Preliminary fast diagnosis of severe fever with thrombocytopenia syndrome with clinical and epidemiological parameters

Objectives To identify specific clinical and epidemiological parameters for clinical diagnosis of SFTSV infection with relatively higher accuracy. Methods 231 suspected cases of SFTS were reported by various medical institutions from 2011 to 2013 in Jiangsu Province, China. They were followed with SFTSV diagnosis tests and interview-administered questionnaires about demographic characteristics, clinical symptoms and epidemiological exposure factors. Univariate and multivariable logistic regression analysis were used to examine the diagnostic value of these parameters. Results SFTSV infection occurred only from April to October annually and usually in hilly areas of specific regions. Three prediction models of SFTSV infection were constructed. Model 3 with clinical and epidemiological parameters combined the benefits of both Model 1and Model 2, which was optimal and had an overall accuracy of 80.2%. Independent indicators for clinical diagnosis of SFTSV infection in Model 3 were as follows: lymphadenopathy (P = 0.01), leucopenia (P<0.01), age >50 years (P = 0.01), tick bites (P<0.01), raising domestic animals in the residential areas (P<0.01) and farming (P = 0.03). Conclusions Our results show that using a combination of clinical and epidemiological parameters may be a feasible strategy to provide preliminary fast diagnosis as differentiating SFTSV infection from SFTS-like diseases, thus reducing the risk of misdiagnosis.

Introduction onset of fever (!38˚C) and other symptoms(e.g., gastrointestinal symptoms, bleeding), epidemiological exposure factors (e.g., being exposed to ticks 2 weeks before illness onset) and laboratory data showing thrombocytopenia or leukopenia. Laboratory-confirmed cases are defined as suspected cases who meet one or more of the following criteria: (1) detection of SFTSV RNA, (2) seroconversion or !4-fold increase in IgG against SFTSV between two serum samples collected at least 2 weeks apart, and (3) isolation of SFTSV. Others of the suspected cases are defined as misdiagnosis cases of SFTS.

Sample collection and epidemiological investigation
When suspected cases were detected by medical institutions, the serum samples were collected and sent to Jiangsu Provincial Center for Diseases Control and Prevention (CDC) for testing within 24 hours. At the same time, an epidemiological investigation was initiated. A standardized questionnaire including demographic characteristics, clinical symptoms and epidemiological exposure factors (living and working environment, and exposure history within the previous 2 weeks prior to illness onset) was used (S1 Appendix). Professionals from local CDC interviewed suspected cases, their family members and the clinicians in charge. Additionally, all cases' medical records were reviewed one week after their discharges or deaths. To further analyze the indicators for clinical diagnosis of SFTSV infection, the suspected cases were categorized into laboratory-confirmed cases and misdiagnosis cases according to results of laboratory testing.

Ethical consideration
The study was approved by the Ethics Committee of the Jiangsu Provincial CDC. All aspects of the study complied with the Declaration of Helsinki. The Ethics Committee specifically approved verbal informed consent as an alternate to written informed consent, because data were planned to be analyzed anonymously. Moreover, a small amount of blood samples were acquired from human clinical samples, which were collected by following medical institutions' approved procedures and only used for laboratory diagnosis of SFTSV infection.

Statistical analysis
Data were double entered into an EpiData 3.1 (the EpiData Association, Denmark, Europe), and the database consistency check was performed. SPSS version 18.0 (Statistical Product and Service Solutions, Chicago, IL, USA) was used for all statistical analyses. Statistical tests were two-sided with a significance level of 0.05. Chi-square or Fisher's exact test was used to compare dichotomous variables. Indicators for clinical diagnosis of SFTSV infection were identified using univariate and multivariable non-conditional logistic regression models. The multivariate non-conditional logistic regression models were constructed and included univariate analysis of variables with significance level of 0.05. The forward stepwise elimination procedure was applied to exclude the variables with P>0.10 in the multivariable model. Consequently, we presented odds ratios (ORs) with 95% confidence intervals (CIs) for various indicators.

Results
The spatial and temporal distribution A total of 231 suspected cases of SFTS were reported by medical institutions from 2011 to 2013 in Jiangsu Province, China. Over 85% suspected cases were reported by clinicians from tertiary hospitals. The average period from illness onset to confirmation of SFTSV infection was 8.5 days, ranging from 1 to 50 days. Of 231 suspected cases, 89 (38.5%) were positive for SFTSV, and 19 died. The case fatality rate was 21.3% (19/89). Fig 1 showed that all laboratory-confirmed cases were reported from April to October annually. The incidence peak occurred around July. No cases were detected from November to next March. On the contrary, misdiagnosis cases occurred year-round although a relative peak occurred from May to July.
For the geographical distribution, laboratory-confirmed cases were concentrated in hilly areas around the boundary of Anhui Province and Jiangsu Province, while misdiagnosis cases were scattered (Fig 2).

