Conceived and designed the experiments: VS WH JMS UN WS SK DA KU AMF. Performed the experiments: VS WH JMS UN WS SK DA KU AMF. Analyzed the data: VS WH WS AMF. Contributed reagents/materials/analysis tools: VS WH JMS UN WS SK DA KU AMF. Wrote the paper: VS WH JMS UN WS SK DA AMF.
The authors have declared that no competing interests exist.
The National Avian Influenza Surveillance (NAIS) system detected human H5N1 cases in Thailand from 2004–2006. Using NAIS data, we identified risk factors for death among H5N1 cases and described differences between H5N1 and human (seasonal) influenza cases.
NAIS identified 11,641 suspect H5N1 cases (e.g. persons with fever and respiratory symptoms or pneumonia, and exposure to sick or dead poultry). All suspect H5N1 cases were tested with polymerase chain reaction (PCR) assays for influenza A(H5N1) and human influenza viruses. NAIS detected 25 H5N1 and 2074 human influenza cases; 17 (68%) and 22 (1%) were fatal, respectively. We collected detailed information from medical records on all H5N1 cases, all fatal human influenza cases, and a sampled subset of 230 hospitalized non-fatal human influenza cases drawn from provinces with ≥1 H5N1 case or human influenza fatality.
Fatal versus non-fatal H5N1 cases were more likely to present with low white blood cell (p = 0.05), lymphocyte (p<0.02), and platelet counts (p<0.01); have elevated liver enzymes (p = 0.05); and progress to circulatory (p<0.001) and respiratory failure (p<0.001). There were no differences in age, medical conditions, or antiviral treatment between fatal and non-fatal H5N1 cases. Compared to a sample of human influenza cases, all H5N1 cases had direct exposure to sick or dead birds (60% vs. 100%, p<0.05). Fatal H5N1 and fatal human influenza cases were similar clinically except that fatal H5N1 cases more commonly: had fever (p<0.001), vomiting (p<0.01), low white blood cell counts (p<0.01), received oseltamivir (71% vs. 23%, p<.001), but less often had ≥1 chronic medical conditions (p<0.001).
In the absence of diagnostic testing during an influenza A(H5N1) epizootic, a few epidemiologic, clinical, and laboratory findings might provide clues to help target H5N1 control efforts. Severe human influenza and H5N1 cases were clinically similar, and both would benefit from early antiviral treatment.
With the emergence of an epizootic of highly pathogenic avian influenza A(H5N1) virus among poultry in Southeast Asia in December 2003, Thailand's first patients with influenza A(H5N1) virus infection (hereafter referred to as H5N1 cases) were identified in January 2004
Thailand's National Avian Influenza Surveillance System (NAIS) was a laboratory-based surveillance system to detect (human) H5N1 cases. NAIS was established by Thailand's Ministry of Public Health (MOPH) in December 2003 following Thailand's first outbreaks of influenza A(H5N1) among poultry. Laboratory and epidemiological response components of the surveillance system have been previously described
Suspect H5N1 case detection and management occurred in the following way: once a suspect H5N1 case was identified by a clinician, the patient was isolated, diagnostic respiratory specimens were obtained, antiviral therapy with oseltamivir was initiated, local and provincial public health authorities were notified, and the patient was referred to a provincial or regional hospital for advanced medical care if clinically warranted
We conducted a retrospective cohort study of H5N1 and human influenza cases captured in Thailand's NAIS system between January 2004 and December 2006. A description of the study design has been reported previously in a companion study
Medical charts of sampled patients were reviewed by supervised teams of trained medical and public health professionals from the MOPH and the Thailand MOPH-U.S.CDC Collaboration (TUC) and a form was used to abstract detailed demographic, epidemiological, clinical, treatment, and outcome data from medical charts. Data were analyzed using STATA (Version 8, College Station, TX) statistical software. Disaggregated clinical and epidemiological data for 17 of 25 H5N1 cases have been previously reported, however, the data presented here represent a new abstraction and a complete analysis of all 25 H5N1 cases detected in Thailand during the study period
This research was carried out with approval from and in compliance with the standards of the ethical review committees of the MOPH and the United States Centers for Disease Control and Prevention (US CDC). Informed consent was not required as data were analyzed anonymously.
