In the present study, we analyzed the clinical signs and symptoms of 548 patients with suspected CHIKV infection and their association with confirmed CHIKV serology results to formulate a clinical screening tool for use in primary care settings. After univariate analysis, the variables associated with positive CHIKV serology results were discussed with an expert panel. Based on their experience in diagnosis and disease treatment, the most representative variables in CHIKV-infected patients were established.
It is well known that expert consensus defines the most appropriate selection of variables by considering different perspectives and positions of the experts consulted in the process39. Therefore, we included more meaningful and valuable variables for clinicians in the final multivariate analysis. The symptoms independently associated with CHIKV infection were observed to be symmetric arthritis, fatigue, rash, and ankle joint pain. A clinical screening tool was developed, which yielded high specificity (87.4%) and positive predictive value (PPV; 85.6%) with moderate sensitivity (64.4%) and negative predictive value (NPV; 67.8%).
When assessing people´s health, two types of tests are used: diagnostic tests that offer final information on the presence or absence of a condition and screening tests that are less demanding on the healthcare system, more accessible, less invasive, time-consuming, and expensive40. The screening tests display ideal characteristics for countries where arboviral diseases caused by CHIKV, DENV, and ZIKAV are endemic. These tests are evaluated according to their sensitivity, specificity, PPV, and NPV. In brief, sensitivity and specificity refer to the accuracy of a screening test with a reference or gold standard, whereas PPV and NPV indicate the success of a screening test in classifying people as having or not having a condition40. Therefore, in screening situations for individuals in a clinical setting, it is more appropriate to use the PPV and NPV values for evaluating the performance of a screening tool. Our screening tool has a high PPV but a moderate NPV. A high PPV is desirable in situations where the costs of diagnostics, treatment, and services are increased when the studied condition progresses slowly or is not life-threatening40. A moderate NPV might be acceptable if later assessments can be programmed and completed or if the condition is possible to sort out with no treatment40. Since chikungunya disease displays the above-mentioned characteristics, we believe that our screening tool will be useful in diagnosing CHIKV infection in a primary care setting where only clinical variables are at hand.
Other performance indicators of our screening tool, such as the Youden index (YI; 52) and accuracy (75.0%) when applied to our cohort, were higher than the previously developed diagnostic criteria or screening tools for CHIKV infection. The WHO case definition showed lower performance values in our cohort than our screening tool. Due to the ambiguity of the definition (Supplementary Table S3), only the mandatory symptoms (joint pain and fever) were used to calculate performance in our cohort31. Fever and joint pain are commonly included in other arboviral case definitions. For example, the WHO ZIKAV case definition (2016) states that the presence of fever or rash plus at least one more symptom, of which arthralgia is one, constitutes a suspected case41. In addition, the WHO DENV case definition includes fever plus two more symptoms, in which pain and aches are also present42. In fact, the recurring symptoms in the WHO case definitions of CHIKV, ZIKAV, and DENV are fever, aches, and pain (interpreted as arthralgia or myalgia). Furthermore, in the WHO CHIKV case definition, the use of terms like “usually incapacitating” or “usually accompanied by…” could lead to misinterpretation by physicians resulting in over or underdiagnosis and finally poor performance when used in epidemics. Moreover, using fever as a mandatory or inclusion symptom dismisses asymptomatic patients, which increases the percentage of false negatives and selection bias.
Multiple attempts have been made to develop a better screening tool. Sissoko et al. (2010) found the combination of fever and polyarthralgia as the most relevant clinical pattern of CHIKV infection to identify presumptive cases during epidemics, yielding an accuracy of 87% with high sensitivity (84%) and specificity (89%)11. However, when applied to our cohort, the accuracy and sensitivity decreased to 69% and 51%, respectively. A possible explanation could be attributed to the median age of their cohort (24 years) since symptomatic expression of infection is lower in younger age groups11.
In 2013, Thiberville et al. developed a clinical score with fever and arthralgia as mandatory symptoms. They added the presence of specific joint involvement (wrist or hand arthralgia) and the absence of myalgia to improve performance26. Their clinical score had the best performance in our cohort, with similar results as our own (Table 5). We believe that the resemblance lies in the use of specific disease symptoms. Our screening tool requires the inclusion of symmetrical arthritis or ankle joint pain to reach the cut-off point when added to other more generic symptoms such as fatigue and rash. These symptoms are almost unique to CHIKV infection and are rare in other arboviral infections caused by ZIKAV or DENV.
