Risk factors associated with death among influenza A (H1N1) patients, Tamil Nadu, India, 2010SR Balaganesakumar1, MV Murhekar2, KK Swamy1, MR Kumar1, P Manickam2, PRT Pandian1
1 Directorate of Public Health and Preventive Medicine, Govt. of Tamil Nadu, Chennai, Tamil Nadu, India
2 National Institute of Epidemiology, ICMR, Chennai, Tamil Nadu, India
Background: Limited information is available about the risk factors associated with death among patients of influenza A (H1N1) in India. Aims: To describe the epidemiology of laboratory-confirmed influenza A (H1N1) patients and identify risk factors associated with death. Settings and Design: We reviewed the surveillance data of laboratory-confirmed patients in Tamil Nadu, India, for the year 2010. We conducted a case-control study by comparing 70 laboratory-confirmed A (H1N1) patients who died (cases) with 210 A (H1N1) patients who recovered (controls) to identify the risk factors for deaths. Materials and Methods: We interviewed the controls and immediate care-takers of the influenza patients who died to collect information about socio-demographic details and co-morbid conditions. We used an abstraction form to collect the information about the clinical details from the case records of the hospitals where the cases and controls received treatment. Statistical Analysis: We analysed the surveillance data by time, place and person. We conducted univariate and multivariate logistic regression analysis for identifying factors associated death. Results: During 2010, 1302 laboratory-confirmed cases were reported to the Tamil Nadu surveillance unit. Of these, 72 patients died (case fatality=5.5%). About 2/3 of the cases and 40% of the deaths were from three districts. On multivariate analysis, past history of diabetes, treatment in private hospitals, treatment with corticosteroids during illness, visit to >1 healthcare facility before laboratory confirmation and delay of >48 h in starting antivirals were found to be independently associated with the deaths. Conclusions: Influenza patients with previous history of diabetes, who had treatment with corticosteroids during illness, and started with antivirals after 48 h of onset of symptoms, were at higher risk of adverse outcome. In order to reduce the risk of death during future waves of influenza in Tamil Nadu, the physicians need to be sensitised regarding (1) higher risk of adverse outcomes among A (H1N1) patients with diabetes; (2) adherence to the national protocol for categorisation of cases; (3) prompt initiation of antivirals for severe cases; and (4) avoidance of systemic corticosteroids during management.
Keywords: Influenza A (H1N1), death, risk factors, Tamil Nadu, India
Since the beginning of the pandemic of influenza A (H1N1) in India, more than 46,000 laboratory-confirmed cases, including 2728 deaths, were reported till December 2010.  The pandemic had two distinct waves: First during May-December 2009 and the second during June-December 2010. About 63% of the reported deaths occurred during the second wave. , In August 2010, WHO announced the post-pandemic phase.  During this phase, the pandemic virus is expected to continue to circulate as a seasonal virus. Cases and outbreaks due to the virus would continue to occur and the high-risk groups would continue to be affected disproportionately by severe disease.  In this context, identification of risk groups for adverse outcome assumes importance.
Most of the Indian studies on pandemic influenza describe the clinical presentation of disease. ,,, However, little information is available about the risk factors associated with adverse outcomes. With this background, we conducted a study in Tamil Nadu to (1) describe the influenza A (H1N1) patients by time, place and person in Tamil Nadu, and (2) identify the factors associated with death.
Descriptive epidemiological study
A laboratory-confirmed case of influenza A (H1N1) was defined as acute febrile respiratory illness with laboratory-confirmed pandemic influenza A (H1N1) virus infection by real-time reverse transcriptase polymerase chain reaction (RT-PCR), at any of the 13 laboratories designated by the government of Tamil Nadu. 
