Keywords

1 Introduction

Diabetes mellitus occurs when the body, due to certain circumstances, loses the ability to control the blood glucose level. At this stage, the patient needs to rely on external resources—insulin or oral medication—to bring the blood glucose level closer to the normal values (i.e., 79 - 110 mg/dL). For insulin therapy, in particular, there are two common practices that have been available for some time: the Multiple Daily Injections therapy (MDI) and Insulin Pumping therapy. This research focuses mainly on the MDI therapy, and, specifically, concentrating on the adherence to daily insulin doses.

1.1 MDI Therapy Challenges

MDI (Fig. 1) involves taking a number of insulin shots on a daily basis (e.g., 4 or 5 times per day). MDI therapy depends on the management between two types of doses: Bolus, doses that are taken before meals or sometimes for glucose level adjustment, and Basal, doses that are taken to lower the glucose level between meals and during sleep. Overall, MDI therapy is considered practical and flexible compared to its counterpart, the pump therapy. However, it is also considered highly challenging [1]. In order to maximize the outcomes from MDI therapy, the patients need to maintain a high level of adherence to the insulin medication. Timing, for example, is critical for insulin medication. The patient has to match daily routines (meals, exercise, sleep time, etc.) and insulin onset behavior. Some research has shown evidence of patients suffering from poor glucose control as a result of being incompetent with their insulin medication adherence [2, 3].

Fig. 1.
figure 1

Multiple daily injections

Full size image

1.2 Insulin Glargine

Insulin Glargine is a long-acting type of insulin produced by SANOFI under the brand name “Lantus®”. It can be used by both TYPE 1 and TYPE 2 patients for basal doses. Directions state that insulin Glargine should be taken once daily within a 24-hour period in a daily fixed amount. This is similar to hypertension and cardiac medications.

In this research, we choose Insulin Glargine as an example for our application for two reasons: First, it is compatible with the instrument that we are using within this research, and second, its nature and directions can be verified with the proposed application in this research.

1.3 The Research Goal

This research is utilizing a couple of ubiquitous technologies, such as cloud computing and smart devices, to provide a supporting system for MDI patients. The main objective is to provide a solution that can help maintaining a high level of adherence to the insulin Glargine medication. The research is giving an example in how we can utilize modern technologies to support MDI management. It followed the Wizard of Oz prototyping concepts in order to test the usability of the proposed system among a group of diabetic patients.

2 Related Work

2.1 Intelligent Insulin Pens

The first major upgrade was the creation of insulin pens in 1985 by NOVO NORDISK [4]. They could provide more practical and accurate solutions for MDI, The situation remained relatively unchanged until 2007 when Eli Lilly introduced their HumaPen® Memoir™ model [5]. HumaPen® Memoir™ was introduced with a memory feature that can record last few taken doses. The feature aimed to overcome double dosing, one of the major issues in MDI therapy. This was followed by another model from NOVO NORDISK in 2012, which applied the same feature but with a different concept [6]. Pendiq GmbH was the first company that introduced a full new technological solution for MDI instrument. The new Pendiq® intelligent insulin pen was introduced with some features that could overcome certain issues existing within MDI [7], such as insulin sensitivity and noncompliant dosing behavior.

2.2 Reminder Systems for Medications

After conducting some reviews for approximately 80 applications and other literature review, we found that most of these applications can be used as management tools to keep track of diabetic records; however, few of these applications have features like dosing and carbohydrates calculation (e.g., Rapidcalc app) or dose and blood checking reminders (e.g., Glucose Buddy app). Unfortunately, most of these applications lack intuitive features like automated data entry or data sync between devices [8]. Similarly, a lot of the commercial tools available within the market are not open source and lack cross compatibility between each other as well [9].

