Bias Correction in Estimating Proportions by Imperfect Pooled Testing

In the estimation of proportions by pooled testing, the MLE is biased. Hepworth and Biggerstaff (JABES, 22:602-614, 2017) proposed an estimator based on the bias correction method of Firth (Biometrika 80:27-38, 1993) and showed that it is almost unbiased across a range of pooled testing problems involving no misclassification. We now extend their work to allow for imperfect testing. We derive the estimator, provide a Newton-Raphson iterative formula for its computation and test it in situations involving equal or unequal pool sizes, drawing on problems encountered in plant disease assessment and prevalence estimation of mosquito-borne viruses. Our estimator is highly effective at reducing the bias for prevalences consistent with the pooled testing procedure employed.

Reproductive incompatibility between Amblyomma maculatum (Acari: Ixodidae) group ticks from two disjunct geographical regions within the USA

The Amblyomma maculatum Koch group of ixodid ticks consists of three species: A. maculatum, A. triste, and A. tigrinum. However, since Koch described this group in 1844, the systematics of its members has been the subject of ongoing debate. This is especially true of A. maculatum and A. triste; recent molecular analyses reveal insufficient genetic divergence to separate these as distinct species. Further confounding this issue is the discovery in 2014 of A. maculatum group ticks in southern Arizona (AZ), USA, that share morphological characteristics with both A. triste and A. maculatum. To biologically evaluate the identity of A. maculatum group ticks from southern Arizona, we analyzed the reproductive compatibility between specimens of A. maculatum group ticks collected from Georgia (GA), USA, and southern Arizona. Female ticks from both Arizona and Georgia were mated with males from both the Georgia and Arizona Amblyomma populations, creating two homologous and two heterologous F1 cohorts of ticks: GA ♀/GA ♂, AZ ♀/AZ ♂, GA ♀/AZ ♂, and AZ ♀/GA ♂. Each cohort was maintained separately into the F2 generation with F1 females mating only with F1 males from their same cohort. Survival and fecundity parameters were measured for all developmental stages. The observed survival parameters for heterologous cohorts were comparable to those of the homologous cohorts through the F1 generation. However, the F1 heterologous females produced F2 egg clutches that did not hatch, thus indicating that the Arizona and Georgia populations of A. maculatum group ticks tested here represent different biological species.

Evolving viral and serological stages of Zika virus RNA-positive blood donors and estimation of incidence of infection during the 2016 Puerto Rican Zika epidemic: an observational cohort study

BACKGROUND

Puerto Rico began screening blood donations for Zika virus RNA with nucleic acid amplification tests (NAATs) on April 3, 2016, because of an emerging Zika virus outbreak. We followed up positive donors to assess the dynamics of viral and serological markers during the early stages of Zika virus infection and update the estimate of infection incidence in the Puerto Rican population during the outbreak.

METHODS

Blood donations from volunteer donors in Puerto Rico were screened for the presence of Zika virus RNA using the cobas Zika NAAT. Positive donations were further tested to confirm infection, estimate viral load, and identify Zika virus-specific IgM antibodies. Individuals with positive blood donations were invited to attend follow-up visits. Donations with confirmed infection (defined as detection of Zika virus RNA or IgM on additional testing of index or follow-up samples) were assessed for stage of infection according to Zika virus RNA detectability in simulated minipools, viral load, and Zika virus IgM status. A three-step process was used to estimate the mean duration of NAAT reactivity of Zika virus in human plasma from individuals identified pre-seroconversion with at least one follow up visit and to update the 2016 incidence estimate of Zika virus infection.