As part of the surveillance of influenza A (H1N1), public and private hospitals treating suspected cases of influenza were requested by the state health authorities to send nasal or throat swabs for laboratory confirmation to the nearest designated laboratories. These laboratories reported laboratory-confirmed patients to Tamil Nadu Integrated Disease Surveillance Project (IDSP) units at the district and state levels. The district unit contacted the laboratory-confirmed patients and collected information about demographic details, presenting symptoms, treatment details and risk factors. We obtained a line-list of all laboratory-confirmed patients diagnosed during 2010 from the State IDSP unit, and collected information about time, place and person details through the district IDSP units using a structured questionnaire.
For the purpose of analysis, we included only those patients who were residing in Tamil Nadu. We analysed the data to describe the distribution of cases by time, place and person. We constructed an epidemic curve, plotted the number of cases by district and calculated the proportion of patients in different age groups and gender.
Definition of case and control
We conducted an un-matched case-control study to identify the risk factors associated with mortality among the laboratory-confirmed patients. Influenza patients who died during 1 July 2010-31 December 2010 and residing in Tamil Nadu were considered as cases, and those who had the illness during the same period and recovered as controls.
Sample size and sampling
We required 42 cases and 126 controls for assumption of an odds ratio (OR) of 3, 50% exposure among the controls to any underlying medical illness, type-I error of 5%, power of 80% and a case: Control ratio of 1:3. We included 70 patients who died during July 2010-December 2010 as cases. We created a sampling frame of all the laboratory-confirmed patients who had the illness during the same period and recovered, and randomly selected the required number of controls from that sampling frame.
We used two data sources to collect information about the various risk factors associated with deaths. First, we used a pilot-tested, semi-structured questionnaire to interview the controls and immediate care-takers of the patients who died (cases) to collect information about socio-demographic details and co-morbid conditions. Second, we used an abstraction form to collect information about clinical presentation, treatment details and history of any co-morbid conditions from the case records of the hospitals where the cases and controls received treatment. As information about height and weight were not available for all cases and controls, we collected details about the build of the individual from the respondents. The study was conducted during January 2011-April 2011.
We entered the data into the Epi info-3.5.3 version (Centers for Disease Control, Atlanta, Georgia, USA) software. We used univariate and multiple logistic regression (MLR) analysis to calculate crude and adjusted ORs (AORs), and their 95% confidence intervals (CIs) for factors associated with deaths. We calculated the population-attributable fraction (PAF) to estimate the proportion of cases in the population attributable to the various factors.
Human subject protection
The study was initiated after institutional ethics committee approval. We obtained written informed consent from the patients or their legally acceptable representatives. In the case of children aged 7-18 years, we obtained verbal assent and written informed consent from their legal guardians. The procedures for human subject protection were in accordance with the ICMR ethical guidelines for biomedical research on human participants.
During the calendar year 2010, the 13 laboratories tested throat or nasal swabs from 9630 clinically suspected patients. Of these, 1382 were due to influenza A (H1N1). We included 1302 (94%) patients who were residents of Tamil Nadu in the descriptive analysis.
Age and gender distribution
The median age of the patients was 26 years (range: 1-82 years). About two-third of the cases were aged between 15 and 59 years and 51% were female [Table 1]. Seventy-two patients died with a case fatality ratio of 0.7% among the suspected patients and 5.5% among the laboratory-confirmed patients.
Fever (96%) and cough (79%) were the commonest presenting symptoms among the laboratory-confirmed patients. A total of 12% patients had breathlessness [Table 1]. Sixty-one (85%) of the 72 patients who died had breathlessness. Sixty (83%) patients had clinical presentation suggestive of secondary bacterial pneumonia, whereas 51 (71%) had radiological evidence of pneumonia [Table 2].
Treatment details and outcome
Sixty-four percent (n0=828) of the patients sought treatment from private hospitals, whereas 474 (36%) were treated at the public healthcare facilities. Of the 72 patients who died, 64 (89%) were admitted in intensive care units, 60 (83%) received ventilator support and 50 (70%) were treated with corticosteroids [Table 2]. All fatal cases were treated with Oseltamivir; 34% received up to two of the 10 recommended doses, 45% received 3-10 doses, whereas the remaining 21% received >10 doses. A total of 65 (90%) of the 72 patients who died received treatment from private hospitals; however, 11 patients receiving treatment from private hospitals were transferred to public healthcare facilities prior to death.
