Increasing our understanding of the factors affecting the severity of the 2009 A/H1N1 influenza pandemic in different regions of the world could lead to improved clinical practice and mitigation strategies for future influenza pandemics. Even though a number of studies have shed light into the risk factors associated with severe outcomes of 2009 A/H1N1 influenza infections in different populations (e.g., [1–5]), analyses of the case fatality ratio, e.g. the proportion of deaths among all symptomatic infections, spanning multiple pandemic waves, and geographic regions are scarce. Between-country differences in the mortality burden of the 2009 pandemic [6] could be linked to differences in influenza case management, underlying population health, or intrinsic differences in disease transmission. Additional studies elucidating the determinants of disease severity globally are warranted to guide prevention efforts in future influenza pandemics.

In Mexico, the 2009 A/H1N1 influenza pandemic was characterized by a three-wave pattern occurring in the spring, summer, and fall of 2009 with substantial geographical heterogeneity [7]. A recent study suggests that Mexico experienced high excess mortality burden during the 2009 A/H1N1 influenza pandemic relative to other countries [6]. However, an assessment of potential factors that contributed to the relatively high pandemic death toll in Mexico are lacking. Here, we fill this gap by analyzing a large series of laboratory-confirmed A/H1N1 influenza cases, hospitalization, and deaths monitored by the Mexican Social Security medical system during April 1 through December 31, 2009 in Mexico. In particular, we quantify the association between disease severity, hospital admission delays, and neuraminidase inhibitor use by demographic information, pandemic wave, and geographic regions of Mexico.

We have carried out a detailed analysis of the association between case fatality ratio, admission delays, and neuraminidase inhibitor administration patterns in ILI and laboratory-confirmed pandemic influenza A/H1N1 patients. Our data are based on individual-level patient information collected through a prospective influenza surveillance system during Apr-Dec 2009 in Mexico. In our large sample comprising 117,818 patients, the risk of death among A/H1N1 inpatients was significantly associated with admission delay and antiviral treatment, in line with previous studies [2, 11–17]. We also found that age, gender, and geography were significantly associated with risk of death. Our findings suggest significant antiviral effectiveness when administered during the early symptomatic phase (<=2 days) when antiviral treatment is considered clinically meaningful [18–21]. Our results also confirm an increasing case fatality ratio with older age [7, 22, 23]. Finally, our results reveal marked temporal trends in patterns of antiviral use in Mexico, dropping to ~9% by fall 2009 from 50% in earlier pandemic months.

We found the case fatality ratio based on laboratory-confirmed A/H1N1 influenza cases (outpatients and inpatients) to be 2- to 3-fold lower during the period of high antiviral administration during April-July, 2009 compared to the period of low antiviral administration (August-December, 2009). Accordingly, CFR was 1.5 to 1.9 times lower among ILI and A/H1N1 patients treated with antivirals compared to untreated patients [2, 14]. For A/H1N1-positive inpatients with admission delay < =2 days, CFR was twice lower for the treated group than the untreated group, which is in agreement with previous studies [15–17].

We did not find any difference in length of hospital stay by antiviral treatment status. A recent study found that antiviral treatment reduced the length of stay by 18% for hospitalized children with seasonal influenza [21]. No pandemic study is available for comparison with our data.

We did not find evidence of temporal trends in severity of pandemic infections beyond those associated with antiviral use and hospital admission delay, as suggested by the lack of significance of the pandemic wave indicator in our multivariate model. This suggests that there was no meaningful change in severity of circulating pandemic influenza viruses throughout 2009 in Mexico, and resistance to antivirals was not an issue. This is in line with an analysis of the risk of death among hospitalized cases during the summer and fall 2009 A/H1N1 pandemic waves in England [24].

Hospital admission delay was the strongest predictor of death among laboratory-confirmed A/H1N1 influenza inpatients, followed by antiviral treatment, in a multivariate model adjusting for age, gender, and geographic region. Admission delays <2 days significantly decreased risk of death by 55-71% while antiviral treatment reduced risk of death by ~48%, consistent with previous studies [2, 4, 11–17, 25–28].

Our multivariate logistic regression analysis also supported a ~46% (95% CI 21-77%) increased mortality risk of death among A/H1N1 male inpatients compared to females. One study has reported a higher risk of hospitalization among males with pandemic A/H1N1 influenza in South Korea [29]. Our results also indicated that the southeastern region experienced a reduced risk of A/H1N1 inpatient death after controlling for antiviral use and other factors. We do not expect that socio-economic or ethnic differences played a role given that all patients covered through the IMSS health system in Mexico are workers and their families. Overall this suggests that factors beyond admission delays and antiviral treatment could have played a role in the severity of the 2009 A/H1N1 influenza pandemic, including behavioral factors and case management practices. We were not able to ascertain whether the availability of critical case management and intensive care units differed by geographical area or over time.

