Contributors: Francesco Checchi, Zhixi Chen, Zeina Jamaluddine, Greg Barnsley, Sebastian Funk, Pratik Gupte, Tak Igusa.
Funding: UK Humanitarian Innovation Hub
Welcome to this repository. Every sub-folder on this directory corresponds to a cause-specific module of our report and corresponding Methods Annex, published at www.gaza-projections.org . The sub-folder gaza_overall contains code and data required to do final analysis of all the cause-specific modules together.
All of the data contained in this repository are in the public domain, or we have received permission to publish them. The study has received ethics approval from the London School of Hygiene and Tropical Medicine and the Johns Hopkins University Bloomberg School of Public Health.
To replicate the analysis, please download the entire repository and keep the folder structure as it is. Each folder has an inputs-code-outputs structure. All input files are contained in the /inputs sub-folder, while all the code is in the /code sub-folder and all outputs (data files, tables, graphs) are saved to /outputs. Analyses in R can be re-run from gaza_[xxx]/code/00_master_script.R, which will load packages and source all other scripts for a given cause-specific module. The directory for reading and outputting files is set automatically when 00_master_script.R is run. An updated version of R software should be installed (https://www.r-project.org/). We recommend to run the code from the open-source RStudio interface (https://www.rstudio.com/products/rstudio/download/). Both R and RStudio are free and open-source. R package epidemics, needed for the infectious diseases analysis, will likely require installing Rtools43 (https://cran.r-project.org/bin/windows/Rtools/rtools43/rtools.html) first. This rather large bundle can be uninstalled once epidemics is successfully installed and compiled.
gaza_injuries_data.xlsxcontains all the data needed for this module, including daily numbers of deaths and injuries reported by the MoH and UNRWA (worksheetdaily), line lists of injured persons as provided by the MoH (moh_list), extent of internal displacement in shelters (idps) and data on deaths and injuries overall and due to unexploded ordnance/mines during the 2014 war (ordnance). Each worksheet is one dataset and thedictionaryworksheet (not needed for analysis) describes each variable.gaza_injuries_parameters.xlsxcontains parameters needed for the analysis, which can be modified here by the user (worksheetgeneral). Worksheetsurvcontains meta-data from a large cohort of injuries, where variabledayis the number of days since the injury andp_dis the proportion of cases who died during the interval since the previous time point.gaza_noninjury_to_date.xlsxhas been generated manually by binding together deaths due to causes other than traumatic injury, as estimated/projected in all the other cause-specific modules. This is needed here to subtract a non-injury death level from projections based on MoH data.out_adjustment_factors.csvis an output of thegaza_overallcode and is needed here to adjust final injury estimates for the probability of dying from competing causes. See Methods Annex.
00_master_script.Rinstalls/loads packages, sets colour palettes, initialises random numbers, recognises the local directory and calls all the other scripts;01_read_prepare_data.Rreads data and parameters and prepares them for further analysis;02_prepare_simulations.Rvisualises MoH and UNRWA data; fits count models of deaths and injuries, for both the escalation and status quo scenarios; fits a model of the deaths reporting fraction; figures out the age-sex distribution of injury deaths; prepares a template timeline/cohort for further analysis steps; and works out the proportion of injury deaths who die immediately; graphs are produced and outputs saved for next steps;03_estimate_cf_scenario.Rimplements a simulation to project injury deaths and injuries in the ceasefire scenario, including deaths due to injuries sustained previously during the war and due to unexploded ordnance;04_estimate_sq_es_scenarios.Rimplements a simulation to project injury deaths and injuries in the status quo and escalation scenarios;05_analyse_visualise.Rcollects outputs of the above simulations, computes means and uncertainty intervals and tabulates/graphs the findings.
