Improving primary healthcare for non-communicable diseases during annual flooding in Bihar, India: study protocol for a hybrid implementation-effectiveness parallel cluster randomised controlled trial

Abstract

Background

Non-communicable diseases (NCDs) cause 4.7 million deaths annually in India. Floods increasingly disrupt NCD care, especially in Bihar, one of India’s most flood-prone and socioeconomically disadvantaged states. This study aims to strengthen the capacity of the community and primary healthcare systems to improve NCD outcomes during annual flooding in Bihar.

Methods

We describe a type II, hybrid implementation-effectiveness parallel cluster randomised controlled trial with an embedded qualitative process evaluation. A multi-component, co-developed intervention targeting primary care and community settings will be implemented and evaluated over two flood seasons in 26 flood-prone community development blocks in three districts of Bihar. Normalisation Process Theory (NPT) will guide the interpretation and adaptation of the intervention for broader applicability. Data will be collected through baseline and follow-up cross-sectional surveys conducted immediately after and 6 months following each annual flood season. Mixed-methods implementation and effectiveness outcomes will be evaluated using the RE-AIM-QuEST (Reach, Effectiveness, Adoption, Implementation, and Maintenance, Qualitative Evaluation for Systematic Translation) framework. Co-primary outcomes are patient medication adherence and availability of a flood-proof kit with medication and clinical handover information during floods. A preliminary health economic evaluation will estimate the cost-effectiveness of the intervention from a healthcare and societal perspective, with and without the use of early-warning flood forecasting.

Conclusion

The FUSION (Floods, UnderStanding clImate change and nON communicable disease) trial will generate evidence on effective and scalable strategies to enhance health system and community resilience for the continuity of NCD care during recurrent flooding in India. Findings will inform public health policy and disaster preparedness efforts in similar resource-constrained, climate-vulnerable settings.

Trial registration

CTRI/2025/07/092055 [Registered on 30/07/2025] http://www.ctri.nic.in/Clinicaltrials/pmaindet2.php?trialid=138142; ISCTRN18019775 [Registered on 12/08/2025] https://www.isrctn.com/ISRCTN18019775

Supplementary Information

The online version contains supplementary material available at 10.1186/s13063-026-09633-3.

Introduction

Background and rationale

Non-communicable diseases (NCDs) account for 82% of premature deaths in low- and middle-income countries (LMICs) [1], representing a major and escalating public health concern. In India, this burden is particularly severe, with NCDs accounting for 4.7 million deaths in 2017 [2], equivalent to 60% of all-cause mortality [3]. The leading contributors to NCD mortality include cardiovascular diseases (23%), chronic respiratory diseases (9%), cancer (6%), and diabetes (2.4%) in India [2]. India also has the second-highest number of diabetes cases globally [4]. Effective management of these chronic conditions depends on uninterrupted access to medications, regular clinical monitoring, and functional primary healthcare services. However, such continuity is often compromised during natural disasters, particularly floods, increasing the risk of acute complications, hospitalisation, long-term disability, and mortality among people with NCDs [5–10]. Available evidence highlights the significant and often prolonged impact of natural disasters and conflict on individuals with NCDs [7–10]. These effects may persist for months after the event, with both immediate and long-term consequences for health outcomes [7, 11–13]. In response, the World Health Organization (WHO) recommends that primary and community healthcare services be strengthened during disasters, with a focus on maintaining care for priority conditions such as diabetes, cardiovascular diseases, and chronic respiratory diseases [14]. India is highly susceptible to extreme weather events, particularly floods, due to its geographical location and monsoon-driven climate. Seasonal monsoons influenced by the Indian Ocean, Arabian Sea, and Bay of Bengal cause widespread flooding across multiple states each year [15]. Climate change has further disrupted rainfall patterns, leading to increased frequency, intensity, and severity of floods [15, 16]. According to the INFORM Risk Index, India scores 8.5 out of 10 for flood-related humanitarian risk, among the highest globally [17]. Flood emergencies occur not only during predictable annual events but also as catastrophic, large-scale floods that can affect both previously impacted and new regions. These events often lead to mass displacement and widespread damage to infrastructure, severely limiting access to routine healthcare services [14, 18]. People living with NCDs are especially vulnerable, frequently losing access to essential medications, prescriptions, and medical devices. Damaged roads and health facilities further restrict emergency healthcare delivery. In addition, the psychological stress of displacement and exposure to poor environmental conditions during displacement can further exacerbate chronic conditions [19].

Despite the known risks, there is limited evidence of evaluated interventions aimed at ensuring the continuity and resilience of NCD care during natural disasters. This gap is particularly pronounced in LMICs, including in flood-prone regions of South Asia. A recent review by ELRHA highlighted that most studies were conducted in conflict zones or refugee settings, with very few focused on natural disasters, including recurrent or catastrophic flooding [20]. One notable exception is the 2018 flood response in Kerala, India, where limited efforts were made to implement the WHO guidelines for NCD management [14]. These included basic stockpiling of essential medications, use of simplified treatment protocols, minimal mapping of referral facilities, and some health education initiatives [21]. Even in Kerala, where NCDs’ prevalence is high and health literacy and care seeking are relatively strong, more than half of the over 36,000 individuals who accessed relief camps during the floods sought care for exacerbations of chronic conditions, which indicates the patients’ and health system’s lack of preparedness [22]. This highlights the urgent need for disaster preparedness strategies that explicitly address the needs of people living with NCDs.

