Introduction

Stroke is one of the leading causes of premature mortality and significant disability, particularly in low- and middle-income countries (LMICs), where more than 85% of stroke-related fatalities and 89% of disability-adjusted life years occur globally [1]. Primary care-based stroke management for community-dwelling patients such as regular visits by primary care providers with therapeutic lifestyle recommendations is a pragmatic and promising strategy to improve patient outcomes. A limited number of cluster-randomized controlled trials (cRCTs) conducted in LMICs have demonstrated the benefit of such interventions for stroke management [2–4]. Typically, these trials have only evaluated short-term effectiveness for cardiovascular risk factors. Their long-term benefits beyond the active intervention phase on endpoints such as mortality remain unclear.

Evidence regarding long-term benefits on hard endpoints has crucial implications for sound policy making and improving population health. From 2017 to 2018, we conducted a 12-month cRCT of a primary-care based system-integrated mobile-health technology-enabled model of care (SINEMA) for stroke management in rural China [5]. This model strengthened the existing primary care workforce through training and support embedded in the entire healthcare system and innovatively integrated both provider-side and patient-facing mobile health (mHealth) technology. Compared with usual care, the intervention significantly reduced the primary outcome of systolic blood pressure at 12-month when the intervention ended [4].

We conducted post-trial assessment and followed up all participants until April 30, 2023 with ascertainment of vital status and causes of death. In this paper, we report the findings on long-term outcomes of all-cause and cause-specific mortality beyond the active intervention phase of participants post SINEMA trial.

Methods

Results

A total of 1,299 participants from 50 villages (mean age 65.7 years, 42.6% females) were recruited at baseline with 637 (from 25 villages) and 662 (from 25 villages) randomly allocated into the intervention arm and control arm. Among them 1,269 were alive at 12 months and eligible for post-trial follow-up (Fig 1). Most of participants (86.1%) had been diagnosed with ischemic stroke at baseline with a median duration of 5.3 years since the first stroke event. The baseline characteristic of participants in the intervention and control arms were generally comparable at baseline (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Baseline characteristics of participants who recruited in the SINEMA trial (n = 1,299) and those who survived at 1-year post-trial for long-term follow-up participants (n = 1,269).

https://doi.org/10.1371/journal.pmed.1004564.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Study profile of the SINEMA trial and the post-trial follow-up.

SD, standard deviation.

https://doi.org/10.1371/journal.pmed.1004564.g001

Vital status was confirmed for all participants (Fig B in S1 File). During the initial 12-month trial phase, 30 deaths (2.3% of all participants) were recorded and the baseline characteristics of 1,269 surviving participants at 12 months was presented in Table 1. Additional 246 deaths (18.9% of all participants) were identified during post-trial period until 70-months post-baseline, resulting in a total of 276 deaths (21.2% of all participants, whose baseline characteristics was presented in Table D in S1 File). These resulted to a total of 6,890 person-years of follow-up (3,421 for intervention and 3,469 for control) for vital status. The median follow-up duration for the 276 deaths was 43.0 months (quantile 1–quantile 3: 26.7–56.8, range 2.3–70.0 months).

For the survived participants (n = 1,023, whose baseline characteristics presented in Table E in S1 File), duration of follow-up for vital status ranged from 69.0 to 70.2 months because all of them were surveyed during a 42-day period of recruitment and baseline survey in 2017 and followed-up until April 30, 2023. Notable differences of baseline characteristics were observed between the survivors and the deceased. Specifically, those who died were older on average, predominantly male, engaged in less physical activity at baseline, had poorer mobility, and exhibited a higher level of disability (Table F in S1 File).

Among the 276 deaths, 121 occurred in the intervention arm compared to 155 in the control arm (Table 2). About three in four deaths (75.7%) (n = 209; 92 in intervention versus 117 in control) had the underlying cause of cardiovascular diseases, and 32.6% of all deaths (43.1% of the cardiovascular cause-specific deaths) were due to stroke (n = 90; 38 in intervention versus 52 in control). As shown in Fig 2, between-arm differences in cumulative incidence favoring the intervention arm began to emerge around 12-months post-baseline (after the active intervention phase ended, and were previously reported) [4] and became more pronounced as time progressed for all-cause (Fig 2A), cardiovascular cause-specific (Fig 2B), and stroke cause-specific mortality (Fig 2C).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. The cumulative incidence of mortality and hazard ratios (95% CIs) of intervention on mortality at 70 months post-baseline (n = 1,299).

https://doi.org/10.1371/journal.pmed.1004564.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Cumulative incidence for (A) all-cause, (B) cardiovascular cause-specific, and (C) stroke cause-specific mortality from baseline to 70 months.