Demographic and clinical characteristics of SFTSV infection
All of suspected cases ranged from 16 years old to 83 years old and the median age of them were 56.5 years old. 55.8% of them were farmers. In order to facilitate the analysis, we divided suspected cases into two categories by age: >50 years and 50 years, and two categories by occupation: farmers or non-farmers. For laboratory-confirmed cases and misdiagnosis cases, respectively, 75.3% and 52.8% were older than 50 years; 70.8% and 46.5% were farmers; 56.2% and 63.4% were male. Univariate analysis indicated that, there were significant differences in age and occupation between two groups (Table 1).
By univariate analysis, laboratory-confirmed SFTS cases presented clinical manifestations such as diarrhea (P = 0.02), lymphadenopathy (P<0.01) and leukopenia (P<0.01). In contrast, misdiagnosis cases presented "chilly" more frequently (P = 0.03). In addition, the highest body temperatures (39.1±0.7˚C) during course in the laboratory-confirmed cases were lower than that (39.3±0.8˚C) in misdiagnosis cases significantly (P = 0.01). There were no significant differences in other clinical symptoms between laboratory-confirmed cases and misdiagnosis cases (Table 1).

Epidemiological exposure factors of SFTSV infection
For diagnosis of SFTS as a tick-borne zoonosis, the exposure history within two weeks prior illness onset is certainly important. By univariate analysis, exposure to ticks (P<0.01), tick bites (P<0.01), presence of ticks in the residential areas or workplace (P<0.01), tea-picking (P<0.01), grazing (P<0.01), farming (P<0.01), raising domestic animals in the residential areas (P<0.01), mowing (P<0.01), presence of rats (P<0.01) and contact with wild animals (P = 0.04) were correlated to laboratory diagnosis with SFTSV infection ( Table 2).

Prediction model of SFTS etiology diagnosis
Firstly, with demographic and clinical parameters of P 0.05 by univariate analysis, a multivariable logistic-regression analysis showed that age >50 years (P = 0.01), being a farmer Then, only with risk factors of P 0.05 by univariate analysis, a multivariable logistic-regression analysis showed that farming (P<0.01), raising domestic animals in the residential areas (P<0.01), presence of ticks in the residential areas or workplace (P = 0.06) and tick bites (P = 0.04) were independently associated with SFTSV infection (Table 2). This model (Model 2) showed a sensitivity of 76.3% and specificity of 78.1%, with an overall accuracy of 77.4%. The probability predicted for SFTS etiology was obtained from the following Eq (2) Finally, with demographic parameters, clinical parameters and epidemiological exposure factors, a multivariable logistic-regression analysis showed that lymphadenopathy (P = 0.01), leucopenia (P<0.01), age >50 years (P = 0.01), tick bites (P<0.01), raising domestic animals in the residential areas (P<0.01) and farming (P = 0.03) were independently associated with SFTSV infection (