Bihar, one of India’s most flood-prone and socioeconomically disadvantaged states, provides a critical context for this study. Annual floods impact 28 of the state’s 38 districts and approximately 73% of its land area with particularly severe impacts in districts such as Muzaffarpur, Darbhanga, and East Champaran [23, 24]. Bihar accounts for 30–40% of India’s total flood-related damages [25, 26], yet its health infrastructure remains challenged. Our study aims to demonstrate the health outcomes of strengthening both the health system and community resilience in managing common long-term conditions, specifically hypertension, diabetes, and chronic respiratory diseases, during flood events.

Objectives

We will co-design, implement, and evaluate a complex intervention via a hybrid cluster randomised controlled trial (cRCT) to improve existing NCD service delivery during floods by strengthening local health system preparedness by involving healthcare providers, communities, and stakeholders. This protocol primarily describes the hybrid implementation-effectiveness cRCT. Other objectives in parallel to this trial to enrich the overall study are to co-develop an integrated meteorological/hydrological flood early-warning system to provide timely forecasting in the future, integrated with the needs of the health system and to provide a preliminary estimate of the potential costs and benefits of the complex intervention.

Methods

Community engagement and involvement

Community engagement and involvement (CEI) will be a strong component of this study. Working with a local non-governmental organisation (NGO), we will engage with communities within study sites to build relationships and trust, identify community leaders, and support trial participation. We will form community advisory groups in neighbouring districts to review study materials, co-develop interventions, and support dissemination activities.

Trial design

We will conduct a 1:1 hybrid implementation-effectiveness, parallel with baseline cRCT [27], guided by the RE-AIM-QUEST (Reach, Effectiveness, Adoption, Implementation, and Maintenance, Qualitative Evaluation for Systematic Translation) framework [28]. We will incorporate process evaluation using a mixed-methods framework to identify real-time implementation barriers and explain how the specific implementation context may influence translation to additional settings [28].

The intervention comprises a multi-level package, with components targeting healthcare facilities (HCFs) such as community health centres and health and wellness centres, sub-HCF levels such as wards, smaller communities, and individual participants. This design makes individual-level randomisation infeasible and raises a high risk of contamination or spillover effect. As such, a cRCT with repeated cross-sectional surveys will be employed to capture population-level changes at two time points.

Setting and study participants

Setting and cluster design

Clusters will be located within three flood-prone districts of Bihar State, in North Central India. The unit of randomisation (which is a cluster) will be the community development block (CDB), a subdivision of a district, administratively earmarked for planning and development by the state government. CDB covers a population of approximately 100,000–150,000, with a community health centre (CHC). The CDBs will be selected based on the eligibility criteria listed below. A CDB can cover a vast geographic area, some of which will not experience floods, while several small areas may be prone to floods. We will choose flood-prone areas to be designated as our ‘study areas within our CDB clusters’. These areas within the study clusters from which we will recruit our participants will meet the criteria described below.

Eligibility criteria

CDBs

Each CDB will necessarily: be in Muzaffarpur, Darbhanga, or East Champaran districts of Bihar State; have a primary care facility responsible for service delivery, which has at least one doctor and a pharmacist; and include at least one ‘flood-prone area within its coverage’ with a population of at least 4000–6000. A flood-prone CDB is one in which flood-relief camps and/or community kitchens were set up in 3 of the last 5 years during flood seasons.

A door-to-door survey of three villages was conducted from May to July 2024 in Siswara village and Pannapur Chakki village in Muzaffarpur district. Our preliminary analysis suggests that at least 5–10% of the population have been diagnosed with hypertension, diabetes mellitus, and CRDs (asthma and COPD).

Flood-prone areas within coverage of CDB

Only one group of flood-prone areas will be chosen from each CDB cluster, even though some intervention components may need to embrace all flood areas in the CDB catchment, as the same CHC provides the services to all. However, only measurements from the flood-prone pre-selected regions will be included in the study.

The population of flood-prone areas within the CDB study cluster will be selected from several flood-prone wards. Wards are the smallest administrative units with clear geographical boundaries under the local government system. Eligible wards must have experienced flooding requiring camps and/or community kitchens in at least three of the past 5 years, should lack recent or known planned flood mitigation interventions for the next 2 years, and have reasonable access to a health and wellness centre with a community health officer able to prescribe repeat medications.

CDB exclusion criteria

We will exclude CDBs that are either inaccessible from the Muzaffarpur city project office due to travel times exceeding 3–4 h, located within 5 km of an adjacent cluster, or previously involved in piloting the intervention.

Study participants

Participant inclusion criteria

Participants will be eligible if they are adults aged ≥18 years and have resided in the flood-prone study cluster continuously since May 2025 (covering the recent flood period and 2 months prior for pre-flood preparations). Participants should have a self-reported diagnosis made at least 6 months before May 2025 of one or more target NCDs: hypertension, diabetes, or chronic respiratory diseases (asthma and COPD).

Exclusion criteria

We will exclude participants with the target NCDs who cannot converse in Hindi or English or are not cognitively coherent, so they cannot consent, or are likely to move away from their house before May 2026 for baseline and May 2028 for follow-up (self-reported).