ACLED_FATALITIES.xlsxcontains raw fatalities data downloaded from ACLED website. It's at the event level, with location, type of warfare, and estimated or recorded fatalities and injuries.MOH_UNWRA_Death_Rate.xlsxcontains the death rate per 1000 people reported by the Ministry of Health (MoH) and the death rate among UNRWA employees per 1000 people reported by UNRWA.OXFORD_POPULATION.csvcontains population movement in each governorate, based on daily counts of active social media users.
__init__.pyinstalls/loads packages, imports data, imputes the ACLED data using MICE, prepares parameters for bootstrapping and performs bootstrpping.functions.pycontains all the functions called by__init__.py.
gaza_nutrition_parameters.xlsxcontains several worksheets (tabs), all of which butlistsare read into the analysis. Worksheetgeneralallows the user to set parameters for the nutrition analysis. Worksheetscenarioscontains assumptions made for each scenario concerning daily caloric intake from humanitarian aid during months 1-3 and 4-6 of the projection period, by scenario and as a ratio relative to the target daily intake (see worksheetgeneral). Worksheetnon-aidcontains assumptions by month (from 7 october 2023 to the end of the projection period) and scenario regarding the proportion of the pre-war caloric intake that would have been met by existing food stocks or the agriculture/livestock sector. Lastly, worksheetbreastfeedingcontains pre-crisis and crisis period estimate of the proportion of infants exclusively breastfeeding during the first 6 months of life, for other crisis-affected settings in the Middle East.gaza_food_trucks.csvcontains the reported number of food trucks reaching any open crossing into Gaza, per day.gaza_survey2020_kcal_bmi_aggcontains aggregate (by age, sex) data from a 2020 survey of adults aged 40+ years in Gaza. For each age-sex stratum the dataset reports mean values of the weight, height and daily Kcal intake. Thesvy_wtcolumn contains the relative proportion of survey participants in each age-sex stratum, and is used when averaging results to all age groups.gm_anthro_2019_agg.rdscontains aggregate (by age in months, sex, height and weight) anthropometric data from systematic growth monitoring of children aged 6-59 months old in Gaza. The data were collected in 2019. Thewtvariable is the proportion of children within the age-sex-height-weight-stratum, and is used when averaging GAM and SAM prevalences for the entire population aged 6-59 momths old.gaza_nutrition_starvation_lit.xlsxcontains metadata extracted from old studies of people exposed to starvation conditions. The main outcome of interest is weight loss. This dataset is read and analysed by the archived R script02_fit_wt_loss_model.R(see below), but is not used for the analysis.
00_master_script.Rinstalls/loads packages, sets colour palettes, initialises random numbers, recognises the local directory and calls all the other scripts;01_specify_wt_functions.Rspecifies functions needed to model adult weight loss;02_read_prepare_inputs.Rreads data and parameters and prepares them for further analysis;03_estimate_food_aid.Rapplies assumptions and uncertainty ranges to the data on food trucks so as to estimate/project caloric intake to date/over the projection period, by scenario. This is done through a simulation; each simulation run is outputted, and graphs are generated.04_estimate_wt_loss.Rapplies the estimated/projected nutrient intake to a 2020 survey of adult anthropometry and food intake. A mechanistic model is used to estimate/project percent weight loss. The analysis occurs in a simulation to account for parameter uncertainty; each simulation run is outputted, and graphs are generated.05_project_gam_sam.Rapplies the estimated/projected adult weight loss to pre-war anthropometric data from children 6-59 months old, and computed globald and severe acute malnutrition prevalence. This is also done in a simulation. Graphs and tables are outputted.06_project_breastfeeding.Rprojects the reduction in exclusive breastfeeding based on the pre-war level in Gaza and observed reductions in crisis-affected settings in the region.- The
archivesub-folder includes the scripts (i)99_read_prepare_nonpublic_data.Rso users can see how we managed datasets that are not released publicly in their original form. This script cannot be implemented as the datasets it relies on are not made public; and (ii)02_fit_wt_loss_model.R, which fits a model to old studies of weight loss among people exposed to starvation (see above,gaza_nutrition_starvation_lit.xlsx): this model has not been used in the analysis.
