Forecasting malaria incidences in India using the ARIMA model

By Avishek Majumder & Diksha Tomer on October 23, 2018

The present article aims to predict the prevalence of malaria in India post-2017 (2018-2022). Therefore, the article presents a predictive assessment on the basis of secondary data available for malaria incidences in the period of 1998-2017. The predictive assessment uses the Autoregressive integrated moving average (ARIMA) model. Epidemiological analysis help to prepare malaria prevention strategies with the help of forecasted trends in disease incidence. In addition, the model can help to predict mortality or incidence cases in future and plan new treatment or control methods in a specific geographical area.

Time series modelling to forecast malaria incidences

The time series modelling presents the relationship between the observed malaria incidences from past observations, without using any other variables (Promprou, Jaroensutasinee, & Jaroensutasinee, 2006). This study aimed at developing univariate time-series models for the annual malaria incidences in India based on the reported cases since 1998. This forecasting offers the potential for improved contingency planning of public health intervention.

The annual data of malaria incidences in India was collected from NVBDCP (National Vector Borne Disease Control Programme) official site http://nvbdcp.gov.in/  from the year 1998-2017 (Ministry of Health & Welfare, 2017). The first ever official malaria control policy was planned in 1998. Over the years the same plan has undergone multiple updates. The policy is a part of the 5-year development plan, the population distribution is divided for a gap of 5 years. The STATA 13 software was used for developing the autoregressive integrated moving average (ARIMA) models.

ARIMA model has three values to be determined i.e. ‘d’ value representing the differential order of the trend after being stabilised, ‘p’ value that represents the partial autocorrelation between the data and ‘q’ value that represents the autocorrelation between the data, on the basis of which the forecasting is done. The autoregressive order ‘p’ was then determined as 3 after plotting the graph of the differential values of incidence cases. After that autocorrelation, the model was determined as ARIMA (1,0,3).

Observation

Then, the ARIMA regression was done and it was found out that the present values significantly depend upon the past values on the third lag value with significance i.e. p value< 0.05. This was done to prepare the data for forecasting so that the future predicted values will depend upon the trend of the past values of the incidence cases of malaria in India. The p-value will show that the ARIMA regression results are significant and the forecasting will then be reliable for that ARIMA model.

ARIMA regression results for forecasting
ARIMA regression results for forecasting

Results of forecasting for 2018 to 2022

The determined model is used to forecast malarial cases for 5 sequential future years i.e. from 2018 up to the year 2022. This forecasting is based on the last obtainable data points i.e. malarial incidence cases in the year 2017 as the forecasting origin.

2018 604976.3
2019 641659.4
2020 773808.6
2021 576380.9
2022 357356.2

The predicted values for the year 2018 came out to be 604976 malarial incidence cases in India. An increase in the malarial incidence cases is expected in the year 2019-20 with the incidence cases predicted 773808 cases in the year 2020. Decreasing trend is expected to be observed in the year 2021-22, with the 357356 malarial incidence cases in the year 2022 in India.

Graph of the actual malaria incidences trend with the predicted values
Graph of the actual malaria incidences trend with the predicted values

Forecasting malaria in India

The results from this study have confirmed that there will be an increase in the overall incidence cases of malaria in the year 2020 with the total incidence of 773808 cases. This may be possible due to the day to day increase in the population of urban areas that lead to the sub-standards of living conditions. Therefore, a preventive measure must be prepared to minimize the risk of mortality due to malaria incidence at that time.

The focus should be on developing more effective and economical antimalarial drugs to prevent the situation of reemergence. The cases of malaria are expected to decline by the year 2022. Although there is no particular justification as to why this fluctuation may occur in future but, may be linked to either development of new improved drugs or improved policy implications. Healthcare members too will enable to plan and implement awareness methods as a pre-control measure in the specific district or state.

Without malaria control interventions, the predicted values will not adhere to real values in future. Therefore, it is crucial to use the predicted values by the healthcare and pharma industries implement or strategize pre-control measures for reducing malaria cases in India.

Implications

The predicted trends, however, cannot yet possibly claim that the results will be an accurate prediction. The extent of prevalence of malaria in regards to the change in the climate expected to eventually impact the malarial disease pattern. The study, however, lacks in considering impact factors, such as climatic conditions, man-made issues, and natural calamities for the changes in the prevalence of malaria in India.

Time-series forecasting of malaria incidence in India will provide the government to improving planning, control and prevention by public health intervention. In addition, the pharma industries will help the health care members with pre-provision of treatment interventions and drugs as per the rise or fall of malaria incidences. Prediction thereby helps in better implementation of control measures and policies against the prevalence of malaria.

References

  • Ministry of Health & Welfare, G. (2017). Malaria :: National Vector Borne Disease Control Programme (NVBDCP). Retrieved August 28, 2018, from http://nvbdcp.gov.in/index1.php?lang=1&level=1&sublinkid=5784&lid=3689.
  • Promprou, S., Jaroensutasinee, M., & Jaroensutasinee, K. (2006). Forecasting dengue haemorrhagic fever cases in Southern Thailand using ARIMA Models. Dengue Bulletin, 30, 99–106.
  • Sharma, M. K., Sharma, P. K., Kumar, B., Mahajan, S., & Archana, A. (2017). Testing statistical models for forecasting malaria cases in India. International Research Journal of Agricultural Economics and Statistics, 8(1), 8–14. https://doi.org/10.15740/HAS/IRJAES/8.1/8-14.

Discuss