Intervention

The intervention involves co-developing and piloting health system and community/patient preparedness and response components for NCD management during the flood period. A formative qualitative situational analysis using focus groups, interviews, and observations was conducted in 2023 and 2024 and mapped the barriers and facilitators for managing NCDs during annual flooding [29]. The findings from situational analysis are being used in a series of Theory of Change workshops to co-develop the intervention. Final intervention components will be finalised through this co-development process with stakeholders and community engagement by December 2025. The resulting multi-system intervention will be described in detail using the TIDieR framework [30]. However, a summary of key components thus far anticipated are as follows.

Our health intervention consists of three interconnected work streams of community, primary care, and cluster-level health systems, applying a multi-level approach that has demonstrated greater effectiveness than single-level interventions. The intervention is aligned with all pillars of the WHO health system strengthening framework. It covers service delivery, workforce capacity strengthening, information systems, essential medicines, and leadership/governance (at the CDB level).

Key components currently considered for the intervention are illustrated in Fig. 1 and include:

1. Secure, flood-proof storage of stockpiled essential NCD medications, including insulin, adapted from the WHO essential medicines lists, to prevent spoilage during floods, along with basic diagnostic tools for use in relief camps and for outreach to isolated communities.

2. Clinical protocols, algorithms, and checklists for flood preparedness and rapid NCD management during flood emergencies by primary care clinicians and non-physician healthcare workers, including lifestyle advice, rapid NCD assessments, referral pathways, and brief mental health input. These protocols and tools will be based on existing national and regional standard treatment protocols for NCD care and the WHO package of essential non-communicable (PEN) [31] strategy. These documents will be provided in hard copies.

3. Enhance the use of existing phone/telemedicine advice and referral systems from existing models in Bihar (currently exist but are not fully utilised).

4. Support primary and community healthcare workers during floods with minimal monetary incentives in line with the government of India, plus non-monetary incentives, and training on the above components. The training to emphasise the need to hold NCD assessments and advice, as well as repeat prescribing during flood camps or other medical assistance to communities during flood times.

5. Mental health support via community support systems employing evidence-based interventions and community or primary care resources.

6. Support the completion of the government’s recommended line-listing of NCD patients.

7. Engage NCD patients and carers extensively by working with community leaders and primary care staff at health facilities and through community health workers (ASHAs) to strengthen community education and sensitisation. ASHAs to be trained with regular government incentives to contact all NCD patients on their lists and assist them in preparing flood-readiness plans over the next few months before the flood season, through simple, pragmatic preparation strategies for patients and families, including a commonly available low-cost flood-proof 1-month drug supply ‘containers’ with essential emergency items and patient-held clinical handover information records detailing multi-morbidities, NCD diagnoses, and medication.

8. Promote community interventions during floods, through low-cost, locally tailored means of communication, IEC material, or social media methods, refined through CEI consultation.

9. Adaptation and implementation of the UNDRR’s Public Health Scorecard [32] at CDB level in the intervention clusters after baseline, and then the year after intervention, in both control and intervention clusters. The scorecard, the implementation of which constitutes a public health intervention, was launched in April 2020 and is a validated tool used in diverse settings to identify, rank, and prioritise actions for strengthening disaster resilience at the planning level. It follows a workshop method and protocol for a consensus-based approach and will be reviewed with local stakeholders and decision-makers in our clusters. If deemed suitable, it will be adapted in collaboration with CDB cluster-level stakeholders and UNDRR, focusing on NCDs. It has been proven effective in identifying practical, grassroots initiatives and offers a structured mechanism to evaluate and recommend priority actions before and after interventions [33, 34]. The scorecard aligns with the Sendai Framework and the UNDRR Ten Essentials for Making Cities Resilient 2030 (MCR2030) [35], supporting informed decision-making and sustainable improvements in NCD care during flooding.

Fig. 1.

Intervention activities

All the above approaches align with evidence favouring active engagement over passive learning and encourage innovation consistent with community action principles.

Control conditions

In control clusters, routine care will continue without additional intervention. NCD management under usual care is generally unstructured, with patients making limited individual efforts to stockpile medications. Routine flood-relief camps rarely provide tailored guidance on NCDs.

Outcomes

The intervention will be complex and designed to target multiple causal pathways that could impact several health conditions. For our main analysis, we consider two primary outcomes: one reflecting implementation and the other representing effectiveness, serving as a proxy for a clinical health outcome. Table 1 outlines all planned outcomes and Fig. 2 demonstrates the steps involved in the trial.

Table 1.

