Admission Clinicopathological Factors Associated with Prolonged Hospital Stay Among Hospitalized Patients with Dengue Viral Infections

Introduction

Dengue fever is the most important mosquito-borne viral disease in tropical and subtropical countries and is a significant cause of high economic and disease burden (Shepard et al. 2013). Dengue is hyperendemic in Malaysia with all four dengue virus (DENV) serotypes cocirculating, resulting in illness ranging from mild fever to potentially fatal dengue shock syndrome. Dengue cases continue to escalate in Malaysia, with an estimated 143,891 dengue episodes occurring annually and of which 62,256 (43.3%) cases require hospitalization (Shepard et al. 2013). Hospitalization drains resources and places unnecessary financial burden on patients as well as on the national economy, with an average unit cost per-dengue-episode for inpatients at tertiary hospitals estimated at USD$863.21 (Shepard et al. 2013). Despite increasing dengue cases and hospitalizations, studies focusing on factors linked to prolonged hospitalization (PH) are scarce (Khalil et al. 2014, Mallhi et al. 2017), with none reporting the association between DENV serotypes and PH.

Our study aims to identify clinical–laboratory factors present upon admission associated with PH in patients with DENV infections. This information may potentially improve clinical management and support better use of resources in hospitals, especially during outbreaks, which often threaten to overwhelm health care services.

Materials and Methods

This retrospective study was conducted from January to April 2014 at Hospital Sultanah Aminah in Johor Bahru, the primary referral hospital in southern Malaysia. Hospitalized patients (aged >12 years) with positive dengue nonstructural protein 1 (NS1) antigen (Standard Diagnostic, Inc., Seoul, Korea) and satisfying the World Health Organization (2009) clinical diagnostic criteria for dengue were identified. Viral RNA was extracted using QIAamp viral RNA Mini Kit (Qiagen) from serum of these patients. Complementary DNA (cDNA) from viral RNA was synthesized by reverse transcription using AccessQuick RT-PCR System kit (Promega). This cDNA was used for DENV serotyping by multiplex RT-PCR (Yong et al. 2007). To confirm the specificity of the primers to DENV and to rule out cross-reactivity with other flaviviruses, the primers were blasted through the National Center for Biotechnology Information database.

All serum samples were also tested using Panbio^® Dengue IgG Capture ELISA (enzyme-linked immunosorbent assay; Alere, Brisbane, Australia), which provides serological evidence of secondary DENV infections, as it incorporates a cutoff value of >22 Panbio units. This is equivalent to hemagglutination inhibition level of ≥1:2560, which is indicative of secondary DENV infections, thereby distinguishing it from past infections.

Relevant admission clinico-laboratory findings were retrieved from medical records. PH was defined as hospitalization ≥4 days (Khalil et al. 2014, Mallhi et al. 2017). Data were analyzed using IBM^® SPSS^® 25.0. Multivariate logistic regression (MLR) was performed for variables with p value <0.2 to examine the associations between PH and admission parameters. Ethical approval was obtained from MREC, Ministry of Health Malaysia (NMRR-14-617-21061).

Results and Discussion

Of the 253 NS1 antigen and PCR-positive DENV hospitalized patients, 95 (37.5%) had PH (≥4 days). The mean length of hospital stay was 3.43 ± 2.085 days (median = 3 days, interquartile range = 7 days).

Older age, pregnancy, diabetes mellitus, DENV-2, shorter preadmission fever duration as well as high creatinine and leukocyte counts were significantly (p < 0.05) associated with PH upon univariate analysis (UVA) (Table 1). Subsequent MLR revealed that diabetes (adjusted odds ratio [AOR] = 6.261, 95% confidence interval [CI] = 2.130–18.406, p = 0.001), DENV-2 (AOR = 2.581, 95% CI = 1.179–5.650, p = 0.018), and duration of fever ≤4 days (AOR = 2.423, 95% CI = 0.872–6.734, p = 0.09) with shorter preadmission fever duration (AOR = 0.679, 95% CI = 0.481–0.957, p = 0.027) were positively associated with PH.

Patients with diabetes had 6.26 times increased risk of PH. The mean duration of hospitalization was significantly longer (p = 0.002) among patients with diabetes (4.31 ± 1.96 vs. 3.33 ± 2.08 days), possibly due to illness severity, as diabetes is a recognized trigger for dengue hemorrhagic fever and severe organ involvement (Pang et al. 2017). It has been postulated that diabetes results in endothelial and immune dysfunction (Pang et al. 2017). Superimposed DENV infection in patients with diabetes may complicate management, necessitating extended observation in the hospital. Although older age did not independently predict PH in multivariate analysis (MVA), it was significant in UVA (p = 0.004) with a similar trend reported by Khalil et al. (2014).

