Commentary

David Ogilvie takes you through this paper and explains what it means<

Much public health research can be carried out on sets of routine data, and as long as you know what the data mean, it can be a quick way of answering a research question.

Where did the data come from?

The Guthrie card is used for analysing a blood sample taken from babies at the age of 7 days. The information collected includes the postcode of residence, the age of the mother, and whether or not she is breast feeding. Are these data likely to be accurate? We can assume that the mother's age and the initial letters of the postcode were probably recorded correctly. But could the recording of breast feeding be open to bias? Do women always answer a question about this emotive subject honestly?

How did the authors analyse the data?

The analysis was simple. For each postcode area, and for Scotland as a whole, the authors calculated the proportion of babies who were recorded as being breast fed in two different study years. For most areas, there was an increase in the proportion being breast fed. But was this increase significant? The confidence intervals in the fourth column of the table reflect the degree of uncertainty about the estimates of the size of the change. If we take Scotland as a whole, the estimated increase in breast feeding was 6.4%. The confidence interval tells us that we can be 95% sure that the actual increase was somewhere between 6.0% and 6.8%. This is a narrow range, so we can feel confident that there really was an increase of roughly this size. Compare the figures for Shetland: the estimate for this much smaller population is much less precise. The estimated change is 4.7%, indicating a decrease in the proportion being breast fed, but the confidence interval ranges from -10.3% to + 0.9%. This wide range includes both negative and positive values so we cannot rule out the possibility that the rate actually increased in Shetland.

What about maternal age?

The authors point out that the average age of mothers has increased and that this might explain the increase in breast feeding. They explored this possibility by adjusting for the effect of maternal age in a procedure called standardisation (see box). The comparison of standardised rates shows that the breastfeeding rate would still have increased from 35.6% to 39.4%, even if the age of mothers had not changed.

So what?

It is clearly encouraging that breast feeding became more popular during the 1990s. But here are some other questions raised by the study:

• Why did the breastfeeding rate fall in Aberdeen and Shetland?
• Why is there a twofold difference between the areas with the highest and lowest rates Perth and Motherwell?
• The authors mention the baby friendly hospital initiative. Is what happens in maternity units the key to promoting breast feeding, or are attitudes in the wider community more important? How could you investigate this?
• This paper gives us no idea of whether the breastfeeding rate is higher or lower in Scotland than in other countries. It would have been helpful if the authors gave some indication of this.

Standardisation

We know that the breastfeeding rate in 1997­8 is higher than that in 1990­1. But does this difference still persist after we allow for the fact that the mothers in 1997­8 were older?
Here's what the authors did:

1. Calculate the breastfeeding rate for each maternal age band in the 1997­8 sample (a).
2. Find out the number of babies in each maternal age band in the 1990­1 sample (b).
3. For each age band, multiply (a) by (b) to calculate the number of breastfed babies that there would be in a maternal age band of that size.
4. Add up the number of breastfed babies in all the maternal age bands and divide this by the total number of babies to get the overall breastfeeding rate you would expect to see in 1997­8 if the maternal age structure had not changed since 1990­1.

You will often see standardisation used in studies comparing rates of something between different populations. It allows fairer comparisons to be made between populations with different compositions.

David Ogilvie, specialist registrar in public health medicine, Hamilton,Lanarkshire
Email: david.ogilvie@lanarkshirehb.scot.nhs.uk


studentBMJ 2001;09:261-304 August ISSN 0966-6494