Multidisciplinary approach to COVID-19 risk communication: a framework and tool for individual and regional risk assessment

The study addresses how to assess and communicate COVID-19 risk at both individual and regional levels to support targeted, optimized public health and socioeconomic decisions beyond blanket lockdowns. In the context of a rapidly spreading pandemic, with human-to-human transmission and widespread lockdowns (including Nepal), the authors argue that sustained total lockdowns are not feasible given their economic and social consequences. The purpose is to develop an evidence-based, multidisciplinary framework and tool that estimates personal vulnerability and infection probability as well as regional transmission and overall risk in near real-time, and to communicate these risks to guide behavior and policy (e.g., prioritizing essential services, easing restrictions in low-risk zones). The importance lies in enabling nuanced, data-driven decisions that balance disease control with maintaining essential economic and social functions, particularly in low-income countries with limited testing capacity and resources.

The paper references WHO guidance for pandemic response (hospital capacity, public awareness, supplies) and literature on risk perception and communication, noting their role in reducing fear and encouraging protective behaviors during pandemics. Studies from China indicated psychological impacts of COVID-19; research on institutional and individual behaviors highlights the effects of containment, testing, and contact tracing, and the adoption of early strict lockdowns in LMICs like Nepal. Economic literature and UN reports document disrupted supply chains and exacerbation of poverty and hunger due to lockdowns, underscoring the need to balance health measures with economic activity. Additional studies inform occupational risk (drivers, health workers, etc.), the protective effects of masks, distancing, and eye protection, and demographic risk factors (age, comorbidities, gender). Prior risk assessment and governance tools are cited to position COVIRA as a dual risk assessment and communication platform, contributing to risk zoning, personal risk estimation, and targeted mitigation strategies.

The framework integrates three data streams: (1) daily COVID-19 updates (cases, quarantines) to parameterize transmission and update maps; (2) static/regional data (demography; health conditions and facilities; socioeconomic indicators including poverty, literacy, WASH; population density; exposure facilities such as airports and hotels; food production; supply-chain hubs); and (3) individual questionnaire responses for personal risk. Personal risk assessment comprises: (a) individual vulnerability (danger to life if infected), (b) probability of infection given exposure and local risk, and (c) transmission risk behaviors from questionnaire data. Vulnerability is modeled mainly via age, gender, and comorbidities using international datasets (China, Italy, Spain, Germany, USA, UK). An exponential age-risk relation is fit: y = a·e^{βx}, where x is age normalized (mean 45.11, SD 27.32), with a=8.947 (95% CI 5.955–11.94), β=1.492 (1.272–1.712), R^2=0.9538. Comorbidity risk factors (CRF) are derived from reported severe outcomes and mortality by condition; a comorbidity coefficient z is modeled as z = a·x^2 + b·x + c with a = −3.40646579722815e−5 (95% CI −4.477e−05, −2.336e−05), b = 0.00672965343817772 (0.00563, 0.007829), c = 0.635624565570039 (0.6123, 0.6589), R^2=0.9994. The COVID Risk Index (CRI) is computed as CRI(%) = y · [(1 − z) + z · (CRF/100)]. CRF values (scale 0–100) include: asthma/COPD 100; cardiovascular/coronary heart disease 99; renal disease 90; diabetes 69; hypertension 54; obesity 54; cancer 54; cerebrovascular disease 48. A gender adjustment applies: with male risk factor set to 100, female total risk is multiplied by 0.4224. CRI is mapped to qualitative levels using a five-point scale: Very Low (0–6), Low (6–15), Moderate (15–28), High (28–48), Very High (48–100), based on equal area under the risk curve partitions. Probability of infection is modeled as the product of local regional risk and personal exposure: Probability of COVID-19 infection = regional risk × exposure, upper-bounded at 100. Symptoms (e.g., fever, cough, shortness of breath, fatigue) inform advice to seek care but are not used to directly calculate probability due to asymptomatic cases. Exposure is quantified via occupation and activities: occupation exposure values (0–100) include, e.g., drivers/transport 100, nurses/health workers 67, shop salespersons 50, teachers 40, agriculture 20, none 0. Activity-related exposure includes recent travel to infected areas 100, meeting a known infected person 100, close proximity in workplace 90, frequent public areas 80, in-person meetings 70, close proximity to colleagues 60, family members with outside contacts 25–50. Preventive measures reduce exposure (deductions): mask −14, hand hygiene −14, physical distancing −14, eye protection −8; with minimum exposure floor at 50, full adherence can effectively nullify net exposure. Regional risk assessment includes: (1) COVID-19 Transmission Risk (CTR), (2) Public Health Risk (PHR), (3) Socioeconomic Risk (SER), and a separate Regional Importance (REG) mapping for essential services. CTR combines positive case score (PCS), quarantined score (QNT), community exposure (EXP), and population density (POD): CTR = 0.6·PCS + 0.1·QNT + 0.1·EXP + 0.2·POD. PCS is on 0–100 relative scale: zero if no cases; first detected case assigns 50 (reflecting immediate community risk), increasing logarithmically toward 100 as cases approach the number of households. Formula: PCS = 100 − log(a/b) · 50 / log(1/b), where a = active cases; b = number of households (estimated via average household size 4.67). Spatial spillover considers adjacent regions within 10 km as equivalent active cases, and regions within the next 10 km buffer at half weight. Temporal decay reduces PCS by 50% after 60 consecutive days with no cases and nullifies after 90 days. QNT follows a similar scoring logic. EXP at regional scale is based on facilities combinations (international arrivals, domestic airport, 3+ star hotels, other hotels) with scores (e.g., all present = 100; subsets 90, 80, 70, 60, 50). POD is linearly scaled with a minimum score of 20. PHR reflects population with underlying health conditions (UHC) and availability of health facilities (HF) per population (hospitals weighted 100, primary health care centers 10, health posts 1). Both UHC and HF are normalized 0–100; PHR = UHC − 0.2·HF, assuming 20% of severe cases can access critical services. SER integrates poverty (PVI), literacy (LTR), and WASH indices: SER = 0.4·PVI + 0.2·LTR + 0.4·WASH, reflecting greater weight on poverty and WASH as slower-changing determinants of vulnerability. Regional importance maps independently assess food production (FOOD; rice, wheat, maize with seasonal timing) and supply-chain network hubs (SCN) scored as: capital city 100, provincial HQ 90, major hubs 70, district HQ 50, others 30; these inform prioritization during restrictions. Total Risk Score (TRS) combines CTR with modifiers PHR and SER: TRS = CTR · (0.6·PHR + 0.4·SER) / 100. Mapping and analytics are implemented in R (v3.5.3). Data sources include international case/outcome datasets, Nepal national statistics, and daily updates from Nepal’s Ministry of Health and Population; administrative boundaries from Survey Department Nepal. Where precise statistical data were lacking, a Delphi approach was used to parameterize components and equations.

