Utilisation of surveillance data and modelling techniques to inform infectious disease control programs in indonesia
File(s)
Author(s)
Djaafara, Bimandra
Type
Thesis
Abstract
Infectious diseases continue to pose a significant health challenge in Indonesia, despite progress in reducing their burden in the past decades. Furthermore, the emergence of COVID-19 in 2020 has further emphasised the pressing need for evidence-based policies and effective infectious disease control measures. To address this urgent need, this thesis utilises the combination of statistical and mathematical modelling approaches to analyse and explore disease surveillance data with the main objectives of enhancing our understanding of the current burden and transmission dynamics, evaluating the effectiveness of past interventions, and assessing the potential impact of future intervention strategies. This work focuses on two important diseases in Indonesia: COVID-19 and malaria.
This thesis’ first three analysis chapters focus on analysing data related to the COVID-19 epidemic in Jakarta and Java Island. In Chapter 2, I utilised a branching process statistical model to estimate the effective reproduction number (R_t) of SARS-CoV-2 in Jakarta using various data sources, including the daily number of funerals with COVID-19 protocols (C19P), which accounts for people with COVID-19 symptoms who died without receiving COVID-19 tests. The results highlight that R_t estimates based on the C19P funeral data captured the earlier declines in COVID-19 transmission, which were also reflected by the Google community mobility data, and demonstrated a sustained decline in transmissibility throughout the period of strict non-pharmaceutical interventions (NPIs). Additionally, this analysis supports the hypothesis that COVID-19 was circulating before the first cases were detected in the country.
In Chapter 3, I developed a metapopulation model that took estimates of district-level population mobility from call detail records (CDR) data to evaluate the impact of NPIs implemented during the first wave of the COVID-19 epidemic in Java Island. The analysis found that the NPIs implemented not only helped curb the spread of COVID-19 from the urban epicentres in Java but also prevented a potentially larger disease burden in the rural areas where the populations are older, and healthcare is poorer.
In Chapter 4, I utilised an established mathematical model of COVID-19 dynamics, calibrating the model to the confirmed and suspected COVID-19 deaths data at the province level in Java Island. The model calibrations showed differences in the estimated current burden of the epidemic based on those two data, which led to differences in the future scenario simulations of the model. This emphasises the importance of choosing the correct data for model calibrations. Furthermore, the model simulations also demonstrated the importance of restricting transmission during the early period of COVID-19 vaccinations, highlighting the incremental impact of preventing people from getting infectious before receiving protection from the vaccines.
Transitioning to the analysis of malaria data, Chapter 5 showed an analysis and exploration of routine malaria surveillance data in Indonesia from 2010 to 2019. The analysis explored the trends of multiple malaria metrics derived from the routine surveillance data, such as incidence rate, test positivity ratio, demography of malaria cases, and parasite species. By evaluating the context of those malaria metric trends observed, this chapter highlights the national-level success of the malaria control program in Indonesia, halving the total number of malaria incidences over the decade, with ongoing declines observed in most regions of the country. However, this analysis also showed how the progress towards malaria elimination has stagnated in the Papua region since 2015.
In Chapter 6, I used a plausibility design framework incorporating Bradford Hill criteria of causality to evaluate the hypothesis that the introduction of LLIN mass distribution campaigns was the main driver of the decline in malaria incidence across Indonesia from 2010 to 2019. By synthesising the evidence from the literature review, model simulations, and analysis of routine surveillance using the Bradford Hill causality criteria, I found strong evidence supporting the plausibility of LLIN mass campaigns as the main driver of malaria incidence declines across the country. This chapter also highlights the potential contribution of scale-ups of other interventions in maintaining the decline and preventing the resurgence of malaria incidence.
In Chapter 7, I conducted a high-resolution analysis of malaria data in Papua, the region contributing to the highest malaria burden in the country, where progress towards malaria elimination has plateaued. I developed a framework to calibrate the Imperial Plasmodium falciparum malaria mathematical model to malaria incidence data adjusted using the WHO case-adjustment formula at the district level. The modelling exercises highlight the variation in progress towards malaria elimination at both province and district levels, with stagnation and resurgence mainly driven by high-endemic baseline districts in Papua province. The analysis of LLIN distribution and usage data also revealed that lower utilisation of LLINs may have exacerbated the malaria problem in those districts.