Shorter preadmission fever (≤4 days) independently predicted PH, a result that was also demonstrated in a Thai study (Wongchidwan et al. 2018). The median time from fever onset to hospitalization was 3 (1–13) and 5 (1–10) days for patients with and without PH, respectively. To further explore this association, we compared the clinico-laboratory parameters for patients with preadmission fever ≤4 days (early) versus >4 days (late). Patients who presented earlier had higher platelet counts (121.4 vs. 105.6 × 10⁹/L), higher albumin levels (40.1 vs. 39.6 g/L), and lower ALT (87.5 vs. 94.4 IU/L) and AST levels (132.6 vs. 150 IU/L) in contrast to patients who presented late, although statistical significance was only attained for hepatic transaminases. These findings are consistent with comparisons made between patients in the febrile phase. Given that dengue is a dynamic disease with febrile, plasma leakage, and convalescence phases (World Health Organization 2009), it is not unexpected for patients presenting in earlier phases to be hospitalized longer for closer monitoring during the critical plasma leakage phase. Thus, the length of hospital stay may vary depending on the phase of dengue during admission and varying discharge criteria of hospitals (Wongchidwan et al. 2018). Except for vomiting, all admission symptoms (headache, abdominal pain, myalgia, and neurological and hemorrhagic symptoms) were present at higher frequency among patients admitted within 4 days of fever onset, suggesting that these may be the motivators for patients seeking early medical care. However, these differences were significantly higher only for headache (p = 0.014).

DENV-1 (64.8%) and DENV-2 (18.2%) were the predominant serotypes among our patient cohort (Table 1). This was followed by DENV coinfection (12.8%), DENV-3 (2.8%), and DENV-4 (0.4%). However, only DENV-2 was significantly associated with PH; this to the best of our knowledge may be the first study showing such an association. This knowledge becomes especially important in light of the more widespread availability of reliable diagnostic tests such as CDC DENV-1-4 RT-PCR assays for detection and serotyping of DENV upon admission (Santiago et al. 2013). DENV-2 has been associated with hemorrhagic manifestations, atypical symptoms, and unusual complications (Thomas et al. 2008). Moreover, DENV-2 replicates more efficiently, translating to higher viral loads and disease severity (Vaughn et al. 2000, Tricou et al. 2011, Thomas et al. 2008), both of which may contribute toward PH. Interestingly, higher frequency of DENV-2 was noted in patients with preadmission fever ≤4 days (22% vs. 13.6%, p = 0.08). Conversely, patients with preadmission fever >4 days had higher frequency of DENV-1 (73% vs. 59%, p = 0.027). Although secondary DENV infection has been associated with severe dengue (Thomas et al. 2008), concurring with the study of Mallhi et al. (2017), no link with PH was noted. An interplay between prior immunity, infecting serotype/genotype, and viral load may contribute toward differences in clinical manifestations and disease severity (Tricou et al. 2011).

Pathological hallmarks of DENV infections are the presence of thrombocytopenia and plasma leakage. Concurring with the findings from previous studies, neither severe thrombocytopenia, elevated hepatic transaminases, nor elevated hematocrit on admission predicted PH (Khalil et al. 2014, Mallhi et al. 2017). Although elevated creatinine was associated with PH in UVA, it was not included in MVA as the missing values were >5%. Nevertheless, high creatinine level and acute kidney injury were associated with PH in other studies (Khalil et al. 2014).

Based on the 2009 WHO dengue case classification, the final diagnosis of our patients was mainly dengue with warning signs (n = 188, 74.3%) followed by dengue without warning signs (n = 49, 19.4%) and severe dengue (n = 16, 6.3%). PH was significantly associated with a diagnosis of severe dengue (p < 0.001) but not associated with the diagnosis of either dengue without warning signs (p = 0.148) or dengue with warning signs (p = 0.441).

The limitations of this study included the bias inherent to a retrospective study. Moreover, our results should not necessarily be generalized to the other settings since this is a single-center study.

Conclusions

Our findings reveal that more than one-third of patients admitted for DENV infection experienced PH. Diabetes mellitus, DENV-2, and shorter duration of preadmission fever were independently associated with PH. These findings may guide risk stratification and optimize management strategies (i.e., extra vigilance toward patients with comorbidities) in those at risk of PH. This may make a significant difference in areas endemic for dengue, especially during periods of outbreak, which often threaten to overwhelm health care facilities. A larger scale prospective study will be useful to strengthen these findings.