During January 2004 through December 2006, 11,641persons across Thailand were identified as suspected H5N1 cases in the NAIS system; among these, 2,074 (18%) persons in 73 of 76 Thai provinces tested positive for human influenza (A/H1N1, A/H3N2, or type B) and 25 persons in 19 of 76 provinces tested positive for H5N1 virus infection
Our study included the following: all 25 H5N1 cases, all 22 fatal human influenza cases (arising from 15 of 76 provinces), and a sampled subset of 230 (of 2074 total) hospitalized non-fatal human influenza cases drawn from 25 of 76 Thai provinces in which at least one fatal H5N1 or human influenza case had occurred (
Although human influenza cases were detected in virtually every Thai province, H5N1 cases were geographically localized to the central and northern regions where most commercial poultry production was located. H5N1 cases were detected without a consistent seasonal pattern; whereas human influenza cases were most frequently detected during one or two annual periods of peak influenza activity occurring between July and November or between January and March (
Several differences related to H5N1 virus exposure were observed between H5N1 and human influenza cases (
H5N1 Influenza | Human Influenza | |||
(N = 25) | (N = 252) | |||
Male | 16 (64) | 147 (59) | ||
Median, years (range) | 18 (1.6–68) | 14 (.08–102) | ||
0–9 yrs | 10 (40) | 89 (25) | ||
10–19 yrs | 4 (14) | 64 (25) | ||
20–29 yrs | 4 (16) | 19 (8) | ||
30–39 yrs | 1 (4) | 18 (7) | ||
40–49 yrs | 3 (12) | 20 (8) | ||
> = 50 yrs | 3 (12) | 42 (17) | ||
Poultry worker | 2 (8) | 8 (3) | ||
Healthcare worker | 0 (0) | 1 (<1) | ||
Other | 22 (92) | 224 (96) | ||
Consumed sick or dead birds | ||||
Touched sick or dead birds | ||||
Cared for sick or dead birds | ||||
Butchered sick or dead birds | ||||
Sick or dead birds in household | ||||
Contacted another human H5N1 case | 3 (16) | 7 (4) | ||
≥1 direct exposures | ||||
No direct exposure to H5N1 | 0 (0) | 101 (40) | ||
Indirect exposure only | ||||
Indirect plus ≥1 direct exposures | 12 (48) | 72 (29) | ||
*P<0.05.
**Indirect exposures: the presence of sick or dead birds in the patient's neighborhood.
During 2004, the first year of the surveillance, H5N1 cases presented to hospital later than did human influenza cases (median: 5 vs. 2 days, p<.01); this difference persisted during the subsequent two years of surveillance (2005–2006) (data not shown).
Among H5N1 cases, those with fatal and non-fatal illness were similar with regard to age (median: 14 vs. 25 years old, p = .86), presence of an underlying medical condition, presenting symptoms, and radiographic evidence of pneumonia or ARDS on an admission chest X-ray (