Overall, this thesis demonstrates the value of integrating modelling techniques with disease surveillance data to further our understanding of disease burden and transmission dynamics in the population. It highlights the value of using alternative dataset, such as C19P funeral data, for estimating disease dynamics during the early phase of an epidemic when reporting and testing are still unreliable and limited. These are important findings for informing evidence-based policies for infectious disease control programs in Indonesia, with potential implications for other countries facing similar challenges.
This thesis’ first three analysis chapters focus on analysing data related to the COVID-19 epidemic in Jakarta and Java Island. In Chapter 2, I utilised a branching process statistical model to estimate the effective reproduction number (R_t) of SARS-CoV-2 in Jakarta using various data sources, including the daily number of funerals with COVID-19 protocols (C19P), which accounts for people with COVID-19 symptoms who died without receiving COVID-19 tests. The results highlight that R_t estimates based on the C19P funeral data captured the earlier declines in COVID-19 transmission, which were also reflected by the Google community mobility data, and demonstrated a sustained decline in transmissibility throughout the period of strict non-pharmaceutical interventions (NPIs). Additionally, this analysis supports the hypothesis that COVID-19 was circulating before the first cases were detected in the country.
In Chapter 3, I developed a metapopulation model that took estimates of district-level population mobility from call detail records (CDR) data to evaluate the impact of NPIs implemented during the first wave of the COVID-19 epidemic in Java Island. The analysis found that the NPIs implemented not only helped curb the spread of COVID-19 from the urban epicentres in Java but also prevented a potentially larger disease burden in the rural areas where the populations are older, and healthcare is poorer.
In Chapter 4, I utilised an established mathematical model of COVID-19 dynamics, calibrating the model to the confirmed and suspected COVID-19 deaths data at the province level in Java Island. The model calibrations showed differences in the estimated current burden of the epidemic based on those two data, which led to differences in the future scenario simulations of the model. This emphasises the importance of choosing the correct data for model calibrations. Furthermore, the model simulations also demonstrated the importance of restricting transmission during the early period of COVID-19 vaccinations, highlighting the incremental impact of preventing people from getting infectious before receiving protection from the vaccines.
Transitioning to the analysis of malaria data, Chapter 5 showed an analysis and exploration of routine malaria surveillance data in Indonesia from 2010 to 2019. The analysis explored the trends of multiple malaria metrics derived from the routine surveillance data, such as incidence rate, test positivity ratio, demography of malaria cases, and parasite species. By evaluating the context of those malaria metric trends observed, this chapter highlights the national-level success of the malaria control program in Indonesia, halving the total number of malaria incidences over the decade, with ongoing declines observed in most regions of the country. However, this analysis also showed how the progress towards malaria elimination has stagnated in the Papua region since 2015.
In Chapter 6, I used a plausibility design framework incorporating Bradford Hill criteria of causality to evaluate the hypothesis that the introduction of LLIN mass distribution campaigns was the main driver of the decline in malaria incidence across Indonesia from 2010 to 2019. By synthesising the evidence from the literature review, model simulations, and analysis of routine surveillance using the Bradford Hill causality criteria, I found strong evidence supporting the plausibility of LLIN mass campaigns as the main driver of malaria incidence declines across the country. This chapter also highlights the potential contribution of scale-ups of other interventions in maintaining the decline and preventing the resurgence of malaria incidence.
In Chapter 7, I conducted a high-resolution analysis of malaria data in Papua, the region contributing to the highest malaria burden in the country, where progress towards malaria elimination has plateaued. I developed a framework to calibrate the Imperial Plasmodium falciparum malaria mathematical model to malaria incidence data adjusted using the WHO case-adjustment formula at the district level. The modelling exercises highlight the variation in progress towards malaria elimination at both province and district levels, with stagnation and resurgence mainly driven by high-endemic baseline districts in Papua province. The analysis of LLIN distribution and usage data also revealed that lower utilisation of LLINs may have exacerbated the malaria problem in those districts.
Overall, this thesis demonstrates the value of integrating modelling techniques with disease surveillance data to further our understanding of disease burden and transmission dynamics in the population. It highlights the value of using alternative dataset, such as C19P funeral data, for estimating disease dynamics during the early phase of an epidemic when reporting and testing are still unreliable and limited. These are important findings for informing evidence-based policies for infectious disease control programs in Indonesia, with potential implications for other countries facing similar challenges.
Version
Open Access
Date Issued
2023-05-09
Date Awarded
01/09/2023
License URL
Advisor
Walker, Patrick
Churcher, Thomas
Sherrard-Smith, Ellie
Sponsor
Imperial College London
Grant Number
G24038
Publisher Department
School of Public Health
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)