MPH Master pieces

Control of MRSA

Here are two papers about infection control in hospitals.

The first is an impressive account of implementation. In a quasi-military setting (the Veterans Administration), a bundle of packages was mandated in all hospitals. Impressive falls in MRSA followed.The intervention was as follows:

"The bundle consisted of surveillance for nasal colonization with MRSA for all patients admitted to the hospital, all patients transferred from one unit to another within the hospital, and all patients discharged from the hospital; contact precautions for patients who were either colonized or infected with MRSA; hand hygiene; and a change in the institutional culture"

I particularly like the attention paid to institutional culture. Before-and-after comparisons are tricky because so much else can be going on, but in this case the extent of the drop and its timing are pretty convincing. Note also the careful attention to definitions of MRSA colonisation or infection, and to data capture.

The second study used a stronger design (a randomised controlled trial) but a weaker intervention - no signficant changes resulted.

Epidemiology of sleep

Ten years ago I wrote an editorial bemoaning our lack of attention to sleep. It is an important modifiable factor in many accidents, and we know little of its epidemiology. So I have to blog this study!

Exercise: a meta-analysis

I don't normally like meta-analysis because it is so difficult to get a feel for the individual studies  - I worry about garbage in garbage out even if the analysts have applied quality scores. But Today I make an exception: I need to see the wood for the trees. There have been many small studies which tried to get people to exercise more.

Enough metaphors for now. Here is the meta-analysis.Over 350 studies were included (like I say, there have been a lot of them); the authors paid particular attention to heterogeneity i.e. the studies and their results were all different. One key to this is use of the "random effects" model.

Difficult to summarise all those studies but looking at the first and third quartiles of the studies, most were aimed at middle aged women (40 - 50 years old), and consisted of one-hour group sessions, with between 16 and 48 sessions. Overall the answer seems to be that you can increase physical activity, by about 500 extra steps per day - I guess we'd call this modest but useful.

Various other points of interest - using fancy models such as social cognitive theory or the transtheoretical model seemed to make things worse; goal setting, rewards and other behaivoural strategies made things better. The actual amount of exercise during the programme seemed irrelevant:

"Neither the number of intervention strategies
nor the total minutes of intervention content
(including total minutes of supervised physical
activity) was associated with physical activity
outcomes."

This may fit with the theory that self-identity is important to behvaiour change: if you start to think of yourself as 'an exerciser' you will take more exercise; so perhaps participating in a programme achieves that switch in self identity?

BMI aged 17 and coronary disease in middle age

This paper is pretty impressive. It documents the relationship between BMI at age 17 and the chances of coronary heart disease over the next twenty years or so. Plenty of studies have attempted that but in this one (a) height and weight were measured, not self report and (b) the coronary disease was angiography-proven, stenosis of more than 50% in at least one coronary artery. And in case you are worrying that they may have missed one or two cases, (c) every participant had regular treadmill stress tests after their 35th birthday.

You don't get all that, of course, in an ordinary normal population of volunteers - this data was gathered from members of the Israeli Defence Force.

The conclusion was that even slight overweight at age 17 predicted trouble in the coronary arteries in early middle age. The authors make the point that this occurred at a level of BMI considered 'normal'. One of the functions of epidemiology is to give us the full picture of a disease, and as so often happens what we regard as 'normal' is not necessarily healthy.

COPD - overdiagnosis and underdiagnosis

Here are two pieces on chronic obstructive pulmonary disase (COPD). COPD is important because it drives a lot of emergency admissions to hospital, especially in the winter.

In the USA a study of patients over 65 and in the Medicare system showed that some places diagnose a lot of COPD but that does not translate into a lot more COPD deaths. (And likewise for eight other chronic conditions.) More precisely the case fatality ratio drops as the frequency of diganosis increases. This suggests that milder cases are being identified. Tha authors point out that identifying more cases has various economic and other effects - the more diagnoses you record the more you get paid. I guess this unworthy thought arose because the Medicare data analysed were claims for item-of-service fees.

I called this 'overdiagnosis' in the blog title but actually the diagnosis may be appropriate and accurate. At a minimum  it means that a diagnosis of COPD carries a different prognostic import in different places.  Which in turns means that if you are adjusting outcome measures for severity (e.g. co-morbidity) you are not comparing like with like.

The study in England used a different strategy:

Step1: A few years ago, the Health Survey for England included a component which measured lung function - FVC and FEV1. Hence we have an objective measure of lung impairment, which can be used to make an objective diagnosis of COPD (e.g. FEV1 less than 70% of FVC).

Step 2: Analyse for predictors of COPD - age, sex, smoking, ethnicity etc.

Step 3: This allows you, for any given population, and knowing its composition by age / sex/ ethnicity etc, to estimate how many people should have COPD.

Stpe 4 - compare this with the number of GP-diagnosed cases, using data returns in the Quality Outcomes Framework (which incidentally like the Medicare data is linked to payment of doctors, albeit less directly).

Net conclusion, after some unfortunately very complicated spatial statistics: underdiagnosis in London.