All H5N1 (N = 25) | Non-fatal H5N1 (N = 8) | Fatal H5N1 (N = 17) | p-value | UnadjustedOR [95% CI] |
|
≥1 condition | 4 (16) | 2 (25) | 2 (12) | 0.57 | 0.4 [0.04–3.52] |
Fever>38.2°C | 23 (96) | 6 (86) | 17 (100) | 0.29 | |
Cough | 20 (83) | 6 (86) | 14 (82) | 1 | 1.29 [0.11–15.0] |
Sore throat | 9 (38) | 2 (29) | 7 (41) | 0.67 | 0.57 [0.01–3.83] |
Dsypnia |
17 (81) | 4 (57) | 13 (76) | 0.37 | 0.41 [0.06–2.66] |
Headache | 2 (8) | 0 (0) | 2 (12) | 1 | |
Myalgia | 6 (25) | 1 (14) | 5 (29) | 0.63 | 0.4 [0.4–4.23] |
Altered mental status |
2 (8) | 0 (0) | 2 (12) | 1 | |
Abdominal pain | 4 (17) | 1 (14) | 3 (18) | 1 | |
Vomiting | 12 (50) | 1 (14) | 11 (65) | 0.07 | 0.09 [0.01–0.94] |
Diarrhea | 5 (21) | 1 (14) | 4 (23) | 1 | 0.45 [0.05–5.94] |
White blood cell (WBC) count | 5.1 (1.1–18.3) | 7.6 (3.1–13.6) | 4.1 (1.1–18.3) | n/a | |
Neutrophil count | 3.9 (1.2–15.6) | 4.5 (1.6–8.9) | 3.4 (1.2–15.6) | 0.11 | n/a |
Lymphocyte count | 1.2 (.1–4.3) | 1.8 (.8–4.3) | .9 (.1–2.7) | 0. |
n/a |
Platelet count | 172 (77–528) | 254 (178–528) | 152 (77–304) | n/a | |
Hemoglobin | 12.3 (9–16.7) | 10.4 (9–14) | 12.7 (11–16.7) | n/a | |
AST | 148 (17–1032) | 27 (17–306) | 181 (34–1032) | n/a | |
ALT | 50 (7–1048) | 20 (7–82) | 55 (26–1048) | n/a | |
BUN | 12 (4–61) | 9 (4–9) | 22 (7–61) | n/a | |
Creatinine | 1 (.2–2.8) | 0.7 (.2–.8) | 1.4 (.5–2.8) | n/a | |
Pneumonia |
4 (50) | 7 (41) | 11 (44) | 1 | 0.7[0.13–3.79] |
Pneumonia |
4 (50) | 7 (41) | 11 (44) | 1 | 0.7 [0.13–3.79] |
Pneumonia |
15 (60) | 4 (50) | 11 (65) | 0.67 | 1.83 [0.33–10.09] |
Pneumonia |
20 (80) | 4 (50) | 16 (94) | ||
Hypotension | 13 (57) | 0 (0) | 13 (81) | ||
Respiratory failure | 17 (71) | 0 (0) | 17 (100) | ||
ARDS | 14 (56) | 0 (0) | 14 (82) | ||
Need for ICU | 17 (68) | 0 (0) | 17 (100) | ||
Symptom onset to first medical care | 4 (1–14) | 4 (3–14) | 3 (1–10) | 0.2 | n/a |
Symptom onset to hospital admission | 4 (2–18) | 4.5 (3–18) | 4 (2–10) | 0.34 | n/a |
Symptom onset to oseltamivir treatment | 7 (4–21) | 7 (5–7) | 8 (4–21) | 0.39 | n/a |
Hospital admission to oseltamivir treatment | 2.5 (0–14) | 2 (0–3) | 3 (0–14) | 0.28 | n/a |
Hospital admission to death or discharge | n/a | 13 (1–21) | 6 (2–23) | n/a | n/a |
N/a: not applicable;
*unadjusted OR and 95% CI of a fatal H5N1 outcome among H5N1 cases;
dyspnea: defined as any recorded difficulty breathing or shortness of breath in any age group, or in a child less than age 5, the presence of stridor or chest in-drawing;
altered mental status: defined as any recorded presence of altered mental status, confusion, or unconsciousness;
pneumonia: defined as the presence of alveolar infiltrates, interstitial infiltrates, or lobar consolidation on chest X-ray;
