The HMA was developed using the planning, analysis, design, implementation, and evaluation phases of the Web-Roadmap methodology of information science [6].

The data structure is documented in an entity-relation diagram (Figure 5). Separate tables were prepared for the data and knowledge domains, and the data domain tables were further divided according to the change cycle of data values.

By linking the 8 rules for setting the target and the 20 rules for tailored recommendations, 8 algorithms were developed. The algorithms were used to set targets for each intervention item and provide tailored recommendations based on the target and input data. As an example, the algorithm for evaluating the waist circumference is shown in Figure 6. When a user inputs his/her waist circumference, depending on the user’s sex and target waist-circumference evaluation (T7 or T8), the evaluation result (E19 or E20) and recommendation (R19 or R20) will be presented to the user.

In the case of a female user whose waist circumference is 92 cm, a recommendation composed of T8 (“Your target waist circumference is 88 cm”), E20 (“Your waist circumference is 92 cm, 4 cm over the target”), and R20 (“For a hypertension patient, maintaining the proper waist circumference is necessary to lower blood pressure. Reduce your waist circumference to 88 cm”) will be received.

The user interface screens were designed around a menu containing essential functions extracted in the analysis phase. In total, 34 user interface screens were designed.

The HMA was developed using the Android SDK Platform 4.4.2 Java Development Kit and Eclipse. The database management program was developed using MYSQL. The system works on mobile phones running the Android operating system versions 2.3 to 4.4.

The HMA developed in this study was implemented with 5 basic menus at its core (Textbox 2).

In this evaluation, 2 assistant professors teaching adult health nursing and 1 senior nursing researcher participated. All 3 evaluators had more than 10 years of clinical practice experience. There were no inconsistencies in the broader recommendations produced by the evaluators and the HMA. However, the evaluators proposed the following three additional detailed recommendations: (1) adding the option to input the daily amount of alcohol intake in addition to the existing weekly amount of intake, (2) more specific comments on sodium intake, and (3) adding “get advice from a professional” to stress management. These three recommendations were added to the HMA.

The mobile heuristics was evaluated by 5 nursing informatics, medical informatics, and computer programming experts with PhDs. They flagged 33 usability-related problems related to the mobile heuristics items, of which 3 problems that were related to a severity score of 4, and 7 problems that were flagged by more than 2 evaluators were revised after reflecting on the evaluators’ comments.

Initially 38 study participants were recruited, of which 29 (76%) participated in the medication adherence survey and 19 (50%) participated in the perceived usefulness and user satisfaction surveys. The mean age of the subjects was 56.0 years, with the highest percentage (48%, 14/29) of them being in the age group of 50-59 years. Most (66%, 19/29) of the participants were male, 62% (18/29) were educated to higher than a bachelor’s degree, 62 % (18/29) were employed full-time, 62% (18/29) were religious, 93% (27/29) were married, and 72% (21/29) were living with their spouse and children.

Hypertension-related characteristics of the study participants are presented in Table 5. A mean of 106.2 months had passed since the participants were diagnosed with hypertension, and 101.2 months since they had started taking hypertension medication. Most of the participants (79%, 23/29) reported taking a hypertension medication, and 90% (26/29) took medication once a day. Of the participants, only 5 (17%, 5/29) experienced side effects of hypertension drugs, 55% (16/29) had a family history of hypertension, and 66% (19/29) had previously received education about hypertension. The two main sources of hypertension information were hospitals (52%, 15/29) followed by the Internet (28%, 8/29).

In the knowledge extraction phase, it was necessary to decide from which CPG, and to what extent knowledge needs to be extracted, in order to form the knowledge base for the HMA. After selecting multiple CPGs, knowledge on hypertension diagnosis and drug therapy required for health care providers to make decisions was excluded. Instead, the following knowledge relevant to a patient’s self-management of hypertension was included: target blood pressure, lifestyle management intervention, and evidence for tailored recommendations.

The system development stage involved converting the knowledge extracted from the CPG into a computer-interpretable form so that it could be applied to the system. Data and rules were extracted from the text-formatted CPG, then coded so that both the user and the system could understand them. The process of making the CPG knowledge understandable to computers requires cooperation between clinical experts and computer developers [26]. This process would be more efficient if a knowledge modeling tool and system development methodology, specialized for the development of a CPG-based mobile system, are developed.

