Real-Time COVID-19 Forecasting for Four States of India Using a Regression Transmission Model

Introduction: More than a million people are reported to have been infected with COVID in India, since the beginning of the pandemic. However, the epidemic is not the same across the country. Though there are state-level variations rapidly changing disease dynamics and the response has created uncertainty towards appropriate use of models to project for the future. Method: This paper aims at using a validated semi-mechanistic stochastic model to generate short term forecasts. This analysis used data available at the respective state government bulletins for four states. The analysis used a simplified transmission model using Markov Chain Monte Carlo simulation with Metropolis-Hastings updating. Results: Two weeks were used to compare the results with the actual data. The forecasted results are well within the 25th and 75th percentile of the actual cases reported by the respective states. The results indicate a reliable method for a real-time short term forecasting of COVID-19 cases. The 1st week projected interquartile range and actual;reported cases for the state of Kerala, Tamil Nadu, Andhra Pradesh and Odisha were (1064 - 2532) 2234, (17,503 - 50,125) 27,214, (5225 - 11,003) 9563, (2559 - 4461) 3925, respectively. Similarly, the 2nd week projected interquartile range and actual;reported cases were (1055 - 7803) 4221, (18,298 - 73,952) 31,488, (4705 - 23,224) 13,357, (2701 - 9037) 4175 respectively. Conclusion: This real-time forecast can be used as an early warning tool for projecting the changes in the epidemic in the near future triggering proactive management steps.

An Interrupted Time Series Analysis of COVID-19 Positivity before, during and after Lockdown in Four States of India

Objectives: The objective of this study was to examine the impact of large scale non-pharmaceutical interventions on COVID-19 pandemic. Methods: We used interrupted time series analysis (ITS), a quasi-experimental model to evaluate the effect of interventions in four states of India by comparing the COVID-19 positivity before lockdown, during lockdown and opening-up period. Results: The positivity in all the four states declined during lockdown and the trends reversed soon after the lockdown measures were relaxed as the states opened-up. The rate of reduction of positivity was significantly different between states. Between the lockdown and opening-up period, an increase in positivity was recorded in all the states with significant variation between states. Conclusion: The analysis provides conclusive evidence that the lockdown measures had a positive effect in reducing the burden of COVID-19 and establishes a causal relationship.

Urban and Rural HIV Estimates among Adult Population (15 - 49 Years) in Selected States of India Using Spectrum Data

HIV estimation has become a standard tool for understanding the epidemic. Although the majority of India’s population lives in rural areas, to date, an exploration of the urban and rural HIV epidemic has not been undertaken. The objective of this study is to develop HIV estimation based on urban and rural adult populations in selected states of India to understand the difference in HIV related indices geographically. Ten states were selected based on HIV prevalence levels-Andhra Pradesh, Tamil Nadu, Karnataka, Maharashtra, Manipur, and Nagaland, Mizoram, Punjab, Odisha and Jharkhand. Spectrum, version 4.53 beta 19, was used. Data files of Indian national estimation, 2010-11 which included population, HIV Sentinel Surveillance, Integrated Bio Behavioral Assessment and program coverage data, were used and alterations made wherever necessary. The urban and rural sub epidemic structures and their subpopulations were separately configured in the Estimation projection package and curve fitting done. Outputs for each state were separately analyzed. Findings show that HIV prevalence is lower in urban than rural areas in Tamil Nadu and Maharashtra;in Karnataka there is no difference in HIV prevalence in the urban and rural populations;and in the remaining seven states urban HIV prevalence is higher as compared to rural HIV prevalence. In the states of Andhra Pradesh, Tamil Nadu, Maharashtra, Odisha and Punjab, the number of people living with HIV, new HIV infections and deaths among people living with HIV is higher in the rural than in the urban population. An early and lower peak in HIV prevalence and incidence in the urban population was seen in Andhra Pradesh, Tamil Nadu, Karnataka and Naga-land, while in Maharashtra the rural peak was earlier and higher. Mizoram shows an earlier and lower peak in the rural population while Manipur shows an earlier and higher urban peak. In Odisha, the epidemic peaked earlier and was lower in the rural than the urban population. HIV prevalence in the urban population in Punjab was still peaking while HIV incidence was earlier and lower in the rural population. In Jharkhand, both urban and rural HIV prevalence and incidence are still increasing. Our findings indicate lower levels of HIV prevalence and incidence in the urban population as compared to the rural populations in Maharashtra and Tamil Nadu. In the remaining eight states, urban prevalence and incidence are higher than their rural counterparts. Future estimations of the HIV epidemic in the country need to adopt a similar approach to inform the design of appropriate state-level strategies for HIV prevention in urban and rural areas.