Regional Contact Networks and the Pandemic Spread of COVID-19 in India

Predictions to enable State-level Surge Preparedness in moving from Containment to Mitigation

COV-N Study Group: Rupam Bhattacharyya, Shariq Mohammed, Veera Baladandayuthapani (University of Michigan); Sayantan Banerjee (Indian Institute of Management, Indore) and Upali Nanda (HKS Inc and University of Michigan)

Overview

Since the first reported case of COVID-19 on November 17, 2019, in Hubei, China, the virus has spread to 209 countries and territories, afflicting > 2 million people — killing over 135,000 of those [as of April 16th], sending billions into lockdown and ravaging economies and health care infrastructure with each passing day. This is a pandemic at a scale that the living generation has rarely seen. The novel viral strain, SARS-CoV-2, is highly contagious and hence easily spreads when human come in close contact. This motivates a closer ad rem look at the inherent dynamics of the spread at a micro-scale and assess its multi-fold ramifications on the cultural, economic and health infrastructures. Using rapidly evolving COVID-19 individual-level data and a contact network framework, we attempt to answer some critical questions that can inform a phased approach to dealing with the pandemic by appropriate resource prioritization and allocation.

Nuanced modeling can aid more rapid and focused containment and subsequent mitigation with regional specificity, minimal disruption to the overall economy and sensitivity to humanitarian concerns.

We have developed a visualization tool for the general community to obtain, inspect, rerun, or possibly improve on our models available at: https://bayesrx.shinyapps.io/COV-N. We take India as a case study to exemplify this model where an extensive nationwide lockdown is currently in place. We hope that our models will allow India to adapt its approach with surgical efficiency, proportional to risk and resource availability.

Given the lack of tests and the large number of asymptomatic carriers, the strategy for slowing the spread of the COVID-19 pandemic has changed from containment to mitigation focused on slowing the further spread of the virus, reducing the anticipated surge in health care use, providing patients with the right level of care to maximize the likelihood that the majority of patients will only require time-limited home isolation, expanding testing capability to increase available hospital capacity, and tailoring isolation to minimize transmission of SARS-CoV-2.

Chapter 0: Epidemics and Networks- Why Networks Matter

Image for post
Image for post
Figure 1: A simple illustration of an epidemic network across time. Simulated from a network Susceptible-Infected-Recovered (SIR) model, where the nodes are individual patients: blue are susceptible, red are infected, and green are recovered individuals. One time-step in the figure represents 10 units of time in the original scale of transmissions.

Containment procedures such as lockdowns, social distancing, smart quarantines effectively break this human to human “link” to disconnect the infected network, and hence confine and reduce the spread. Subsequently, using contact networks developed with high geo-specificity and resolution, one can learn specific spread mechanisms and inform precise action planning for mitigation and building health care capacities. See Figure 2, where we show a network defined by geographical region for India.

Image for post
Image for post
Figure 2: Network of patients infected with COVID-19 across Indian states and union territories, with each dot representing one patient and the color of the dot representing the state in which the patient was first detected and confirmed as a case. Edges are defined by whether two patients are located within a band of two degrees of latitude and longitude of one another. Note: This map is used for illustrative purpose only. Actual territorial boundaries might differ. We have removed the edges between nodes that are too close for visual clarity

Classical epidemiological models attempt to estimate a basic reproduction number (R0) to describe the intensity of an outbreak which is defined as the average number of infectious individuals generated by one infected individual in a fully susceptible population. Emerging studies have put this number around 2 — essentially doubling per each infected individual — with some studies reporting up to 3.8. The reproduction number depends on the expected degree of the network, the latter being the average number of contacts a node has in the entire network. The transmission rate of the pathogen varies in different regions of a network (for example, in regions with high population density, or port of entries), and as those sub-networks are made more sparse by reducing contact and identification followed by isolation via testing, the chain of contacts break and the average degree comes down, thereby reducing the transmission rate.

To take lessons from the past, eradication of smallpox was vastly possible via network-based policymaking, and mass vaccinations were a failure in areas having higher population density. Containment of epidemic hotspots and identifying nodes with higher contacts played the most vital roles to eradicate the disease. In the absence of a vaccine, policymakers may look into focusing on increased levels of testing and isolation strategies and protocols for hotspots to effectively flatten the curve. Since the effects of a complete lockdown have economic and mental health consequences, India may concentrate on enforcing strategized lockdown in sub-networks having higher infection concentration i.e., in the ‘hotspots’.

