Ensemble forecast of COVID-19 in Karnataka for vulnerability assessment and policy interventions

Sashikumaar Ganesan, Deepak Subramani, Thivin Anandh, Divij Ghose, Giridhara R. BabumedRxiv 2021 · DOI

Abstract

We present an ensemble forecast for Wave-3 of COVID-19 in the state of Karnataka, India, using the IISc Population Balance Model for infectious disease spread. The reported data of confirmed, recovered, and deceased cases in Karnataka from 1 July 2020 to 4 July 2021 is utilized to tune the model’s parameters, and an ensemble forecast is done from 5 July 2021 to 30 June 2022. The ensemble is built with 972 members by varying seven critical parameters that quantify the uncertainty in the spread dynamics (antibody waning, viral mutation) and interventions (pharmaceutical, non-pharmaceutical). The probability of Wave-3, the peak date distribution, and the peak caseload distribution are estimated from the ensemble forecast. Our analysis shows that the most significant causal factors are compliance to Covid-appropriate behavior, daily vaccination rate, and the immune escape new variant emergence-time. These causal factors determine when and how severe the Wave-3 of COVID-19 would be in Karnataka. We observe that when compliance to Covid-Appropriate Behavior is good (i.e., lockdown-like compliance), the emergence of new immune-escape variants beyond Sep ‘21 is unlikely to induce a new wave. A new wave is inevitable when compliance to Covid-Appropriate Behavior is only partial. Increasing the daily vaccination rates reduces the peak active caseload at Wave-3. Consequently, the hospitalization, ICU, and Oxygen requirements also decrease. Compared to Wave-2, the ensemble forecast indicates that the number of daily confirmed cases of children (0-17 years) at Wave-3’s peak could be seven times more on average. Our results provide insights to plan science-informed policy interventions and public health response.

Introduction

The devastating second wave of COVID-19 in India, primarily driven by the Delta variant, highlighted the limitations of single-prediction models in the face of high uncertainty regarding viral mutations, waning immunity, and varying intervention effectiveness. To address this, researchers at the Indian Institute of Science (IISc) developed an ensemble forecast for Karnataka. By simulating 972 distinct scenarios, this study provides a probabilistic framework to help public health officials plan for worst-case requirements in ICU beds, oxygen, and pediatric care, rather than relying on a single, potentially fragile estimate.

Methodology: The IISc Population Balance Model

The study employs the IISc Population Balance Model, a sophisticated partial differential equation (PDE) framework that tracks the infected population across six dimensions: geographic location, infection severity, duration of infection, and age.

To account for uncertainty, the researchers constructed an ensemble of 972 members by varying seven "control variables":

  • Case-to-Infection Ratio (CIR): Estimates of underreporting based on serological surveys (e.g., CIR 40 implies 39 unreported cases for every 1 reported).
  • Immune-Escape New Variant (IENV): Variations in emergence time (July, Sept, Nov 2021) and transmission power.
  • Antibody Waning (ABW): Rates at which natural or vaccine-induced immunity declines.
  • Vaccination Rate (VR): Daily doses ranging from 280k to 560k.
  • Covid-Appropriate Behavior (CAB): Three levels of compliance: Good (lockdown-like), Partial, and No CAB.

Key Drivers of the Third Wave

The ensemble analysis identifies three dominant factors determining the severity and timing of a potential third wave: public compliance (CAB), vaccination rates, and the timing of new variant emergence.

Figure 3: (a) Probability of Wave-3 across different control variables. (b) Heatmap showing the median of active cases at peak for various combinations of interventions and variant scenarios.

As shown in Figure 3(a), compliance with Covid-appropriate behavior is the most critical factor. Under "Good CAB" conditions, the probability of a third wave remains low even with new variants. Conversely, "No CAB" makes a third wave nearly inevitable. Vaccination acts as a "peak-shaver"; while it may not always prevent the onset of a wave, increasing the daily rate to 560k significantly reduces the peak active caseload (Figure 3b).

Peak Timing and Magnitude

The timing of the peak is closely linked to the emergence of new variants. The ensemble shows that once a new immune-escape variant gains traction, a peak typically follows within 45 to 60 days.

Figure 4: Distribution of peak active cases versus peak date across different scenarios of CAB (columns) and variant emergence timing (rows). Shapes represent re-infection percentages, colors represent vaccination rates, and borders represent CIR.

Figure 4 illustrates how the peak magnitude and timing shift based on behavioral compliance. In scenarios with "No CAB" (right column), peaks are both earlier and significantly higher (reaching $10^7$ active cases) compared to "Partial CAB" scenarios. In "Good CAB" scenarios, many simulations result in "No Wave" (NW), particularly if variants emerge later in the year.

Pediatric Vulnerability and Policy Implications

A significant finding is the projected shift in age-wise vulnerability. Since the adult population was prioritized for vaccination, children (0-17 years) are expected to bear a disproportionate burden in a third wave. On average, the number of daily confirmed cases for children could be seven times higher than during the second wave. In worst-case scenarios involving low vaccination and early variant emergence, this factor could rise to twenty times for the 0-11 age group.

Science-informed policy should prioritize maintaining CAB, as even partial relaxation significantly increases wave probability. Furthermore, reaching a threshold of ~500k vaccine doses per day is critical for suppressing peak demand on healthcare infrastructure. Finally, hospitals must prepare for a substantial increase in pediatric ICU and oxygen requirements.

Limitations

The model assumes a fixed vaccine efficacy of 70% and does not account for "pandemic fatigue" potentially leading to a continuous degradation of CAB over time. Future work aims to scale this ensemble approach to a national level to assist other regions in infrastructure planning.

Summary updated

Abstract

Abstract We present an ensemble forecast for Wave-3 of COVID-19 in the state of Karnataka, India, using the IISc Population Balance Model for infectious disease spread. The reported data of confirmed, recovered, and deceased cases in Karnataka from 1 July 2020 to 4 July 2021 is utilized to tune the model’s parameters, and an ensemble forecast is done from 5 July 2021 to 30 June 2022. The ensemble is built with 972 members by varying seven critical parameters that quantify the uncertainty in the spread dynamics (antibody waning, viral mutation) and interventions (pharmaceutical, non-pharmaceutical). The probability of Wave-3, the peak date distribution, and the peak caseload distribution are estimated from the ensemble forecast. Our analysis shows that the most significant causal factors are compliance to Covid-appropriate behavior, daily vaccination rate, and the immune escape new variant emergence-time. These causal factors determine when and how severe the Wave-3 of COVID-19 would be in Karnataka. We observe that when compliance to Covid-Appropriate Behavior is good (i.e., lockdown-like compliance), the emergence of new immune-escape variants beyond Sep ‘21 is unlikely to induce a new wave. A new wave is inevitable when compliance to Covid-Appropriate Behavior is only partial. Increasing the daily vaccination rates reduces the peak active caseload at Wave-3. Consequently, the hospitalization, ICU, and Oxygen requirements also decrease. Compared to Wave-2, the ensemble forecast indicates that the number of daily confirmed cases of children (0-17 years) at Wave-3’s peak could be seven times more on average. Our results provide insights to plan science-informed policy interventions and public health response.
BibTeX
@misc{ganesan2021ensemble,
  author = {Sashikumaar Ganesan and Deepak Subramani and Thivin Anandh and Divij Ghose and Giridhara R. Babu},
  title = {{Ensemble forecast of COVID-19 in Karnataka for vulnerability assessment and policy interventions}},
  year = {2021},
  doi = {10.1101/2021.08.31.21262943},
  howpublished = {medRxiv},
}
Figure