FINDINGS

Between April 3 and Dec 31, 2016, 53 112 blood donations were screened for Zika virus, of which 351 tested positive, 339 had confirmed infections, and 319 could be staged. Compared with IgM-positive index donations (n=110), IgM-negative index donations (n=209) had higher mean viral loads (1·1 × 10⁶vs 8·3 × 10⁴ international units per mL) and were more likely to be detected in simulated minipools (93% [n=194] vs 26% [n=29]). The proportions of donations with confirmed infections that had viral RNA detected only in individual-donation NAATs (ie, not in simulated minipools) and were IgM positive increased as the epidemic evolved. The estimated mean duration of NAAT detectability in the 140 donors included in the follow-up study was 11·70 days (95% CI 10·06-14·36). Applying this detection period to the observed proportion of donations that were confirmed NAAT positive yielded a Zika virus seasonal incidence estimate of 21·1% (95% CI 18·1-24·1); 768 101 infections in a population of 3 638 773 in 2016.

INTERPRETATION

Characterisation of early Zika virus infection has implications for blood safety because infectivity of blood donations and utility of screening methods likely correlate with viral load and serological stage of infection. Our findings also have implications for diagnostic testing, public health surveillance, and epidemiology, and we estimate that around 21% of the Puerto Rican population was infected during the 2016 outbreak.

FUNDING

Biomedical Advanced Research and Development Authority, National Heart, Lung, and Blood Institute.

Duration of seropositivity following yellow fever vaccination in U.S. military service members

BACKGROUND

The United States military regularly deploys thousands of service members throughout areas of South America and Africa that are endemic for yellow fever (YF) virus. To determine if booster doses might be needed for service members who are repetitively or continually deployed to YF endemic areas, we evaluated seropositivity among US military personnel receiving a single dose of YF vaccine based on time post-vaccination.

METHODS

Serum antibodies were measured using a plaque reduction neutralization test with 50% cutoff in 682 military personnel at 5-39 years post-vaccination. We determined noninferiority of immune response by comparing the proportion seropositive among those vaccinated 10-14 years previously with those vaccinated 5-9 years previously. Noninferiority was supported if the lower-bound of the 2-tailed 95% CI for p_10-14years - p_5-9years was ≥-0.10. Additionally, the geometric mean antibody titer (GMT) at various timepoints following vaccination were compared to the GMT at 5-9 years.

RESULTS

The proportion of military service members with detectable neutralizing antibodies 10-14 years after a single dose of YF vaccine (95.8%, 95% CI 91.2-98.1%) was non-inferior to the proportion 5-9 years after vaccination (97.8%, 95% CI 93.7-99.3%). Additionally, GMT among vaccine recipients at 10-14 years post vaccination (99, 95% CI 82-121) was non-inferior to GMT in YF vaccine recipients at 5-9 years post vaccination (115, 95% CI 96-139). The proportion of vaccinees with neutralizing antibodies remained high, and non-inferior, among those vaccinated 15-19 years prior (98.5%, 95%CI 95.5-99.7%). Although the proportion seropositive decreased among vaccinees ≥ 20 years post vaccination, >90% remained seropositive.

CONCLUSIONS

Neutralizing antibodies were present in > 95% of vaccine recipients for at least 19 years after vaccination, suggesting that booster doses every 10 years are not essential for most U.S. military personnel.

Estimation of time-dependent arbovirus infection risk in blood and tissue donations

West Nile virus (WNV) outbreaks raise the concern of WNV infection in donated blood and blood products destined for transfusion. We describe methods we developed to estimate time-dependent risk of WNV infection in donated blood, including improvements not previously detailed. The methods are then extended for use in estimation of the risk of WNV infection in donated cadaveric tissues by introducing stratification and stratum-specific weighting to address novel aspects of this application. Data from the WNV outbreak in Colorado in 2003 are used to estimate risk for donated cardiac tissue.