Chapter 1: What is next and how state-level responses can change the future

(a) Scenario 1: Low intensity spread: assuming there is strict adherence to containment and mitigation protocols (R0=1.5),

(b) Scenario 2: Medium intensity spread: assuming there is sporadic adherence to containment and mitigation protocols (R0=2),

(c) Scenario 3: High intensity spread: assuming minimal adherence to containment and mitigation protocols (R0=2.5).

Under these scenarios we try to answer the following questions:

  1. Forecast the number of affected individuals for each state in India for the next three months and which states emerge as hotspots based on emerging geo-coded data.
  2. What is the potential impact on health care systems and facilities?
  3. What measures can be taken to mitigate and contain the contact networks.

State-level projections for India and how some states got it right!

Figure 3 shows the predicted number of active COVID-19 cases across Indian states and Union Territories between April 15 to July 1, 2020 along with dynamic bar plots exhibiting the growth in active case counts. Some evident trends emerge. Most states see a significant rise in the number of cases starting around mid-May. There are clear hot-spots emerging in a first cluster of states that include Maharashtra, Delhi, Tamil Nadu and Rajasthan with a second cluster (close behind) emerging in Uttar Pradesh, Gujarat, Madhya Pradesh, Telangana, Andhra Pradesh and Kerala. While a number of these initial hotspots could be traced to international travel and port of entries, some newly formed hotspots have emerged which can be potentially attributed to hub-and-child nodes of contact networks through travels and public gatherings.

Image for post
Image for post
Image for post
Image for post
(a) Low intensity spread (R0=1.5)
Image for post
Image for post
Image for post
Image for post
(b) Medium intensity spread (R0=2)
Image for post
Image for post
Image for post
Image for post
(c) High intensity spread (R0=2.5)

Figure 3: Left panel shows the predicted number of active COVID-19 cases across Indian states and union territories based on Network SIR (susceptible-infected-removed) models. The initial network for each state contains the infected individuals, and an additional number of susceptible individuals so that the total size of the network matches the population of each state. Right panel shows the dynamic bar plot of predicted number of active cases across Indian states and union territories between April 15 and July 01, 2020.

While these predicted numbers are in the hundreds over the next few weeks, they rise multi-fold to thousands to lakhs, perhaps even crores — if the current rates continue under a high intensity scenario. This projection is not to be alarmist, but rather to stress the need for sustained response. Figure 3 show the number of affected cases increasing substantially in late May through June, where some states cross 10 lakhs of cases by mid-June. Unsurprisingly, states with high population density (e.g. Maharashtra, Rajasthan, Madhya Pradesh and Karnataka) are seen to be extremely susceptible to see a spike in the number of cases. Some interesting case-by-case analyses follow.

Kerala had reported the first case in the country on January 30, the patient being a student who had traveled from Wuhan, China. The state, which is a favorite among tourists, witnesses a massive foreign footfall every year. Also, it has one of the highest percentages of expatriates and domestic migrant workers. These served as points of concern in the current scenario making Kerala one of the most vulnerable states. To tackle this, the state directed serving a mandatory isolation period for all people traveling from abroad. Kerala aggressively put itself on war-footing to combat the upcoming peril by mobilizing its strong healthcare workforce in efforts towards vigorous testing, contact tracing to finer levels, increasing duration of quarantine, and ensuring that the migrant workforce is sheltered and well-fed. Kerala is showing promising results, with the percentage of new cases on the decline, along with a steady increase in recovery rate.

In comparison, Maharashtra, the worst affected state currently, had registered an almost similar number of cases by the end of March but started performing poorly, primarily owing to insufficient testing, tracking, and isolation. With its capital Mumbai being the commercial hub of the country, a significant portion of infections in the population was left untraced due to inadequate levels of testing.

Another major talking point comes from Bhilwara district in Rajasthan, once deemed as ‘India’s Italy’ by BBC. Bhilwara has been successful in restricting the infection from spreading in less than 2 weeks from its first reported case, with the last reported case registered on March 31. The success story has been attributed to massive testing, extensive tracking and enforcing strict isolation, apart from door-to-door surveys and continued screening of the population in the district. This led to screening a whopping number of 2.2 million people in the district which accounts for 92% of its population. The screening was not a one-off exercise, but multiple rounds were conducted to check upon the health status and well-being of the people, and in parallel ensuring the supply of essentials and prevention of black marketing. The Bhilwara model serves as a paramount example of how the policy of testing-tracing-isolation-treatment can work wonders in containing the spread of the disease.

