MPH Master pieces

Employment and health

Last week we looked at what happened next after a study published in 1995. This time another update, but more complex in its findings, on the relationship between ill health and staying in work. Clear thinking needed!

The original paper showed that, compared to professional and managerial workers, men [sic] whose jobs were semiskilled, personal service, or unskilled manual were more likely to lose their jobs in a recession but also:

"Men with chronic illness in manual occupations were not drawn back into the labour force during the economic recovery of the late 1980s."

A double whammy.

The new study repeats the old one and adds another 16 years of data i.e. covers the period 1973 - 2009. The analysis is actually quite simple, and a series of trend graphs separately for men and women tell the story. The key issue is that 'unemployed' is different from 'economically inactive'. The latter means that people have dropped out of the workforce completely.

The graphs for 'unemployed' show no difference in ups and downs for those who declare long term illness and those who do not. But for 'economically inactive' the rates are higher, and the gap is wider, for those in semi-skilled and unskilled jobs.

Also of interest is the great rise over the past 30 years in employment of women.

The data source for all this is the General Household Survey and its successor the General Lifestyle Survey. The question on limiting long term illness is simple but this study shows the value of consistent, long term data collection. Unfortunate, then, that it was terminated in 2012.

 

 

Teamwork and breast cancer survival

In 1995 Gillis and Hole showed that women with breast cancer survived longer if treated by specialists: a 16% survival advantage compared to women treated by non-specialists. They concluded that

"Survival differences of the magnitude demonstrated have implications for the provision of services for the treatment of women with breast cancer. There is a need to improve equity in the treatment of breast cancer."

A current paper gives the story of  what happened next in Glasgow. Note the careful definition of 'multidisciplinary team' - Definition, definition, definition: without it everything is a fog and you learn nothing!

 

 

Helicopter ambulances

Helicopter ambulances are noisy, shiny and expensive - sexy in fact. So that makes any evaluation hugely emotive and hugely difficult.

Here is a recent study from the USA. The crucial question is whether like has been compared with like. With a seriously injured patient, you are more likely (perhaps) to call for an air ambulance. When hospital is reached, some trauma centres may perform better than others at saving life. The effort is made to sort that out by "propensity matching" and by analysing separately for Level I and Level II trauma centres. Propensity matching is an increasingly used technique which aims to control for severity. The analysis was restricted to severe trauma only. I'm slightly bothered that patients who died before reaching hospital were excluded, because maybe all that flying around may be more likely to cause death in transit - or the reverse: but either way it seems to be that this info should be included.

With such a large study, all differences will be statistically significant - the key question is whether the differences are large enough to be important.

Two final points. Geography matters - the speed of a helicopter is more relevant in the wilds of Nevada than in the suburbs of Surrey. An early evaluation in the UK was pretty downbeat. Part of the reason then was that the speed of the air ambulance was more than offset by huge delays in the hospital. Hopefully now that we are getting trauma care sorted out that is no longer true, but it is only this month that trauma centres have been officially designated (and hence properly staffed and on standby) in the south of England

And when all that is sorted out, there remains the economic question. Helicopters are hugely expensive: would the money have been better spent on putting out more ground ambulances?

Inequalities in sudden infant death

As promised last week, here is another 1 000 000+ data study.

Sudden infant death dropped fast and early in affluent groups, but slower and longer in deprived areas. So an inequality opened up and later closed.

It is quite a long write-up but worth a careful read. Notice the use of record linkage to relate the infant death to characteristics at birth; also note that in Scotland all stillbirths and neonatal deaths are classified with a version of the Wigglesworth classification. Few (no?) other jurisdictions do this, so the study is only possible in Scotland. The Wigglesworth is simple but crude; various sophistications have been proposed if you ask a 'single medically qualified individual' im to code over 50 000 records per annum, simple is probably good.

Hip replacements and registers

With over 1 000 000 events in a database, you can answer some questions with certainty. It isn't a randomised controlled trial or a meta-analysis, but this study showed beyond doubt that stemmed metal-on-metal prostheses fail sooner than other types. Note also that the size of the database enabled the researchers to prove that this was a general design problem not a feature of one notoriously poor performing brand - there was still ample data to analyse after excluding that brand.

And finally: there was a good theoretical reasons for thinking that large, stemmed prostheses would have lower failure ares. Nature thought otherwise. Theory must always be tested by data from the real world.

Next week - another 1 000 000+ database!

