Received: June 13, 2018; Accepted: July 6, 2018; Published: July 9, 2018
Correspondence
to: Willy Chou, Ncphrology Department, Chi-Mei Medical Center, 901 Chung Hwa Road, Yung Kung Dist., Tainan 710, Taiwan; Email: hsienyiwang@gmail.com
1 Medical Research Department, Chi-Mei Medical Center, Tainan, Taiwan
2 Department of Paediatrics, Chi-Mei medical center, Tainan, Taiwan
3 National Taiwan University School of Medicine, Taiwan
4 Department of Sports Management, College of Leisure and Recreation Management, Chia Nan University of Pharmacy and Science, Tainan, Taiwan
5 Ncphrology Department, Chi-Mei Medical Center, Tainan, Taiwan
Citation: Chien TW, Chow JC, Chang Y, et al. Detecting Dengue Fever in Children: Using Sequencing Symptom Patterns for An Online Assessment Approach. Adv Health Behav, 2018, 1(1): 12-16
Copyright: © 2018 Willy Chou, et al. This is an open access article distributed under the terms of the Creative Commons Attribution License which permits unrestricted use, dis- tribution, and reproduction in any medium, provided the original author and source are credited.
1. Introduction
Dengue fever (DF) is one of the most common arthropod-borne viral diseases worldwide,[ 1] especially in South East Asia, Africa, the Western Pacific, and the Americas.[ 2, 3 ]
There is, however, no accurate and speedy diagnostic screening test for DF at an early stage because its signs and symptomse.g., fever, headache, and myalgiaare similar to those of other illnesses.[4–6] Some studies[ 4, 5 ] that used a univariate analysis report that the presumptive diagnosis of DF is imprecise. Multivariate logistic regressions also do not significantly distinguish patients with Copyright: c 2018 Tsair-Wei Chien, et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. dengue from those with other febrile illnesses.[ 7 ] The multivariate discrimination analyses reported a sensitivity and a specificity 0.76, and an area under the receiver operating characteristic (ROC) curve (AUC) of 0.93, but costly laboratory tests (Dengue Duo IgM & Rapid Strips; Panbio, Queensland, Australia)[ 8–11 ] were needed before DF was serologically confirmed.
DF symptoms are usually assessed using a dichotomous (i.e., absent versus present) evaluation. The dependent variable (DF+ versus DF ) predicted using independent evaluations with a weighted summation score is more accurate than that using simple evaluations with an unweighted summation score. So far, there has been no published study that has reported using the specific sequence of symptoms reported or observed in specific patients suspected of having DF. All published studies to date still report using only a standard group of symptoms with an unweighted summation score that apply to a general group of patients that might have DF.
The non-parametric HT fit statistic has been used in education and psychometrics to identify aberrant test respondents.[ 12, 13 ] It is a transposed formulation of a scalability coefficient for items (e.g., symptoms in this study) and evidently the best among 36 person fit statistics for detecting abnormal behaviors.[ 14 ]
In the present study, we used the HT coefficient combined with weighted and unweighted variables to examine whether these combinations provide a valid and reliable approach for the early detection of DF in children.
2. Materials and methods
2.1 Sample and clinical symptoms
The sample of 177 pediatric patients (≤ 16 years old; DF+:69; DF- :108) was the same as in our previous paper.[8] Guided by the literature,[5–7 ] we collected nineteen DF-related clinical symptoms from the patients medical records to develop the initial set of itemsdesignated as 0=absent or 1=present to screen for DF infection: (i) personal history of DF, (ii) family history of DF, (iii) mosquito bites within the previous 2 weeks, (iv) fever ≥ 39 C, (v) biphasic fever, (vi) rash, (vii) petechiae, (viii) retro-orbital pain, (ix) bone pain (arthralgia), (x) headache, (xi) myalgia, (xii) abdominal pain, (xiii) anorexia, (xiv) occult hematuria, (xv) stool occult blood, (xvi) cough, (xvii) sore throat, (xviii) soft (watery) stool, and (xix) flushed skin. Data from these patients charts were obtained and approved by the Research Ethics Review Board of the Chi-Mei Medical Center.