Discussion
SFTS is an emerging tick-borne infectious disease, which has caused comprehensive public health concerns due to its expanding epidemic areas, the capability of human-to-human transmission and a high case fatality rate [4]. The clinical symptoms of SFTS are nonspecific and consistent with many infectious pathogens including bacteria and viruses. It should be mentioned that Anaplasma phagocytophilum was considered to be the causative agent of SFTS in China before the discovery of SFTSV in 2010, because patients with human granulocytic anaplasmos is also presented fever, thrombocytopenia and leucopenia [1,15]. In fact, in most cases viral laboratory confirmation might not always be available due to unaffordability of test kits, lack of instruments and shortage of trained laboratory technicians in rural areas. At present, SFTS misdiagnosis is very common, posing a challenge both to clinicians and public health officials. On the one hand, our findings also disclosed the problem of SFTS misdiagnosis, that 38.5% of 231 suspected SFTS cases reported by medical institutions in Jiangsu Province from 2011 to 2013, were positive for SFTSV. On the other hand, the median time from illness onset to confirmation was 8.5 days in our study, which suggested that clinicians fail to make the timely diagnosis at present. Therefore, we summarized the first attempt to identify clinical and epidemiological parameters as useful indicators for clinical diagnosis of SFTSV infection, which could provide preliminary fast diagnosis as differentiating SFTSV infection from SFTS-like diseases, thus reduce the risk of misdiagnosis. As a natural foci disease, SFTSV is maintained in a special ecosystem that is best suited to a natural transmission cycle involving tick vectors and animal reservoir hosts. Natural foci can change under climate, geographical and human influence, but they are stable with a long-time activity [21]. From the results, we can conclude that hilly areas around the boundary of Anhui and Jiangsu Province were the potential natural foci of SFTSV. For the temporal distribution, cases with SFTSV infection occurred only from late April to October, which were in agreement with corresponding data from other studies in China, consistent with the life cycle of ticks. Therefore, even if one patient with SFTS-like syndromes was reported from November to next March, we must think carefully diagnosis of SFTS without laboratory testing.
In our study, we constructed three prediction models. Model 1 was based on the demographic and clinical parameters had a higher specificity than Model 2, while Model 2 was only based on the epidemiological exposure factors had a higher sensitivity than Model 1. However, Model 3 which combined the benefits of both Model 1 and Model 2 was optimal and had a highest overall accuracy. At present, the diagnosis of SFTSV infection is often directly derived from clinical characteristics of patients, then pathogenic detection. However, our findings suggest that the epidemiological information of patients plays an important role in SFTS diagnosis, and the clinicians shouldn't over-depend on pathogenic detection, neglecting the inquiry of epidemiological information after learning about clinical syndromes of patients. Our study indicated that persons older than 50 years were readily infected with SFTSV, and was consistent with previous conclusion [22]. The high proportion of elderly patients may be due to physiological factors related to aging, such as decreased immune function. Previous study showed that prevalence of antibodies against SFTSV increased with age [22,23]. However, this may also be related to demographic features of residents in endemic rural areas. Generally, young adults from wooded or hilly areas go to cities to earn better money and the actual residents are just elderly people and children.
Interestingly, cases with SFTSV infection presented some special clinical features. For instance, laboratory-confirmed cases developed more likely gastrointestinal symptoms. In particular, the proportion of diarrhea among laboratory-confirmed cases was significantly higher than that of misdiagnosis cases. But hemorrhagic tendency showed no difference between the two groups. In addition, confirmed cases were more likely to present lymphadenopathy and leucopenia, which became major indexes of the SFTS etiologic diagnosis and retained in the Model 3. Reproduction of virus happened first in neighboring lymph nodes after invasion into the host, which might result in lymphadenopathy. The animal model demonstrated that SFTSV then went into the spleen and liver in which white blood cells were devoured by macrophages [24]. But our study showed that the count of platelet had no significant difference between laboratory-confirmed and misdiagnosis cases although platelets were also devoured in the process.
In our study, epidemiological exposure factors such as tick bites, farming outside two weeks before illness onset and raising domestic animals in the residential areas were helpful to distinguish SFTSV infection from SFTS-like diseases, which were also proved to be risk factors for SFTS in other researches. Ticks, particularly Haemaphysalis longicornis, were thought to be the primary vector for SFTSV. First, a high proportion of patients diagnosed with SFTS reported a history of tick bite [1,[25][26][27][28]. Second, spatial and temporal distributions of human cases were consistent with the fluctuation of certain species of ticks in a given endemic area [1,25,26,29,30].Third, several research groups had detected SFTSV-specific nucleotide sequences, or isolated virus from ticks collected from animals, or the environment, which had high homology (93-100%) with SFTSV isolated from patients [1,2,16,25,29,31,32]. Finally, ticks could acquire the virus by feeding on blood of the SFTSV-infected host animal and transstadially and transovarially transmit it to other developmental stages of ticks [2,3]. Farming activities became one index of the SFTS etiologic diagnosis. The reason for this maybe that SFTS is endemic in rural areas and farming activities around weeds and shrubs increase the exposure opportunities for ticks. It is notable that 82% of laboratory-confirmed cases in our study bred domestic animals before illness onset. We postulated that domestic animals living in close contact with humans were reservoirs for SFTSV, whereas ticks as the vector could transmit the virus from animal hosts to human. Some studies concluded that domestic animals, especially free-range goats, cattle and dogs, had a high seroprevalence of antibodies against SFTVS and confirmed this hypothesis indirectly [23,33,34].
One strength of our study was that the median time between suspected diagnosis and information collection for demographic, clinical and epidemiological parameters was less than one hour. Therefore, suggestive diagnosis offered by these parameters would be available much earlier than laboratory tests. Another strength was that using a combination of these parameters may be an only feasible strategy to improve SFTS diagnosis in resource-poor setting, such as rural areas in China, where viral laboratory confirmation might not be available due to no highly specialized laboratories, no trained laboratory technicians or expensive PCR machines.
There were three major flaws of this study. Not all clinical and epidemiological parameters were considered into our study due to difficulty in data collection. For instance, serum abnormal findings such as elevated liver-associated transferase and prolonged activated partial thromboplastin time could be found in majority cases with SFTSV infection. We don't know whether these abnormities could be used to distinguish SFTSV infection from other pathogen infection. In the future, other clinical and epidemiological parameters related to SFTSV infection will be furtherly identified. Secondly, although study design was perspective, clinical and epidemiological parameters were obtained mainly by interviewing suspected cases, their family members and the clinicians in charge. So there might be memory bias. Thirdly, univariate analysis showed the highest body temperatures (39.1±0.7˚C) in the laboratory-confirmed cases were lower than that (39.3±0.8˚C) in misdiagnosis cases (P = 0.01). From the biological significance, the difference was not significant. Maybe it's related the small sample size.
Our findings suggest that specific clinical and epidemiological parameters may be useful indicators to assist clinicians to make preliminary fast diagnosis as differentiate SFTSV infection from SFTS-like diseases, thus reduce the risk of misdiagnosis.
Supporting information S1 Appendix. The questionnaire of suspected cases with SFTSV infection. (DOC) S1