Trends in pneumonia and coding

Carpenters need to understand woods - how teak differs from cherry, or oak from balsa. We need to understand routine data - where its strengths lie, where the weaknesses and traps are, what we can use it for and what we can't.

So here is an interesting story about pneumonia. Pneumonia is a particularly squishy disease, the balsa wood of epidemiology (maybe I'm pushing this analogy too far...). You don't get coding problems with diseases as distinctive as, say, leukaemia. You also wouldn't get coding problems with pneumonia in a young person. But in older people with multiple morbidities, it can be very difficult to work out what is going on, and hence to code it. Heart failure likewise.

NRT in pregnancy

Here is a good randomised controlled trial on an important topic. I remember discussing the need for this trial in some committee or other a few years ago so it's nice to see it in print.

One of the key issues with any medication during pregnancy is whether any damage is being done to the foetus. With (hopefully) such a rare event, it requires enormous numbers to be sure that a drug really is safe. But actually, all drugs confer both benefit and risk of harm: in this case the trial is ethical because as the authors say:

"There is a general consensus that nicotine-replacement therapy is probably less harmful than smoking."

 

 

International comparisons

If you are minister for health, I guess it's embarassing to find that your country ranks low in the league table of international comparisons. And such comparisons are, for lobby groups, a great way to bash governments. In England the commonest bashing is our position in the European league for cancer survival.

But here is a nice worked example on perinatal mortality which shows how careful you have to be. There are three key problems.

• It is well known that different countries have different attitudes to extremely small babies who die at birth. Some countries will regard this as a stillbirth (and so counts towards perinatal mortality) whereas other countries will record it as a miscarriage (does not count). The more general problem is a table of data where each row is from a different data system. Here each row is a different country, with its own way of collecting mortality data. Sometimes each row is a hospital, with its own way of collecting activity data. You can only be really confident of a true comparison when you design a single data system which covers many sites (as happens for multicentre trials).
  
• Secondly, it is ridiculous to put a country as populous as the USA in the same table as a tiny country like Malta.
  
• Thirdly, league tables are inherently unstable statistically speaking i.e. a small change in the metric can result in huge leaps up or down the league table. As a covering editorial points out, on one comparison
  

"although the US ranked 11th, its rate ratio did not differ significantly from that of most of the 10 countries that ranked ahead of it"

So next year a small random change - and the US jumps 10 ranks. This problem has been known for years.

Lobby groups love league tables, but we should hate them.

 

 

Latest trends in acute myocardial infarct

The latest paper from the excellent Michael Goldacre and team brings up-to-date information from the world of coronary heart disease. It's still our #1 killer in the UK despite strong declines since about 1980.

Many points of interest:

1. note the distinction between 'mortality' and 'case fatality'; these terms are often used loosely but the distinction is crucial to this paper;

2. note also that the paper focusses very precisely on 'acute myocardial infarction' not some broader, looser category;

3. finally note that this paper uses data on hospital admission which is not true incidence.

The method requires record linkage - a database record of a hospital admission for AMI has to be linked to a record for the same person in a (different) database of death records. Such linking can be done on a probablility basis (if it's the same date of birth and the same postcode it's probably the same person) or through a personal identifier (NHS number). The latter is obviously stronger but often impossible. Probabilistic matching is good enough for many epidemiological uses.

The original record linkage scheme covered only the Oxford region. Back in those days there were only two or three computers in the whole country powerful enough to do the work, and the research team had to borrow processing time on the weather forecasters Cray super-computer. Now it's no big deal to do a matching run for an entire national database - they probably used an iPhone. (Just joking.)

Smoking laws

Laws which restrict certain actions provoke controversy so it's up to us to know the evidence. (Notice I said 'restrict certain actions' not 'curtail certain freedoms' - words matter. As scientists we should try to remain as objective as possible and 'freedom' is one of the least objective words in the dictionary.)

This is low grade evidence but it's an important question. England and Ireland have passed strong smoke-free laws; in the Netherlands the laws are weaker. So that sets up a natural experiment to compare comprehensive versus partial legislation. The data in this paper are from annual surveys of a cohort which (if I have read the report correctly) maintains response rates over 75%.

The statistical handling is complex, with 'generalised estimating equations' to handle the problem of repeat analysis on the same set of people, and also as the authors point out, of the central question:

Of key interest in this study is the period by country interaction, as it allows for formal testing of whether quit attempts and/or success evolved over the two time periods differently in each of the three countries.

Comprehensive legislation also seems to create more support among smokers than partial bans.