2.2 The HT fit statistic
HT is defined for the persons of a dichotomous dataset with L items (in columns) and N persons (in rows),[ 12 ] where Xni is the scored (0,1) response of person n to item i, and Pn = Sn/L. Here, Sm is the raw score for person m, and Sn is the raw score for person n.
HT is the sum of the covariances between person n and the other persons divided by the maximum possible sum of those covariances, so that the range of HT is -1 to +1. When the responses by person's are positively correlated with those of all the other persons, then HT (n) will be positive. In contrast, when the responses by person n are negatively correlated with those of all the other persons, then HT (n) will be negative. When person's responses are random, HT (n) will be close to zero[ 11 ]. We hypothesized that DF+ patients have different HT coefficients than do DF patients. All DF+ group members were sequenced to the DF group members to obtain an HT coefficient using equation (1).
2.3 Selecting symptoms and determining predictor variables
All symptoms were examined by the probability of Type I error using the following three steps in Figure 1 to determine predictor variables. First, each symptom was separately examined by the univariate approach using a X2 test and logistic regression, respectively, for identifying a significant association with DF. Second, two models (i.e., the univariate and the multivariate approaches) were investigated for determining valid predictor variables associated with DF when the probability of Type I error is less than 0.05. Third, the predictor variables were used in a weighted combination for discriminating patients suspected with dengue virus infection.
Figure 1.Overall study concept and the flow chart
2.4 Detecting dengue fever: a comparison of three models
The efficacy of three models (A, B, and C) for detecting dengue fever was examined: (i) A comparison was made using univariate logistic regression in Model A to examine effects through the AUC yielded by Unweighted (i.e., summed item) scores, Weighted (i.e., logistic regression) scores, and HT coefficients, respectively, (ii) Multivariate logistic regression with the three aforementioned factors combined was used in Model B, (iii) after selecting the significant variables in Model B, the combined predictive variables were analyzed using multivariate logistic regression in Model C to obtain effective weighted coefficients, and (iv) finally, we wanted to use a single continuous variable yielded by the combined predictive variables in Model C to compare the AUC with the counterparts in Model A and C.
2.5 Statistical tools and data analyses
SPSS 15.0 for Windows (SPSS Inc., Chicago, IL) and MedCalc 9.5.0.0 for Windows (MedCalc Software, Mariakerke, Belgium) were used to calculate (i) the probability of false positives (Type I error) using a X2 test and logistic regression, (ii) Youden J index (the higher, the better), AUC (area under the ROC curve), sensitivity, specificity, and the cutoff point at maximal summations of specificity and sensitivity, (iii) correlation coefficients among variables of unweighted, weighted, and HT scores.
3 Results
Sixty-nine pediatric patients clinically diagnosed with DF and 108 with no evidence of DF infection were included in this study (Table 1). A X2 test and logistic regression analyses showed that only six symptoms (Family History, Fever ≥ 39°C, Skin Rash, Petechiae, Abdominal Pain, and Weakness) were significant for assessing the likelihood of DF (Table 2).