Study outcomes and measurement timepoints

Category	Measure	Description	Outcome type, data collection method	Time of data collection	Control data	RE-AIM	Available if no flood
Primary outcome	Patient self-report of medication adherence during the flood period	Patient report of adherence and refill during the flood using the ARMS validated questionnaire	- Dichotomised as ≤16 or >16 (max 48 scores) - Patient questionnaire	- Baseline - Post-intervention - Immediate post-flood period (S1 and S3)	Y	E	Y
Primary outcome	Observed patient’s flood-proof home kit with or without content in the kit or available to the patient	Including observed handover document/record availability with information completed in the kit during the survey, e.g. sealable bag/container—plus handover card/prescription (The availability of routine medication (patient report) and observation of handover card (with below content) in or outside kit will qualify. The card must include the prescription date, diagnosis, age, and BP measurement.)	- Binary (Having the following criteria: a waterproof kit, with a handover card with defined components; name, diagnosis, list of medication within 6 months, BP measurement within 6 months; and a report that medicines were in the kit in the flood period) - Patient questionnaire including observation component	- Baseline - Post-intervention - Immediate post-flood period (S1 and S3)	Y	I	Y
Secondary outcomes (patient-related health)	Self-report of inpatient admission (all cause and NCD condition)	Patient report of inpatient admissions or under observation (>4 h) from a private or public healthcare facility from the start of the first flood in the season (index flood) until 6 months post-flood	- Discrete variable - Patient questionnaire	- Baseline - Post-intervention - Immediate post-flood period (S1 and S3) - 6-month post-flood period (S2 and S4)	Y	E	N
	Self-report of outpatient consultation exacerbations plus inpatient for NCD exacerbation	Patient report of outpatient (including primary care) consultations for worsening/flare-up of NCD signs/symptoms with change in medication or titration (i.e. excluding visits for regular check-ups/repeat prescriptions) from private or public healthcare facility from the start of the first flood in the season (index flood) until 6 months post-flood Patient report of inpatient admissions (>4 h) admissions or under observation from private or public healthcare facility from the start of the first flood in the season (index flood) until 6 months post-flood	- Discrete variable - Patient questionnaire	- Baseline - Post-intervention - Immediate post-flood period (S1 and S3) - 6-month post-flood period (S2 and S4)	Y	E	N
	Self-reported mental health	Mental health standardised scores for depression and anxiety Patient Health Questionnaire (PHQ-9) and Generalised Anxiety Disorder Questionnaire (GAD-7) (for all participants)	- PHQ-8 score - GAD-7 score Continuous variable based on protocol for scores - Patient questionnaire	- Baseline - Post-intervention - Immediate post-flood period (S1 and S3) and 6-month post-flood period (S2 and S4)	Y	E	N
	Measured blood pressure	Blood pressure ≥140/90 mmHg (average of 2nd and 3rd measures during HV)	- Continuous variable - Average of 2nd and 3rd measurements using WHO STEP criteria	- Baseline - Post-intervention - Immediate post-flood period (S1 and S3) - 6-month post-flood period (S2 and S4)	Y	E	N
	Chronic disease respiratory symptoms score	Standardised scores Modified Medical Research Council (mMRC) breathlessness scale plus Chronic Airways Assessment Test (CAAT) (for CRD patients only)	- mMRC score - CAAT score, continuous variable - Patient questionnaire	- Baseline - Post-intervention - The immediate post-flood period - 6-month post-flood period	Y	E	N
Secondary outcomes (healthcare level)	Proportion of HF healthcare workers trained in intervention	Ratio of trained HF healthcare providers; HF HCPs attending complete training on counselling/advising NCD patients pre-flood to prepare and during flood to be assessed and managed (using protocol to be developed) for their level of healthcare work	Ratio: total attended fully vs total invited - Training attendance register	During intervention training	N	A	Y
	Proportion of ASHAs/community health workers trained in intervention	Ratio of trained community healthcare providers; community HCPs attending complete training on counselling/advising NCD patients pre-flood to prepare and during flood to be assessed and managed (using protocol to be developed) for their level of healthcare work	Ratio: total attended fully vs total invited - Training attendance register	During intervention training	N	A	Y
	Post-training knowledge of intervention and training content	Knowledge score post-training survey for healthcare providers for ASHAs and HCP (their test will be tailored to their level)	- Composite test score - Written multiple-choice test	- Pre- and post-training questionnaire for all HCP trainings (HF/ASHA) - If a control survey is conducted, it is to be done within a month before our training	N	I/A	Y
	Observed the availability of the health facility flood-proof kit with contents	Availability of HFs flood-proof kit and defined NCD drugs (CHC/PHC/APHC/HWC) and devices	A score based on the quantity of drugs and devices	Observed before the flood period in July–September	Y	I	Y
	Self-reported home visits by community HCPs/ASHAs	Community/ASHA reported pre-flood home visits in our clusters	- Proportion of NCDs visited before the flood period	After the flood period in October	Y	A	Y
	Preparedness and response policy and planning change score from policy makers	Adapted version of CADRI UNDDR Public health resilience scorecard for emergency preparedness, asked of decision-makers and senior health workers	Composite score Adapted UNDDR Public health resilience scorecard	Administered at community block-level at baseline and post-intervention	Y	I	Y
Secondary outcomes (patient-reported implementation)	Patient self-report of taking the flood-proof kit to camp	If patients carried a flood-proof kit when they moved to camps or anywhere away from home during the flood period	- Binary - Patient questionnaire	- Baseline - Post-intervention - Immediate post-flood period (S1 and S3)	Y	I	N
	Patient self-report of taking the handover card/prescription to camp	If patients carried handover information when they moved to camps or anywhere away from home during the flood period	- Binary - Patient questionnaire	- Baseline - Post-intervention - Immediate post-flood period (S1 and S3)	Y	I	N
	Patient self-report of knowledge or use of camp services for NCD	Patient report and/or use of availability/receipt of NCD-specific assessment or related services	- Binary - Patient questionnaire	- Baseline - Post-intervention - Immediate post-flood period (S1 and S3)	Y	R&A	N
	Attendance/use of mental health intervention	Knowledge and/or use of MH intervention (to be defined) provided by the health system (including ASHAs/community, HCPs)	- Binary - Patient questionnaire	- Baseline - Post-intervention - Immediate post-flood period (S1 and S3) - 6-month post-flood period (S2 and S4)	Y	R&A	Y
	Patient self-report of medication availability during the flood period	Patient report of availability of his/her medication during the flood period from home kit or healthcare provider, or any route	Binary and categorical (3 or 4 score) - Patient questionnaire	- Baseline - Post-intervention - Immediate post-flood period (S1 and S3)	Y	R	N
	Patient report of receiving advice prior to the flood for preparedness	Patients report receiving advice from any healthcare provider to explain about the flood-proof home kit and the importance of medication continuity during floods. (ASHA’s or other healthcare providers in community or health facility face-to-face or calls)	- Binary - Patient questionnaire	- Baseline - Post-intervention - Immediate post-flood period (S1 and S3)	Y	I	Y

