This study was designed as a prospective single-arm pilot study. The study sample was recruited from consecutive HF patients who were referred to the outpatient clinic of our institute since July 2014. Inclusion criteria were as follows: (1) age ≥18 years; (2) left ventricular (LV) systolic HF diagnosis for >3 months, regardless of etiology according to current guidelines [3,9]; (3) LV ejection fraction (EF) <40% on transthoracic echocardiography of a duration of at least 3 months; (4) New York Heart Association (NYHA) class II or III; and (5) clinical stability including no change in medications for at least the past 4 weeks. Exclusion criteria were the following: (1) hospitalization within 3 months or a condition expected to require hospitalization; (2) a history of acute myocardial infarction or unstable angina within 6 months; (3) a history of significant valvular heart disease requiring surgical or interventional correction; (4) renal replacement therapy; (5) severe hepatic dysfunction defined as aspartate transaminase and alanine transaminase ≥3 times the upper normal limits; (6) malignancy with life expectancy <6 months; (7) childbearing potential or breastfeeding; (8) functional impairment related to severe musculoskeletal or neurological problems; and (9) refusal to participate. The study protocol was in accordance with the Declaration of Helsinki and approved by the institutional review board of our institution. Written informed consent was obtained from all the enrolled subjects.

Patients were educated regarding the full range of HF self-care behaviors, including diet, exercise, and medication adherence. All participants underwent the following assessments at baseline and after 12 weeks of the ICT-based telehealth program: medical history, physical examination, laboratory tests including urine sodium concentration (uNa) and N-terminal prohormone of brain natriuretic peptide (NT-proBNP), 6-min walk test, NYHA functional class, and Minnesota Living with Heart Failure Questionnaire (MLHFQ) scores [17-19]. The study endpoints included the changes in uNa, MLHFQ scores, NT-proBNP, and 6-min walk distance after 12 weeks of intervention. The adherence to the telehealth program was also analyzed, and a good adherence was defined as the use of the program for 100 times or more during the 12-week intervention period. The study flow is illustrated in Figure 1.

The ICT-based telehealth center comprised multidisciplinary team members, including experienced cardiologists, professional nurses, a dietitian, and computer programmers. Participants were instructed to use the ICT-based telehealth program with voice recognition technology (AIMMED Co Ltd) that was developed for HF patients to provide data and to communicate with clinicians more conveniently through ordinary communication channels via their handheld mobile phone or landline. Patients used their voice or touch-tone telephone keypads to enter essential information for the assessment of HF, including daily weight, diet, exercise, medication adherence, overall symptom change, and home blood pressure. Figure 2 illustrates the schematic diagram of the ICT-based telehealth program.

Direct voice input of health-related data was performed via landline or mobile phone through the automatic call and response system. The automatic call system was programmed to automatically dial the participants’ registered phone numbers once daily 5 days per week for 12 weeks, and when a person answered the phone, it asked prespecified questions to collect narrative responses. Additionally, patients were also allowed free calls into the system during the study period to collect health-related data from participants who did not answer the call from the automatic call system. Patients, guided by a series of prespecified verbal questions, were requested to speak or to enter a preset range of inputs, including body weight, blood pressure, heart rate, and HF-related symptoms. Examples of possible answers were also presented for patients’ convenience and to reduce the error of the data collection. To further ensure the accuracy of data, the program provided patients the opportunity to either confirm or correct the initial extracted data. An example flow diagram for voice recognition is shown in Figure 3.

A clinical decision support system was implemented to provide immediate tailored feedback by voice or text messages (short message service, SMS) to patients according to a predetermined algorithm following current guidelines [3,20]. When data entries were outside predefined ranges or symptoms were reported, an automatic phone call was made to patients to confirm the data and to their attending physicians and research nurses to notify them of the occurrence of these events. Patients received timely self-care feedback from their physicians and were monitored to ensure resolution. Attending physicians were able to tailor HF management and adjust clinic visit schedules, taking into account the clinical status of each patient. After an out-of-range value occurred, the choice of whether to modify the limits of the range was at the discretion of the physician, who made the decision weighing the individual risk and benefit. Patients were instructed on the use of the website or mobile app created for this study, which provided individually tailored feedback messages based on data entered by the patients (Multimedia Appendix 1). By analyzing the cumulative data of each individual, patients also received personalized guidance on long-term self-care management of HF via the website or as a mobile report. We summarized the functionalities of the ICT-based telehealth program used in our study compared with previous programs in Multimedia Appendix 2.