Odisha has been dynamic as well from the very beginning of the outbreak, starting with locking down its capital Bhubaneshwar from March 12, before the state-wide lockdown from March 24. The state is closely following the Bhilwara model and has also announced an extension of the lockdown in the state till April 30, even before the announcement of the second phase of country-wide closure. The strong health workforce has synchronized in complete harmony with numerous self-help groups and policy-makers to effectively take benefit from the lockdown.

Chapter 2: How can India build healthcare capacity?

At a national level, several steps for containment and mitigation have already been put in place by the government including an extended nationwide lockdown and increasing awareness about preventive and behavioral measures. This will immensely contribute towards flattening the curve, which is absolutely essential to prevent a situation where the healthcare infrastructure of the country is no longer able to handle the burden. From our predictions, the number of people who could potentially get infected over time shows alarming counts which could overburden the existing healthcare infrastructure in the country. Specifically, the public healthcare facilities might see a huge surge in patients, as the majority of the Indian population depends on public healthcare resources. Hence, it is immensely important to assess if the infrastructure is ready to handle a huge surge of COVID-19 confirmed cases.

Here we look at the healthcare capacity in India across different states and analyze the overall availability of hospital beds in India. We first obtain data about the hospital beds across the country from a press release of the Ministry of Health and Family Welfare. Additionally, we obtain the population for each state from the Open Government Data Platform of India. The levels of healthcare resources required will vary based on the different clinical outcomes of the infection, and the existing availability of healthcare infrastructure. We capture the former aspects using a compartmental epidemiological model based on the classical SEIR model, which describes the spread and clinical progression of COVID-19. The infection levels are broadly categorized as mild, severe and critical, and we focus on the medium intensity spread scenario with the reproduction number R0=2 under intervention (for further data and details see our app here).

The predictions based on the SIER model indicate a peak for the number of cases around mid-May. In Figure 4, we show these numbers of additional hospital beds required (sorted from the lowest requirement to highest) in a waterfall plot assuming the percentage of prior occupancy at 25%, 50%, and 75%, respectively. Assuming that half of the cases who are hospitalized will need ventilators, we plot the requirement of ventilators across these states. These additional hospital bed requirements are also plotted on a map of India for all the states and union territories in Figure 5. Clearly, we see that many smaller states such as Arunachal Pradesh, Goa, Mizoram and Sikkim, and the union territories have sufficient healthcare infrastructure to deal with a surge under the assumptions we have considered. Whereas states such as Uttar Pradesh, Bihar, Maharashtra, and West Bengal could potentially be in desperate need of additional hospital beds and healthcare infrastructure to successfully deal with a pandemic of such scale.

Our results indicate that there is an immediate need for the administrators to mobilize resources and infrastructure in hotspot areas and acquire the appropriate number of hospital beds (permanent or makeshift), ventilators, personal protective equipment, and the accompanying personnel to support the huge surge which lies ahead.

The time is right to take such decisions while the extended lockdown is in effect and combined with other mitigation strategies, India can significantly flatten the curve.

Image for post
Image for post
Figure 4: Waterfall plots showing the shortfall of hospital beds and ventilators across Indian states and union territories on the peak day of incidence of new cases. These estimates are based on a SEIR model with the reproduction number R0=2 under sporadic adherence to containment and mitigation protocols. The normal occupancy rate of the hospital beds is varied as 25% (Panel a), 50% (Panel b), and 75% (Panel c).
Image for post
Image for post
Figure 5: Map of India showing the shortfall of hospital beds across Indian states and union territories on the peak day of incidence of new cases. These estimates are based on a SEIR model with the reproduction number R0=2 sporadic adherence to containment and mitigation protocols. The normal occupancy rate of the hospital beds is varied as 25% (Panel a), 50% (Panel b), and 75% (Panel c). Note: These maps are used for illustrative purpose only. Actual territorial boundaries might differ.

Effect of co-morbidities We have also analyzed published worldwide data from over 260,000 COVID-19 cases to assess the global prevalence of co-morbidities in the available cases. In Figure 6, we plot a word cloud corresponding to the prevalence of these co-morbidities. As can be seen, diabetes, hypertension, cardiovascular and respiratory diseases have high rates of occurrence in these patients. This has major implications for India as it has one of the highest rates of diabetic patients (close to 12%) which amounts to close to 15.6 crore people who are at risk. Hypertension rates are close to 30% — amounting to 39 crores and the number of people with cardiovascular issues is estimated to be 54.6 million (5 crores). Additionally, issues with respiratory illness (asthma, COPD, lung diseases) are relatively high in certain regions given various factors such as air quality and pollution.