The evaluation of the knowledge-base accuracy indicated that the recommendations proposed by the experts were more detailed than those suggested by the HMA. Since the accuracy of the knowledge base was evaluated after the HMA had been developed, coding needed to be performed twice to reflect this change. This extra coding step could be avoided if the expert evaluation of the knowledge base was performed in the algorithm development stage.

In the mobile heuristics evaluation, diverse expert opinions were sought from evaluators with diverse backgrounds, such as nursing, medical informatics, and computer programming. The various evaluators made different numbers of comments on different heuristics items. Since the capability of each evaluator acts as an important variable in mobile heuristics evaluation [27], including diverse experts from several fields facilitated the identification of usability problems.

The users found the HMA useful. According to Davis [12], there is a positive correlation between system acceptance and perceived usefulness. Although system acceptance was not directly measured in the present study, we assume that users will have a positive attitude toward using an app for measuring hypertension.

Users rated that they were highly satisfied with the HMA’s blood pressure and medication recording functions. This is in accordance with a previous study on a chronic disease management mobile app [28], which found patients suffering from chronic diseases to be very interested in its recording functions. Recording functions are especially useful for those who need to record and monitor their health information on a chronic basis.

The education provided about medicine information also received a high score of 3.8 in the user satisfaction survey, which is attributed to the HMA being connected to an external database and provides the latest information on hypertension medicine in real-time, including visual information. The quality of information is an important factor in determining the success of information systems [29]. This is consistent with the present study where users were satisfied with the well-constructed knowledge base because it provided quality information on medicine.

On the other hand, users showed a lower user satisfaction for the data sending function. The purpose of this function was not only to store, but also to send the users’ records to the users’ health care providers for feedback. However, because the HMA manages the users’ blood pressure records, provides tailored recommendations, and offers analysis of their medications, the users may have not found it necessary to send their records to their health care provider, resulting in a lower score for the data sending function. The HMA was rated as useful but inefficient for hypertension care in the survey of the perceived usefulness. The users were satisfied with the monitoring functions themselves (eg, blood pressure and medication recording), but rated the efficiency aspect of the HMA poorly due to the additional act of recording, since the users previously did not have to do this.

This study found that the HMA improved medication adherence. However, an attrition bias [30] may have been present since 9 of the 38 participants (24%) dropped out during the 4-week study period. Those who answered the posttest survey may have represented a high-adherence group, and thus not representative of the entire survey cohort. Results that are more accurate are likely to be obtained in a future study that includes both control and experimental groups with similar initial adherence rates.

Implications of the study findings are limited, as we did not measure the direct effects of the HMA on parameters such as actual medication adherence, lifestyle improvement, and change in blood pressure. A future study should measure the actual medication adherence using digital adherence measuring equipment [31], and evaluate lifestyle improvements and blood pressure changes over the long term. A future app with wireless medical sensors and technologies for better data collection and transmission will allow this future study possible.

There is the possibility that those with low medication adherence refused to participate in the survey. Thus, it is necessary to explore the reasons for the withdrawals to evaluate selection bias on the medical adherence survey. In addition, it is arguable that the 4-week intervention was too short and the number of participants involved was not sufficient to measure the effectiveness of the HMA. This will certainly introduce bias in the measurements. A future study with a longer intervention period and a larger study cohort is recommended.

Finally, we did not assess patient engagement and use of the HMA for their hypertension management, which should be taken into consideration in future studies.

This study describes the development of a mobile HMA based on CPGs that allows patients with hypertension to conduct appropriate self-management. CPGs were used as the knowledge base of the HMA, and critical functions were constructed to aid 2 factors of hypertension management: lifestyle improvement and drug treatment. The developed HMA was evaluated based on the accuracy of the HMA’s knowledge base and its mobile heuristics, was applied to a cohort of patients with hypertension patients, and evaluated for its perceived usefulness and user satisfaction. The HMA scored 3.7 out of 5 for perceived usefulness, and with respect to user satisfaction, it scored 4.3 for blood pressure recording, 3.8 for medication recording, 3.1 for data sending, 3.2 for alerting, 3.4 for recommending, and 3.8 for educating about medication. The findings also demonstrated that medication adherence increased significantly after using the HMA (P=.001).

This study is significant not only in that it developed a mobile app for hypertension management based on CPGs, but also evaluated the developed system and its effects on patients with hypertension over a 4-week period. Further studies are necessary to determine the direct effects of the HMA on patients with respect to actual medication adherence, lifestyle improvement, and change in blood pressure.