**not possible to calculate an unadjusted OR.
Similar proportions of fatal and non-fatal H5N1 cases received treatment with the antiviral agent oseltamivir (71% vs. 57%, p = .65), an antibiotic agent (100% vs. 100%, p = 1), and or a corticosteroid (29% vs. 43%, p = .65). Clinical complications noted in H5N1 cases were more frequently present among fatal compared to non-fatal cases, including respiratory failure requiring intubation, hypotension requiring inotropic agents, development of acute respiratory distress syndrome (ARDS), and need for intensive care unit (ICU) management (
Fatal human influenza cases compared to fatal H5N1 cases tended to be older (median age: 39 vs. 14 years, p = .14), although not statistically significant, and were significantly more likely to have a chronic medical condition (
Fatal H5N1 | Fatal Human Influenza | p-value | UnadjustedOR [95% CI] |
|
(N = 17) | (N = 22) | |||
≥1 condition | 2 (12) | 15 (68) | ||
Fever>38.2°C | 17 (100) | 9 (41) | ||
Cough | 14 (82) | 15 (68) | 0.46 | 0.46 [0.1–2.13] |
Sore throat | 7 (41) | 1 (5) | ||
Dsypnia |
13 (76) | 16 (73) | 1 | 0.82 [0.19–3.54] |
Headache | 2 (12) | 0 (0) | 0.18 | |
Myalgia | 5 (29) | 1 (5) | ||
Altered mental status |
2 (12) | 6 (27) | 0.37 | 0.36 [0.06–2.04] |
Abdominal pain | 3 (18) | 0 (0) | 0.07 | |
Vomiting | 11 (65) | 4 (18) | ||
Diarrhea | 4 (23) | 2 (9) | 0.37 | 0.33 [0.05–2.04] |
White blood cell (WBC) count | 4.1 (1.1–18.3) | 8.7 (0.9–34.7) | n/a | |
Neutrophil count | 3.4 (1.2–15.6) | 6.0 (.05–29.5) | n/a | |
Lymphocyte count | .9 (.1–2.7) | 1.5 (.4–5.0) | 0.09 | n/a |
Platelet count | 152 (77–304) | 156 (90–392) | 0.61 | n/a |
Hemoglobin | 12.7 (11–16.7) | 12.2 (8.5–17.7) | 0.41 | n/a |
AST | 181 (34–1032) | 161 (48–2289) | 0.88 | n/a |
ALT | 55 (26–1048) | 70 (18–444) | 0.47 | n/a |
BUN | 22 (7–61) | 25 (7–92) | 0.46 | n/a |
Creatinine | 1.4 (.5–2.8) | 1.3 (.2–10.4) | 0.81 | n/a |
Pneumonia |
7 (41) | 18 (82) | ||
Pneumonia |
7 (41) | 19 (86) | ||
Pneumonia |
11 (65) | 18 (82) | 0.28 | 0.4 [0.09–1.77] |
Pneumonia |
16 (94) | 19 (86) | 0.62 | 2.53 [0.24–26.72] |
Hypotension | 13 (81) | 20 (91) | 0.63 | 2.3 [0.34–15.75] |
Respiratory failure | 17 (100) | 22 (100) | n/a | n/a |
ARDS | 14 (82) | 7 (32) | ||
Need for ICU | 17 (100) | 15 (68) | n/a | |
Symptom onset to first medical care | 3 (1–10) | 2 (0–7) | 0.22 | n/a |
Symptom onset to hospital admission | 4 (2–10) | 2 (0–7) | n/a | |
Symptom onset to oseltamivir treatment | 8 (4–21) | 4 (4–7) | n/a | |
Hospital admission to oseltamivir treatment | 3 (0–14) | 1 (0–3) | 0.1 | n/a |
Hospital admission to death or discharge | 6 (2–23) | 3.5 (0–68) | 0.09 | n/a |
N/a: not applicable;
*unadjusted OR and 95% CI of a fatal H5N1 outcome versus a fatal human influenza outcome;
dyspnea: defined as any recorded difficulty breathing or shortness of breath in any age group, or in a child less than age 5, the presence of stridor or chest in-drawing;
altered mental status: defined as any recorded presence of altered mental status, confusion, or unconsciousness;
pneumonia: defined as the presence of alveolar infiltrates, interstitial infiltrates, or lobar consolidation on chest X-ray;
**not possible to calculate an unadjusted OR.
Among the two groups, fatal H5N1 cases were more likely to receive oseltamivir that fatal human influenza cases (71% vs. 23%, p<0.01). However, fatal H5N1 cases experienced greater delays in the number of median days between symptom onset and hospital admission, and even more significant delays between symptom onset and oseltamivir treatment (median: 8 vs. 4 days, p<.02)(
During the 2004–2006 influenza A(H5N1) epizootic in Thailand, we identified few differences in admission clinical or epidemiologic characteristics between fatal and non-fatal H5N1 cases; however, the presence of laboratory abnormalities in white blood cell counts and liver enzymes, hypotension, or ARDS may suggest a poor prognosis. Compared to patients hospitalized for human influenza infections, all H5N1 cases had a history of direct exposure to sick or dead birds. H5N1 cases had a high case fatality ratio in Thailand, unlike patients with human influenza infection identified from NAIS. However, severely ill patients with human influenza infection who subsequently died looked clinically similar to patients with H5N1 infection. The presence of underlying medical conditions, lack of measured fever, normal WBC counts, and the absence of ARDS were the best predictors in severely ill patients with suspect H5N1 that the infection was due to human influenza viruses versus H5N1 virus.