Light purple area indicates 12-month active intervention phase. Light yellow area indicates post-trial observational phase. (C) indicates control arm receiving usual care (blue lines) and (I) indicates arm receiving the SINEMA intervention, a primary care-based mobile health intervention for stroke secondary prevention (orange lines).

https://doi.org/10.1371/journal.pmed.1004564.g002

By 70-months post-baseline, the cumulative incidence of all-cause mortality reached 19.0% in the intervention arm versus 23.4% in the control arm, associated with a HR estimated from the Cox model of 0.73 (95% CI [0.59, 0.90]; P = 0.004; Table 2). Reduced risk in cardiovascular cause-specific mortality (HR, 0.73; 95% CI [0.58, 0.94]; P = 0.013) was also observed in the intervention arm, compared with control arm. Although lower cumulative incidence was observed in the intervention arm (6.0%) compared with the control arm(7.9%), for stroke cause-specific mortality, such differences was not statistically significant(HR was 0.71 (95% CI [0.44, 1.15]; P = 0.16)). The results remained robust after adjusting for variables identified as different between intervention arms at baseline (Table 2). In the subdistribution model, the results for the three mortality outcomes were consistent with the cause-specific model that treated competing events as censoring (Table G in S1 File).

The subgroup analysis on all-cause mortality showed that greater between-arm differences were observed with stronger association among participants who were female, younger than 65-years old, less educated, had lower household incomes, were divorced or widowed, had shorter stroke duration, had systolic blood pressure lower than 140 mmHg or had more severe disabilities at baseline (Fig 3). Significant differences in subgroups were observed for marital status (P for interaction = 0.002), education level, and disability status (both P for interactions < 0.10). The observed patterns for subgroup analyses of both cardiovascular (Fig C in S1 File) and stroke cause-specific mortality (Fig D in S1 File) mirrored those seen for all-cause mortality.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. All-cause mortality in each pre-specified and post-hoc subgroup, minimally adjusted model.

Hazard ratios were adjusted for age, sex, and township, with villages considered as clusters. CI, confidence interval.

https://doi.org/10.1371/journal.pmed.1004564.g003

To better understand mortality outcomes, we evaluated the intermediate outcomes among participants who survived at long-term follow-up, assessed in 2022−23 (median follow-up of 66.6 months post baseline). Out of the 1,023 confirmed survivors, loss to follow up was low (n = 44, 4.3%) and 979 successfully completed the survey and were included in the intermediate outcome analysis. Over 66 months, the SINEMA intervention was associated with a sustained and significant net reduction in systolic blood pressure of 2.9 mmHg (95% CI, −5.4, −0.4; Table 3). Observed differences favoring intervention was observed also for diastolic blood pressure and medication adherence with higher proportion of individuals with perfect adherence to medication, but not for physical activity or functioning abilities (Table 3; risk differences are reported in Table H in S1 File). Post-hoc analysis on changes in potential lifestyle factors, including smoking status, body mass index (BMI), and waist circumference were reported in Table I in S1 File, but no significant differences in two arms were observed.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Effects of the SINEMA intervention on intermediate outcomes from baseline to 70 months among living patients (n = 979).

https://doi.org/10.1371/journal.pmed.1004564.t003

Discussion

During a maximum follow-up of 70.2 months from the baseline of the SINEMA trial, individuals who received the SINEMA intervention showed significant long-term relative reductions of 27% for all-cause mortality, 27% for cardiovascular cause-specific mortality. Although lower incidence in stroke cause-specific mortality was observed in the intervention arm, the between-arm difference was not statistically significant. These between-arm differences were descriptively more pronounced among females, less educated, divorced or widowed, disabled, or lower-income people.

Our study is one of the few long-term observational follow-up studies of cRCTs on patients experiencing stroke in LMICs. The 70-month follow-up analyses extend earlier findings on the effectiveness of primary care-based mHealth strategies to promote stroke secondary prevention in resource-limited settings [2–4] and further proven the potential sustained impact of this time-limited intensive intervention on mortality.

Stroke was responsible for 6.55 million excess deaths in 2019 worldwide[1], with direct healthcare expenditures reaching US$183.13 billion [18]. We also found a cumulative incidence of overall mortality as high as 21.2% at 5.5 years, comparable to the 21% death rate at 6 years in a large Chinese population study [19]. The long-term outcomes in mortality observed in this study address such global health challenges and shed light for effective strategies to reduce stroke mortality especially in resource-poor settings.

This long-term follow-up of SINEMA trial addresses the gap in assessing the sustained association of interventions with mortality. Previous trials typically reported within-trial findings or focused on the immediate outcomes of cardiovascular risk factors such as blood pressure [20,21] or cardiovascular events [22]. While a few trials on hypertension control, such as COBIN [23] and COBRA [24] reported 5-year post-trial effects on the outcome of blood pressure, they yielded mixed findings [23]. In addition, the long-term effects of interventions involving mHealth technology on cardiovascular disease management remain unreported. Our long-term follow-up study observed sustained association of the intervention with some intermediate outcomes [13], as well as on long-term mortality. The significant relative reductions of over 20% in deaths provide evidence for policy-makers to scale up the SINEMA intervention to reap larger population health benefits. Of note, the delivery cost was relatively low at US$24 per capita [4], with high cost-effectiveness during the trial (incremental cost-effectiveness ratio of US$792.9 per quality-adjusted life year gained) and cost savings post-trial (protocol published; results under review) [25], promising for transfer and broader implementation. Discussions are ongoing to engage policy-makers and other stakeholders in potential scale-up.

