The model included a hypothetical population of 1000 patients with an average age of 65.3 years and consisted of 12 health states: <120, 120‐129, 130‐139, 140‐149, 150‐159, 160‐169, ≥170 mm Hg, CV event, recurrent event, postrecurrent event, all-cause death, and CVD-related death. The initial distribution and demographics (Multimedia Appendix 1 [8,18-23]) of patients over the systolic blood pressure states were based on the Blood Pressure Lowering Treatment Trialists’ Collaboration study [18], reflecting a real-world distribution of patients with hypertension and no history of CVD. Patients could transition from higher blood-pressure states to lower blood-pressure states on an annual basis, based on drug therapy, until the patients were on target (120‐129 mm Hg). A reverse transition was not possible. Each year, patients could experience a CV event, after which the patients returned to the pre-event blood-pressure health state, as no direct blood pressure lowering effect due to the CV event was expected. A CV event was a composite event consisting of either a myocardial infarction (MI), a cerebral hemorrhage, or an ischemic cerebrovascular event, and the event risk was blood pressure dependent [19]. Patients could experience a recurrent event after which they progressed into the postrecurrent event health state. The risks of all-cause death and CVD-related death were assumed to be blood pressure independent.

Standard of care (SOC) was based on the current practice of hypertension management in the Netherlands. The care provided via the hospital outpatient department (OPD) was based on the latest European Society of Hypertension guidelines on the management of arterial hypertension [24]. In the model, patients in the SOC group would be managed with drug therapy and lifestyle interventions. Patients would on average have 3 in-person OPD consultations in the hospital with their clinician during each 1-year cycle, based on the standard diagnosis-treatment combination for patients with hypertension in the Netherlands [25].

Patients in the HBPT group were similarly managed in terms of drug treatment compared to the SOC group. The HBPT intervention was based on the HBPT program developed by the Maasstad Hospital in Rotterdam, the Netherlands, and adopted and studied throughout the region [26]. In this program, which is conducted in a hospital setting, patients measured their blood pressure during a complete week with 2 measurements in the morning and 2 in the evening. Measurement weeks were scheduled depending on the level of blood pressure control (eg, weekly in case of very uncontrolled blood pressure [>180/110 mm Hg] and monthly in case of controlled blood pressure [<140/90 mm Hg]), but on average occurred once every month. The monitoring platform used was the Luscii [6] application, which is the most widely used platform in the Netherlands for remote patient monitoring. The most frequently used patient monitoring setup in the Netherlands includes a “hospital-based telemonitoring center” with specialized e-nurses. Blood pressure data are automatically synchronized via the monitoring platform to a special health care provider dashboard, integrated into the electronic health record. The e-nurses in the telemonitoring center assess all the alarms generated by the monitoring platform based on the blood pressure data and discuss these alarms with clinicians if needed. A schematic overview of the HBPT processing steps is included in Multimedia Appendix 2. The clinicians supervising the e-nurses are internal medicine specialists, residents, or nurse practitioners who would also be involved in the SOC for patients with hypertension. They are also responsible for remotely adjusting blood pressure medication if needed.

Multimedia Appendix 1 provides an overview of the baseline blood pressure distribution, annual event probabilities, and efficacy model inputs. Each systolic blood pressure state corresponded to a risk of a CV event, which was based on a large prospective real-world study [19]. The transition probability from a higher blood pressure state to a lower blood pressure state was based on a decrease of 5.1 mm Hg per year, which corresponded to the clinical effect of pharmacological therapy reported in the latest available meta-analysis [20] and applied to both groups. Patients in the HBPT group had an additional decrease of 12 mm Hg in systolic blood pressure in the first year due to HBPT. This additional effect was based on the latest available literature on the clinical effectiveness of HBPT [8]. A notable proportion, 19.7% of the patients had resistant hypertension resulting in the absence of blood pressure reduction [21]. The probability of dying from a CV event (CVD-related death) was based on the Blood Pressure Lowering Treatment Trialists’ Collaboration study [18]. The probability of all-cause mortality was based on the age-based population mortality in the Netherlands [22] and was corrected for CVD-related deaths [22]. The probability of suffering a recurrent CV event was derived from a large study assessing the 10-year risk of recurrent vascular events [23].

The utilities of the model health states were derived from published literature (Multimedia Appendix 3 [27-30]). The baseline utility value was 0.96 for patients with hypertension [27], which declined to 0.79 following an MI [28], 0.64 following a cerebral infarction [28], and 0.59 following an intracranial hemorrhage [29]. The weighted average of these event utilities was 0.67 and was used in the model as the “postevent” utility for the year after the event occurred. The conservative assumption was made that after 1 year of the event, the utility would equal the baseline utility.

A recurrent intracerebral infarction or MI corresponded with a utility of 0.74 and 0.62, respectively [30]. For a recurrent intracranial hemorrhage, the utility value was considered equal to the utility value of a first intracranial hemorrhage, which was 0.59 [29]. The weighted average utility of a recurrence was 0.64 and was used for the year the recurrent event occurred and the subsequent years the patient was in the postrecurrent health state [27].

Costs were divided into direct medical costs and nonmedical costs (Multimedia Appendix 4 [16,25,31-42]). For the HBPT group, direct medical costs consisted of a one-time out-of-pocket purchase of a blood pressure device [31], costs for remote monitoring [32], standard drug costs [33,43], additional drug costs [44], and in-person OPD consultations. The remote monitoring costs were based on an official Dutch tariff [32] for patients who are part of a remote monitoring program. A hospital can claim this tariff 3 times a year as a flat fee for a patient who is remotely monitored to cover costs for the license of telemonitoring software, salaries for the involved health care workers, and development costs. In the SOC group, direct medical costs only consisted of standard drug costs and costs for the in-person OPD consultations. Direct medical costs for a stroke (infarction and hemorrhage), MI, or CV-related death were based on data available from the Dutch National Health Care Institute and Ministry of Health, Welfare and Sport [34,35]. Total costs for each event were based on the overall reported expenses divided by the weighted incidences of both stroke and MI.

