Arbo-Score: A Rapid Score for Early Identification of Patients with Imported Arbovirosis Caused by Dengue, Chikungunya and Zika Virus

Background: Chikungunya (CHIKV), Dengue (DENV), and Zika (ZIKV) viruses present significant clinical and epidemiological overlap, making an accurate and rapid diagnosis challenging. Timely activation of preventive vector control measures is crucial to avoid outbreaks in non-endemic settings. Diagnosis is based on combination of serological and molecular assays which could be time consuming and sometimes disappointing. Methods: We report the results of a retrospective case-control study carried out at a tertiary teaching hospital in Italy, including all febrile subjects returning from tropical countries during the period 2014–2019. Controls were travelers with other febrile illnesses who tested negative in laboratory analysis for CHIKV, DENV, ZIKV arbovirosis. A score weighted on the regression coefficients for the independent predictors was generated. Results: Ninety patients were identified: 34 cases (22 DENV, 4 CHIKV, and 8 ZIKV) and 56 controls. According to our results, myalgia, cutaneous rash, absence of respiratory symptoms, leukopenia, and hypertransaminasemia showed the strongest association with arbovirosis. Combining these variables, we generated a scoring model that showed an excellent performance (AUC 0.93). The best cut-off (>=2) presented a sensitivity of 82.35% and specificity of 96.43%. Conclusion: A handy and simple score, based on three clinical data (myalgia, cutaneous rash and absence of respiratory symptoms) and two laboratory results (leukopenia and hypertransaminasemia), provides a useful tool to help diagnose arboviral infections and appropriately activate vector control measures in order to avoid local transmission.


Study Design, Setting and Inclusion Criteria
This was a retrospective, unmatched case-control study.
The study was carried out at the Infectious and Tropical Disease Unit of Careggi Hospital, a tertiary teaching hospital in Florence, Italy, in the period from 1 January 2014, to 31 December 2019.
We included all febrile subjects returning from the tropics referred to our outpatient or inpatient service (depending on the severity of their symptoms and comorbidities) for suspected imported arbovirosis. The inclusion criteria were as follows: Fever at the time of the medical visit, or history of fever (measured axillary temperature above 37.5 • C) developed within 2 weeks after returning from a tropical or subtropical country endemic for at least one of the three arboviruses (DENV, and/or CHIKV, and/or ZIKV).
(1) presentation to the service no later than 2 weeks from the first day of fever (2) being tested for DENV, and/or CHIKV, and/or ZIKV We excluded patients diagnosed with malaria, as the disease was usually ruled out at the beginning of the diagnostic process in travelers returning from malarial areas through blood smear and/or molecular tests. The study was conducted under the provisions of the Declaration of Helsinki and in accordance with the International Conference on Harmonization Consolidated Guideline on Good Clinical Practice.

Definitions
To identify DENV, CHIKV, and ZIKV infection cases, we used the case definition proposed by the Italian Ministry of Health arbovirosis surveillance plan, published in 2020 [32]. According to these definitions, we included in the "case group" in our study only probable or confirmed cases. Detailed information about these definitions is reported in Supplementary Table S1.
In all febrile subjects returning from the tropics, molecular tests on serum and/or urinary and/or saliva samples were performed in case of the onset of symptoms from less than 7 days. Otherwise, serological tests (IgG and IgM) were performed. In the case of DENV, we also performed the antigen NS1 test within 7 days from the beginning of clinical symptoms. Detailed information about tests performed at our center is reported in Supplementary Table S2. Controls were travelers with other febrile illnesses (OFIs) who tested negative in laboratory analysis for all three arboviruses.
For both cases and controls, main epidemiological and clinical data and laboratory values were collected in a database. A tourist was defined as any person who had crossed an international border to travel outside the country where they had settled; if the length of travel lasts more than 6 months, they are called expatriates. A migrant was defined as any person arriving in a country different from their own to settle in the new country. A traveler who had returned to their country of origin to visit relatives and/or friends was defined as a VRF.

