Is Extensive testing of Susceptible Population Critical to Control an Epidemic?

Evidence Based Medicine

I am an admirer of Dr. David Sackett, who is regarded as the Founder of the discipline of clinical epidemiology. His major contributions to the public health discipline are applying scientific and simple mathematical methods in clinical appraisals and measuring the outcome of clinical interventions to understand efficacy.  I happen to attend one of his lectures out of interest when he was teaching at McMaster University, Canada, while I was taking courses in epidemiology at Guelph.  His book, ‘Clinical Epidemiology’, was a reading reference material given to us. I valued this book so much that after my return to India when someone failed to return this book, I had to send an SOS to Dr. Wayne Martin, my epidemiology teacher, who ultimately shared an extra personal copy.  Ask any Ph.D. holder (maybe from MIT or Harvard or IIT), about his Ph.D. research, and the majority would answer that the thesis had a very short shelf-life (sometimes ends even before convocation).  What one learns during the Ph.D. is to ‘ask the right questions’ and search for ‘evidence-based answers’ to it.  The context of my introductory paragraph is in this blog. I am going to ask a question; ‘Is conducting laboratory tests on the majority of the susceptible population critical to control an epidemic?  Moreover, I will attempt to answer by relying on a simple method used by Dr. Sackett (and others) in their clinical research.

Clinical signs more powerful than laboratory tests

The clinical epidemiologists consider even symptoms or clinical signs as one of the important tests- evidence for arriving at a proper diagnosis. Experienced clinicians who are adept at taking history and physical examination will find that at times the presenting clinical signs are more powerful than many laboratory tests. There are easy ways to measure the diagnostic strength of symptoms. First, record symptoms in suspected patients (a mixture of diseased and non-diseased) and also carry out confirmatory laboratory test(s) in the same set of patients.  The confirmatory test should be the ‘Gold Standard’ in diagnosing the disease (such as X-ray, MRI, CT Scan, PCR, blood tests, etc.).   The data obtained then can be arrayed in a simple 2 x 2 table.

Let me explain this by using mock data.

The symptoms here are in series (that is when all the three signs are present). You can also do this exercise with a single symptom and combine ones to see which set is better. The Gold Standard Test for the disease is PCR, with an assumption that when the PCR test in a case is positive, the diagnosis is confirmed.  From this table, ‘sensitivity’ and ‘specificity’ of symptoms as measures of diagnostic strengths can be calculated. Even other indices, such as likelihood ratios, Gain in Certainty, Odds Ratio, etc., can also be calculated, but I would restrict to only two basic indices. The sensitivity is 95 % (a/a+b) and the specificity is 87 % (c/c+d).

This means that if a diagnosis is made based on the presence of symptoms alone only 5 % positive cases are likely to be missed, and around 13% of negative cases are likely to be falsely implicated. Is this a big deal? The acceptable threshold of sensitivity and specificity would vary from disease to disease, morbidity, mortality, available medical interventions, cost, and logistics of the confirmatory test. In case the medical intervention is surgical both sensitivity and specificity should be high (> 99%)  If there is no vaccine or specific treatment available in that case, high sensitivity but moderate specificity is acceptable.

For the sake of understanding let us consider that the disease is highly contagious (expected morbidity more than 80%) but mortality is low (less than 1%), and the intervention options for positive cases are:

(a) Self-quarantine for 14 days

(b) Quarantine plus home treatment and

(c) hospitalization.

In the above case, relying only on symptoms will lead to missing 5 % patients who would remain in the population undetected but would be ultimately hospitalized if they become serious.  Second, 13% of the patients who are actually negative will be advised self-isolation but ultimately will be released since those are negative. By including a PCR in the protocol only 5 % diagnostic strength will be added.  So, a considered decision whether to let trained volunteers record people with a set of symptoms and prescribe self-isolation and hospitalization of serious cases is to be given a priority over the creation of expensive testing facilities needs to be taken by the planners.  There are also ways to refine symptom-recording.  For example, instead of the only cough, it could be refined to ‘cough with a frequency every three minutes’; instead of fever, ‘between 101 or more than 103 ^oF’ may further enhance sensitivity and specificity. Recording of clinical signs with a scoring system also enable developing algorithm to support e-diagnosis.

PCR a Gold-Standard A Myth or Reality

Many times, a laboratory test, in the real sense, does not significantly alter diagnostic precision or medical intervention.  For example, when we see a case of milk fever, the diagnostic precision on symptoms alone is very high, almost 99%, and in such a case if serum calcium is ordered it will not alter even the treatment protocol as irrespective of serum Ca level treatment dose would remain same.  So, a laboratory test should not only significantly add to diagnostic precision but should also help in altering the medical intervention rationally and to the benefit of the patient.

Although in the above example, PCR is taken as a ‘Gold Standard’ test, in reality, it is not.  It is a ‘Gold Standard’ for detecting the presence of the organism (bacteria or virus) but not the disease.  In the event of an epidemic in the initial stages, doing such a confirmatory test is essential for confirmation of the diagnosis to give credence to the association of symptoms to a disease.  But, in later stages, screening the majority of the population in the midst of a pandemic or epidemic is not necessary if the symptoms alone can provide diagnostic strength (sensitivity) of > 90%.  As veterinarians, we all know that when there is a suspected outbreak of FMD, few initial cases go through the laboratory tests, and based on symptoms in later stages, cases are diagnosed and treated.

There is also another constraint with PCR tests. During an outbreak, many people or animals are likely to harbor pathogens, but that does not mean these cases would necessarily run a clinical course. In fact, they might become immune due to this exposure!  In the PCR test, there is an induced chemical amplification of the organism’s genomic material; hence it would not measure the infection load a person is carrying. The pro-testing lobby’s argument that a PCR test will detect infected people before there are symptoms would appear rational because these could be potential shedders of the virus. There are unequivocal case studies that proved that even very serious proportion viral outbreaks came under control only when people (or animals) attain immune status.  This is possible only when the population undergoes vaccination (if available) or when > 60 % get a low dose of virus from the environment.

This blog is specially meant for the veterinarians who, unfortunately (or fortunately) don’t have access to the diagnostic laboratories like their counterpart.  They can build up their diagnosis confidence by rationally considering clinical presentation and appraisals by following such simple methodologies.

(The blogger practices Herd Medicine) 

Dr. Abdul Samad