Table 1.Demographic characteristics of the study sample
Table 2.Logistic analysis of symptoms for the patients suspected with dengue virus infection using the univariate approach
P-values were determined by the test and the Wald test of Logistic regression
Table 3.Comparisons of AUC for the study models
a: coefficient of Logistic regression
b: Youden J index
c: item-score summation method
d: multiplying item-score with the weighted regression coefficient
e: the Ht coefficient
f: using the two combined variables to predict patients DF
*: p < 0.05
Comparisons of the AUCs for the three study models (A, B, and C) showed that the weighted variable (derived by the Logistic regression) and the HT coefficient can be jointly used for predicting DF risk using equation (2):
The risk probability can be computed using the transformed equation (3):
A cutoff point of 1.03 (P = 0.26) was determined using the combined predictive variables in Model C: sensitivity = 0.91, specificity = 0.76, and AUC = 0.88 (Figure 2 and Table 3). Predictive power was better: specificity was 5.56% (i.e., 75.93-70.37 shown in Table 3) higher than when using traditional logistic regression; however, the AUC was slightly lower (0.72) than when using the unweighted (0.84) and the weighted (0.87) variables ((Table 2)). The HT coefficients related to the weighted and unweighted scores were 0.26 and 0.22, respectively. The weighted score has a higher correlation coefficient than does the unweighted score to the HT coefficients.
Figure 2.Four models plotted by ROC curves
Figure 3.Figure 3 Snapshots on a smart phone responding questions (top) and the result (bottom) for assessing the patient DF
A snapshot on a smart phone responding to questions (Figure 3, top) was generated and the results for assessing whether the patient has DF (Figure 3, bottom) were determined, which indicated that patients suspected of having DF can directly scan the QR-code to obtain their DF logit scores (or the risk probability) and examine whether these 6 symptoms are useful for predicting a high DF risk (>1.03 logits or P> 0:26 = exp( -1.03 logits)/(1 + exp( -1.03 logits)).
4 Discussion
We found that using the HT coefficient yielded predictions that were 5.56% more specific (i.e., 75.93-70.37 shown in Table 3) than those of traditional logistic regression. The HT index is promising when the patient sequence symptom pattern is compared with the DF+ group to detect dengue fever in children. It can be combined with the weighted summation score to jointly predict the DF risk and then to report that risk on smartphones.
The HT coefficient has been used in education and psychometrics to identify aberrant test respondents.[ 12, 13 ] Although some have used item response theory (IRT) fit statistics (e.g., outfit mean square error > 2.0) to select abnormal responses that indicate cheating, careless responding, lucky guessing, creative responding, or random responding,[ 15 ] our literature review revealed no published papers that reported using the HT coefficient in medical settings, especially for detecting individual aberrant response patterns different from the study reference sample, or, like the current study, identifying the DF risk by comparing their sequence symptom pattern to that of the DF+ group.
A diagnosis of DF is usually confirmed by three steps: (i) observing DF-related symptoms, (ii) testing laboratory data such as white blood cells (WBCs) and platelets (PLTs), and (iii) serologically verifying DF using dengue IgM and IgG antibodies, polymerase chain reaction (PCR) analysis, and virus isolation tests. The latter two are relatively expensive. It is needed to develop a self-assessment approach (e.g., scanning QR-code, responding questions, and obtaining the DF risk on his/her smart phone) (1) helping patients for consultation at an earlier stage, (2) prompting doctors sampling patient laboratory data when he/her DF risk reaches a cutpoint of P=0.26=exp(-1.03 logits)/(1+exp(-1.03 logits)).
We found that the weighted score was a better predictor than was the unweighted score (see Model A and Model B in Table 3). However, we still see so many scales in medical setting using unweighted summation scores to determine the presence or absence of disease. Along with the smartphones popularly used in the technical age, the way of obtaining the DF risk on smartphones using the combined HT coefficient and weighted scores is available and worth recommending to healthcare providers to use for detecting the risk for DF.
This study has some limitations. First, the DF cutpoint based on the symptoms of our study sample might be biased toward that population. Moreover, we did not remove abnormal data when the HT coefficient was less than the critical value of 0.22, which best identifies aberrantly responding examinees.[ 14 ] Second, although the sample size was small, using the Rasch HT coefficient combined with the AUC yielded highly accurate discriminatory screening. This finding, however, requires confirmation in prospective studies of other regions with a substantial incidence of DF.