S1 and S2 are baseline surveys immediately after the flood period, and S3 and S4 are after the follow-up, respectively; see the Data collection procedures section for clarification

CEI community engagement and involvement, FGDs focus group discussions, PHC primary healthcare, SSI semi-structured interviews. Baseline refers to immediate (S1), and six months after flood period (S2) pre-intervention. Follow-up refers to post-intervention immediate (S3), and six months after flood period (S4)

Fig. 2.

Steps in the hybrid implementation effectiveness parallel cluster randomised control trial

Primary outcomes

The effectiveness primary outcome will be the patient self-reported medication adherence and refill as measured by Adherence to Refills and Medications Scale (ARMS)−12, a licensed, validated scale [36] assessed at immediate post-flood patient surveys (baseline survey 1 (S1) and post-intervention survey 3 (S3), Fig. 2). ARMS-12 is a tool developed for use in low-literacy settings [36]. It is a 12-question scale with a maximum score of 48, and lower scores indicating better adherence. We will use this as a binary variable with a cut-off score of 16.32 for adherence [36, 37].

The implementation primary outcome will be the status of the observed patient’s flood-proof kit. It will be assessed at immediate post-flood patient surveys (baseline survey 1 (S1) and post-intervention survey 3 (S3), Fig. 2) through observation and back-up questions. The indicators to be assessed and fulfilled include the observed presence of a flood-proof storage method (e.g. bag or box), with availability of drugs or confirmation from the patient that the storage bag or box held medications during the flood period; the presence of an NCD card or other handover document within the storage method with documentation of minimum required information, such as patient name, diagnosis, a dated medication record; and a blood pressure reading within the last 6 months. If any one of the above criteria is not fulfilled, the kit will be classified as absent or inadequate.

Study sample size and power

Interference approach

The study will assess two co-primary outcomes, and the intervention will be considered successful only if both demonstrate a measurable effect. This approach mitigates the need for alpha multiplicity adjustment but has been allowed in the power calculation (below).

Power

The power calculations accounted for the clustered design of the study and assumed a two-period decay correlation structure, which is suitable for a cRCT, where the correlations between observations made within the same clusters, but at different time periods, are expected to decay [38, 39]. The clustered design and correlation structure have been incorporated in the sample size calculations using the intra-cluster correlation (ICC) and cluster-autocorrelation (CAC), respectively.

The co-primary outcomes are binary. There is no information available to determine the prevalence of each outcome, but it is not expected to be extremely low or high. Power calculations assumed a prevalence of 50% as a conservative approach (meaning if the prevalence is higher or lower than 50%, we will have increased power at the same sample size to detect a difference of the same magnitude). Target effect sizes for both outcomes are a 20% absolute increase (for example, 50 to 70% or 30 to 50%). We assumed a cluster size at each cross-section of 70 participants, and 26 clusters. All clusters will be included in the primary analysis, irrespective of whether a flood occurs or the intervention is implemented. The anticipated total sample size is 3640 (cluster size × number of clusters × observation points). There is expected to be minimal variation in cluster sizes, as within each cluster, 70 participants will be recruited from a much larger eligible sample (cross-sectional sampling), and recruitment will continue until 70 participants are reached. There is no information available to guide ICCs or CACs. We therefore considered anticipated power across a range of plausible values. The assumed values for these correlations have been guided by correlations for similar outcomes in similar settings and general patterns in the literature [40, 41]. To this end, we have assumed an ICC of 0.05 but considered sensitivity across the range 0.03 to 0.1. To allow for the repeated cross-sectional measures, we have allowed for a cluster-auto correlation coefficient of 0.8 and 0.9, again guided by the literature [42].

In all power calculations, we assumed a 5% significance level, and the power for each outcome was multiplied by the power for the other outcome (to determine the power for both being statistically significant). All power calculations assumed the same absolute effect on both outcomes. Power calculations have been implemented in https://clusterrcts.shinyapps.io/rshinyapp/ [39]. Across all scenarios (with CACs of 0.8 and 0.9 and ICCs of 0.03 to 0.1), we have over 90% power to detect an absolute difference of 20% (see online supplement S1).

Randomisation and sampling

Sequence generation, allocation concealment, cluster allocation, and blinding

In the first year of the study (completed by October 2024), 34 CDBs were identified from the three districts using a checklist (see Fig. 3). This process captured accurate data on cluster characteristics such as population size, a functioning community health centre staffed by a clinical healthcare professional, and compliance with predefined inclusion and exclusion criteria. Prior to baseline data collection, 26 CDBs will be selected from the identified 34, in consultation with local community leaders. All 26 clusters will be randomly assigned to the intervention or control arm at a 1:1 ratio and following baseline assessments in clusters and before intervention implementation revealed to the health system. Allocation will be concealed, and the randomisation will be performed by an independent statistician using a computer-based restricted randomisation algorithm (using the Stata version v19 or higher) to ensure balance between arms in terms of population size. After randomisation, the allocation will be revealed to the health system and community stakeholders so that intervention delivery can be implemented in the selected clusters. A restricted randomisation algorithm will create a balance across intervention and control clusters on CDB population size served and average distance of wards from the nearest prescribing health facility. The trial is open-label, and double blinding is not possible due to the nature of the health system intervention. No unblinding procedures are required. The participants and the healthcare providers will be aware of the interventions. Data will be collected by independent research assistants. The statistician conducting the analysis will remain unaware of the intervention and control group allocation codes.