gaza_infections_parameters.xlsxcontains various worksheets with parameters for the analysis, which the user should modify here if needed:generalcontains various general parameters;immunity_assumptionscontains assumptions made regarding baseline susceptibility to infection and disease for various diseases not included in Gaza's routine vaccination schedule;epidemic_parameterscontains ranges, relative values and age-specific values for different epidemic-prone infectious disease parameters including the basic reproduction number, the case-fatality ratio, the duration of the infectious and pre-infectious periods, and the proportion of infectious that become symptomatic;endemic_parameterscontains assumed values of age-specific mortality, relative share of infectious disease mortality, and seasonality for various endemic infections;list_diseasesis a table of all infectious diseases considered;other_listsis not used.gaza_infections_endemic_data.xlsxcontains data on age-specific and proportional mortality due to endemic infections between 2016 and 2022, with assumptions for 2023-2024; thedictionarytab describes variables.gaza_infections_see_data.xlsxcontains responses by an expert eliciation panels to questions related to the likely values of infection-related parameters in Gaza. Thedictionarytab describes variables, andlistsis not used.see_distributions.rdsis produced during the analysis itself and used in subsequent steps, and contains empirical distributions of parameters, derived from expert elicitation.out_adjustment_factors.csvis an output of thegaza_overallcode and is needed here to adjust final injury estimates for the probability of dying from competing causes. See Methods Annex.fit_model_endemic_deaths.rdsis a model of annual endemic infectious deaths, fit during the analysis itself and used in subsequent steps.digaale_pop.rdsanddigaale_svy.rdsare data from a social mixing survey done among IDPs in Somaliland, used as assumption of the social contact structure for Gaza. These datasets are downloaded by the code from their source online, if not already in the folder.- the sub-folder
immunity_projectionscontains estimates/projections of various immune/susceptible classes, by disease, period and scenario, as produced by the immunity-tracking model. In practice only two of the datasets are used in subsequent steps.
- the sub-folders
immunity_projections_modelandimmunity_projections_scriptsforward to repositories for an R package (IVODE) and R scripts developed for this project to track immunity/susceptibility status for diseases included in Gaza's routine vaccination programme. Please see the Readme sections for those repositories. This can be viewed as a sub-module of the infectious diseases module. 00_master_script.Rinstalls/loads packages, sets colour palettes, initialises random numbers, recognises the local directory and calls all the other scripts;01_specify_functions.Rdefines functions used to compute scores and weights for expert elicitation data, to implement SEIR models of each epidemic pathogen, and to distribute social mixing data into finer age groups;02_read_prepare_data.Rreads and prepares datasets; analyses and graphs expert elicitation data; prepares compartments and values for SEIR models, including social mixing data.03_prepare_simulations.Rprepares all the random simulations for the SEIR / epidemic projections;04_run_simulations.Rruns simulations for epidemic-prone and endemic infections, and collects outputs;05_analyse_and_visualise.Ranalyses outputs, producing graphs and tables.04b_run_simulations_to_date.Ris not needed for the analysis, but produces estimates of endemic infection deaths for the war period to date.
gaza_mnh_list_outputs.xlsxis the output of the LiST model as applied in this project, and is used to produce graphs and tables in R. The file contains four worksheets. Worksheetsmaternal,neonatalandstillbirthscontain LiST-projected maternal deaths, neonatal deaths and stillbirths (excluding those due to traumatic injury), by scenario and period (months 1-3, months 4-6 in the projection period); the columnsmean,lcianducicontain point estimates and 95% uncertainty intervals, while columnd_crisis_excessindicates whether the value refers to the counterfactual baseline ('d_baseline'), the projection under scenario assumptions ('d_crisis'), or excess deaths ('d_crisis' - 'd_baseline'). Worksheetcumcontains excess maternal deaths, neonatal deaths and stillbirths by month, including pre-war, the period to date and the three scenarios: only the point estimate is included.gaza_MNH-LiST_pre war_indicators.xlsxis a collection of pre-war indicators / data relevant to the MNH module, only included for reference.gaza_MNH_LiST_final_reduction_2024.xlsxcontains the scenario-specific assumptions fed into the LiST model. For more details on how LiST was used, please see the Methods Annex or write to info@gaza-projections.org .