A study by Macpherson et al. (2014) found that a patient with joint pain and any combination of fever, myalgia, or rash was in 85% agreement with a positive CHIKV serological test result12. However, when applied to our cohort, the combination of arthralgia and fever yielded the best accuracy but sacrificed YI. Combining arthralgia with myalgia or rash increased specificity at the expense of sensitivity.
Other authors have elaborated on screening tools using simple clinical laboratory parameters. For example, the performance of Thiberville screening tool increases if lymphopenia is present26. Godaert et al. (2017) used lymphopenia in the presence of fever and ankle arthralgia and the absence of neutrophil leukocytosis for CHIKV infection screening in elderly people16. Laboratory studies improve diagnosis; however, even simple laboratory tests are sometimes unavailable to primary care physicians. Therefore, developing a diagnostic tool based on clinical parameters was our primary goal.
With the appearance of ZIKAV epidemics, the clinical symptoms that help differentiate CHIKV from ZIKAV or DENV have been studied. Cleton et al. (2015) found that arthralgia, arthritis, and rash were associated with CHIKV infection, whereas DENV-positive patients had increased odds ratios for rash, fever, and hemorrhagic symptoms43. In our cohort, the syndromic combination of arthritis and rash yielded a high specificity but moderate to low sensitivity, yet a similar PPV and NPV to our screening tool.
Sahadeo et al. (2015) compared patients with confirmed DENV and CHIKV infections to obtain clinical and laboratory features that could help distinguish between the two diseases27. The combination with the best performance to differentiate between DENV and CHIKV infection was rash, joint pain, and leukocyte count < 7x103/µl. However, the PPV (58%) was less than optimal.
Another study by Lee et al. (2012) designed decision tree models for discriminating between DENV and CHIKV infections using clinical symptoms (presence of fever and its duration, bleeding, and illness) or laboratory tests (presence of thrombocytopenia)22. Interestingly, fever was associated with DENV infection and absent in CHIKV infection. A similar feature was found in our screening tool, where fever was not a mandatory symptom for suspected CHIKV diagnosis.
In a noteworthy study by Braga et al. (2017), a case definition was developed from a cohort of patients where CHIKV, DENV, and ZIKA were co-circulating17. A score ≥ 7.5 granted the diagnosis of ZIKV from the following symptoms: pruritus, rash, conjunctival hyperemia, and the absence of fever and anorexia. This is consistent with the decision tree of Lee et al. (2012) and our screening tool, where fever as a symptom was not included. Half of the patients in our cohort with confirmed CHIKV would test positive for ZIKAV according to Braga’s ZIKAV case definition using a combination of no fever and rash (true positives: 58.9%), rash and pruritus (true positives: 44.7%), and no fever or pruritus (true positives: 55.9%). This can be explained by the fact that rash, a frequent symptom in our CHIKV-confirmed patients (87.4%), was awarded a high score (7 points from a cut-off of 7.5) in Braga’s case definition. Other studies on CHIKV epidemics have reported similar findings44,45. One could argue that if the CHIKV sample in Braga et al. (2017) study was larger, the appointed value for rash would be smaller and would have less preponderance in the score.
The same exercise yielded similar results when the WHO DENV case definitions were applied to our cohort. This exemplifies the need to assign importance to cardinal and specific symptoms of each infection. Although fatigue and rash are present in other definitions of DENV and ZIKAV, the presence of joint involvement (symmetrical arthritis or ankle joint pain) is mandatory in our screening tool to reach the cut-off point of 5.5. With this in mind, we proposed an algorithm for the clinical approach to CHIKV, ZIKAV, and DENV infections (Fig. 7).
Our study has some limitations. First, given that our study resulted from the structure of a COPCORD approach, there is a selection bias regarding MSK symptoms. Second, since there was no physical exam when the symptoms started in each patient, these symptoms could not be validated by a physician; therefore, recall bias could be present. Third, PCR was not performed to confirm CHIKV infection because of its high cost. Lastly, our screening tool has not been validated in other cohorts.
Nevertheless, our study has several strengths. The country’s population is well represented in the study, bearing in mind the number of samples and the six geographic regions in which the study was performed. In addition, the accuracy of physical examination, especially of the musculoskeletal system, was warranted because all the patients were evaluated by a trained or in training professional of rheumatology. Finally, since the patients were evaluated in their homes and not in a medical setting, we could find even asymptomatic patients who otherwise would not attend a physician.