The Pubmed Index and lessons from ethnicity and stroke
I have started using what I have termed the “Pubmed index” to show trends on a particular topic in medical research. In essence, after an unrestricted search in Pubmed to find how many articles mention my search terms, I look at how many hits I get by restricting the search to 10, 5 and 1 years, to find out how much of that research has been done recently, and how “hot” the topic is. “Ethnicity” and “stroke” gives 2411 references, of which 1719 hits are in the last 10 years, 1067 in the last 5 years, and 182 in the last year. In other words, most of the work on ethnicity and stroke has been done in the last 10 years, and the topic is still being researched.
This makes sense because 10-15 years ago, data from the UK and North America showed that there were differences between black and white populations in stroke. For example, there was an excess of haemorrhagic stroke in black populations. In America, this ethnic group had reduced access to stroke services and higher mortality rates, leading to the so-called “stroke belt” in the South-eastern region of the US. The research base for racial or ethnic disparities in stroke has increased across ethnic groups, including Hispanics and South Asians. More recently, with epidemiologic transition and rising prevalence of chronic diseases, data is showing differences in stroke subtypes and risk factors between black and white populations in sub-Saharan Africa.
Ethnicity research has faced hurdles, partly because inadequate definitions and partly due to concerns about political correctness, and ethnic minorities are still under-represented in clinical trials in stroke. With increasing migration within and between populations, a greater consensus is required regarding future studies of ethnic disparities. More comparisons need to be made between ethnic groups as “immigrant populations” versus the ethnic groups in “home populations”, to consider what the effects of migration on disease in its own right are.
Prospective studies which follow populations over time, like the REGARDS(REasons for Geographic And Racial Differences in Stroke) Study in the US, are a preferred methodology than retrospective studies. There has been surprisingly little research about ethnicity in relation to other “non-modifiable risk factors”, such as family history or genome-wide scans, which will help in characterising how ethnicity is associated with risk factors, behaviour, treatment and disease outcomes.
This week, the BMJ published a study online showing that black patients and patients from higher socioeconomic groups were more likely to be admitted to stroke units than white patients and poorer patients in an area of London. These results are interesting because historically data about ethnic disparities, regardless of the country or the setting, has invariably shown that ethnic minorities have less access to services. A rapid response to the article, rightly concludes that we need to know more accurately what is happening overall in the population, rather than having isolated studies. As unwarranted variations in population medicine are increasingly studied and linked with changes in health policy, proactive, rather than reactive studies of ethnicity are required. In other words, it is not enough to just describe variations and disparities, we need to move towards explanations and potential actions which can reduce disparities, and need better population-wide surveillance.
Variation and inequality-what are the causes?
Yesterday the NHS Atlas of Variation in Healthcare was launched. It aims to “address variations in activity and spend within the NHS” and “search for un-warranted variation”. Unwarranted variation is defined as “Variation in the utilization of health care services that cannot be explained by variation in patient or patient preferences”, and addressing it may “maximise health outcome and minimise inequalities”. The media coverage, as expected, has focused on the shocking “postcode lottery” of NHS healthcare with a 14-fold difference in hip replacement rates and a three- to four-fold variation in the percentage of patients getting the best possible stroke care. Across countries and across disease areas, there has been a flurry of research to show both VARIATION and INEQUALITIES. What do these words mean?
Variation, variability and statistical dispersion are terms often used interchangeably, but they all describe the spread of a variable. Variation can be described using measures such as the standard deviation, the range and the coefficient of variation (CV). For example, the CV is defined as the ratio of the standard deviation to the mean. CV, unlike the standard deviation or the range, does not have units-ie. It is dimensionless.
Variability can occur due to random measurement errors. For example if we assume the outdoor temperature to be fixed, the variation between measurements is due to observational error. With people, such assumptions are false: observed variation is because distinct members of a population differ greatly. For example, the way we measure blood pressure has been called into question by recent research about blood pressure variability
The Longman’s dictionary defines “inequality” as “an unfair situation, in which some groups in society have more money, opportunities, power etc than others”. So “inequalities” are “unwarranted variation”. Probably the most famous recent studies of health inequalities are Sir Michael Marmot’s Whitehall Studies, first started in 1967, showing that men in the lowest employment grades in the civil service were much more likely to die prematurely than men in the highest grades, and led to the study of “socioeconomic inequalities in health”. The WHO set up a Commission for Social Determinants of Health, led by Marmot, which has published several reports on how to address social health inequalities. Another example of health inequalities research is the Global Burden of Disease project of the WHO has studied variations and inequalities in global disease distribution.
But the difficult part is characterising what causes these variations. The Right Care programme, led by Sir Muir Gray, has for the first time attempted to aggregate what the NHS spends on particular groups of disease. Perhaps surprisingly, this list is topped by more than £10bn spent on mental health in England, £7.5bn on circulatory diseases, and £5bn on cancers. The unpacking of this kind of data is where the real inequalities will get tackled.