Fig. 3.

Participant timeline

Participant recruitment at baseline and follow-up, cross-sectional survey rounds post-flood

A baseline survey will be conducted in study clusters at the end of October 2025, or earlier if flooding ends sooner (typically from July to early October; see the Data collection procedures section below) (designated as baseline survey 1 or S1 as per Fig. 3). Participants will be recruited via door-to-door household surveys. Consecutive houses on a randomly selected street or lane within the selected cluster will be approached for the survey. In selected households, eligible adults will be identified and consented. Where multiple eligible individuals are present, a maximum of two will be randomly selected using simple randomisation (e.g. drawing lots). Two attempts will be made to collect data from any eligible individual absent during the survey. Where a carer takes care of medication and other arrangements or costs of a patient’s healthcare, the carer will be invited to help with survey responses. Recruitment will continue until the target sample size is reached. The same participants will be surveyed again 6 months later (baseline survey 2, S2). Post-intervention, another group of cross-sectional survey participants will be recruited with a 6-month follow-up of that cohort (first follow-up survey designated S3 and the second follow-up survey designated S4; Fig. 3). Baseline and post-intervention follow-up surveys are cross-sectional and therefore participants will be selected independently at each time point at S1 and S3, with a unique ID collected for tracking repeat measures at 6 months for S2 and S4. Figure 3 (SPIRIT) shows details on the schedule of enrolment, interventions, and assessments.

Data collection procedures

All tools will be field-tested and piloted. Pilot of the patient survey tool in one village after translation into the local language was done with 60 participants, before electronic implementation on tablets with a REDCap database. Fieldworkers (FWs) will also measure blood pressure (see Table 1).

FWs will undergo a minimum 2-week training, including three or more supervised field visits. Data collection will occur twice annually for each baseline and post-intervention follow-up flood year to minimise recall bias and capture both short- and medium-term flood impacts. Within each year, the first survey will occur within 2–6 weeks (over approximately 4 weeks, S1 and S3) after state authorities close community kitchens/camps, and all houses in the ward can be reached on foot. A follow-up survey, conducted about 6–9 months later (over 4 weeks, S2 and S4), before June 2026, when floods may begin again, will assess NCD exacerbations and healthcare visits since the last flood period.

FWs will conduct home visits in pairs. During S1 and S3, trained FWs will use the developed tools to identify patients with NCDs, collecting data on their conditions, patient and family characteristics, and outcomes during and since the recent flood period. The S2 and S4 visits will focus on visits to healthcare facilities due to exacerbations (Table 1) and preparation for the next flood. Figure 4 describes each data collection round with approximate dates.

Fig. 4.

Data collection rounds with approximate dates

Informed consent

Written informed consent (Box S1 Model consent and information sheet; Online supplement) will be obtained from all study participants for each data collection (surveys/interviews). Each participant will receive a participant information sheet detailing the study purpose, procedures, and contact information for both the principal investigator and the Member Secretary of the local Institutional Ethics Committee prior to providing consent. Necessary approvals for authorising the cluster-level interventions have already been secured from the Bihar State Government, relevant district administrations, and local self-government bodies. Participants cannot opt in or out of cluster-level health system improvements.

Data processing and final dataset

Statistical analysis

A complete analysis, taking into account the clustered nature of the design, will be conducted in accordance with the CONSORT reporting guidelines [43], illustrating the flow of clusters and participants throughout the trial. Characteristics will be compared across study arms; however, no statistical testing for differences will be undertaken.

All analysis and reporting will allow for the clustered nature of the design and will be analysed, preserving the randomisation irrespective of adherence both at the individual and cluster level. Clustering will be allowed for in all analyses through the use of random effects [44]. We will employ the constrained baseline approach [45]. Under a constrained baseline analysis, each observation (for an individual) has a binary indicator for time, indicating whether the observation was made during the baseline phase or the post-randomisation phase (reflecting change in the control arm over time), and a binary indicator for treatment, indicating exposure to the intervention. The constrained part of the model assumes no baseline differences, as is recommended in cluster trials, where baseline assessments of outcomes are conducted [38]. Furthermore, given only 26 clusters, we will use methods that have been proven to estimate the standard errors accurately [44]. The Satterthwaite small sample correction will be used to correct the standard errors for all analyses, as there are expected to be 26 clusters in the trial [46]. For binary outcomes, we will report both relative and absolute measures of effect. To this end, for the binary outcomes, we aim to fit a mixed-effect logistic model to estimate the odds ratios and then use marginal standardisation to obtain the risk difference. If a participant discontinues during data collection, all data collected up to that point will be retained for analysis under the intention-to-treat principle. If a participant withdraws consent for use of their data, all of their data will be removed from the dataset and excluded from analysis. For continuous outcomes, we will report mean differences and use generalised linear mixed models, with random cluster effects, using REML methods. Plans will be put in place in case of model non-convergence, which can arise when using generalised linear mixed models, especially with uncommon binary outcomes. Firstly, we will fit the generalised linear mixed effects model without the cluster-by-period random effect. If this analysis does not converge, we will use a weighted cluster-level approach analysis. Moreover, for the analysis of uncommon binary (secondary) outcomes, we will use methods that increase the likelihood of model convergence and increase stability through the use of propensity score-based adjustment [47]. All analyses will adjust for factors used in the restricted randomisation procedure as per recommendations and baseline measures of the outcomes [48]. A secondary covariate-adjusted analysis will also be presented, using a pre-specified set of covariates which are deemed to be of clinical importance. Missing data will be reported and explored. We will not conduct imputations for outcomes but will investigate the need for imputation for other characteristics based on the number of missing values. For missing covariate data, the fully adjusted covariate analysis will plan to use multiple imputation, with the number of imputations dependent on the percentage of missing data and again allowing for the clustered nature of the design [49]. For a small number of flood contingent outcomes (see outcomes Table 1), analysis will be undertaken in the subset of clusters which have been affected by flooding (see definition in section on primary outcomes; i.e. where community kitchens/camps have been established by authorities). Sub-group analyses and details of the analysis will be described in the statistical analysis plan. The primary outcome will be regarded as significant at the 5% level, and effect sizes along with 95% CIs will be reported. Results concerning secondary clinical, healthcare, and implementation outcomes will be interpreted with caution. For this reason, treatment effect measures with confidence intervals will be presented solely as supporting evidence.