The most intriguing finding of our study was perhaps the sharp drop in antiviral use from 50% in the spring and summer wave to 9% in the main fall pandemic in Mexico. Antiviral use was similar among inpatients and outpatients in the fall, suggesting that decision to treat was not related to symptoms severity. This pronounced change in clinical practice could be explained in part by the great uncertainty surrounding pandemic severity in the early pandemic stages, with early data suggesting atypically severe disease [30, 31]. By the end of the summer 2009 pandemic wave, it became clear that the severity of the 2009 pandemic virus was comparable to that of contemporaneous seasonal influenza epidemics [32]. Accordingly, antivirals were administered much more conservatively although antiviral availability was not an issue at IMSS facilities. Overall, our findings suggest that higher rates of timely antiviral treatment during the 2009 fall pandemic wave in Mexico could have led to a substantially lower death toll. As the novel pandemic A/H1N1 virus continues to circulate around the world, antiviral treatment should be considered for the great majority of severe ILI patients requiring hospitalization.

Case fatality ratio estimates for the 2009 A/H1N1 influenza pandemic have differed by one to two orders of magnitude between countries [33–40]. Our case-based severity estimates for Mexico are relatively high at 2% for A/H1N1-positive cases and 14% for hospitalized A/H1N1-positive patients. These high estimates likely reflect a bias of the Mexican IMSS influenza surveillance system towards the higher levels of the severity pyramid due to difficulties in identifying asymptomatic or mild cases [35]. In general it is difficult to compare CFR estimates between countries due to differences in patient care and probability of seeking care. However, estimates of case fatality ratio are useful comparative measures of severity across regions of the same country, pandemic waves, age groups, and patterns of neuraminidase inhibitor administration.

Lower CFR estimates in other countries could reflect in part higher rates of neuraminidase inhibitor treatment and prophylaxis, compared to those in Mexico [41–44]. In the US, rates of antiviral treatment among hospitalized A/H1N1 patients remained high throughout the pandemic at 50-82% [4, 45, 46]. In contrast, the distribution of admission delays in our study is generally similar to that in other studies [13, 44, 47, 48]. One study reported a short median admission delay of one day in emergency departments, antiviral treatment rates >99%, and no pandemic-associated death in an upper middle to high socioeconomic population in Chile [42].

Several strengths and caveats of our study are worth noting. We used data on ILI and laboratory-confirmed influenza cases reported to the IMSS, where about one-third of all ILI cases were consistently tested for influenza in all regions and throughout the main pandemic period [7]. We compared the case fatality ratio across regions, pandemic waves, age groups, and patterns of neuraminidase inhibitor administration in Mexico. Of note, there was no evidence of weaker disease surveillance in smaller states, and testing rates for novel A/H1N1 influenza remained stable throughout the pandemic period [7]. Moreover, only a small fraction of the records had missing data. Specifically, 0.2% of all records missing the date of symptoms onset, 4% of inpatient records missing admission delay, and antiviral treatment was missing in 0.1% of records. We also note that the exact timing of start of antiviral treatment was not available. Data on antiviral treatment was available for those patients treated upon initial consultation or admission in hospitals, but were not able to ascertain antiviral treatment for patients well initially and therefore untreated, and deteriorated subsequently during hospitalization. The study patients, who visited the IMSS facilities, were likely to be representative of those patients with more severe disease, given the representativeness of the IMSS system. We note that antiviral treatment patterns among IMSS-affiliated populations may have been slightly higher than among non-IMSS populations, given the slightly higher socio-economic status of IMSS affiliates. Further, due to lack of information on the sensitivity and specificity of the ILI and ARI definitions used in the IMSS system [49], we considered both ILI/ARI and laboratory-confirmed outcomes in our analyses.

Another limitation of our analysis of disease severity was to disregard the impact of underlying chronic conditions. Our study was focused on relative comparisons of disease severity over time, geography, and treatment groups, and we assumed similar underlying conditions in comparison groups. Lastly, a subset of our A/H1N1-positive patients may have had secondary bacterial infection, but no data was available to evaluate the role of bacterial coinfections or antibiotic treatment on severity of A/H1N1 infection [50, 51].

We found an association between the case fatality ratio of the 2009 A/H1N1 influenza pandemic in Mexico and hospital admission delays, neuraminidase inhibitor treatment, demographics, and geography. Our results suggest that differences in antiviral treatment rates and admission delays could partly explain the reported variability in pandemic mortality burden between high and middle-income countries. Severity patterns in low-income regions, especially in Africa, may differ from those reported here and be related to frequency of underlying chronic conditions or access to care. More information on disease severity in low-income regions is warranted. Overall, our study underscores the potential impact of decreasing admission delays and increasing antiviral use in the inpatient setting to mitigate the mortality burden of future influenza pandemics.