You’ve probably heard that pneumonia kills more children than any other disease, but have you ever wondered how we ‘count’ pneumonia cases? As we look at disease burden figures from around the world, how do we distinguish pneumonia from other illnesses? There are several ways to count cases, but as it turns out, it’s harder than you might imagine.
Participants learned the subtleties of interpreting chest x-rays at PERCH workshop.
Why? Because we do not often get samples from the infected site (the lung) as this is very invasive. Instead, we can look at children who meet a clinical definition such as “rapid, shallow breathing,” but we know that kind of definition overestimates pneumonia cases – in other words, you get many false positives. We can also look to see if any bacteria grows in a blood sample from a child with likely pneumonia, but we miss many cases that way too, this time because of many false negatives, because bacteria aren’t always found in the blood and bacteria are not the only cause of pneumonia. The best approach has traditionally been looking at a chest X-ray or CXR, which is a frequently used tool for identifying pneumonia in clinical studies. But this leaves us with another not-so-simple question: what, exactly, does pneumonia look like on a CXR?
This is an important question for the Pneumonia Etiology Research for Child Health (PERCH) study team. With seven sites across Africa and Asia, PERCH is enrolling hospitalized children with severe and very severe respiratory symptoms according to WHO clinical definitions. The study aims to better understand which infectious pathogens are now causing childhood pneumonia in a variety of settings – rural, urban, HIV prevalent or not, differing vaccine coverage levels, and limited resource availability – so we can better plan for the interventions to prevent and treat pneumonia in the future. But to do this, we first need to know which children we think have pneumonia actually do, and which do not.
Defining pneumonia on a CXR can be more troublesome than it sounds. First, interpretation of a CXR involves subtleties of skill and experience than can make it a somewhat subjective process when comparing two doctors’ results. Doctors in different countries, or even just on different floors of the same hospital, may read CXRs differently. So, it is important that we have a standardized approach to radiologic diagnosis. In research (and for PERCH in particular) we are also very interested in those children who definitely do not have pneumonia, as understanding what pathogens are carried by sick children both with and without pneumonia helps us to know which pathogens are causing the greatest burden of disease. So we aim for specificity in our CXR interpretations, which is different from day-to-day clinical practice, where the focus is to identify definite cases of pneumonia in order to quickly treat the child.
Participants examined chest x-rays on laptops during the workshop.
To help achieve the goal of a uniform diagnosis, the WHO developed a methodology for the standardized interpretation of pediatric CXRs. The methodology encourages simple definitions for pneumonia and aims for a high level of agreement between CXR readers. This approach was developed for vaccine impact studies, however, which are importantly different from PERCH’s questions relating to the causes of disease. Beyond PERCH there is continued demand for radiological standardization in environmental studies, antibiotic treatment studies, and epidemiologic studies of pneumonia.
Although a complete review of the WHO methodology is not a primary objective of the PERCH project, the PERCH team knew it was important to organize a comprehensive training program before we began interpreting CXRs in the study. In late August 2012, 14 physicians and radiologists (each associated with one of the seven study sites) met in London, along with collaborating colleagues from the CDC’s International Emerging Infections Program, the Drakenstein Child Health Lung Study in South Africa, and the PCV (Pneumococcal Conjugate Vaccine) Impact Study in Kenya. They were joined by members of the PERCH core team and four technical experts from Australia, Kenya, and the United Kingdom, who served as trainers for the meeting. The purpose of this meeting was to realign the WHO methodology to current study goals, identify ways in which it can be optimized, and ensure PERCH CXR readers have a harmonized approach to diagnosing radiological pneumonia.
X-rays may be one of medicine’s oldest technologies, but our challenge was to improve their interpretation to help a very contemporary study. This gathering allowed the team to think critically about how CXR diagnosis can best serve PERCH’s needs, and the outcomes will help not only the PERCH project, but broader pneumonia diagnosis and research efforts as well. We were able to refine the outcomes we are gathering from CXRs, to be clear on which children have normal CXRs and which definitely have pneumonia, while at the same time calibrate CXRs from PERCH against the WHO methodology to extend these definitions beyond the vaccine impact setting. We now have a group of highly trained CXR readers who will be providing assessments on all of the CXRs from cases within the PERCH study, and we have established a feedback mechanism so that the quality and safety of CXRs at each site is maximized both within the study and beyond.
With this important step now complete, we are armed to interpret the subtle CXR findings that a study like PERCH may identify. So now, for PERCH, the ‘counting’ begins.
Nick Fancourt is a PhD candidate at the Johns Hopkins Bloomberg School of Public Health and member of the PERCH study team. He is also an International Fulbright Science and Technology Fellow.
The views and information expressed within this blog are the author’s alone and do not represent the Fulbright Program or the U.S. Department of State.