01_visualise_outputs.Rcreates graphs based ongaza_mnh_list_outputs.xlsx. Note that the MNH analysis was done using the LiST model, which does not run on R software.
gaza_NCD_baseline_mortality.xlsxcontains annual mortality figures attributable to NCDs along with age-specific mortality data due to NCDs, as reported by the MoH.gaza_NCD_treated_untreated.xlsxcontains the relative survival rates for patients with NCDs from literature, focusing on the acute phase survival and long-term survival up to 10 years post-diagnosis. It is important to note that the survival rates presented are conditional upon survival of the acute phase for stroke and ischemic heart disease, and acute presentation data for other NCDs have been omitted. The columntreated_untreateddisplays survival data labels. Labels accompanied by lb or ub indicate a lower bound or upper bound, respectively.gaza_NCD_coverage.xlsxcontains assumed ranges of treatment coverage for specific NCD-relevant treatment services. The assumptions are based on pre-war data, expert knowledge, and conversations with health actors within Gaza.HR_new_logistic_normal.xlsxcontains pre-calculated hazard rates (acute phase up to 40 years post-diagnosis) for each NCD based on survival curves fitted using the Log-normal, Log-logistic, and Exponential survival distributions.gaza_NCD_treated_untreated_survival.xlsxincludes all the information (gaza_NCD_baseline_mortality.xlsx,gaza_NCD_treated_untreated.xlsx,gaza_NCD_coverage.xlsx) in a format that is readable by code.
01_survival_curve_fitting.pycontains functions for fitting survival curves using four candidate distributions (i.e., Weibull, log-normal, log-logistic, and gamma). Fitting results were stored for further distribution selection. Detailed distribution selection and parameters can be found in the Methods Annex.02_hazard_rate_pre_generator.pycontains the process of generating hazard rates for various Non-Communicable Diseases (NCDs) across different survival periods (0-10 years). Results can be found in the input folder with the file name ofHR_new_logistic_normal.xlsx.03_ncd_main_simulation.mcontains the main simulation function for NCD projection model. Before run the main simulation please make sure you change the name into valid name for MATLAB and select the NCD for simulation on line 82.04_sim_out_processing.pyanalyses the simulation outputs and processes the data into wanted format.05_NCDs visuals.Rcreates graphs for the outputs.
gaza_overall_parameters.xlsxcontains key dates (generaltab), age- and sex-specific population (poptab) and the list of infectious diseases analysed (list_diseasestab).gaza_overall_data.xlsxcontains projected deaths by cause (specific disease for infections and NCDs), scenario, subperiod (months 1-3 and 4-6 of the projection period) and age. Means, medians and lower/upper bounds of the uncertainty interval are provided; 'd_base_[xxx]' means counterfactual baseline estimates, 'd_crisis_[xxx]' projections based on the scenario assumptions, and 'd_excess_[xxx] the difference between 'd_crisis' and 'd_base'. At present this dataset is assembled semi-manually from cause-specific module outputs.
00_master_script.Rinstalls/loads packages, sets colour palettes, initialises random numbers, recognises the local directory and calls all the other scripts;01_read_prepare_data.Rreads parameters and aggregated output from the cause-specific modules; and solves discrepancies in age and sex categories;02_analyse_data.Rcomputes and applies adjustment factors for the probability of deaths from concurrent causes, and generates tables and graphs of all-cause mortality. The outputout_adjustment_factors.csvis then manually copy-pasted into the inputs sub-folders of cause-specific modules, as needed.