Severe Respiratory Diseases – Hidden in plain sight

Neglected Generations · Post 1 of 3

The hidden burden behind a bad flu season.

In the week to 16 August 2026, 177 people were admitted to four Auckland hospitals with severe acute respiratory illness (SARI). It was the highest weekly SARI admission rate recorded in Tāmaki Makaurau since 2015.

Influenza was the most commonly detected virus. Emergency departments were under pressure, wards were full, and for a few weeks, respiratory illness gained prominence in the news.

What we could see was a fraction of what was there. It is a consequence of how respiratory disease is counted, and the gap turns out to be wider than most people realise.

There are several quite different ways the burden drops out of view. Sometimes the infection is never identified, because the patient does not match the definition being used, or because no test was done. Sometimes the illness is recorded perfectly well, but the infection’s contribution to it is not. And sometimes the infection has resolved by the time its consequence arrives, so the consequence is filed under an entirely different diagnosis. Influenza, RSV and pneumococcal disease each go missing in their own way. The heart attacks and lost independence that follow them go missing in another way again.

Not all of those 177 admissions were influenza. SARI stands for severe acute respiratory infection, and SARI surveillance counts people admitted with a respiratory illness of some kind. It is a smoke alarm rather than a diagnosis: quick to tell you something is burning, not designed to tell you what.

New Zealand data give some sense of the scale. A modelling study covering 1994 to 2008 estimated that influenza contributed an average of about 2,260 hospitalisations a year across all medical illness, against about 822 a year in the narrower pneumonia-and-influenza category. Both are influenza-attributable estimates, which means they come from modelling how admissions rise and fall with influenza activity rather than from counting discharge records with influenza written on them. Many of those 2,260 admissions were coded as something else. The authors concluded that influenza’s real contribution to hospital admissions was around nine times greater than routine discharge data suggested.

Why the Numbers Miss So Much

Two systems generate most of what we know about respiratory disease in hospitals, and neither was built to measure total burden.

Surveillance tracks disease occurrence and trends; its job is to tell public health teams what is circulating and whether it is going up. Hospital administrative coding does something else again. It records what happened during a clinical episode and supports billing, planning and the running of the system. Both are useful, and neither on its own tells you how much disease there actually is. The gap between the two takes a different form in each of the winter pathogens.

Why doesn’t influenza always look like influenza in an eighty-year-old?

In a healthy forty-year-old it usually does: sudden fever, cough, aches all over. In an eighty-year-old the fever may be blunted or absent altogether, and what brings them in might be confusion, a fall, breathlessness, or a chronic condition that has abruptly worsened. Look for influenza using a description written around a fit adult and you will walk past it in the people most likely to end up seriously unwell.

The CIRN Serious Outcomes Surveillance Network in Canada put a number on this, comparing standard influenza-like-illness criteria against laboratory-confirmed infection in hospitalised adults across several seasons. The measure that matters is sensitivity: of the people who genuinely had influenza, what proportion did the criteria pick up? In patients under 65 the criteria caught 51.1% of confirmed cases. In patients 65 and over, sensitivity fell to 44.6%. More than half the laboratory-confirmed influenza in older hospitalised adults did not meet the definition being used to count influenza.

Coding adds a second layer. A US modelling study asked what happens if you count influenza-associated hospitalisations only where a pneumonia diagnosis code is present. Across adults, that approach missed around 31% of influenza-associated respiratory and circulatory hospitalisations. The precise figure varies by age group and by modelling assumption. The direction does not.

RSV goes missing earlier, at the testing stage

With RSV the problem is not whether the patient fits a definition. It is whether anyone looked. Older adults admitted with a respiratory illness are not routinely tested for RSV, and the tests we do use are less sensitive in adults, where viral loads run lower than in infants. An infection nobody tested for cannot appear in any count.

A global systematic review and modelling study tried to correct for both problems at once, adjusting reported RSV hospitalisation rates in adults 65 and older for how often testing happens and how sensitive those tests are. The unadjusted rate was 157 per 100,000. After adjustment it was 347 per 100,000, more than double. Scaled across high-income countries, that implies roughly 787,000 RSV-associated hospitalisations in a single year.

How big that correction should be depends on the assumptions built into the model, and reasonable people argue about them. The direction is not in dispute. European analyses using entirely different national datasets arrive at the same conclusion by another route.

New Zealand has less to work with here than it should. A scoping review of Australian and New Zealand data found substantial gaps in the local literature. That is not an undercount in itself, but a disease whose local burden has never been properly characterised is an easy one to leave out of the conversation.

Pneumonia is recorded, just not where anyone looks for it

When someone is admitted, the record carries one main diagnosis, formally the most responsible diagnosis, meaning the single condition considered most responsible for the admission and the length of stay. Everything else going on at the time is recorded as secondary. Researchers estimating disease burden usually count only the admissions where their condition of interest is the main diagnosis, because it is the cleanest and most consistent thing to count. Which is a bit like working out what is in someone’s shopping trolley by looking only at the largest item.

A Canadian analysis of a decade of hospital data found that between 55% and 58% of community-acquired pneumonia hospitalisations in adults 50 and older recorded pneumonia as something other than the most responsible diagnosis. These were not trivial admissions. They had longer stays and higher in-hospital mortality than the ones where pneumonia was the headline.

Here the disease has not vanished from the record at all. Its place in the story has.