- The COVIRA tool operationalizes a dynamic, near real-time risk assessment at personal and regional levels, producing scores on a 0–100 scale and qualitative categories for communication. - Personal risk: Age exhibits an exponential relationship with mortality risk (y = 8.947·e^{1.492x}, R^2=0.9538). Comorbidities substantially elevate risk; derived CRF values (0–100) include asthma/COPD 100, cardiovascular disease 99, renal disease 90, diabetes 69, hypertension/obesity/cancer 54, cerebrovascular 48. A comorbidity coefficient z (quadratic in age) adjusts the contribution of CRF to total risk; female risk is scaled by 0.4224 relative to male. CRI thresholds for qualitative risk levels are: Very Low 0–6, Low 6–15, Moderate 15–28, High 28–48, Very High 48–100. - Probability of infection for an individual is estimated as the product of local regional risk and personal exposure, with exposure derived from occupation and activities and reduced by preventive behaviors (mask, hygiene, distancing, eye protection). Symptom profiles (fever, cough, shortness of breath most common) prompt clinical advice. - Regional risk: Pre-pandemic maps show Kathmandu at high transmission and overall risk; southern border regions with India display higher CTR owing to cross-border exposure and population density. As cases evolved (May 10, 20, 30; June 10, 2020), observed spread aligned with modeled higher-risk regions, indicating predictive coherence between base risk maps and subsequent case emergence. - Public health and socioeconomic risk layers highlight areas where limited health services, higher poverty, lower WASH, and literacy compound vulnerabilities. - Regional importance mapping identifies critical food production areas and supply-chain hubs to prioritize during movement restrictions. - A publicly accessible web application (http://www.covira.info) was launched (June 21, 2020) to deliver personalized risk results, local transmission risk, and tailored guidance; primarily configured for Nepal but generalizable globally.

The framework addresses the need for actionable, fine-grained risk information to guide both individual behaviors and regional policy decisions, especially where extensive testing is infeasible. The strong age-risk relationship, significant comorbidity effects, and exposure-based infection probability allow tailored individual guidance (e.g., heightened precautions for high CRI; feasible continuation of work for individuals with low vulnerability but high transmission risk if appropriate measures are taken). Regional CTR, PHR, and SER layers, combined into a total risk score, enable authorities to prioritize interventions, maintain essential services, and judiciously ease restrictions in very-low-risk areas while maintaining vigilance. Temporal and spatial dynamics (buffers, decay of PCS after periods without cases) reflect transmission realities and mobility patterns, improving situational awareness. Comparisons between pre-pandemic base risk maps and later case distributions show that higher observed incidence occurred in regions flagged as higher risk (e.g., Kathmandu, border areas), supporting the framework’s validity for forecasting and planning. The tool’s risk communication component provides concise, personalized protective advice, potentially improving risk perception and adherence to preventive behaviors, and facilitating safer mobility and reopening strategies.

COVIRA provides a multidisciplinary, holistic risk assessment and communication system for COVID-19, integrating personal vulnerability and exposure with regional transmission, public health capacity, and socioeconomic vulnerability. Validated against evolving case patterns in Nepal, it supports targeted public health measures, prioritization of essential services, and informed individual decisions, offering a pathway to modulate restrictions without compromising public health. The framework and tool are readily adaptable globally, as they rely on widely available data types and modular components. Future work could incorporate additional demographic and contextual factors (e.g., ethnicity, vaccination status as applicable), refine parameters using country-specific datasets, enhance validation across diverse settings, integrate mobility data streams, and extend the approach to other infectious diseases and disaster scenarios.

- Data availability and timeliness: Regional/zonal risk is dynamic and depends on frequently updated inputs (cases, quarantines, immigration), which may be delayed or incomplete. - Use of international datasets: Personal risk models were calibrated using data from China and several high-income countries, which may not fully generalize to low-income contexts like Nepal. - Crowdsourced inputs: Personal risk relies on self-reported questionnaire data; incorrect entries affect individual scores (though regional risk is independent of these inputs). - Parameterization: Some components lacked precise statistical bases and were parameterized using a Delphi method, introducing expert-judgment uncertainty. - Symptom-based infection inference: Asymptomatic transmission limits the diagnostic value of symptoms; the tool therefore uses symptoms primarily for advising clinical contact rather than probability estimation. - Unmodeled factors: Potential contributors such as detailed occupation-specific practices, household structure heterogeneity, and real-time mobility patterns are simplified or not explicitly modeled.