Reverse Transcription-Polymerase Chain Reaction Testing on Filter Paper-Dried Serum for Laboratory-Based Dengue Surveillance-American Samoa, 2018

Laboratory-based surveillance for arboviral diseases is challenging in resource-limited settings. We evaluated the use of filter paper-dried sera for detection of dengue virus (DENV) RNA during an outbreak in American Samoa. Matched liquid and filter paper-dried sera were collected from patients with suspected dengue and shipped to a reference laboratory for diagnostic testing. RNA was extracted from each sample and tested for DENV RNA by real-time reverse transcription-polymerase chain reaction (RT-PCR). Of 18 RT-PCR-positive liquid specimens, 14 matched filter paper-dried specimens were positive for a sensitivity of 78% (95% CI, 55-91%). Of 82 RT-PCR-negative liquid specimens, all filter paper-dried specimens were negative for a specificity of 100% (95% CI, 96-100%). Shipping of filter paper-dried specimens was similarly timely but less expensive than shipping liquid sera. Using filter paper-dried serum or blood can be a cost-effective and sustainable approach to surveillance of dengue and other arboviral diseases in resource-limited settings.

Evaluation of Patient IgM and IgG Reactivity Against Multiple Antigens for Improvement of Serodiagnostic Testing for Early Lyme Disease

Serologic testing is the standard for laboratory diagnosis and confirmation of Lyme disease. Serodiagnostic assays to detect antibodies against Borrelia burgdorferi, the agent of Lyme borreliosis, are used for detection of infection. However, serologic testing within the first month of infection is less sensitive as patients' antibody responses continue to develop. Previously, we screened several B. burgdorferi in vivo expressed antigens for candidates that elicit early antibody responses in patients with Stage 1 and 2 Lyme disease. We evaluated patient IgM seroreactivity against 6 antigens and found an increase in sensitivity without compromising specificity when compared to current IgM second-tier immunoblot scoring. In this study, we continued the evaluation using a multi-antigen panel to measure IgM plus IgG seroreactivity in these early Lyme disease patients' serum samples. Using two statistical methods for calculating positivity cutoff values, sensitivity was 70 and 84-87%, for early acute and early convalescent Lyme disease patients, respectively. Specificity was 98-100% for healthy non-endemic control patients, and 96-100% for healthy endemic controls depending on the statistical analysis. We conclude that improved serologic testing for early Lyme disease may be achieved by the addition of multiple borrelial antigens that elicit IgM and IgG antibodies early in infection.

Seroprevalence and Symptomatic Attack Rate of Chikungunya Virus Infection, United States Virgin Islands, 2014-2015

When introduced into a naïve population, chikungunya virus generally spreads rapidly, causing large outbreaks of fever and severe polyarthralgia. We randomly selected households in the U.S. Virgin Islands (USVI) to estimate seroprevalence and symptomatic attack rate for chikungunya virus infection at approximately 1 year following the introduction of the virus. Eligible household members were administered a questionnaire and tested for chikungunya virus antibodies. Estimated proportions were calibrated to age and gender of the population. We enrolled 509 participants. The weighted infection rate was 31% (95% confidence interval [CI]: 26-36%). Among those with evidence of chikungunya virus infection, 72% (95% CI: 65-80%) reported symptomatic illness and 31% (95% CI: 23-38%) reported joint pain at least once per week approximately 1 year following the introduction of the virus to USVI. Comparing rates from infected and noninfected study participants, 70% (95% CI: 62-79%) of fever and polyarthralgia and 23% (95% CI: 9-37%) of continuing joint pain in patients infected with chikungunya virus were due to their infection. Overall, an estimated 43% (95% CI: 33-52%) of the febrile illness and polyarthralgia in the USVI population during the outbreak was attributable to chikungunya virus and only 12% (95% CI: 7-17%) of longer term joint pains were attributed to chikungunya virus. Although the rates of infection, symptomatic disease, and longer term joint symptoms identified in USVI are similar to other outbreaks of the disease, a lower proportion of acute fever and joint pain was found to be attributable to chikungunya virus.