The relative reduction in mortality observed in our study may seem larger than what can be expected from the reduction in blood pressure. Although the net reduction of 2.9 mmHg in systolic blood pressure appeared modest, this reduction was sustained over the long term of 6 years. Although multiple factors and mechanisms could contribute to the changes in mortality and our complex intervention is not fully designed to illustrate such mechanisms of changes, this sustained association was likely associated with the maintained behavior changes in medication adherence in the intervention arm, which may provide some clues. A meta-analysis of 18 clinical trials demonstrated that interventions enhancing medication adherence reduce mortality in patients with cardiovascular diseases [26]. Evidence also indicated that even a 1 mmHg reduction in systolic blood pressure significantly decreases the risk of fatal and non-fatal cardiovascular events and stroke, thereby reducing mortality [26].

The substantial reduction in mortality may be partially attributable to the higher death risk in older rural Chinese patients experiencing stroke. Despite improvements in healthcare, stroke mortality in China still remains higher than in many high-income countries [27,28]. In addition, urban-rural disparities in stroke mortality have persisted—14% versus 21% in 2020 [19]. Therefore, our study has the potential to reduce urban–rural inequity in health outcomes for patients experiencing stroke. Beyond this, the SINEMA intervention also shows promise in reducing health inequity by socioeconomic status or sex. Our study observed stronger between arm differences than the overall effect size among vulnerable groups with poor access to optimal care, including females, the less educated, divorced or widowed individuals, the disabled, and those with lower income [29]. These results highlight the greater value of investment in primary care to reduce health inequalities linked to residence, socioeconomic status, or sex [30].

Interestingly, we also observed reduction in non-stroke cardiovascular cause-specific mortality, indicating the potential broader cardiovascular benefits beyond stroke prevention. Such observed reduction in non-stroke cardiovascular cause-specific mortality may be attributable to the control of shared risk factors between stroke and other cardiovascular diseases in blood pressure, improvement in medication adherence to statins and antiplatelet use and increase in physical activity, though more studies are needed to investigate such broader benefits.

Several key features of the SINEMA intervention may have contributed to its long-term sustained association: (1) the design of the intervention—based on comprehensive field-based contextual research [5,7,8] and streamlined clinical guidelines suitable for resource-limited settings—increased the likelihood of sustained effectiveness; (2) delivered by village doctors, the intervention integrated into routine care with low marginal cost of human resources. Moreover, motivations from non-financial incentives, as well as the trust building with patients as reported in the process evaluation, may have lasting association beyond the trial period [31]; (3) the intervention was human-centerd with equal emphases on behavior changes for both patients and providers instead of only one group. Monthly follow-up visits reinforced provider–patient interactions—a core principle of the chronic care model [32]; and (4) mHealth technology was crucial to our intervention [33]. The SINEMA app, free for village doctors, featured user-friendly functions could be used after the trial stopped.

Key strengths of our study include extensive field-based contextual research and rigorously implemented cRCT design, extended post-trial follow-up, intention-to-treat analyses with independent biostatistical verification, and standardized endpoint assessment from multiple sources to ensure objectivity and complete (100%) follow-up on vital status. Our study also has several limitations. First, the study was conducted in a single county in Northern China, which limits its generalizability. Future research using a qualitative approach is being planned to provide insight on adaptable strategies for other settings. Second, although we have tried to infer the roles of several factors through intermediate outcome analyses, we were unable to fully uncover the exact mechanisms by which the multicomponent intervention reduced mortality. Moreover, given the observational nature of the post-trial follow-up phase and the potential influence of unobserved interventions or unmeasured factors on mortality, we are not able to fully build the causal inferences, and caution should be exercised when interpreting the findings. Lastly, not all causes of death have accurate clinical diagnostic records in rural areas of China, but our multi-source verification partially mitigates this concern. More importantly, the assessment methods remained identical for both the intervention and the control arms, ensuring their compatibility.

In conclusion, this study demonstrated a sustained reduction in long-term mortality of the primary care-based mHealth intervention beyond the trial phase among patients experiencing stroke in rural China. These results provide clinically and practically implication for evidence-based policymaking, suggesting the potential of improving stroke secondary prevention in resource-limited areas and reducing health inequity via strengthening primary care and applying digital health solutions. Such intervention strategies are potentially relevant to not only stroke but other non-communicable diseases. To aid its expansion and wider implementation, future research investigating the transferability and adaptation of the implementation strategies in China and other LMICs are warranted. When scaled up, the intervention has the potential to contribute to achieving the sustainable development goals of reducing premature non-communicable disease mortality by 2030.

Supporting information

Acknowledgments

We thank Zhenli Xu, Mobai Hou, and the team at the Nanhe District Centre for Disease Prevention and Control, Li Wang, Huayun Zhang and the team at the Renze District Centre for Disease Prevention and Control, and all staff members from township healthcare centres and village clinics, and all patients who participated or supported the study. We would like to also acknowledge Jun Wang (Department of Otorhinolaryngology, Head, and Neck Surgery, The First Affiliated Hospital of Nanchang University)’s help for data visualization.