Qualitative/mixed-methods analysis using RE-AIM-QuEST framework

We will use the RE-AIM-QUEST framework [28] and NPT [50] to guide a mixed-methods formative and embedded process evaluation. We will conduct semi-structured in-depth interviews, key informant interviews, and focus group discussions with a diverse range of stakeholders involved in planning and implementing the NCD intervention package for floods. Participants will include patients enrolled in the intervention cluster (adherent and non-adherent groups), frontline health workers such as Accredited Social Health Activists (ASHAs) and Auxiliary Nurse Midwives (ANMs), community health officers, medical officers, and district and state-level health and disaster management officials. A separate qualitative process protocol, detailing the study design, data collection methods, and analysis plan, will be produced separately.

We will integrate baseline and follow-up cluster RCT data with concurrently collected qualitative data guided by the NPT [50] and RE-AIM frameworks [28]. NPT will guide analysis of barriers and facilitators for embedding the intervention routine practices among patients and primary care providers, supporting replication and scaling of the intervention to enable a resilient community and primary healthcare systems during flood periods. RE-AIM will be used as a framework to evaluate implementation outcomes, with qualitative methods employed alongside quantitative measures to examine contextual factors influencing the intervention’s reach, adoption, implementation, fidelity, and sustainability.

Health economic analysis

Exploratory health economic decision modelling will be used to estimate the cost-effectiveness of the complex intervention compared with usual care over the longer term, adopting both a healthcare and societal approach. Outcomes will be in terms of incremental cost per life year gained, quality-adjusted life year gained, disability-adjusted life year avoided, and major clinical event averted. The model will be populated with data from the trial and published studies and supplemented with estimates from expert opinion. Specific trial data required by the model are participant data on demographics and pre-existing NCDs, and survey information at all time points on healthcare utilisation (inpatient admission, outpatient consultations, medications) and out-of-pocket expenses. Trial data will also inform the full costing of the complex intervention.

A separate health economics protocol and analysis plan with full details of the model structure, data inputs, and methods for analysis will be produced separately.

Ethics

Research ethics approval

In India, full ethical approval was granted by the Institutional Ethics Committee, Sree Chitra Tirunal Institute for Medical Sciences and Technology, India (Ref: SCT/IEC/2142/DECEMBER/2023). The study also received full ethical approval from the University of Birmingham Science, Technology, Engineering, and Mathematics (STEM) Research Ethics Committee (ERN_2178-Dec2024).

Protocol amendments

Amendments to the protocols can only be made by the chief and/or principal investigators or their designees, with agreement from the sponsors. Substantial amendments will also be reviewed by the project’s oversight/steering committees, stakeholders, and must be approved by the Indian ethics committee. Non-substantial or administrative amendments will be documented but do not require formal approval.

Confidentiality

To safeguard patient privacy and confidentiality, all surveyors will receive training in Good Clinical Practices and relevant regulatory guidelines. Data collection is conducted using unique identification numbers and secure data collection platforms that comply with established data protection standards in India. All datasets undergo deidentification and data sanitisation processes to remove or obscure personally identifiable information. A restricted set of deidentified and sanitised data will be shared with consortium partners.

Data management, retention, and storage

Detailed data management plan is available online [51]. Briefly, survey data will be entered in a structured electronic case report form and will be stored in REDCap hosted at one of the partner institutions in India. Data access will be restricted to the investigator team and encrypted both during transmission and at rest. Data quality will be ensured through built-in validation rules, routine data quality checks for completeness and missing data, double verification of all out-of-range and inconsistent values, central monitoring, and ongoing supervision and training of field staff. Data exported from REDCap during analyses will be stored in a secure folder on the internal network drive. All electronic files and the final cleaned dataset will only be accessible to the investigator team. Data will be retained in accordance with institutional policies and stored in a secure format to ensure confidentiality and integrity throughout the retention period.