Image for post
Image for post
Figure 6: Word cloud representing the global prevalence of co-morbidities in the COVID-19 infected patients. The font size of the word is proportional to the square-root of the number of times a given comorbidity is observed.

Chapter 3: What lies ahead…

The Prime Minister of India has announced the extension of the country-wide lockdown till May 3, with extensive plans to mitigate the crisis both from the health front and also on the economic front. We give a snapshot of the pro-active measures taken by India in the last couple of weeks.

  1. Increasing levels of testing across states, especially in hotspots
  2. Increasing the number of quarantine facilities, hospital beds and isolation units
  3. Extending financial help towards daily wage workers
  4. Ensuring the supply of PPEs and adequate medicines across the country
  5. Exporting potential drug Hydroxychloroquine to affected countries in need
  6. Creating SAARC COVID-19 Emergency Fund to tackle and fight the pandemic in South Asia
  7. Implementing drone surveillance systems to ensure social distancing, and improving sanitization processes
  8. Strictly monitoring zonal-level implementation of lockdown measures
  9. Launching the mobile app, Aarogya Setu, in a bid to keep the citizens informed about the COVID-19. Through a location-based social network construction, it provides information on whether the user has interacted with anyone who has tested positive for the disease.

While an impressive immediate response, is this response sustainable?

In a country like India density is ubiquitous. We have more than a billion people in a relatively small country. We have extensive family networks and societal networks. Even our technology interactions have a mediated human interface. The challenge is that this immediate response, the rapid lock-down, has been by deploying some draconian measures that will have a lasting effect on many classes of society, especially the urban poor.

What is needed is a resilient and sustainable approach, with regional sensitivity and national connectivity.

Moving forward, a strategy is needed which makes India’s density its advantage. We will need to continue to leverage the massive mobile network of India — ensure that the reminders about social distancing, hand-washing & hygiene are not overlooked. Next, we have to ensure that testing is strategic [not possible to test everyone], contact tracing is immediate [once a person is tested positive it is relatively easy to trace previous contacts and isolate them immediately], and quarantine is smart [right level of quarantine based on the level of risk]. In addition, we have to ensure that healthcare facilities are ready for the surge by decanting non-critical services, rapidly converting existing facilities, ramping up medical equipment and supplies, and leveraging alternate care facilities. The biggest challenge will be to do this in a humane, culturally sensitive and respectful way. There is a fine line between citizenship and vigilantism that must be carefully monitored.

To summarize, to contain and mitigate SARS-CoV-2, India must continue to manage contact networks by investing in the following:

● Continuous reinforcement of social distancing parameters augmented with mask use and hand-washing to reduce risk — with empathy to the different living conditions across different strata of society,

● Extensive literacy across social strata around basic hygiene and following protocols,

● Developing and leveraging extensive digital contact tracing capabilities; Google + Apple, Aarogya Setu app, etc., while protecting civil rights,

Smart quarantine strategies that leverage regional resources in a proportional way,

● Hyper-vigilance on migration and commute routes,

● Landscape of therapeutic interventions; clinical trials of hydroxychloroquine, vaccines, etc.,

● Rapid deployment plan for health facilities and alternate care facilities based on bed & ventilator need that leverage regional resources,

● Early management of supply chain for medical equipment, supplies and protective gear for staff [potentially planning for a flow of resources between states as they reach their peaks at different times],

● Appropriate post-COVID surge in basic health, rehabilitation, and mental health resources,

● State level Government + private industry partnerships for a phased and proportional response.

The strongest response to contagions is through building resilient communities that can manage the contact networks. The data and analyses in this article along with the interactive app can aid regional/state level specificity to allow a proportional response, prepare for the surge by planning for different scenarios, and create a roadmap so the country can eventually come back to normalcy in a phased manner that is socially responsible, equitable, and sustained.

Authors: Rupam Bhattacharyya is a Doctoral student in Biostatistics, Shariq Mohammed is a Postdoctoral research fellow and Veera Baladandayuthapani is a Professor of Biostatistics at University of Michigan. Sayantan Banerjee is Assistant Professor, Operations Management & Quantitative Techniques Area, at Indian Institute of Management, Indore. Upali Nanda is a Principal and Director of Research at HKS Inc and an Associate Professor of Practice at University of Michigan.

Please contact Veera Baladandayuthapani (veerab@umich.edu) for questions, comments and enquiries.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store