In general, recommendations from healthcare firms always suggest matching the medication time, such as insulin Glargine or hypertension medication, with daily routines (e.g., mealtime or bedtime); however, practices suggest also using reminder systems as well in order to act as a backup to the daily routine [10]. There are several methods that have been used as reminder systems. Some examples are calendars, smartphone apps, alarms and text messaging. Nevertheless, studies found that the most popular method among them is the use of smart devices [11]. Results from different research concluded that the use of short messages (SMS), calendars, apps and alarms managed to raise the adherence level among patients [12, 13]; on the other hand, the use of smart devices as reminder systems has its own drawbacks. First, they encourage the users to rely on them rather than promoting habitual behavior. Second, most of the apps and alarms in current smart devices are not customizable and they might not fit with the users’ needs. Lastly, they lack a post-completion check or acknowledgment features. Even if the patient could respond to the reminder correctly, there is an issue of the patient forgetting whether he/she did respond to the notification correctly. Most of the studies, which evaluated the influence of reminder systems on adherence improvements, were relying on patients’ self-reporting. Studies found that self-reporting can be overestimated and liable to alteration and human-errors [14]. For this reason, it is very important to include a reliable post-completion check system that can keep track of taken medications. For example, in the study that applied gamification concepts through the utilization of social media [15], patients are competing against each other in order to keep the highest rank of adherence level within a list. Other research recommends wearable computing and wireless networks to send reminders for medication and refilling and also keeping track of these actions [16].

3 Research Method

Following the above examples, we are proposing similar approaches to create a reminder system for basal MDI doses (Fig. 2). In order to make the reminders effective and encourage habitual directions, we would like to include the following features:

  • Automated-snooze capability repeated within a period of time

  • Automated post-completion check with log data.

  • Cloud-based communication

Fig. 2.
figure 2

Cloud-based reminder system

Full size image

In order to evaluate the interaction between the suggested system and patients, we conducted two usability studies; each one of them lasted 3 weeks with 4 participants in each study (i.e., Total of 6 weeks with 8 participants). Details of the two studies are explained below.

3.1 Design

The two studies were composed of two parts. The first part was an oral interview. The questions focused on personal diabetic management, the use of technology within diabetic practices, and opinions about current technologies. The second part was a practical experiment divided into two parts. The first part of the experiment lasted 10 days. It focused on the interaction of the patient with the usual systems using intelligent insulin pens, and then measured its influence on the medication adherence level. The second part of the experiment also lasted 10 days; however, this time the focus was on the interaction between the patients and the suggested cloud-based reminder system, and then measuring its influence on the adherence to medication. The main equipment used for this experiment were:

  1. 1.

    4 x Intelligent insulin pens – To keep data logs of administered doses

  2. 2.

    4x portable laptops with pre-installed diabetic management software and remote access agent software – To verify the administration of daily doses

  3. 3.

    Smart reminders with cloud feature installed in the patient’s personal devices – The device was pre-set to the patient’s dose time and shared with the administrator through a shared cloud account.

3.2 Participants

Participants (Table 1) were recruited through either representatives from King Fahd Hospital in Jeddah, Saudi Arabia, or through personal communications from the Saudi diabetic community. There were 3 criteria for participant selection. First, the participant had to be between the age of 21 and 75. Second, they had to maintain a good level of physical and mental health. Third, the participant had to be under insulin Glargine therapy (i.e., Lantus®). Any other brands were excluded from this experiment. Finally, the participant had to have at least basic knowledge of using PCs and smart devices (i.e., Apple’ iPad or iPhone, Android based devices, etc.). Any cases that did not meet these conditions were not considered for the study. The main reasons for having a small number of participants were the limitation in the number of equipment and amount of budget, and, also, all the participants were required to continue administration to minimize risk and provide any support in case of any malfunctions with the experiment equipment.

Table 1. List of participants
Full size table

After getting approval from their primary care provider to proceed with the study, participants were provided with consent forms. After agreeing, they were enrolled in the study. All the participants completed the first part, the oral interviews; however, two participants withdrew from the second part of the study due to some concerns regarding the complexity of the intelligent insulin pens. Except for the withdrawn participants, all the other participants were rewarded with almost $ 20 for each week. In case there was any violation within the stated rules, a deduction penalty had to be applied on the given reward. The other withdrawn participants were rewarded with complimentary rewards for their participation with the interviews. All the participants were provided a sufficient amount of insulin Glargine covering the whole period. All the participants were provided with Internet data subscription in their smart devices to cover the experiment period.

3.3 Procedure and Data Collection

One day was dedicated for the introduction of the whole study, the oral interviews and tutorial for conducting the experiment. Participants were taught how to use the intelligent pens for dosing and changing the insulin cartridge; any other features were not covered in this experiment.

