Evaluation of potential risk of transmission of avian influenza A viruses at live bird markets in response to unusual crow die-offs in Bangladesh.

Abstract In response to unusual crow die‐offs from avian influenza A(H5N1) virus infection during January‐February 2017 in Dhaka, Bangladesh, a One Health team assessed potential infection risks in live bird markets (LBMs). Evidence of aerosolized avian influenza A viruses was detected in LBMs and in the respiratory tracts of market workers, indicating exposure and potential for infection. This study highlighted the importance of surveillance platforms with a coordinated One Health strategy to investigate and mitigate zoonotic risk.

investigation assessed possible sources of HPAI A(H5N1) virus and potential spillover risks to humans.

| INITIATI ON OF OUTB RE AK INVE S TIG ATION
In response to a report of unusual crow mortality around central Dhaka, on January 14, 2017, 8 the Institute of Epidemiology, Disease Control and Research (IEDCR) initiated a multidisciplinary investigation from January 21 to February 12, 2017. The wildlife team identified 124 crow deaths within 7 km of the initial reported crow roosts. 8 Crow samples tested positive for A(H5N1) virus by realtime reverse transcription PCR (rRT-PCR). 8 The hypothesis that crows might have acquired A(H5N1) virus infection after consuming infected dead poultry and/or their waste products 3 was based on the observed crow die-offs near LBMs, where crows were observed feeding on poultry offal. To assess occupational risk of infection to exposed humans, we conducted a cross-sectional survey, examined air samples in LBMs, and collected respiratory samples from workers for influenza testing at 10 LBMs near the crow die-offs.

| ME THODS
For markets with ≤20 workers, we recruited all employees; while in markets with >20 workers, we randomly selected 20 workers per market. We collected nasal and throat swabs, and information on illness symptoms and workers' practices using a semi-structured questionnaire. In each market, air samples were also collected by a liquid cyclonic air sampler, 9 together with market-level hygiene assessments through observation. Based on a previous study, one air sampler was used for 30 minutes, placed in the center of each LBM, 1.3 meters from the ground and approximately 0.5 m from poultry housing 10 during 10:30 to 11:30 AM. At each LBM, the animal health team collected swabs from fresh fecal droppings beneath the poultry cages and accumulated offal samples; four samples from random sites were pooled together as one environmental sample per LBM.
All samples were tested using rRT-PCR with appropriate positive and negative controls to exclude contamination. Nasal and throat swabs were tested for influenza A and B viruses with subtyping of M-genepositive influenza A viruses for seasonal H3, H1N1pdm09, and avian H5/H7/H9. 11 Influenza A positive air and pooled environmental samples were subtyped for avian H5/H7/H9 only. Descriptive analysis was conducted using statistical software STATA (version 14.2).

| RE SULTS AND D ISCUSS I ON
Of the 10 markets, three had <8 poultry stalls, and the rest had 9-16 stalls. The average number of workers per stall was 3.0 (SD 1.1).
Almost all stalls slaughtered and defeathered birds. Multiple species of poultry were sold, including chickens, pigeon, geese, quail, and ducks at seven markets, and three LBMs only sold chickens. In all LBMs, birds not sold at the end of the day were kept in the same stall.
No LBMs practiced market closure days or rest days without poultry.
All LBMs had visible poultry feces on the ground, but no dead birds. Six LBMs reported market cleaning more than once daily, but only one used disinfectant (eg, bleach). Three LBMs reported disposal of solid waste at least twice daily. Half of the markets had open drains. About 60% of stalls experienced poultry deaths in the week before the investigations, and some workers reported discarding poultry carcasses as garbage or giving them to other workers.
Temperature and relative humidity were not significantly different among LBMs during air sampling.
We enrolled 151 workers from 81 stalls, with mean age of 31.3 (SD 11.8) years and median work experience of 9.0 years, interquartile range (IQR) 4-16 years, and all except one were male. Nearly 40% of the workers reported one or more of the following signs and symptoms in the previous 10 days: fever/feverishness (11.3%); measured temperature of ≥100.4°F (2.0%); cough (15.2%); sore throat (6.0%); runny nose (23.8%); eye redness (2.0%); diarrhea (0.7%); difficulty breathing (4.6%); headache or body ache (11.3%); and 60.9% were asymptomatic. Three workers reported febrile respiratory symptoms, and all three tested negative for influenza viruses by rRT-PCR. Overall, 21 (13.9%) LBM workers had respiratory specimens that tested positive for influenza A (12.6% of nasal swabs, 4% of throat swabs), of whom 62% were asymptomatic and only four (19%) reported respiratory symptoms (runny nose and/or cough) without fever. Six LBMs had at least one worker who tested positive for influenza A virus. Most of the influenza A positive samples were either H9 or non-subtypeable (Table 1).
All 10 LBMs had pooled environmental specimens that tested positive for influenza A, including five H5, one with both H5 and H9, and four markets with non-subtypeable specimens. AIVs were detected by air sampling at nine of 10 LBMs, including four with H9, and five with co-detections of both H5 and H9 ( Table 2) and ducks generated infectious droplets and aerosols. 12,13 Overall, LBMs were reported to be highly contaminated with AIVs by multiple studies in Bangladesh and other countries. 14,15 Taken together, these studies suggest that the most likely source of the aerosolized AIVs in LBMs in Bangladesh is from slaughtering of infected poultry. The crows as carrion eaters were likely infected from the offal or wastage of infected poultry. Third, we did not evaluate workers' practices for the occupational risk assessment because they had similar work practices as all participated in slaughtering and defeathering. Fourth, several influenza A positive respiratory samples were non-subtypeable by RT-PCR. We hypothesized that (a) viral load was too low as most of the non-subtypeable specimens were collected from nonill workers that yielded high Ct values, or (b) the virus subtype was different from those that we tested for. Fifth, detection of AIV RNA in respiratory specimens of healthy LBM workers cannot distinguish asymptomatic infection from contamination. We believe it is more likely that such detection represents the latter, but additional studies are needed. For example, testing to detect viral replication in serial respiratory specimens and serological testing of paired sera to as-

CO N S E NT TO PA RTI CI PATE
All the participants provided informed written consent prior to their inclusion in the study.