Venous blood and urinary samples were obtained from all patients at baseline and after completion of 12 weeks of ICT-based telehealth program. Data including uNa, NT-proBNP, 6-min walk test, and MLHFQ scores were collected to assess the impact of this program on patient outcomes. Specifically, uNa and NT-proBNP were used to assess the adherence to a low-salt diet and the progression of HF, respectively. To assess functional capacity, 6-min walk test was conducted using a standardized protocol between 11 AM to 2 PM on usual medications [21,22]. Briefly, patients were required to perform a 6-min shuttle walk test with markers placed every 25 m. During the 6-min walk test, patients were accompanied by trained research nurses, who were independent of this study, for the safety of participants as well as for the accuracy and reliability of the results. We also used MLHFQ scores to quantify and compare health-related quality of life outcomes before and after the telehealth program [17-19]. The MLHFQ is an established questionnaire, which was developed to measure the subjective influence of HF and HF treatments on an individual’s quality of life, including physical, socioeconomic, and emotional aspects. It comprises 21 items with scores ranging from 0 to 105, and higher scores indicate a lower quality of life in HF patients [17,19].

Descriptive statistics were used to characterize the study population and measurements. Data were presented as numbers and percentages for categorical variables and mean ± standard error for continuous variables. Differences between continuous variables were compared by the Student t test or Mann-Whitney U test for independent samples, and those between categorical variables were analyzed by the chi-square test or Fisher exact test, as appropriate. A P value of <.05 was considered statistically significant. The Wilcoxon signed-rank test was used to compare values before and after ICT-based telehealth program. In addition, subgroup analysis was also performed according to the adherence to intervention.

Of the 40 chronic HF patients who were screened, 31 subjects met the inclusion and exclusion criteria and agreed to participate in this study. Furthermore, 4 patients (4/31, 13%) were dropped from the study because of hospitalization for acute decompensated HF (n=2) and withdrawal of consent (n=2). Two patients hospitalized for acute decompensated HF withdrew within 1 week and 4 weeks of starting the study, respectively, and the cause of HF hospitalization was considered to be a viral infection in both patients. The final analysis was performed in 27 patients (27/31, 87%) who completed the baseline and 12-week measurements. Ten patients (10/27, 37%) showed good adherence to the program, defined as the frequency of its use ≥100 times during the 12-week intervention period. Table 1 describes the baseline characteristics of the total study population and subgroups according to adherence to the intervention. The mean age of the total study population was 63.4 years, and one-third were females (9/27, 33%). Nearly three-fourths had been hospitalized at least once because of HF before the enrollment (20/27, 74%). The number of prior HF hospitalizations per individual was 1.15; 14 patients had 1, 2 patients had 2, and 4 patients had 3 or more previous HF hospitalizations. Before the program started, 70% of patients were already undergoing treatment with diuretics, including mineralocorticoid receptor antagonists (19/27, 70%). There were no changes in HF medications or dosages during the study period. The etiology of 8 patients was ischemia. The baseline mean values of LV ejection fraction, MLHFQ scores, and 6-min walk distance were 30%, 22.2 points, and 408.4 m, respectively. When comparing the baseline characteristics between patients according to adherence to the program, there was no significant difference between the groups. Patients with good adherence tended to have worse baseline parameters, including significantly more frequent hospitalizations related to HF, greater levels of NT-proBNP and uNa, higher MLHFQ scores, and shorter 6-min walk distances, compared with those without (Table 1).

Twenty-one patients utilized the program via their mobile phone or landline, and about 59% (16/27) of them also accessed the website. The accuracy of voice recognition was 93% and 95% in patients using mobile phone and landline, respectively. Of the patients in whom the user experience with the voice recognition system could be evaluated, ≥85% patients (23/27) indicated that they were satisfied or neutral with the service provided, as shown in Multimedia Appendix 3. Patients were interested in the following services: blood pressure (27/27, 100%), followed by body weight (26/27, 96%; 951 cases/week), and medication (26/27, 96%; 776 cases/week). Table 2 shows the utilization according to the method and the components accessed.

In the total study population, there was no significant interval change in laboratory and functional outcome variables after 12 weeks of ICT-based telehealth program (Figure 4): uNa (91.9 to 90.0; P=.45), MLHFQ scores (22.2 to 20.3; P=.18), 6-min walk distance, (408.4 to 410.6; P=.59), and NT-proBNP (288.6 to 318.1; P=.91). When we stratified patients into two groups by adherence to the ICT-based telehealth program, there was a significant reduction in the value of uNa (103.1 to 78.1; P=.01) and a marginal decrease in the MLHFQ scores (27.5 to 21.4; P=.08) in HF patients with good adherence (Figure 5). However, no significant changes were observed in those without good adherence regarding these variables (Figure 6). On the other hand, 6-min walk distance neither significantly improved in patients with good adherence nor in those without (Figure 6). No significant change in NT-proBNP was observed in both groups of patients (Figures 5 and 6). The MLHFQ scores decreased after 12 weeks of ICT-based telehealth program in 18 patients (18/27, 67%). When we assessed the change in the subscores of physical (8 items), emotional (5 items), and socioeconomic (8 items) dimensions of the MLHFQ, although statistically insignificant, patients with good adherence showed numerically greater improvements in all dimensions. Specifically, the improvement in the total score of MLHFQ was largely driven by the decrease in emotional dimension measures among three quality of life dimensions (94%, 17/18 patients). The change in emotional domain scores was −2.1 in patients with good adherence and −0.2 in those without (P=.48). Results in terms of physical dimension measures revealed that the change was −2.5 in patients with good adherence and 0.35 in those without (P=.13). Regarding socioeconomic dimension measures, the change was −1.5 in patients with good adherence and 0.5 in those without (P=.11).