Autocidal gravid ovitraps protect humans from chikungunya virus infection by reducing Aedes aegypti mosquito populations

Background

Public health responses to outbreaks of dengue, chikungunya, and Zika virus have been stymied by the inability to control the primary vector, Aedes aegypti mosquitos. Consequently, the need for novel approaches to Aedes vector control is urgent. Placement of three autocidal gravid ovitraps (AGO traps) in ~85% of homes in a community was previously shown to sustainably reduce the density of female Ae. aegypti by >80%. Following the introduction of chikungunya virus (CHIKV) to Puerto Rico, we conducted a seroprevalence survey to estimate the prevalence of CHIKV infection in communities with and without AGO traps and evaluate their effect on reducing CHIKV transmission.

Methods and findings

Multivariate models that calculated adjusted prevalence ratios (aPR) showed that among 175 and 152 residents of communities with and without AGO traps, respectively, an estimated 26.1% and 43.8% had been infected with CHIKV (aPR = 0.50, 95% CI: 0.37–0.91). After stratification by time spent in their community, protection from CHIKV infection was strongest among residents who reported spending many or all weekly daytime hours in their community:10.3% seropositive in communities with AGO traps vs. 48.7% in communities without (PR = 0.21, 95% CI: 0.11–0.41). The age-adjusted rate of fever with arthralgia attributable to CHIKV infection was 58% (95% CI: 46–66%). The monthly number of CHIKV-infected mosquitos and symptomatic residents were diminished in communities with AGO traps compared to those without.

Conclusions

These findings indicate that AGO traps are an effective tool that protects humans from infection with a virus transmitted by Ae. aegypti mosquitos. Future studies should evaluate their protective effectiveness in large, urban communities.

Aedes species mosquitos transmit pathogens of public health importance, including dengue, Zika, and chikungunya viruses. No tools exist to control these mosquitos that sustainably and effectively prevent human infections. Autocidal gravid ovitraps (AGO traps) have been shown to sustainably reduce Aedes populations by >80%. After chikungunya virus was introduced into Puerto Rico, we conducted serosurveys in communities with and without AGO traps. We observed a two-fold lower prevalence of chikungunya virus infection among residents of communities with AGO traps compared to communities without. Among infected residents of communities with traps, a significant proportion likely had been infected while outside their community. These findings indicate that AGO traps are an effective tool that protects humans from infection with pathogens transmitted by Aedes mosquitos.

Bloodmeal Host Selection of Culex quinquefasciatus (Diptera: Culicidae) in Las Vegas, Nevada, United States

St. Louis encephalitis virus (SLEV) and West Nile virus (WNV) have recently emerged in the southwestern United States. Surveillance for arboviruses in Las Vegas, NV, detected a surge of SLEV activity in the southern house mosquito (Culex quinquefasciatus Say) during 2016. To identify candidate avian amplifiers, we assessed the identification, viral infection, and immune status of vertebrate hosts for 195 blood-engorged Cx. quinquefasciatus mosquitoes collected in August and September 2016. Bloodmeals were identified from 164 engorged abdomens, representing 19 species of birds and three species of mammals. No SLEV or WNV viremia was detected, but one mosquito tested positive for Culex flavivirus. House finch (Haemorhous mexicanus) (Muller) was the most common bloodmeal, followed by domestic chicken (Gallus gallus) (Linnaeus), American robin (Turdus migratorius) L., house sparrow (Passer domesticus) (L.), great-tailed grackle (Quiscalus mexicanus) (Gmelin), northern mockingbird (Mimus polyglottos) (L.) and mourning dove (Zenaida macroura) (L.). SLEV-reactive antibodies were detected in six identified bloodmeals and WNV-reactive antibodies were detected in 33. House sparrow and house finch were the most likely hosts to show previous exposure to SLEV and WNV, respectively. Over-utilization by Cx. quinquefasciatus for bloodmeal hosts was observed primarily among robin, finch and sparrow, all species that roost communally. House finch stands out as a candidate important amplifier for both SLEV and WNV because of its preference by mosquito vectors, and high competence for closely related virus strains. While implicated in previous outbreaks as an important mosquito vector, Cx. quinquefasciatus feeds infrequently on mammals in Las Vegas, indicating a low risk for bridge transmission to humans.