For the following 10 days, the first 3 days were considered a trial for the new system. After the three days, the participants were given the option to withdraw if they could not continue. After that, we proceeded with the experiment. Basically, the patients were asked to take their doses in the manner they had been doing. The directions stated that as soon as the participants would take their doses, they had to return the pen to its station and connect it through the provided laptop. This allowed for data collection on a daily basis through a remote agent.

For the next 10 days, the first three days were also set aside as a trial for the new setup. After that, participants were directed to follow the same dosing process as in the first experiment, but this time using the reminder system notifications (Fig. 3). The directions stated that participants should take a dose as soon as the reminder was activated, within 30 min after the reminder alarm. They were also allowed to take the dose within 30 min before the reminder alarm.

Fig. 3.
figure 3

2nd experiment process

Full size image

The administration of the dose was checked remotely through the management software. If the required dose was administered, the reminder (or the snooze) would be disabled, and then a confirmation message would be sent to the patient. If the administration of the dose could not be confirmed, snooze was activated for every 10 min until the confirmation of the dose could be accomplished, or until a period of 30 min would pass. After that, a message would be sent to the patients stating that “the dose was not administered within the appointed time and it should be administered as soon as seeing the sent message”. Please note, for the second round of experiments, all system components were automated from the patients’ prospective. Only the connectivity between the pen and laptop was not. The reason for this was the absence of the supporting technology within the intelligent pens. The doses verification and snooze activation processes were done manually through the administration side (i.e., It was blinded from the patients). Also, all the reminder system control was done totally through the administration side. Any attempt from the patient’s side was considered a violation of the experiment rules. The connectivity between the intelligent pens and the provided laptop had to be working at all times, except during dosing time. As soon as the patient would finish dosing, the pen had to be returned to its designated station. Failing to follow this was also considered a violation of the experiment directions.

After finishing the two experiments, we conducted one more oral interview to collect patients’ feedback regarding the usability of the whole system, and also to understand the patient’s behavior while interacting with the system.

3.4 Scoring

We measured the level of adherence during both experiments. We based these values on SANOFI’s directions for insulin Glargine [17], “once a day” and “within 24-hour”. If the patient managed to take the dose within the assigned period, a full score would be given for that day. If the patient took the dose outside the assigned time, a half point score would be given for that day. If the dose was completely absent during the whole day, no score would be given for that day. We made another measurement to evaluate the level of awareness among patients. We based these values on how fast the patient could respond to the reminder alarm (Table 2).

Table 2. Scale for the patient’s awarness
Full size table

4 Results

4.1 Results from the Oral Interviews

The first question asked about the administration of daily doses. All the patients were taking at least 4 shots per day (i.e., one shot for the basal). All the patients reported no regular mistakes while administering the doses. Two of them were associating their dose time with one daily routine, while two others were using phone alarms to remind themselves about the dose. The others were just trying to keep the doses closer to an assigned time. We asked the participants about the other technologies for insulin delivery, such as insulin pumps. Only two of the participants said they would consider switching to another technology if it would be suitable for them and could offer more advanced features. The others preferred to stay on the MDI therapy. We asked about the experience in using intelligent insulin pens. All the participants reported that they never saw or used these types of pens before this interview. Four participants did not welcome the idea of replacing standard insulin pens with intelligent ones as they were concerned about complexity. The other four participants were willing to switch to these types of pens if they could assure a better administration for insulin doses. We asked about the use of smart devices and their utilization for diabetic management. Only one participant was using smartphones for diabetic management, on an irregular basis, as it required extra work from the patient’s side.

4.2 Results from the Two Experiments

Observed data. Only two patients scored around the middle level in the adherence scale (i.e., above 3.5 in the scale). The other patients maintained a good level of adherence (i.e., above 6 in the scale). After applying the cloud-based reminder system (Fig. 4), we noticed a good improvement with the patients who were scoring low with in the first experiment. For the others, performance was either the same or slightly better.

Fig. 4.
figure 4

Comparison between the adherence levels in both experiments

Full size image

As per the level of awareness (Fig. 5), three patients typically responded immediately after the 1st alarm. Only two cases required more than one alarm (i.e., two or three) to respond to the notification. We had only one participant who kept administering the medication before the alarm issuance (i.e., within 30 min before the alarm).