Distribution of cases by month of occurrence
The cases of influenza A (H1N1) were reported from all months of 2010. The number of cases reported started increasing from June, peaked in September and then declined during December 2010 [Figure 1].
Distribution of cases by district
The cases of A (H1N1) were reported from 29 of the 32 districts. Sixty-eight percent of (888/1302) of the cases were residents of Chennai (29%), Vellore (20%) or Coimbatore (19%) [Figure 2]. Additionally, 234 (18%) patients from neighbouring districts also sought treatment from health facilities in Chennai, Vellore or Coimbatore. These three districts together accounted for 29 (40%) of the total deaths.
We included 70 patients who died during July 2010-December 2010 and 210 laboratory-confirmed patients who had the illness during this period and recovered in the case-control study. All the cases and 57% of the controls were hospitalised. On univariate analysis, the odds of fatal outcome was higher among cases who were diabetic (OR=3.7, 95% CI=1.9-7.2), reportedly obese (OR=2.7, 95% CI=1.4-5.5), received treatment from private hospitals (OR=9.9, 95% CI=4.1-23.8) and those who were treated with corticosteroids (OR=23.8, 95% CI=11.9-47.5) [Table 3]. Such patients were also more likely to have consulted >1 health facility before laboratory confirmation (OR=11.4, 95% CI=5.4-23.7) and received antivirals after 48 h of onset of symptoms (OR=25.5, 95% CI=6.1-106.8) [Table 3]. On stratified analysis, the odds of death associated with treatment with corticosteroids were not different among patients attending private or public health facilities (OR=20.4, 95% CI=9.2-45.5 and OR=13.4, 95% CI=2.27-79.2, respectively; χ2 =0.177, P=0.67). The guidelines of the Ministry of Health and Family Welfare for case categorisation and management were followed up among 58% of patients treated in private and 90% in public health facilities.
On multivariate analysis, past history of diabetes (AOR=4, 95% CI=1.2-13), treatment in private hospitals (AOR=7.8, 95% CI=2.4-25), treatment with corticosteroids during illness (AOR=17.5, 95% CI=6.8-45), visit to >1 health facility before laboratory confirmation (AOR=10.2, 95% CI=3.1-33) and delay of >48 h in starting of antivirals (AOR=18.5, 95% CI=3.7-92.6) were independently associated with death [Table 4].
In Tamil Nadu, most of the cases during the second wave of influenza pandemic were aged 15-59 years and mostly from three districts of the state. About 6% of the laboratory-confirmed patients died. History of diabetes and certain treatment-related factors such as treatment from private hospitals, treatment with corticosteroids and delay in starting antivirals were independent predictors of deaths.
The age distribution of influenza cases, as well as the case fatality ratio observed in our study, were similar to those reported in other studies describing the epidemiology of the disease. ,, About two-thirds of the laboratory-confirmed cases and 40% of the deaths were among patients residing in three cities of Tamil Nadu. This may be on account of the fact that 10 of the 13 laboratories designated by the state health department for diagnosis of pandemic influenza were located in these three cities. Further, these cities have several tertiary care hospitals that serve as referral centres for the neighbouring districts.
Most of the cases caused by the influenza A (H1N1) virus have been mild and self-limiting in nature, with higher risk of adverse outcome among certain risk groups.  In Tamil Nadu, 46% of the laboratory-confirmed patients and 61% of the fatal cases had at least one of these known risk factors. However, presence of diabetes was the only known risk factor that was found to be associated with increased risk of death. In a cohort study among 1479 laboratory-confirmed and hospitalised cases of pandemic A (H1N1) influenza in Canada, risk of severe outcome was found to be highest among patients with diabetes.  Similar findings were also reported by other investigators. 