Among H5N1 cases, previous reports have described several admission findings associated with fatal outcome that might help clinicians risk stratify patients. Among H5N1 cases in Thailand during 2004–2006, we found that most admission signs and symptoms, chest radiograph findings, and epidemiologic factors were not useful in estimating risk for fatal disease. However, multiple laboratory abnormalities on admission were associated with a fatal outcome among H5N1 cases. Similar to other reports, we found that fatal H5N1 cases compared to those who survived, were more likely to have a lower median white blood cell count (WBC) or leukopenia
We were able to directly compare patients infected with influenza A (H5N1) virus or human influenza viruses identified from the same surveillance system. Several epidemiologic factors may offer clues that the infection is a human influenza infection versus H5N1 virus infection. Human influenza viruses were most common during annual peaks in human influenza circulation, while H5N1 cases had no seasonality. Also, H5N1 cases had a history of direct exposure from dead or sick birds and were from provinces with commercial poultry operations; many patients with human influenza infection had only indirect exposure to dead or sick birds. Finally, underlying medical conditions that are associated with complications (including death) from human influenza infection were present in less than 15% of fatal H5N1 cases. In contrast, almost 60% of fatal human influenza infections had chronic medical conditions. These differences may offer some clues when laboratory testing is delayed or lacking to target control efforts to identify and prevent secondary human transmission H5N1 cases during a large epizootic outbreak.
Clinically, patients with severe human influenza infections that resulted in death were generally similar to patients with influenza A (H5N1) infection; shortness of breath was commonly observed on presentation and platelet counts and liver enzymes levels on initial laboratory assessment were similar. Also, similar proportions of cases in both groups had disease progression complicated by circulatory and respiratory failure. Although these clinical characteristics have been previously described in large series of severe and fatal H5N1 cases
Although the NAIS system was developed to find H5N1 cases, it eventually detected 80 times more human influenza cases
Our study was subject to several limitations. NAIS was a passive surveillance system, and H5N1 cases may have been missed as a consequence of under-reporting and or under-detection. In addition, the characteristics of the patients with human influenza infection from NAIS are different from those identified from surveillance for human influenza infections. Consequently, adults ≥65 years of age and very young children with human influenza, with no poultry exposure, were not represented. Clinical and epidemiologic data on individual cases varied according to the completeness of medical records. Also, information on potential H5N1 exposures were extracted retrospectively and may have been more complete for confirmed H5N1 cases. Due to our small sample size, we may not be able to detect some differences in clinical characteristics between fatal and non-fatal H5N1 cases and human influenza cases. Finally, inferences regarding the clinical presentation of H5N1 cases are limited to the clade one H5N1 viruses circulating in Thailand and the surrounding region during the study period and may not be representative of other clades of H5N1 virus known to cause human disease in other parts of the world.
As a result of Thailand's strong public health response to the emergence of influenza A(H5N1) virus infection in humans, we were provided a rare opportunity to compare H5N1 cases and human influenza cases identified by the same surveillance system. In lieu of laboratory diagnostics, history of direct exposure to dead or sick birds and lack of underlying medical conditions that are associated with complications due to human influenza may provide clues for targeting H5N1 control efforts. Fatalities from human influenza infection were uncommon. However, patients with severe human influenza infection had a clinical presentation that overlapped considerably with H5N1 cases. All of these patients, those with H5N1 infection and those with severe human influenza infection, would benefit from early empiric antiviral treatment.
We would like to acknowledge the technical support of colleagues in the Thailand Ministry of Public Health, the Thailand-US CDC Collaboration (TUC), and the Influenza Division at the Centers for Disease Control and Prevention in Atlanta, Georgia.