Fig. 5.
figure 5

Level of awareness

Full size image

Patients’ feedback. We asked the patients about the overall evaluation of the system. All the patients gave positive feedback, especially those with busy schedules and who could not match dosing with daily habits. For those who were able to match doses and routines, they insisted that the system could raise the assurance factor within themselves. One comment pointed out that the system was more reliable than usual habits because daily habits may change. Also, the system has the ability to confirm and acknowledging the administration of doses, while daily habits still depend on a patient’s awareness and self-management. We also inquired about the main reasons for not administering the doses within the assigned time during both experiments. All the comments indicted that they were aware of the situation, but they could not proceed with administration. Either it was because of being outdoors or out of reach, or being involved with some social activities. Only one patient reported an oversleeping case, which prevented both the administration of the dose on time and hearing the alarm as well. When we asked about the replacement of standard systems with this kind of digital system (i.e., the use of intelligent pens along with smart devices), two patients were open to the idea. They appreciated how the digital solution can assure a proper administration of the dose, including the dosing process, but other patients were concerned about the complexity of the intelligent pens. They thought that standard solutions were more practical and faster.

5 Discussion

Judging from the performance of the system among the patients, we found all the people who were associating the dose time with an appointed time or a daily habit had a slightly better performance in the 2nd experiment. Nevertheless, based on the comments, the system kept the patients alerted and raised the assurance factor within themselves. On the other hand, people, who had busy routines, or had some difficulties with following an assigned time, had a better advantage from the system. So we can assume here this type of system can be an alternatively good method for those who are concerned about any sudden changes within their daily routines. However, for those who are maintaining a schedule that prevents them from being strict with their medication time, this kind of system can be essential for them. Since all the confirmation and snoozing processes are done automatically out of the user’s control, it has the potential to keep the patient alerted and prevent missing doses incidents.

The patients appraised the use of regular smart devices rather than using specialized reminder devices. We noticed, within the data, the more the patient would be attached to his\her smart device, the faster his/her response would be. This gave us an idea that, in the future, this kind of system should have a customized level of alerts. For example, people who keep their devices away sometimes might need a shorter period between snooze alarms; similarly, people who suffer from deep sleep might need a louder or more noticeable notification to wake them up. Also, since this alarm is utilizing cloud services, it can be installed within multiple devices. The advantage here is that this kind of setup can overcome problems such as battery outage or inaccessibility.

Finally, while not directly related to our motivations, we noticed some negative feedback regarding the replacement of the standard MDI system with the intelligent pens. The use of intelligent pens was not greatly welcomed by some of our participants. The intelligent pens were criticized for being more complex than the standard pens. Although this feedback might not concern our scope, it does indirectly affect this system. The system requires digital instruments that can exchange data with smart devices for data processing. It cannot be accomplished with standard pens. Future developments of these kinds of pen should find a solution that can maintain a balance between simplicity and advanced operations. Future developments should also focus on creating solutions that would allow smoother communication through Bluetooth and wireless connectivity for data processing purposes.

6 Conclusion and Future Work

This paper proposed a cloud-based reminder system to be used by patients who are following MDI therapy for diabetic management. The features of the system were based on some aspects of the literature review done for this research. In order to maximize the outcomes, we believe the system should maintain an automated snooze feature, a post-completion acknowledgment and cloud-based services. The components of the system were composed mainly from intelligent insulin pens associated with smart devices. The study was divided into two parts: an oral interview and a practical experiment. The practical experiment first tests the regular setup for the patients, and then tests the setup after applying the cloud-based system. The results from the usability study showed there were some improvements after applying the system when comparing both setups. We concluded that the system could be a good alternative for people who rely on associating their dose time with daily routines or regular alarm/notification systems. On the other hand, people whose schedules are very busy would benefit greatly from this kind of system.

We would like to point out that this study had some limitations. First, both the number of participants and time period were insufficient for a complete diabetic study, which relates the glycemic control with the system outcomes. These types of study require a longer period of time (i.e., between 2 to 3 months). Also, collaboration with diabetes related practitioners might be necessary. Second, the study was based on the Wizard of Oz prototyping method; we hope in the future we would able to make a full-automated system, which can be tested and used in an actual/real setting without restrictions or adjusted scenarios.