The study was the work of Dr Carrie Reed from the US Centers for Disease Control and Prevention (CDC) in Atlanta, Georgia, and colleagues from the CDC and Harvard School of Public Health in Boston, Massachusetts, and is due to be published in the December print issue of Emerging Infectious Diseases, although an Epub ahead of print version was put online earlier this week.
Reed and colleagues used a modelling approach to work out what the likely prevalence of infection in the overall US population was likely to have been during the spring outbreak, based on the officially reported lab-confirmed and hospitalization cases.
Using this method they estimated there were between 1.8 to 5.7 million cases of swine flu, including between 9,000 and 21,000 hospitalizations, in the US in the period April to 23 July 2009.
In a press briefing this week, Dr Anne Schuchat, director of CDC's National Center for Immunization and Respiratory Diseases, said the study confirmed what the CDC had been saying all along, that the laboratory confirmed reporting was just "the tip of the iceberg".
In their study, Reed and colleagues explain that after human cases of H1N1 2009 swine flu were first identified in the US in April this year, by the end of July there were over 40,000 lab-confirmed cases, representing only a fraction of total cases. However, there are many reasons why this figure does not show the overall disease burden in the population.
" ... not all ill persons seek medical care and have a specimen collected, not all specimens are sent to a public health laboratory for confirmatory testing with reverse transcription-PCR (rapid point-of-care testing can not differentiate pandemic (H1N1) 2009 from other strains), and not all specimens will give positive results because of the timing of collection or the quality of the specimen," explained the authors.
So they designed a simple "probabilistic multiplier model" that adjusted the count of lab-confirmed cases at each of the following steps: seeking medicare, collection of specimens, submission of specimens, lab-detection of the virus, and reporting of confirmed cases. The same approach has been used to work out the underrecognized impact of foodborne illness in the United States, they wrote.
For the analysis they used the fact that up to 23 July 2009, a total of 43,677 laboratory-confirmed cases of pandemic (H1N1) 2009 were reported by the 50 states of the US and the District of Columbia, including 5,009 hospitalizations and 302 deaths.
They based their assumptions in the "multiplier" on what has been observed in other studies, surveys and investigations of (H1N1) 2009 swine flu, for instance one study examined the health seeking behaviour of people with flu-like illness.
They also made other assumptions, such as "given recommendations for testing, patients hospitalized with pandemic (H1N1) 2009 would more likely have been tested and their cases reported than would outpatients".
The model separated hospitalized from non-hospitalized cases. For hospitalized cases they used larger estimates of the proportion of specimens collected, tested, and reported.
Reed and colleagues also adjusted for trend changes over time, such as the fact that early in the pandemic, doctors and health departments were encouraged to collect specimens from all patients with flu-like illness and send them for testing, whereas later the focus shifted to hospitalized patients, so for that period the model used a lower estimate for the proportion of specimens collected from patients with mild illness.
Using the model, these and other assumptions, they estimated that for the period April to July 2009:
- The median multiplier of reported to estimated cases was 79 (ie every reported case of pandemic (H1N1) 2009 probably represented 79 total cases; with a 90 per cent probability that this figure lay between 47 to 148).
- There was a median of 3.0 million symptomatic cases of pandemic (H1N1) 2009 in the US (90 per cent probability range of 1.8 to 5.7 million).
- Every hospitalized case of pandemic (H1N1) 2009 that was reported probably represented a median of 2.7 total hospitalized persons (90 per cent range 1.9 to 4.3).
- This gives a median estimate of 14,000 (range 9,000 - 21,000) hospitalizations and therefore an estimated ratio of hospitalizations to total symptomatic cases of 0.45 per cent (90 per cent probability range of 0.16 to 1.2 per cent).
- The incidence of pandemic (H1N1) 2009 over the first 4 months of the pandemic in the US ranged from a median of 107 per 100,000 in persons aged 65 years and over, to 2,196 per 100,000 in persons aged between 5 and 24 years.
- The incidence of hospitalization was highest in young children under 5 years of age (median 13.0 per 100,000, with 90 per cent probability of being the range 8.8 to 20.2 years).
"We demonstrate that the reported cases of laboratory confirmed pandemic (H1N1) 2009 are likely a substantial underestimation of the total number of actual illnesses that occurred in the community during the spring of 2009."
"We estimate that through July 23, 2009, from 1.8 million to 5.7 million symptomatic cases of pandemic (H1N1) 2009 occurred in the United States, resulting in 9,000 - 21,000 hospitalizations," they added.
They explained that they did not estimate the number of deaths directly from the model, but took into account that among the reports of lab-confirmed cases up to 23 July, the ratio of deaths to hospitalizations was 6 per cent.
They applied this percentage to the number of hospitalizations calculated by the model, which essentially assumed that deaths and hospitalizations were under-reported to the same extent. The result was a median estimate of 800 deaths (90 per cent probability ranging from 550 to 1,300) between April and 23 July.
However, the authors cautioned that this method has several limitations and they are already developing more sophisticated models that take into account for instance illness severity and intensive care unit admissions and deaths.
Models like this are increasingly being used by epidemiologists to estimate population-wide figures of outbreaks ahead of surveillance data, which can take months and even years to collect and collate.
"Using this tool, health officials and policy makers could adjust the model parameters to represent their local experience, which may provide useful estimates of the prevalence of pandemic (H1N1) 2009 in their areas and help plan for a subsequent wave of the epidemic in the fall and winter months of 2009-2010," wrote Reed and colleagues.
A spreadsheet version of the multiplier model that Reed and colleagues used is available on the CDC website at http://www.cdc.gov/h1n1flu/tools.
"Estimates of the prevalence of pandemic (H1N1) 2009, United States, April-July 2009."
Carrie Reed, Frederick J. Angulo, David L. Swerdlow, Marc Lipsitch, Martin I. Meltzer, Daniel Jernigan, and Lyn Finelli.
Emerg Infect Dis (PDF) Dec 2009, [Epub ahead of print, 29 Oct 2009]
Additional sources: CDC Press briefing.