Variables and Statistical Approach
We also collected clinical data regarding the most relevant presented signs and symptoms, as well as laboratory test results including counts of leukocytes, neutrophils and platelets, prothrombin time (PT), ALT and CRP values. Statistical analysis was made using STATA software (v. 14.0).
Continuous variables were summarized as medians and interquartile ranges (IQRs) and were compared using Mann-Whitney U tests for two-group comparisons. Categorical variables were expressed as frequencies and percentages and were analyzed using Chi-squared or Fisher's exact tests, as appropriate. Leukocytes, platelets, and ALT values were also interpreted as categorical variables using the reference values available in our laboratory. A univariate logistic regression analysis was performed. Any variable with a p-value equal or inferior to 0.2 was considered potentially significant and was further analyzed in multivariate logistic regression (CRP and neutrophil count were excluded because they were available only for limited number of observations). Therefore, we created a simplified score based on the regression coefficients for the independent predictors. Specifically, as previously reported, we divided the smallest regression coefficient by the lowest factor in the model and rounded this quotient to the nearest whole number [33]. We assessed the performance of the score model with the area under the curve (AUC) of receiver operating characteristic (ROC) curve [34]. Therefore, we found the best cut-off by calculating the Youden Index.

Results
Detailed information about the selection process of cases and controls is described in the flow chart in Supplementary Figure S1.
In Table 2, we report the characteristics of different arbovirosis.  Regarding DENV infection, myalgia (n = 17, 77.3%) was the most common symptom, and leukopenia (n = 17, 77.3%) was the most common laboratory finding. Interestingly, there was only one case of severe dengue, which also presented leukocytosis (29,800/µL). Arthralgia and arthritis were the most common clinical features present in CHIKV infections, in 100% and 75% of patients, respectively. The most frequent symptoms of ZIKV patients were rash (100%) and headache (75%). Regarding the number of neutrophils, the median value resulted smaller in DENV infection (1418/µL). CRP was positive in 8 of 24 tested cases (33.3%): the highest value (189 mg/dL) was registered in the case of severe dengue, where no bacterial superinfection was demonstrated. Among cases with hypertransaminasemia, 11 (91.7%) were DENV infections; there were no ZIKV patients presenting with ALT elevation. Median platelet level was higher in CHIKV (349 × 10 3 /mcL) than DENV (142 × 10 3 /mcL) and ZIKV (158 × 10 3 /mcL). According to our multivariate analysis, myalgia, rash, absence of respiratory symptoms, leukopenia and hypertransaminasemia showed the strongest association with arbovirosis (Table 3). Therefore, using regression coefficients, we generated a scoring model including +2 points for leukopenia, +1 point for hypertransaminasemia, +1 point for rash, and +1 point for myalgia and −1 point for respiratory symptoms. The receiver operating characteristics (ROC) curve showed an AUC of 0.93 (Figure 1). The best cut-off point resulted in greater than or equal to 2 (Table 4). Table 4. Youden index, sensitivity, and specificity of different cut-off points of a score for the early identification of imported arbovirosis infections in febrile ill travelers returning from an endemic country, based on three clinical signs (myalgia +1 point, rash +1 point, respiratory symptoms −1 point) and two laboratory values (leukopenia +2 points, hypertransaminasemia +1 point).

