CHAPTER 3 DATA GATHERING: STRUCTURED LITERATURE REVIEW ON
3.3 Steps of the structured literature review
3.3.4 Capturing data from literature to the dataset
Following the analysis and determination of the relevance of literature pieces, the information gleaned from the review is captured to the dataset. The categories are informed by the scope delimitation previously discussed in §3.2.1 and are discussed below.
Data source
The data source used within the modelling application is noted and categorised according to the following categories and criteria, namely:
None (e.g. no data source used);
Case data (e.g. data on confirmed cases of disease infection); Travel data (e.g. data on movement of individuals);
Parameters from literature (e.g. data on transmission parameters previously formalised in the literature);
Population estimates (e.g. census data); and
Assumed (e.g. data which assumes important transmission characteristics).
The data source categories noted above are not necessarily mutually exclusive, therefore a literature instance can be classified as using more than one of these data sources.
Method of model fit
Due to the vast number of model fitting methods employed, it is not practical to research each available one in order to define a set of fitting methods to use when analysing the dataset. Instead, the various methods used to fit models to the data source are noted and added to the dataset as these are serendipitously uncovered during the capturing process.
Modelling scope
The modelling scope is determined from the scope of the modelling application, which in turn is affected strongly by the scope of the data source used. The scope is noted and categorised according to the following categories and criteria, namely:
General (i.e. a general modelling application with no indication of the scale of the application, typically a theoretical model for a specific disease instance);
Global (i.e. disease transmission between more than two countries); Intercountry (i.e. disease transmission between two countries); Country (i.e. disease transmission within a single country); Provincial (i.e. disease transmission within a province); and
Rationale of article
The rationale of an article is typically not explicitly mentioned, but it is useful to note the goals of the research question (mentioned previously in §3.1.4) which are implicitly part of the modelling approaches followed in literature. The set of rationales given in the list that follows are formulated to capture the specific focus of the research questions being addressed in each article:
Model disease transmission dynamics (develop a model to study disease transmission dynamics);
Investigate causal relationships (develop a model to investigate the effect of factors which affect the chain of infection and correlates to changes in disease propagation or prevalence); Investigate super spreading events (develop a model to analyse instances of unusually high
secondary infections emanating from a few individuals);
Forecast disease instance (develop a model to not only fit data or parameters, but to explicitly forecast future disease prevalence from the model);
Develop a model and analyse behaviour (develop a theoretical model of disease transmission and investigate behaviour of the model in the context of varying parameter values); and
Evaluate interventions (develop a model to evaluate one or more of the treatment strategies or vaccination strategies).
The modelling rationales defined here are not necessarily mutually exclusive, therefore more than one modelling rationale may be used in a literature instance to guide the modelling process.
Compartmental classification
During the analysis of the literature, it is determined whether compartmental classification (discussed previously in §2.3.3) is incorporated within the modelling application. If it is incorporated, the compartmental classification categories that are used, as well as their descriptions, are noted and added to the dataset as these are serendipitously uncovered during the capturing process.
Modelling approaches
Similar to the reasoning provided for the methods utilised to fit models, it is not practical to research each available modelling approach in order to define a predetermined set. Such an approach could also bias the analysis by causing articles to be placed into categories that may be only a reasonably accurate classification. It is, however, practical to organise the modelling approaches into the three broad categories discussed previously in §2.3.4, namely:
Simulation models; Network models; and
Mathematical models.
Within each of these broad categories, the specific modelling approaches are noted and added serendipitously to the dataset as they are uncovered. Methods that do not fit the criteria for either the simulation or network model category, are incorporated within the mathematical category.
Mentioned transmission modes
As the transmission mode is expected to play a significant role in the disease dynamics, it is noted whether the disease transmission mode is explicitly mentioned, either within the contextualisation section or during a description of the modelling process. The reasoning is that this may indicate increased awareness of the dynamics of the specific disease transmission mode when selecting modelling approaches and incorporating contextual considerations for the particular disease outbreak.
Theoretical transmission modes
The theoretical transmission modes for each disease (as captured from the GIDEON database and mentioned in Table 3.3) are noted. In many cases, only some of the theoretical transmission modes of a disease may be explicitly mentioned in the article. Especially in cases where only some of the potential theoretical transmission modes of a disease are mentioned, it is reasonable to assume that the model was constructed to only include those transmission modes which are explicitly mentioned. When investigating the relationship between various modelling approaches and modelling considerations and the transmission mode, however, it is useful to consider both the full set of theoretical transmission modes for a disease as well as only the sub-set of transmission modes that are explicitly mentioned.
Alternative mixing patterns
The default mixing pattern in a modelling approach is the homogenous mixing of contacts, however, the incorporation of alternative (i.e. non-standard) mixing patterns are captured to the dataset as these are serendipitously uncovered during the capturing process.
Intervention strategies
The set of intervention strategies discussed in §2.5.2 are used as an initial template to categorise the incorporation of intervention strategies in the dataset. As additional intervention strategies are uncovered serendipitously during the capturing process, these are added to the dataset.
Contextual factors
During the analysis of the literature piece, it is determined whether contextual factors are incorporated in the modelling application. The two categories of contextual factors considered within the analysis are highlighted in Table 3.5 and the nature of the incorporation is noted according to the following criteria, namely:
Mentioned (a counter to keep track of contextual factors present in modelling applications); Linked to disease transmission (when the link between a contextual factor and the effect
on disease propagation is investigated); and
Modelled (when a contextual factor is explicitly modelled or included in a modelling application).
Table 3.5: Predetermined categories used to capture the contextual factors.
Environmental (criteria discussed in more detail in §2.4.1)
Demographics (criteria discussed in more detail in §2.4.2) Climate Seasonality Demography Population density Migration Socio-economic factors
Table 3.6: Summary of the omissions and deviations to the steps of the ‘iterative filtering’ process and ‘capturing data from the literature to the dataset’ process.
Omission and related section Deviation and related section
Transmission modes omitted from
the review. §3.2.2
Additional keyword exclusion as part
of the iterative filtering process. §3.3.3
RI diseases and non-RI diseases not
included in the review. §3.3.2
Additional timeframe exclusion as
part of the iterative filtering process. §3.3.3
Pay-per-view articles not included in
the dataset. §3.3.3
Mathematical vs simulation
classification assumption. §3.3.4
Referencing literature instances of
the dataset in the bibliography. §3.3.3
Assumption relating to capturing