5 Conclusions
Analyzing six simple symptoms using logistic regression is useful and valid for the early detection of DF risk in children. Combining the Rasch HT coefficient with the weighted score yields a prediction that is 5.56% more specific than does traditional logistic regression. A selfassessment app using patient smartphones is available to help people suspected of having DF, and it might eliminate the need for costly and time-consuming laboratory tests.
6 Competing interests
The authors declare that they have no competing interests.
7 Authors contributions
T.-W.C. and S-C.K. conceived and designed the study, performed the statistical analyses and were in charge of recruiting study participants. W.-S.L. and T.-W.C. helped design the study, collected information and interpreted data. All authors read and approved the final article. This research was supported by grant ChiMei Foundation Hospital research CMFCR10593 from the Chi-Mei Medical Center. The authors have no other funding or conflicts of interest to disclose.
References
[1] World Health Organization. Dengue and Dengue Haemorrhagic
Fever. Fact sheet No. 117. Geneva: World Health Organization,
2002.
[2] Henchal EA, Putnak JB. The dengue viruses. Clin Microbiol
Rev, 1990, 3(4):376-396.
https://doi.org/10.1128/CMR.3.4.376
[3] Gubler DJ. Dengue and dengue hemorrhagic fever. Clin Microbiol
Rev, 1998, 11(3):480-496.
https://doi.org/10.1016/S1045-1870(97)80003-9
[4] Phuong HL, de Vries PJ, Nga TT, et al. Dengue as a cause
of acute undifferentiated fever in Vietnam. BMC Infect Dis,
2006, 6:123.
https://doi.org/10.1186/1471-2334-6-123
[5] Nunes-Arajo FR, Ferreira MS, Nishioka SD. Dengue fever
in Brazilian adults and children: assessment of clinical findings
and their validity for diagnosis. Ann Trop Med Parasitol,
2003, 97(4):415-419.
https://doi.org/10.1179/000349803235002263
[6] Hammond SN, Balmaseda A, Prez L, et al. Differences in
dengue severity in infants, children, and adults in a 3-year
hospital-based study in Nicaragua. Am J Trop Med Hyg,
2005, 73(6):1063-1070.
https://doi.org/10.4269/ajtmh.2005.73.1063
[7] Potts JA, Rothman AI. Clinical and laboratory features that
distinguish dengue from other febrile illnesses in endemic
populations. Trop Med Int Health, 2008, 13(11):1328-1340.
https://doi.org/10.1111/j.1365-3156.2008.02151.x
[8] Lai WP, Chien TW, Lin HJ, Su SB, Chang CH. A screening
tool for dengue fever in children. Pediatr Infect Dis J, 2013,
32(4):320-324.
https://doi.org/10.1097/INF.0b013e31827e111e
[9] Lai WP, Chien TW, Lin HJ, Kan WC, Su SB, Chou MT.
An approach for early and appropriate prediction of dengue
fever using white blood cells and platelets. HealthMed,
2012, 6(7):806-812.
[10] Kittigul L, Suankeow K. Use of a rapid immunochromatographic
test for early diagnosis of dengue virus infection.
Eur J Clin Microbiol Infect Dis, 2002, 21(3):224-226.
https://doi.org/10.1007/s10096-001-0691-z
[11] Vaughn DW, Nisalak A, Kalayanarooj S, et al. Evaluation
of a rapid immunochromatographic test for diagnosis of
dengue virus infection. J Clin Microbiol, 1998, 36(1):234-
238.
[12] Sijtsma K. A coefficient of deviant response patterns. Kwantitatieve
Methoden, 1986, 7(22):131-145.
[13] Linacre JM. A comment on the HT person fit statistic. Rasch
Meas Trans, 2012, 26(1):1358.
[14] Karabatsos G. Comparing the aberrant response detection
performance of thirty-six person-fit statistics. Appl Meas
Educ, 2003, 16(4):277-298.
https://doi.org/10.1207/S15324818AME1604_2
[15] Linacre JM. Optimizing rating scale category effectiveness.
J Appl Meas, 2002, 3(1):85-106.