Tuesday, 24 October 2017

What have political interference in research & Harvey Weinstein got in common

I don't suppose anyone will have trouble guessing the answer: neither the existence of sex pests in show business (for heavens sake) nor the existence of political considerations in academic work are anything new.

In the case of abuse in show business, what has changed is the awareness of it as a problem that should not be tolerated. This is a big step forward. Universities, it also turns out, are far from innoocent in this respect. Some senior academics have this weird mixture of power and charisma based on their knowledge and communication skill. In medicine, this "transference" problem is so well known that there are severe sanctions for taking sexual advantage of patients. Even if it might seem like something consensual, the power imbalance makes it abusive. In academe, however, sleeping with your  students seems to be more accepted. At least there is no strict code of ethics, or wasnt the last time I looked. A BSA working group on this issue, 20 or so years ago now, decided not to develop such a code, to the dismay of someone I know with training in medicine who was a member of the working group. This is not to say we thought a lecturer should never develop an emotional relationship with a student, only that it must no take place at the same time as the power relationship of teacher-student. As another friend once said of an academic we both knew "Lets face it dear, if he was the window cleaner you wouldn't look twice".

In the case of academic work, it is to be greatly welcomed that we now have things like "Retraction Watch" working to expose misconduct. And that people are speaking up more about harassment of all kinds. I think this is a similar develoment. It comes at a time when the incentives for misconduct are massively inceasing, but I have remarked on Twitter that it is nothing completely new. So here are some stories about the way things have been over the last 30 years of my career.

In the 1980s there was a lively debate on whether or not unemployment (during the Thatcher recession) was harmful to health. Could it be that sicker or less "fit" people were more likely to become and remain unemployed? A series of 3 studies seemed to show that this was not true, that unemployment itself was harmful.  There were not many holds barred in this competition between what you might say were more versus less politically sophisticated groups and individuals. I was told during this time that one way an academic could get research money was to say "Dear Minister, we can prove that unemployment does not cause ill health or mortality". Other people admitted to have taken up the topic because it made it easier to get published. The topic was a bandwagon. There is more detail on this story in my book "Authorities and Partisans". But the denoument was remarkable: an MRC programme grant that had produced some of these findings was cut from 5 years to 2. That had never happened before and I don't think it ever happened again. This was in 1987.

Most people may not remember that Thatcher commissioned a review of the (then ) Social Science Research Council. The result was a change in title of the UKs major social science funding body to "Economic and Social Research Council" (the ESRC of today). Political interference? Naaahhh.

Some colleagues set out to try & make me understand that a lot of interference in research was not "Political", it was small-p stuff that was done by  those who sat on powerful funding bodies. There were funding bodies, and indeed, journals, that one just did not bother with during the incumbency of certain people because you knew it would be a waste of time. A Scottish Office civil servant admitted that during the Thatcher years there were some eminent Profs who just would not be granted funding from government sources. But other eminent figures would vote down research projects or papers because their own results might be threatened.

When I applied for money to do my own first project on unemployment & health, I had come to a more subtle understanding of this kind of thing. I included a good-ish chunk of text about how, now that we had longitudinal data, we could investigate more closely the selection processes that might produce the appearance of poor health among the unemployed. But silly me! I had not anicipated that there would be assessors of the proposal who were very much outside this ideological camp. One fair minded senior colleague asked to be a referee was a little shocked (he had supervised my PhD, I was mortified). But being fair minded, he did not turn it down. This was around 1993. We got all our papers published, which began to show that unemployment and health needs to be seen not as "either selection or causation" but rather that many unemployed people had had previously more adverse life courses. This work met with very little response, either hostile or friendly. At that time (mid 1990s) Cox regression with time  varying co-ordinates was a bit heavy going for the average reader . More important perhaps was that there was a change of government. When, much later, I went to present some of the results at the Social Exclusion Unit I was asked why all our models adjusted for the height of the parents. I explained this was the only way we had to control for any genetic influences on who became unemployed. The politicans present, to their everlasting credit, were absolutely horrifed at the notion anyone would think unemployment was genetic.

Which brings me to the last story I have energy to talk about today. I am privileged to be involved in a large EU funded project on health inequality during the life course . One of the "work packages" concerns policy implications and policy impact. So the project leader, a distinguished public health academic, who also knows a lot about "omics", invited a participant from a very well known health policy institute. At our last meeting our policy expert was critical of the little genetic work that had yet been done (it will be, it just takes awhile) and of the strong social justice theme in the project overall. "All the policy makers want to hear is about genetic influences" we were told. This story illustrates the importance that has to be given to things like a totally oversimplified view of genetics now in the age of Impact. Admittedly 20 years ago we had to show we had "taken account " of genetic influences in order to get past referees, but our Unemployment & Health project was not assessed on Impact (it got the highest possible grade, which it would not today). And what will happen to the EU projects Impact when it does publish results of proper genetic research which will show miniscule influences? We shall see.