Seroprevalence of Heartland Virus Antibodies in Blood Donors, Northwestern Missouri, USA

We estimated the seroprevalence of Heartland virus antibodies to be 0.9% (95% CI 0.4%–4.2%) in a convenience sample of blood donors from northwestern Missouri, USA, where human cases and infected ticks have been identified. Although these findings suggest that some past human infections were undetected, the estimated prevalence is low.

Reassessing Serosurvey-Based Estimates of the Symptomatic Proportion of Zika Virus Infections

Since the 2007 Zika epidemic in the Micronesian state of Yap, it has been apparent that not all people infected with Zika virus (ZIKV) experience symptoms. However, the proportion of infections that result in symptoms remains unclear. Existing estimates have varied in their interpretation of symptoms due to other causes and the case definition used, and they have assumed perfect test sensitivity and specificity. Using a Bayesian model and data from ZIKV serosurveys in Yap (2007), French Polynesia (2013-2014), and Puerto Rico (2016), we found that assuming perfect sensitivity and specificity generally led to lower estimates of the symptomatic proportion. Incorporating reasonable assumptions for assay sensitivity and specificity, we estimated that 27% (95% credible interval (CrI): 15, 37) (Yap), 44% (95% CrI: 26, 66) (French Polynesia), and 50% (95% CrI: 34, 92) (Puerto Rico) of infections were symptomatic, with variation due to differences in study populations, study designs, and case definitions. The proportion of ZIKV infections causing symptoms is critical for surveillance system design and impact assessment. Here, we accounted for key uncertainties in existing seroprevalence data and found that estimates for the symptomatic proportion ranged from 27% to 50%, suggesting that while the majority of infections are asymptomatic or mildly symptomatic, symptomatic infections might be more common than previously estimated.

Evaluation of in vivo expressed Borrelia burgdorferi antigens for improved IgM serodiagnosis of early Lyme disease

Improved serologic tests are needed for accurate diagnosis and proper treatment of early stage Lyme disease. We evaluated the 3 antigens currently used for 2-tiered IgM immunoblot testing (FlaB, OspC, and BmpA) in combination with 3 additional antigens (BBA65, BBA70, and BBA73) and measured the sensitivity and specificity against a serum repository of positive and negative controls. Using 3 statistical methods for positivity cutoff determinations and scoring criteria, we found increased sensitivities for early Lyme disease when 2 of 6 antigens were positive as compared with the 2 of 3 antigen IgM criteria currently used for second-tier immunoblot scoring. Specificities for negative controls were comparable or superior to using 2 of 3 antigens. These results indicate that IgM sensitivity and specificity of serological testing for Lyme disease in the early stages of illness can be improved by employing antigens that target the initial host antibody responses.

Zika Virus Infection in Patient with No Known Risk Factors, Utah, USA, 2016

In 2016, Zika virus disease developed in a man (patient A) who had no known risk factors beyond caring for a relative who died of this disease (index patient). We investigated the source of infection for patient A by surveying other family contacts, healthcare personnel, and community members, and testing samples for Zika virus. We identified 19 family contacts who had similar exposures to the index patient; 86 healthcare personnel had contact with the index patient, including 57 (66%) who had contact with body fluids. Of 218 community members interviewed, 28 (13%) reported signs/symptoms and 132 (61%) provided a sample. Except for patient A, no other persons tested had laboratory evidence of recent Zika virus infection. Of 5,875 mosquitoes collected, none were known vectors of Zika virus and all were negative for Zika virus. The mechanism of transmission to patient A remains unknown but was likely person-to-person contact with the index patient.