Nonmedical costs consisted of travel costs, parking costs, and costs related to productivity losses in both the SOC and HBPT groups. Productivity losses were based on work absence resulting from the in-person OPD consultations (1 hour for each visit) or due to an event (17.7 absent working days) and were based on data from the Dutch National Healthcare Institute [36] and the Trimbos Institute [37]. The costs of productivity losses were based on the average labor participation [38], average hourly wage [39], and average working week in the 65‐75 years age group corresponding with the average age of 65.3 years used in the current analysis (based on the Blood Pressure Lowering Treatment Trialists’ Collaboration study [18]. Friction costs following a death were calculated based on the friction costs method [40].

Discounting rates were 3% for the costs and 4% for the health outcomes based on the Dutch Economic Evaluation guidelines [15].

One important assumption of the current model is related to the number of in-person consultations for patients in the HBPT group. It was assumed that in the HBPT group, the number of annual consultations dropped from 3 to 1.5 per year. A further decrease in the number of consultations could result in a further reduction of the ICER. Based on the results of scenario 1, HBPT will become cost-effective (<€20,000 per QALY) with the current reimbursement of €504 per year at 1.48 in-person OPD consultations per year and will become cost-saving at 1.18 in-person OPD consultations per year (Figure 3).

The duration of the effect of HBPT comes with great uncertainty and was assumed to last for only 1 year in the base-case analysis, which could be considered conservative. Scenario 2 indicates that the ICER will reach €11,154 per QALY in case the effect of HBPT lasts for 2 years (a total blood pressure reduction of 24 mm Hg after 2 years) and further declines to €9204 per QALY in case HBPT reduces the blood pressure with 12 mm Hg for 3 years (a total blood pressure reduction of 36 mm Hg after 3 years).

The results of scenario 3, in which the model was run over an age range of 30 to 75 years (Figure 4), indicate that the younger the patient’s age at the start of HBPT, the lower the ICER. If HBPT is started at the age of 64 years or below, HBPT could be considered a cost-effective intervention.

To the best of our knowledge, this is the first study to provide an early cost-effectiveness analysis of HBPT in patients with hypertension without previous CV events. Based on the current early cost-effectiveness model, which reflects a societal perspective and includes both short- and long-term costs and benefits, HBPT showed the potential to be cost-effective following realistic reductions in SOC for patients with hypertension. Specifically, a reduction in the number of OPD consultations in the HBPT group will make HBPT cost-effective or even cost-saving. These findings underscore the potential of HBPT and with that the importance of genuine digital transformation in health care, advocating for the substitution of traditional OPD care with digital and remote care, rather than providing digital care as an add-on to standard OPD care. Additionally, we found that early (ie, younger age) and sustained telemonitoring further improves the cost-effectiveness of HBPT.

Policy makers should make use of the fact that HBPT can be cost-effective following a reduction in the number of OPD consultations. The current reimbursement structure provided by the Dutch Health Care Authority [32] does not include any requirements in terms of reducing standard care and therefore seems unsuitable in its current form. Outcome-based health care contracts [45], characterized by performance fees linked to predefined shared objectives between hospitals and insurance companies, are ideally suited for HBPT to pursue genuine digital transformation efforts. By setting a shared objective in terms of physical care replacement with HBPT, a sustainable system could be established that allows for efficient (with fewer resources) and cost-effective care delivery along the lines of value chain optimization and significant displacement effects.

Hypothetically, HBPT should be able to realize short-term benefits through greater efficiency with regard to care organization and long-term benefits, resulting from a greater level of blood pressure control, which translates into a reduction in CV events and CV-related deaths. We found that short-term benefits realized through a substantial reduction in OPD consultations had a major impact on the ICER, but the impact of long-term benefits appeared to be limited. The 1-year effect of HBPT on the blood pressure of patients resulted in minor between-group differences in CV events or CVD-related deaths. Extending the effect of HBPT to 2 or 3 years substantially reduced the ICER, but clinical evidence supporting this assumption is lacking, as follow-up durations in clinical trials are usually no longer than 12 months [8]. Therefore, short-term benefits resulting from more efficient care delivery are expected to become the major driver for a sustainable business model for HBPT. The further upward potential of remote monitoring comes with a multimorbidity perspective (eg, hypertension and diabetes). Many patients have a variety of comorbidities and multiple consultations with different specialists. If one remote monitoring program reduces the number of consultations across multiple medical specialties, remote monitoring is more likely to result in cost savings.

The only available comparable study [46] evaluates the cost-effectiveness of HBPT in a poststroke population using a Markov cohort simulation. The HBPT intervention was cost-saving in the base case and cost-effective in the scenario analyses with an ICER of US $1200-4700 per QALY. In contrast to our study, the benefit in terms of blood pressure reduction due to HBPT was modeled as a continuous effect (year after year). Even though our scenario analyses with 2 and 3 years of clinical benefits of HBPT resulted in HBPT being cost-effective, the results of the previously described study [46] should be considered as optimistic as evidence on a sustained (year after year) effect is lacking. Other studies that reported a positive effect of remote monitoring include heart failure monitoring [47-49] or monitoring of patients with COVID-19 [50]. These studies [46,47,50] highlight the importance of reducing short-term care consumption to come to a favorable ICER. This advocates for the substitution of traditional care with digital and remote care, rather than providing digital care as an add-on to standard care.