Oh, yes, cant resist one last tale. Someone I had worked with for years was on the programme for a workshop in Manchester on Health INequality, to talk about thge genetics of health inequality. I had been invoted to this but didnt feel like going to Manchester (tilting trains make me sick). So I phoned to ask if he would like me to turn up & give some moral support. "You misunderstand" he responded "I am talking in favour of a genetic effect". I was stunned. "If I get an OR of 1.4 for some social determinant" he explained" I only get that paper in JECH. If I get that same result for a gene, I get it in Nature". Go figure.

Monday, 31 July 2017

What happened to health inequality in England & Wales after 1991?



Someone asked me what has happened to health inequality between social class groups (not areas) between 1991 and the present time. I said I would post something about this, taken from my book "Health Inequality: An Introduction".

To start with, lets have a look at what happened to social class difference in mortality between 1931 an 1991. This is a unique series of data, the likes of which does not exist anywhere else in the world. If you want to know what a Standardized Mortality Ratio is you will need to buy the book :-). You culd look it up i Wikipedia I guess, but my explanation is specially designed for people who don't like algebra formulas (because I don't either). At a very simple level, an SMR of 100 means that group has around the average level of mortality for the whole population of that age and sex. More than 100 is higher mortality (bad) and less than 100 means lower mortality (good).

These figures only refer to England and Wales because of the way in which health statistics and censuses are organised in the United Kingdom.


Table 1: Health  inequality i n  England  and Wales,  1931-1991 : Standardized Mortality Ratios by Registrar-General's Social Class (RGSC) i n men aged 15-64
RGSC
1931
1951
1961
1971
1981 *
1991 *
I: Professional
90
86
76
77
66
66
II: Managerial
1991 Il l routi ne
94
92
81
81
76
72
IIIN: Routine non-
manual (1991)





100







Ill : Routine non manual & skilled manual (1931-1981)

97

101

100

104

103








IIIM: Skilled manual










117
IV semi-skilled manual
102
104
103
114
116
116
V  unskilled manual
111
118
143
137
1 66
189
*ages 20-64
Source: (Wilkinson, 1986) ( 1 986: 2, table I . I ); (Drever, Bunting and Harding, 1997): 98, table 8.2)

As you can see, in 1931 mortality in professional men aged 15-64 was 90, i.e.about 10% lower than the average, and for unskilled manual workers it was around 11% higher than the average for all men. By 1971, the SMR for professoinal men was 77, that is, 23% lower than averege and for unskilled manual men it was 37% higher than the average. This dramatic increase in inequality in mortality was what gave rise to the Black Report on health inequality. However, the publication of the Black Report in 1980 did not prevent further sharp increases in the difference in mortality risk between working age men in different social classes up to 1991.

What happened after 1991? Since then more information on health inequality in England and Wales has appeared, but in a rather different form. The numbers that provide the closest we have to a comparison between the series 1931-1991 and 2001 were provided by the Office for National Statistics for England and Wales and are shown in the next Table, which needs to be read in conjunction with the account of the NS-SEC given in the book. Table 2 also gives standardised death rates, not ratios. This is also explained in the book.

Table 2: Trend in inequality in mortality between 1970s and 2001-2 using old and new social class measurements. England and Wales, men age 25-64. Directly age standardized rates per 100,000
1970-72
1979-83
1991-93
2001-03
2010
RGSC I
500
373
280

RGSC V
897
910
806

Rate Ratio
1.8
2.4
2.9


NS-SeC 1.1

182
128
NS-SeC 7
513
458
Rate Ratio
2.8
2.8
Sources: (White, 2007; Office for National Statistics, 2012)


The second table is obviously very different to the first one. . It has been simplified drastically by only including the most and least advantaged occupational classes at each time point. Like table 1 however, what makes it possible to calculate these figures is having a numerator (numbers who die) taken from official death records, and a denominator taken from Censuses or other official statistics.

What the figures seem to show is that the difference in mortality risk between the most advantaged (RGSC I and NS-SeC 1.1) and the least advantaged (RGSC V and NS-SeC 7) social classes slightly fell and then stabilised to a situation where working age men in the least advantaged social class had around 2.8 times the risk of early death of those in the most advantaged class. We do need to be cautious about these digures of course because both the definition of social class and the sources from which the data have been derived have changed.


Drever, F., Bunting, J. and Harding, D. (1997), Male mortality from major causes of death. In F. Drever and M. Whitehead (eds), Health Inequality, 122-142. London: HMSO.
Office for National Statistics (2012), Intercensal Mortality Rates by NSSEC, 2001-2010. London: Office for National Statistics.
White, C., Glickman, M., Johnson, B. and Corbin, T. (2007), Social inequalities in adult male mortality by the National Statstics Socio-economic classification, England and Wales, 2001-03. Health Statistics Quarterly 36, 6-23.
Wilkinson, R.G. (1986), Income and mortality. In R.G. Wilkinson (ed), Class and health: research and longitudinal data, London: Tavistock.