Discussion
Physicians who manage febrile, returning travelers must always place priority on the differential diagnosis of conditions that are treatable, that may cause serious sequelae or death, and pose a risk to public health [35]. Arbovirosis fulfills all the aforementioned priority conditions. A subject returning from an endemic area represents a risk for the emergence of autochthonous cases in areas where competent vectors are present; hence, an appropriate diagnostic approach is crucial to limit this risk. Very recently, small clusters of autochthonous DENV infection have been reported in northeast Italy and southern France, highlighting that, despite current travel limitations imposed by the COVID-19 pandemic, imported arbovirosis may still represent a challenge for clinicians and public health officers in temperate regions [36][37][38][39]. In 2016, researchers from Lausanne Infectious Disease Unit proposed one diagnostic algorithm for travelers with nonspecific febrile illnesses returning from regions experiencing simultaneous outbreaks of DENV, CHIKV, and ZIKV infections, based on serology results [40]. In our opinion, this approach could present some limitations. Firstly, serology in the first week from the onset of symptoms could be negative. Secondly, NAAT and serological ELISA tests could be available only in referral centers [41]. Therefore, patients who have been previously vaccinated for another Flavivirus such as Yellow Fever Virus, Japanese Encephalitis Virus, and Tick-borne Encephalitis Virus, could present false positive serology for DENV or ZIKV. Considering that the cost of routine surveillance preventive measures, such as comprehensive larviciding, over whole urban areas could overcome health benefits, especially in larger municipalities, rapid identification of imported arboviral infections is crucial to rapidly activate preventive measures localized in a specific area to avoid local transmission [42]. Regarding DENV infection, another aspect to consider is the variability of the viremic period, which could last on average nine days in travelers after the onset of symptoms, compared to seven days in endemic settings [43]. At the same time, the extrinsic incubation period (EIP) in competent vector mosquitos, generally referenced to be 8-12 days, could be more variable and inversely correlated with temperature [44]. In a study published in 2019, lower leukocyte and platelet levels resulted in significant associations with DENV, CHIKV, and ZIKV in the differential diagnosis of imported fever [45].
According to our results, leukopenia, cutaneous rash, hypertransaminasemia, respiratory symptoms (at least one among cough, respiratory distress, sore throat and rhinorrhea) and myalgia were the best parameters to predict arbovirosis in a febrile traveler returning from a tropical country. A simple score, called ARBO-SCORE, based on three clinical data values v(myalgia, cutaneous rash and absence of respiratory symptoms) and two easily obtainable laboratory results (leukopenia and hypertransaminasemia), has been shown to provide a useful tool to help diagnose arboviral infections and to effectively activate vector control measures in order to avoid local transmission, with an accuracy of 91%. The score we generated is straightforward and applicable in peripheral centers. Using our best cut-off point, we obtained a sensitivity of 82.35% and specificity of 96.43%. Regarding different arbovirosis, it is important to emphasize that our case of severe dengue presented leucocitosis, which is a parameter prognostic for this disease [46]. Therefore, there was a notably different median for the platelet count: in a previous study, a lower platelet level was considered to be a prognostic value for DENV versus CHIKV [47]. The differences among these three arbovirosis reflect distinctive characteristics already previously described [48]. The study has several limitations. We excluded many patients because of a lack of information about clinical history or laboratory data. Selection biases could have been additional limitations in this study design. Cases and controls were not matched, so sex and age distribution were different in the two groups. Most cases were DENV infections (64.7%). Presence or history of fever could have limited the number of ZIKV cases. In fact, we excluded three afebrile ZIKV cases. We did not perform any statistical analysis to differentiate among the three arbovirosis, because we estimated too small a number of cases.

Conclusions
We elaborated a predictive score named ARBO-SCORE based on three clinical signs (myalgia, rash and respiratory symptoms) and two laboratory values (leukopenia and hypertransaminasemia). The score showed an excellent performance in our population (AUC 0.93), but another validation cohort is necessary to confirm its predictive value. This score could be performed as soon as suspected febrile travelers are encountered, due to its straightforward nature. Therefore, the score is not meant to replace traditional tests for diagnosis, but could help to appropriately activate surveillance systems and perform preventive measures against vectors to avoid autochthonous cases. Further studies are necessary to find clinical predictive scores for each of these arbovirosis.

Supplementary Materials:
The following are available online at http://www.mdpi.com/2076-2607/8/11/1731/s1: Figure S1: Flow chart of the selection process of cases and controls; Table S1: Definition of possible, probable and confirmed CHIK, DENV and ZIKV cases according to the 2020-2025 Italian National arbovirosis prevention and surveillance plan; Table S2: DENV, CHIKV, ZIKV serological test performed at Careggi University Hospital, Florence, Italy; Table S3: Diagnostic results of cases.

Conflicts of Interest:
The authors declare no conflict of interest.