, persuasion and roles in these quadrants in the following sections. The quadrants themselves are informed by the affinity analysis we undertook to process ethnographic data. We focused on the specifics of the remote and situated care and the emotional and interactional differences among and between filial and conjugal caregivers. These relationships have been mapped in the form of the proposed quadrant.

Figure 4. Trust-Persuasion Quadrant in Caregiving Scenarios