MPH Master pieces

Breast cancer screening

Screening programmes should, as the UK national screening committee puts it, do more good than harm, and at a reasonable cost. Screening for breast cancer has become more controversial recently as some of the harms have been emphasised: here is a recent example.

It is very difficult to keep up with the enormous literature, so this summary is welcome. It comes from a committee set up in the UK to offer an 'independent' view. 'Independent' meant 'has never published on breast screening'. That tends to rule out everyone who knows anything about the subject, so the people selected for this high profile work were very great and very good in their own fields.

I think the key numbers to carry in your head are as follows:

20% reduction in mortality

1 death prevented per 235 women invited for screening

10% - 20% overdiagnosis.

Those are the numbers you need for conversations with your friends. but for written briefings you need much more precision:  what do we mean by 'reduction in mortality', by 'deaths prevented' and by 'overdiagnosis' ? The paper sets all this out with great clarity.

Note also that this is for screening as delivered by the UK programme - three yearly, start at age 50, predominantly ethnic white population. Different parameters, different numbers!

Reducing hospital mortality

A couple of weeks ago I commended a paper from Martin Roland and team. He's been keeping busy: here is another one.

This time the focus is a hospital quality improvement programme. It aimed to improve quality of  care for five conditions, using 28 indicators. This paper looks only at mortality, which is a clear and unarguable end point, but rules out the two low-mortality conditions (hip and knee surgery).

It would of course have been better for science if the initiative had been set up as a randomised comparison. But it wasn't. So as always the problem is finding a suitable comparator. Two types of comparator were used - diseases not targetted in the programme, and hospitals not taking part. There is an interesting section on how the comparator diseases were selected.

Even the largest reduction in mortality was less than 2 percentage points, but as the authors point out, the programme adds up to 890 fewer deaths over 18 months. A 2% is not clinically obvious: sometimes you have to do things which you know will work even if you cannot see the benefit for yourself. But you only know they work if people do studies like this.

There is a nice discussion of monetary incentives, which seem to work in some settings but not in others. My own feeling (sorry - it is just a feeling which I can't reference) is that the money decides what you will focus on, and stay focussed on; it isn't the actual cash reward which is the motivator. There are always easier ways to get money, particularly if you are fund-raising for a hospital!

Mumps

More communicable disease this week - a large outbreak of mumps in an almost fully vaccinated community. 

[I initially posted this with the wrong link but I believe I have now fixed it - thanks to 'Jilla' for pointing this out.]

The authors explain how teaching practices in orthodox Jewish religious schools contributed to the outbreak. This emphasises the importance of knowing the social context of any disease - how people live their lives. And that means getting out and about; what used to be called 'shoe leather epidemiology'.

You'll be pleased to see that the statistical analysis uses nothing more complicated than Fisher's exact test, the chi square and Wilcoxon's rank-sum test. A pleasant relief after the complicated statistical models of some recent blogs!

Headlice

This week a nice little randomised controlled trial. Interesting that the humble head louse gets to star in the mighty New England Journal.

The definitions of infested or not seem rigorous to me - three live lice or no lice after 15 minutes of searching. Must have required some patience from both researcher and subject alike! The account of how ivermectin works made me feel almost sorry for the poor louse.

Notice also how commercial research (this is for a product that will sell) is done these days: the 'data were gathered' by a contract research organisation, and 'copy editing support' i.e. the writing done by 'Word Consulting Group'. Makes you wonder what was left for the 'authors' of the paper to do. The answer is, of course, that they give the pharmaceutical company access to patients.

Free bus passes

The standard way to announce your 60th birthday in England nowadays is letting friends know that you qualify for a free bus pass. But do free bus passes results in people walking more (because they get out and about) or less (because they ride instead of walking)?

Here is an attempt to answer that question, using such data as we have. The key source is a routine dataset - the National Travel Survey. As you'd expect from a government survey, the basic method is pukka: a random household survey. The main limitation is low response (60%) and questionnaire method, though this is enhanced by a 7-day diary.

A different problem is assessing socioeconomic status. This is obviously important in a study of an economic intervention: on the one hand poor people appreciate free travel; on the other hand, as 60 years of the NHS have shown us, rich people are much better at asserting free benefits. But th usual markers (employment, house ownership etc) don't work well in people over 60.

I'd like to see the study repeated in teenagers, who also (in London) now get a free pass. Observation on my way into work suggests they use this gift to travel just one or two stops, to avoid the effort of a 400m walk.