Rapid colorimetric detection of Zika virus from serum and urine specimens by reverse transcription loop-mediated isothermal amplification (RT-LAMP)

Zika virus (ZIKV) has emerged as a major global public health concern in the last two years due to its link as a causative agent of human birth defects. Its rapid expansion into the Western Hemisphere as well as the ability to be transmitted from mother to fetus, through sexual transmission and possibly through blood transfusions has increased the need for a rapid and expansive public health response to this unprecedented epidemic. A non-invasive and rapid ZIKV diagnostic screening assay that can be performed in a clinical setting throughout pregnancy is vital for prenatal care of women living in areas of the world where exposure to the virus is possible. To meet this need we have developed a sensitive and specific reverse transcriptase loop-mediated isothermal amplification (RT-LAMP) assay to detect ZIKV RNA in urine and serum with a simple visual detection. RT-LAMP results were shown to have a limit of detection 10-fold higher than qRT-PCR. As little as 1.2 RNA copies/μl was detected by RT-LAMP from a panel of 178 diagnostic specimens. The assay was shown to be highly specific for ZIKV RNA when tested with diagnostic specimens positive for dengue virus (DENV) and chikungunya virus (CHIKV). The assay described here illustrates the potential for a fast, reliable, sensitive and specific assay for the detection of ZIKV from urine or serum that can be performed in a clinical or field setting with minimal equipment and technological expertise.

Bias Correction in Estimating Proportions by Pooled Testing

In the estimation of proportions by pooled testing, the MLE is biased, and several methods of correcting the bias have been presented in previous studies. We propose a new estimator based on the bias correction method introduced by Firth (Biometrika 80:27-38, 1993), which uses a modification of the score function, and we provide an easily computable, Newton-Raphson iterative formula for its computation. Our proposed estimator is almost unbiased across a range of problems, and superior to existing methods. We show that for equal pool sizes the new estimator is equivalent to the estimator proposed by Burrows (Phytopathology 77:363-365, 1987). The performance of our estimator is examined using pooled testing problems encountered in plant disease assessment and prevalence estimation of mosquito-borne viruses.

Use of Blood Donor Screening Data to Estimate Zika Virus Incidence, Puerto Rico, April-August 2016

Puerto Rico has been heavily impacted by Zika virus, a mosquitoborne flavivirus that emerged in the Americas during 2015. Although most persons with Zika virus show no symptoms, the virus can cause neurologic and other complications, including fetal microcephaly. Local Zika virus transmission in Puerto Rico has been reported since December 2015. To prevent transfusion-associated transmission, local blood collection ceased in March 2016 but resumed in April 2016 after Zika virus screening of blood donations became available. Using data from screening of blood donations collected by the 2 largest blood centers in Puerto Rico during April 3-August 12, 2016, and assuming a 9.9-day duration of viremia, we estimated that 469,321 persons in Puerto Rico were infected during this period, for an estimated cumulative incidence of 12.9%. Results from blood donation screening during arboviral outbreaks can supplement routine clinical and surveillance data for improved targeting of prevention efforts.

Estimated Incubation Period for Zika Virus Disease

Information about the Zika virus disease incubation period can help identify risk periods and local virus transmission. In 2015-2016, data from 197 symptomatic travelers with recent Zika virus infection indicated an estimated incubation period of 3-14 days. For symptomatic persons with symptoms >2 weeks after travel, transmission might be not travel associated.