Early therapy with Oseltamivir has been found to reduce the duration of hospitalisation and the risk of progression to severe disease requiring admission to intensive care unit or resulting in death. , According to the Government of India's guidelines for categorisation of influenza A H1N1 cases for home isolation, testing, treatment and hospitalisation, patients with milder symptoms should be isolated at home (category A); patients with influenza-like illness with known risk factors or high-grade fever are treated with Oseltamivir with home isolation (category B); whereas patients with severe symptoms such as breathlessness, chest pain, drowsiness, fall in blood pressure, sputum mixed with blood, bluish discolouration of nails, etc., should be hospitalised and treated with Oseltamivir (category C).  In Tamil Nadu, all laboratory-confirmed patients were given antivirals irrespective of the severity of illness, but such therapy was started after 48 h in the majority of the cases who died. Delay in initiating treatment was an independent predictor of mortality among the influenza patients, accounting for 91% of the deaths in the state. On account of this delay, 79% of the cases who died received antivirals for less than 5 days. The association between visits to >1 health facility before laboratory confirmation and death also indicates the delay in the starting antivirals.
Corticosteroids are occasionally used as an adjunctive therapy for treatment of acute respiratory distress syndrome (ARDS) in severe influenza. , Observational studies among the patients of influenza A (H1N1), however, showed that corticosteroid treatment was associated with a higher likelihood of admission to an intensive care unit, death as clinical outcome, slower viral clearance, as well as longer duration of viral shedding. , The WHO guidelines for pharmacological management of pandemic influenza strongly recommend that patients having severe illness, including viral pneumonitis, respiratory failure and ARDS due to influenza virus infection, should not be given systemic corticosteroids unless indicated for other reasons.  In Tamil Nadu, use of corticosteroid was more among patients who died, and such treatment was associated with increased risk of death.
Treatment in private hospitals was associated with increased risk of death with a PAF of 79%. Several factors could explain this finding: (1) The majority (64%) of the influenza cases in the state received treatment from private hospitals; (2) about two-thirds of the patients treated at private hospitals received antivirals after 48 h of the onset of symptoms, which was an independent risk factor for death; (3) the national protocol for categorisation of influenza cases was followed for 58% of the patients treated at the private hospitals, as compared with 90% of the patients treated at the public hospitals. Sensitising the health providers in private hospitals could substantially reduce the number of deaths due to influenza in the subsequent seasons of transmission.
Our study had certain limitations. First, we included laboratory-confirmed patients reporting to the state IDSP unit in the descriptive study. Ten of the 13 designated laboratories were in three districts; there may be under-reporting from other districts due to non-availability of laboratory facilities. Second, the line-list maintained at the district IDSP units did not have information about case categorisation, as well as details of co-morbid conditions, including their BMI. Third, all cases, and 57% of the controls included in the analytical study, were hospitalised. We did not use hospitalisation as a criterion for selection of cases and controls as no uniform criteria was adopted by the hospitals for hospitalisation of A (H1N1) patients. Inclusion of non-hospitalised controls however is likely to result in biased estimates. To address this, we conducted the analysis by including only hospitalised patients as controls. The risk factors identified were similar to the analysis that included hospitalised as well as non-hospitalised controls except past history of diabetes [Table 5]. In conclusion, factors such as previous history of diabetes, treatment with corticosteroids during illness and delay in starting antivirals were associated with fatal outcome. The protocol for categorisation and management of influenza patients was not strictly followed in the private hospitals. Based on these findings, we propose recommendations to reduce case fatality during subsequent waves of influenza in the state. The treating physicians need to be informed about higher risk of adverse outcomes among influenza patients having a history of diabetes. They also need to be educated about adherence to the Health Ministry's protocol for categorisation of influenza cases and prompt initiation of treatment for category B and C cases with antivirals, and avoid the use of systemic corticosteroids in the management of influenza cases.
[Figure 1], [Figure 2]
[Table 1], [Table 2], [Table 3], [Table 4], [Table 5]