This study is a prospective pilot study examining the possibility of ICT-based telehealth program with voice recognition technology in HF and building up the foundation for further large-scale randomized controlled studies. Our study demonstrated that chronic HF patients with good adherence to ICT-based telehealth program showed improvement in uNa and in symptoms evaluated by the MLHFQ scores. To the best of our knowledge, this is the first study to implement this novel technology to improve self-care of HF. Our study may pave the way for further large-scale clinical outcome trials to determine the efficacy and feasibility of ICT-based interventions for long-term management of HF patients.

The key characteristics of HF include an increase in sodium avidity and a tendency toward congestion [23,24]. Therefore, the maintenance of sodium and water balance is a critical element of HF management, and among several noninvasive methods proposed to assess this balance, uNa has been increasingly used as a simple, inexpensive, and noninvasive indicator for the dietary sodium intake [25-29]. Specifically, it is well established that the urinary sodium excretion is precisely regulated to match dietary sodium intake [25,30]. Although the assessment of 24-hour urine sodium excretion is considered the gold standard method, the measurement of spot uNa could provide an approximate estimate of sodium excretion, with increased convenience and low cost [25,29,31-34]. On the other hand, a previous study demonstrated that spot uNa was reduced by decongestive therapy for advanced HF, which might provide insightful information to titrate the diuretic dose [35,36]. Hence, in general, decreased uNa in HF patients can be a result of low salt intake or effective diuretic therapy [35]. In this study, HF patients with good adherence to ICT-based telehealth program showed reduced uNa after 12 weeks of intervention. Given that patients were clinically stable without any evidence of acute destabilized HF and change in diuretic dose, decreased uNa might be a reflection of better control of dietary salt intake. As recent data emphasize nonpharmacological as well as pharmacological interventions for HF because of its progressive nature [37], an ICT-based telehealth program can be a useful tool for improving self-care and, potentially, prognosis in HF patients [37].

Improving health-related quality of life is not only a major treatment goal in HF but also an effective surrogate marker of appropriate HF therapy and improved prognosis in HF patients [17,38-40]. Among various methods of assessing health-related quality of life, we used the MLHFQ, which has been extensively validated in HF patients [18,19,38,40,41]. In our study, the mean MLHFQ scores numerically decreased by about 2 points in all patients (P=.18) and by 6.1 points in patients with good adherence (P=.08) after 12 weeks of ICT-based telehealth program. Although there is no absolute cutoff for the MLHFQ scores to discriminate between normal and impaired quality of life, a decrease of more than 5 points is considered as a clinically important change [42,43]. Previous studies have suggested that the changes in the MLHFQ scores can translate into differences in clinical outcomes of HF patients. For example, EPICAL trial showed that a 10-point decrement in the MLHFQ scores was associated with a 23% decrease in mortality and a 31% decrease in readmission for HF [17]. Hoekstra et al [42] also reported that each 10-point increment in the MLHFQ scores was associated with a 7% increase in mortality. Given that the short duration of intervention in this study led to a 6.1 point decrease in the MLHFQ scores among patients with good adherence, an ICT-based telehealth program seems promising to enhance quality of life and consequently survival of HF patients. Since the MLHFQ comprises three subscores (physical, emotional, and socioeconomic dimensions), we also compared the changes in each of the subscores between groups. In this analysis, the reduction in the MLHFQ scores mostly stemmed from the improvement in emotional aspects in patients with good adherence, although the between-group difference was not significant. These findings suggest that intense monitoring and automated feedback using an ICT-based telehealth program can play a pivotal role in providing emotional support for HF patients. Considering the close relationship between emotional health and hard clinical outcomes in HF [44,45], ICT-based telehealth program has the potential to improve the prognosis of HF patients. The improvement in physical subscores was also numerically greater in patients with good adherence without reaching statistical significance. Considering that patients with good adherence to ICT-based telehealth program showed reduced uNa, a surrogate marker of better control of sodium intake, it seems plausible that this program can also be helpful in improving the physical aspects of quality of life in HF patients. A future study with a large sample size is needed to validate these findings.

