Using AI to answer clinical questions: COPD Case Studies

David Reeves
published on May 02, 2019


Chronic Obstructive Pulmonary Disease (COPD) describes a group of conditions which cause breathing difficulties, and includes chronic bronchitis and emphysema. The condition mainly affects older people who smoke and is typically managed by clinicians and other health professionals in a primary care setting.

Whilst common, COPD presents healthcare professionals with a number of diagnosis and management challenges. The symptoms are non-specific and occur gradually, with sufferers often initially presenting with normal spirometry. This means that COPD is under-reported and is subject to a difficult differential diagnosis, especially with regards to asthma, as well as other more serious conditions. It is therefore crucially important that clinicians have a full understanding of COPD symptoms, causes and therapy options, and the evidence base that informs this, in order that patients can receive appropriate treatment.

Point of care tools are now able to leverage Artificial Intelligence technologies to assist clinicians in this process, and we shall explore exactly how by utilising the Causaly AI engine to find answers to two COPD related clinical scenarios.

Case Study 1: Symptoms

“A patient, smoker, presents with wheezing and cough, alongside fatigue, sleep disturbance and depression. As the patient is a smoker with wheeze and cough, the clinician suspects COPD, but wants to quickly determine whether sleep disturbance and depression are also associated with COPD”

The clinician needs to determine whether there is evidence to suggest that sleep disturbance and depression are symptoms of COPD. From the Causaly home screen, we can directly search for COPD:
This brings us to the main COPD page (indexed as chronic obstructive airway disease in Causaly). From here, we can use the filters to show evidence only of symptoms:

We can view the symptoms using the Causaly dendrogram, providing a visual overview of every COPD symptom mentioned in the biomedical evidence base:
Quickly scanning the list of symptoms reveals that there is in fact evidence linking COPD with both depression and sleep disturbance. We can now click on the symptom, for example ‘Depressed –symptom’, to be taken directly to all of the contextual evidence from within the biomedical literature linking the two conditions:
There is evidence from 98 studies describing the relationship between COPD and depression, and the clinician is able to quickly explore each piece of relevant evidence. We can see, for example, that the first citation notes that ‘COPD severity was significantly correlated with anxiety and depression’. The health professional is able to explore the entire body of evidence linking the two conditions, should they desire, and can quickly build up a detailed and nuanced picture of the association.
In this example, the clinician suspected COPD because of wheeze, cough and fatigue, but was unsure of the links to depression and sleep disturbance. The information found using Causaly therefore enables a more confident and informed diagnosis, reducing diagnosis error rates and improving patient outcomes.

Case Study 2: Risk Factors

“A patient has recently been diagnosed with COPD. The patient is obese and asks the clinician about the association between COPD and Obesity. The clinician knows that the evidence linking COPD and Obesity is complex and tentative, and wants to quickly learn more about the literature supporting the association, in order to best inform the patient”

In this example, the healthcare professional would like to explore whether obesity is a risk factor for COPD. Again we conduct a search for COPD, but this time we filter by Disease, rather than Symptom as before.
Once again, a quick scan of the results reveals that Obesity is in fact associated with COPD, with 76 pieces of evidence linking the two:
On this occasion, we might want to only look at the best evidence for the relationship, rather than all of the evidence. In this case, we are able to utilise the filters once more, perhaps restricting the evidence to that published in last 5 years, and restricting the publication type to Review, revealing the most up to date and highest quality evidence.
We can now see recent evidence, published in systematic review articles, linking Obesity to COPD, with the second citation highlighting that “obesity is more common in COPD patients”. Once again, the clinician is able to fully explore the relationship and can quickly provide informed answers to the patient, confident that the advice is based on best evidence.


As we have seen from these case studies, the AI model employed by Causaly can give healthcare professionals direct access to deep, complex and relevant literature.

The AI approach is powerful because the model can find the relevant evidence for any relationship, no matter how esoteric.

Currently, if a healthcare professional wants to find the evidence behind a particular condition or therapy, or simply is unsure of how certain conditions or concepts are related, they will often turn first to a point of care clinical decision tool, such as UpToDate or NHS Evidence. Although these tools are a powerful and essential part of modern evidence based medicine, they provide a focused overview of a topic, and will not necessarily cover the specific detail that a clinician needs. In this case, health professionals would probably turn to the vast biomedical databases, such as MEDLINE™ or PubMed, at which point the search for information becomes complex, time consuming and often requiring specialist assistance.

The AI model solves this problem. No matter how obscure the clinical question, or how deep the answer is hidden within the biomedical literature, Causaly AI can find the answer within seconds, allowing the clinician to make evidence based decisions and to provide evidence based information to patients immediately.

David Reeves is a freelance clinical librarian and medical writer.

AI for Clinical Decision Support – What conditions cause female infertility?
use case

AI for Clinical Decision Support – What conditions cause female infertility?

Top causes of female infertility using AI supported clinical decision systems

Can AI enhance traditional clinical literature research methods?

Can AI enhance traditional clinical literature research methods?

The process of finding and evaluating existing clinical research is central to all areas of biomedicine, providing the foundations upon...

Understanding Clinical Outcomes of Spinal Muscular Atrophy
use case

Understanding Clinical Outcomes of Spinal Muscular Atrophy

The objective of this study was to evaluate all possible symptoms of SMA to identify relevant research articles and to define SMA prevalence comprehensively. We asked the question: What are the disorders and syndromes associated with SMA?

Be the first to know

Sign up for Causaly Newsletter