Comparison of Vector Efficiency of Ixodes scapularis (Acari: Ixodidae) From the Northeast and Upper Midwest of the United States for the Lyme Disease Spirochete Borrelia mayonii

Borrelia mayonii, a recently recognized species within the Borrelia burgdorferi sensu lato complex, has been detected in host-seeking Ixodes scapularis Say ticks and found to be associated with Lyme disease in the Upper Midwest. This spirochete has, to date, not been documented from the Northeast, but we previously demonstrated that I. scapularis ticks originating from Connecticut are capable of serving as a vector of B. mayonii In this follow-up study, we compared the vector efficiency for B. mayonii (strain MN14-1420) of I. scapularis ticks originating from Minnesota in the Upper Midwest and Connecticut in the Northeast. CD-1 outbred white mice previously infected with B. mayonii via tick bite were exposed to simultaneous feeding by Minnesota and Connecticut larvae contained within separate feeding capsules. We found no difference in the ability of Minnesota and Connecticut larvae to acquire B. mayonii from infected mice and pass spirochetes to the nymphal stage (overall nymphal infection rates of 11.6 and 13.3%, respectively). Moreover, the efficiency of transmission of B. mayonii by single infected nymphs was similar for the Minnesota and Connecticut ticks (33 and 44%, respectively). We conclude that the examined I. scapularis ticks from the Upper Midwest and Northeast did not differ in their efficiency as vectors for B. mayonii.

Non-human primate antibody response to mosquito salivary proteins: Implications for dengue virus transmission in Puerto Rico

An important step to incriminate a mosquito as a vector of a disease pathogen is finding evidence of direct contact between the mosquito and humans. Typically, this is accomplished through landing/biting catches, or host blood meal analysis in engorged mosquitoes via immunologic assays. An alternate approach is to identify the presence of specific mosquito anti-saliva protein antibodies in the blood of exposed hosts. Following the discovery of dengue infected, free roaming non-human primates in Puerto Rico, we investigated which mosquito species had bitten these primates using a serologic assay. Serum samples from 20 patas monkeys (Erythrocebus patas) and two rhesus macaques (Macaca mulatta) were used to evaluate mosquito bite exposure to Aedes aegypti, Aedes mediovittatus, Aedes taeniorhynchus, and Culex quinquefasciatus mosquitoes. Of 22 non-human primates examined 20 (90%), 17 (77%), 13 (59%), and 7 (31%) were positive for exposure to Ae. mediovittatus, Cx. quinquefasciatus, Ae. taeniorhynchus, and Ae. aegypti, respectively. Our findings indicated that free-roaming primates in Puerto Rico were exposed to the bites of one proven dengue vector, Ae. aegypti and one potential dengue vector, Ae. mediovittatus.

Effect of a Dengue Clinical Case Management Course on Physician Practices in Puerto Rico

BACKGROUND

 Prior to 2010, the clinical management of dengue in Puerto Rico was inconsistent with World Health Organization guidelines. A 4-hour classroom-style course on dengue clinical management was developed in 2009 and mandated for Puerto Rico medical licensure in 2010. Fifty physicians were trained as "master trainers" and gave this course to 7638 physicians. This study evaluated the effect of the course on the clinical management of hospitalized dengue patients.

METHODS

 Pre- and post-course test responses were compared. Changes in physician practices were assessed by reviewing medical records of 430 adult and 1075 pediatric dengue patients at the 12 hospitals in Puerto Rico that reported the most cases during 2008-2009 (pre-intervention) and 2011 (post-intervention). Mixed-effects logistic regression was used to compare key indicators of dengue management.

RESULTS

 Physician test scores increased from 48% to 72% correct. Chart reviews showed that the percentage of adult patients who did not receive corticosteroids increased from 30% to 68% (odds ratio [OR], 5.9; 95% confidence interval [CI], 3.7-9.5) and from 91% to 96% in pediatric patients (OR, 2.7; 95% CI, 1.5-4.9). Usage of isotonic intravenous saline during the critical period increased from 57% to 90% in adult patients (OR, 6.2; 95% CI, 1.9-20.4) and from 25% to 44% in pediatric patients (OR, 3.4; 95% CI, 2.2-5.3).

CONCLUSIONS

 Management of dengue inpatients significantly improved following implementation of a classroom-style course taught by master trainers. An online version of the course was launched in 2014 to expand its reach and sustainability.