Pneumococcus then adds a layer of its own. Even when the pneumonia is properly recorded, the organism responsible is frequently never identified, because non-bacteraemic pneumococcal pneumonia is difficult to confirm with routine testing. The pneumonia is real, the admission is real, and the cause stays unattributed. We know this happens. We cannot put a reliable number on how often.

None of which means these systems are failing. They do the job they were built for. The difficulty comes afterwards, when their outputs are read as though they were measurements of how much disease exists.

When the Consequence Is Counted and the Cause Is Not

Everything so far has been about infections that were missed, or recorded without their contribution being recognised. There is a further problem.

An acute infection can set something off that arrives days later under a completely different diagnostic heading. By the time the consequence turns up (say a heart attack), the infection may be resolved, undocumented, or simply beside the point to whoever is coding the new admission. The burden is counted. The cause is not.

Did the influenza cause the heart attack, or do sick people simply get sick?

This is genuinely hard to study. People who catch influenza and people who have heart attacks differ in all sorts of ways, including age, frailty, existing heart disease and smoking, and no amount of statistical adjustment removes that problem entirely.

So the trick is to stop comparing people with each other and start comparing each person with themselves. Take patients who had both a confirmed influenza infection and a heart attack, then ask whether the heart attacks cluster in the days immediately afterwards or scatter evenly across the rest of the year. It is the epidemiological equivalent of checking whether you trip over more often when the hallway light is off, rather than comparing yourself with your neighbour. Because every patient acts as their own comparison, fixed characteristics cannot explain the result. The design is called a self-controlled case series.

Kwong and colleagues applied it to influenza and acute myocardial infarction. Hospitalisation for a heart attack was roughly six times more likely in the seven days after a positive influenza test than during the same patients’ control periods, and after day seven, the elevation vanished. That disappearance matters as much as the spike, because it is what you would expect from a trigger rather than from a group who were simply at higher risk all along. A later study of senior US veterans, using the same design, found much the same thing.

This is strong evidence that influenza can act as a trigger for a heart attack. It is not evidence about any particular patient, and no individual heart attack can be pinned on a preceding infection on the strength of these studies.

The heart attack gets coded as a heart attack. The influenza sitting upstream of it rarely gets mentioned in the same sentence, and almost never in the same statistic.

Nor are cardiac complications some rare curiosity at the edge of the data.

The harm that never gets a diagnosis code

A heart attack is comparatively easy to count. It produces a diagnosis, a procedure, an admission, a code. Loss of function produces none of those things, which is why it slips out of routine statistics.

What it looks like in practice is an older person who is managing at home, goes into the hospital and comes out unable to any more. Someone who could get to the letterbox and now cannot. Someone who was living independently and is discharged into residential care. For a great many people in their eighties, that is the outcome that matters most.

In a Canadian cohort of adults 65 and older hospitalised with influenza or another acute respiratory illness, around one in five survivors still had persistent functional decline a month later, and one in ten met the threshold for catastrophic disability. Comparable patterns follow pneumonia in nursing home residents, RSV hospitalisation in older adults, and invasive pneumococcal disease.

Where that decline is recorded at all, it appears in an aged-care assessment or a discharge summary that says nothing about the respiratory infection a few weeks earlier.

Who Carries the Burden?

So far this has been one question: how much respiratory disease disappears from view? Of the burden that does stay visible, who is most impacted?

In New Zealand the answer has been consistent for a long time. Auckland surveillance running from 2012 to 2023 found the highest influenza-associated hospitalisation rates among adults aged 80 and over, among Māori and Pacific peoples, and among people living in low socioeconomic status areas. Rates climbed sharply after the pandemic across most groups, reaching 196.7 per 100,000 among Pacific adults in 2023.

That is recent data, but the pattern is not new:

• Community-acquired pneumonia. A hospital study in Christchurch and Hamilton found the overall rate was 3.03 times higher among Māori than non-Māori, with significant differences in every 10-year age band from 45 to 74. Pneumococcal pneumonia was 3.23 times higher overall, though the age-specific comparisons there did not reach statistical significance.
• Influenza mortality. National modelling found Māori aged 65 to 79 dying from influenza at 3.6 times the rate of European and Other New Zealanders in the same age band, and Pacific peoples at 2.4 times.
• Deprivation, measured separately. People living in the most deprived areas died at around 1.8 times the rate of those in the least deprived. Ethnicity and deprivation overlap, but they are not the same variable and were not analysed as one.
• Invasive pneumococcal disease. Auckland surveillance found the highest rates among Pacific peoples, at 33 per 100,000.

These findings describe unequal outcomes, not unequal surveillance. Nothing here suggests case-finding works less well for some populations than others. That would be a different claim needing a different study design.

What the datasets do show, consistently across incidence, mortality and invasive disease over roughly two decades, is that Māori, Pacific peoples and people living in the most deprived areas carry a disproportionate share of severe respiratory disease.

Before We Talk About Value

When a health problem turns out to be bigger than the headline numbers suggest, the reflex is to jump to the economics. Is prevention worth what it costs? I will discuss this topic in Part 2, but ‘is it worth it’ cannot be easily answered because the answer depends entirely on what you count as the thing being prevented.

On the evidence above, that includes influenza infections that never met the case definition used to record them, RSV admissions where no test was done, and pneumonia sitting behind another main diagnosis. It also includes heart attacks arriving in the week after an infection that goes unmentioned in the cardiac record, and the loss of independence that follows a bad chest infection and is never traced back to it. And of course, none of it falls evenly across the population.

A cost-effectiveness model that counts only laboratory-confirmed influenza admissions is answering a much smaller question than the one we are actually asking.