Epidemiology is the study of the causes, patterns and distributions of health and disease across populations, and is crucial to modern healthcare provision, informing evidence based healthcare policy and public health best practice. Central to epidemiological analysis is the twin concepts of incidence and prevalence. Prevalence describes the proportion of a population with a specific disease, and incidence describes the rate at which new cases are acquired in the population. Together, these statistics allow public health professionals to shape clinical and policy responses to disease outbreaks based upon important evidence.
Because of the usefulness of such statistics, incidence and prevalence data appear frequently in published research. This is especially true for systematic review, since research synthesis studies often have a powerful effect on policy and practice decision making. Currently, researchers and public health professionals who are looking for such statistical data will use the large biomedical bibliographic databases to search for this information in primary or review studies, or perhaps use collated services such as those provided by the Office for National Statistic or the Center for Disease Control for example.
Finding this information is currently a time consuming and resource intensive task, but Causaly offers a new approach, placing this data directly at the fingertips of epidemiology researchers.
Using AI technologies, Causaly is able to machine-read the biomedical literature corpus, understand and extract epidemiology data from within the literature and present this directly to the researcher in an easy to read and interactive way.
We shall explore how Causaly delivers this data by exploring the incidence and prevalence statistics of systemic lupus erythematosus, a common autoimmune disease associated with joint pain, fatigue and characteristic rash. We will then compare the data delivered by Causaly with those found in a systematic review of the topic.
Using Causaly’s Epidemiology Module
Finding epidemiology data using Causaly is simple and quick. From the home screen, we navigate to the Epidemiology Module, where we simply type in the disease or condition of interest:
Causaly first presents a graphical representation of incidence and prevalence data, providing an overview of the epidemiological data found in the literature corpus. In the case of Lupus, we can see right away in the histogram that the majority of the literature reports incidence/prevalence data in the 0-10% range, with data in other ranges being evenly distributed at much lower levels.
The adjacent evidence count bar graph gives further insight in the data, illustrating the number and proportion of studies that report epidemiology data as a single value, as a range value, that report no value but may be relevant and those that have a lot of noise but that could still be useful.
Causaly allows researchers to filter the results according to their specific needs. Clicking on the filter button reveals the filter menu, which contains a number of population context filters, including age range, geographic location, ethnicity, country and so on:
This type of population information is crucial to epidemiology research, and Causaly allows users to interact with and restrict the evidence with the touch of a button, removing the need to spend time writing and rewriting database search queries in order to find this information.
As an example, we can restrict the search to show only evidence related to adult age groups in Europe. In this case, we now have 97 articles with 127 extracted incidence/prevalence values:
Causaly brings the researcher right to this evidence, highlighting the article in which the data is found and displaying the actual sentence context from within the paper, as well as extracting the retrieved data values. Users can sort the results, share the search strategy and export the articles to Excel at the touch of a button or can explore each piece of relevant literature in more depth by clicking on the appropriate title.
Within seconds, researchers are able to find and interact with epidemiology data that is otherwise hidden within the massive biomedical literature corpus, and can customise and tailor the information exactly to their needs.
Causaly not only delivers information to researchers very quickly, it also delivers information of high quality. We can explore the quality of this data by comparing information retrieved by Causaly to that presented in a systematic review of the topic. The systematic review represents the gold standard for clinical research methodologies, providing trusted and useful data to practitioners, researchers and policy makers. As an example review, we will compare Causaly with the findings presented in the systemic review by Rees et al. (2017).
It is first worth pointing out that Causaly has found 929 articles that are potentially relevant, which compares favourably with the database searches used in the study. After the title and abstract screening process, the systematic review found 168 potentially relevant studies and this is after the authors had screened 4936 articles retrieved in the initial search, the vast majority of which were not relevant. Causaly finds this information immediately, and finds many more potentially relevant articles, highlighting the fact that systematic literature searching can be improved by including Causaly in the search methodology.
To compare the actual data, we can take an example population reported in the systematic review and compare this with the same population, found using the Causaly filter system. Taking incidence of SLE in the UK population as our example, we can see that Causaly retrieves all of the studies included in the systematic review, as well as a number of others not included:
Without examining the review in depth, we cannot know whether these extra studies were excluded deliberately or were simply not found in the review methodology, but we can be certain that Causaly is providing extra epidemiological data. Researchers can explore and examine these other studies to determine whether or not they are useful, and therefore have a more comprehensive overview of the topic than if they had relied on the review alone. The important difference however, is that a systematic review can take months or even years to complete, whereas Causaly is able to find comparable information in seconds.
Systematic review methodology is complex and nuanced, with strict exclusion criteria and quality appraisal processes and so we are of course not comparing like for like in our analysis here. However, Causaly does offer a remarkably simple, quick and cost efficient synthesis tool that offers a reasonable approximation of review findings, at a fraction of the resource cost.
This could be especially useful if a researcher wants to understand what the body of literature says about epidemiology data, but doesn’t have access to a systematic review, or a systematic review doesn’t exist. The Causaly epidemiology module can be an excellent and easy to use starting point. The module can also be used to add value to existing systematic review methodologies, providing research synthesists and clinical librarians with another tool to use alongside the bibliographic databases.
Causaly is able to offer this quick and efficient solution by leveraging AI technologies to machine read the actual full texts of biomedical literature, not relying simply on title and abstract parsing as is currently routine. As Causaly grows, more full text documents will be added to the text corpus, making the tool ever more powerful.
Rees, F., Doherty, M., Grainge, M., Lanyon, P., & Zhang, W. (2017). The worldwide incidence and prevalence of systemic lupus erythematosus: A systematic review of epidemiological studies. Rheumatology, 56(11), 1945-1961.