Oversight and monitoring

Data monitoring committee

We will not specify any ‘stopping rules’ for the trial as the intervention is community-based, non-invasive, and strengthens the application of evidence-based health systems interventions and presents no risk of harm to the study participants. The role of the data monitoring committee is therefore to monitor trial conduct and data quality rather than identify safety events. The Trial Steering Committee (TSC), with an independent chair and members, will constitute the data monitoring committee for the trial, comprised independent experts in climate change, NCDs, and health systems. The steering committee will meet 5–7 times over the study and trial duration. Prior to trial commencement, it will advise on the design of the study and analysis. Following the completion of baseline data collection, the committee will review a summary of the data collected and the data quality indicators to be specified. The committee will also meet to repeat the review following the completion of each subsequent wave of the study.

Dissemination

Primary endpoint data analysis will begin immediately after the last survey participant. The outcomes and results of this trial will be published. An extensive list of stakeholders, including policymakers and community members, will also receive updates via newsletters and the results at the end of the study. Authorship for all trial-related publications will be determined in accordance with the International Committee of Medical Journal Editors (ICMJE) criteria. All authors will contribute to manuscript development, will review and approve the final version, and professional medical writers will not be used.

Discussion

The FUSION trial will use a hybrid effectiveness–implementation design with mixed methods, including process and economic evaluations, to improve continuity of NCD care during the annual flood season in Bihar. The intervention will be co-designed with healthcare providers, community members, and local authorities, and seeks to strengthen health system preparedness through improved clinical pathways, supply chain resilience, communication systems, and integration of meteorological and hydrological flood forecasts into service delivery planning.

This will be one of the first trials to integrate a health system–driven NCD service delivery model during floods in India. If effective, the model has potential for scale-up to other flood-prone regions nationally and internationally. The qualitative process evaluation will identify facilitators and barriers to implementation, assess fidelity, and inform adaptation for broader application. The economic evaluation will provide preliminary cost and benefit estimates to support policy and scale-up decisions. By addressing both service delivery and disaster preparedness, this intervention has the potential to improve health outcomes, reduce complications, and enhance quality of life for people living with NCDs in high-risk flood settings.

The intervention is designed for feasibility and sustainability within existing primary healthcare structures, using local governance and community health workers. If successful, it could be embedded in state disaster management and health service delivery plans, adapted for other climate-related emergencies, and extended to other chronic conditions.

Trial status

V1 29.09.25—the trial protocol was submitted to BMC Trials.

Recruitment for baseline survey will begin by October 2025 and will finish in June 2026.

Intervention implementation is to begin in July 2026.

Abbreviations

NCDs

Non-communicable diseases

LMIC

Low and Middle income countries

NPT

Normalisation Process Theory

RE-AIM QuEST

Reach, Effectiveness, Adoption, Implementation, and Maintenance Qualitative Evaluation for Systematic Translation

CVDs

Cardiovascular diseases

CRDs

Chronic respiratory diseases

CHB

Community health blocks

FW

Field workers

Authors’ contributions

SMH, PJ, SG, SB (1), RA, SB (2), NB, DP, AG, SH, SK, RJ, SJ, SM, PP, BR, RS (1), JVT, LJ, MMS, AS (1), AS (2), MW, and BKY conceived the project. SMH drafted the first version of this protocol; PJ, RJ, SG, SB (1), SJ, RS (2), PB, and LJ reviewed and revised the methodology. SMH, PJ, RA, SB (1), SB (2), NB, DP, AG, SG, SH, SK, RJ, SJ, NK, SM, PP, BR, RS (1), MMS, GR, TS, AKS, AS (1), AS (2), MW, JVT, LJ, and BKY developed the initial design, data acquisition, and analysis plan. AS (2) and JYT provided the statistical analysis plan. SMH, PJ, SG, SB (1), RS, PP, RJ, SJ, JYT, SH, MW, and LJ revised subsequent versions of the protocol. All authors read and approved the final version of the manuscript. *SB (1)—Steven Blackburn, SB (2)—Sunny Borgohain, RS (1)—Ravikant Singh, RS (2)—Rahul Shidhaye, AS (1)—Ashuthosh Sharma, AS (2)—Alice Sitch.

Funding

A research grant from the National Institute for Health and Care Research (NIHR) UK funded this work (NIHR204871). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. PJ received research grants from the National Health and Medical Research Council (NHMRC, Australia), the National Heart Lung and Blood Institute (NHLBI, USA), the Indian Council of Medical Research (ICMR, India), the Department of Science and Technology-Government of India (DST), and the Trivandrum Nephrology Club. PJ is also a senior clinical and public health fellow of the DBT-Wellcome Trust India Alliance (IA/CPHS/20/1/505229). The views expressed in this publication are those of the authors and not necessarily those of the NIHR or the UK Department of Health and Social Care.

Data availability

The custodian of the final dataset will be the Joint-PI from India. Raw data will be controlled by selected staff in India with the approval of the Joint-PI from India. Anonymised and sanitised data will be shared with selected staff in the UK for analysis. The India and UK teams will maintain exclusive use of the data until the data are published in an open-access journal. Afterwards, data will be made available by formal request to the custodian.

Declarations

Ethics approval and consent to participate

In India, full ethical approval was granted by the Institutional Ethics Committee, Sree Chitra Tirunal Institute for Medical Sciences and Technology, India (Ref: SCT/IEC/2142/DECEMBER/2023). The study also received full ethical approval from the University of Birmingham Science, Technology, Engineering, and Mathematics (STEM) Research Ethics Committee (ERN_2178-Dec2024).

Consent for publication

This manuscript does not contain individual personal data from patients.

Competing interests

The authors declare that they have no competing interests.