Previous studies demonstrated that self-care management using telemonitoring systems did not significantly improve hard endpoints in HF patients, such as all-cause death or HF hospitalization [46,47]. The low rate of adherence to self-care management programs has been suggested as one of possible reasons for these suboptimal results. In the tele-HF trial, only 55% of study patients utilized the telemonitoring system at least 3 times per week during 6 months and, furthermore, 14% of patients never used this system during the study period [46]. In our study, 23 of 27 patients (85.2%) utilized ICT-based telehealth program at least 3 times per week during 12 weeks, and there were no patients who did not use the program. Furthermore, 10 of 27 patients (37.0%) used the ICT-based telehealth program ≥100 times during the 12-week period, indicating the use of this program more than once daily. These patients with good adherence to the program showed the improvement in sodium intake and quality of life. This improved adherence of patients may be partially due to the enhanced integration of ICT-based telehealth program into their daily lives, with the easier use of such systems.

The user experience was positive with ≥85% patients indicating satisfaction or neutral response with the voice recognition system provided. This favorable user experience might have been achieved by the acceptable success rate of voice recognition in our study, considering that 95% accuracy is nearly on par with that of human speech recognition. In this study, however, the accuracy of voice recognition was affected by adverse conditions, including the use of nonstandard speech, voice alteration during walking, and the presence of substantial background noise. Specifically, the success rate of voice recognition was 91% and 86% for the patients who used nonstandard language and for those who were walking during the process of voice recognition, respectively. Furthermore, the success rate was 87% in settings with increased background noise level, such as when using public transportation. Hence, the user experience can be further optimized by improving the accuracy of voice recognition under these unfavorable circumstances.

HF patients performing and maintaining self-care management behaviors, such as control of sodium intake (uNa), is one practical implication of this technology from a health care perspective. Although future studies with larger number of HF patients are warranted, this technology also has the potential to improve health-related quality of life (MLHFQ scores). Current implications for improving functional capacity (6-min walk test) or HF severity (NT-proBNP) are limited. Further studies are also needed to determine whether the use of this technology has workforce implications in the management of HF patients.

This study has several limitations that should be considered. First, our sample size was small and the duration of intervention was short. This limitation made it difficult to show the benefit of the ICT-based telehealth program for clinical outcomes in HF patients, especially given the chronic and progressive nature of HF. However, in the case of pilot studies, sample sizes of 10 to 30 are generally considered large enough to test the null hypothesis and small enough to ignore weak treatment effects [47]. Furthermore, even though our study population largely comprised patients with moderate to severe systolic dysfunction (LVEF <40%) whose mortality is substantial (25.6%-41.7%) [48], there was no mortality and only 2 patients withdrew because of hospitalization for acute decompensated HF at the early phase of the study. Notably, the most likely cause of HF exacerbation in these patients was viral infection, not a high-salt diet or poor adherence to medication. In addition, there were no rehospitalizations or emergency department visits for HF after 4 weeks of intervention in our study participants, approximately 75% of whom had a history of prior HF hospitalizations, suggesting that clinical outcomes can be improved by applying this strategy to HF patients. Considering that a prior history of HF hospitalizations has been reported as a risk factor for repeat hospitalizations in HF patients [49,50], our findings suggest that clinical outcomes can be improved by the appropriate application of ICT-based telehealth program in HF patients, although larger studies with longer follow-up are definitely warranted. Second, the definition of good adherence was arbitrary. We assumed that the total frequency of program use of more than 100 times during the follow-up of 12 weeks could be regarded as accessing the system at least once per day. Third, we did not have data on baseline education level, including literacy, health care literacy, digital literacy, and numeracy skills of the study population, which can be critical components for successful technology uptake [51]. Fourth, although spot uNa provides an approximate estimate of sodium excretion, it has limited usefulness as an indicator of long-term dietary behavior change because of its short-term relevance. Thus, a future study using a more accurate surrogate marker, such as 24-hour urine sodium excretion, is warranted to more clearly discern the beneficial effect of ICT-based telehealth program on sodium intake. Finally, because of the lack of an appropriate control arm, study findings should be interpreted with caution. A prospective randomized study comparing an ICT-based telehealth program with conventional HF care is necessary to define the true clinical benefit from this therapeutic strategy in HF patients.

Short-term application of ICT-based telehealth program in HF demonstrated the potential to improve control of sodium intake and quality of life in chronic HF patients with good adherence to this program. Our results suggest that an ICT-based telehealth program can be an effective strategy to achieve better patient care and clinical outcomes in HF.