Inspired by a Twitter exchange with some medical colleagues here is a re-cap of a paper that we published many years ago now on some of the problems that can arise when you try to get an accurate measure of social inequality in ill-health.
The overall term for these in medical sociology is "Illness Behaviour". You can divide this general idea up into 4 more concrete categories
1. The lay referral network. Almost everyone has one of these. It is the bunch of people you talk to when you have a health worry. "Do you think I should take the day off work?" "Should I take vitamins?" "Should I go to the GP/casualty?" "Do I need to call an ambulance?" are some of the questions that family and friends might be asked. We know that people with a more biologically knowledgeable lay referral network are less likely to turn up at A&E unless they really need to. They are also more likely to get medical advice when they do really need it. But there are also more subtle effects. Women whose lay networks are made up more of friends consult less often, those whose networks are more dominated by family consult more often. You get 2 guesses why this might be, answers to @melb4886.
2. The demands of everyday life. People with more arduous jobs will need to seek care (and get themselves defined as "ill") at a level of symptoms that might not affect people in other kinds of jobs. If you are a roofer with vertigo you will feel "ill", if you are a librarian with vertigo you will probably not. So there will be biases in our measures of inequality in this kind of condition. There are plenty of other examples of this, and they have most intriguing implications for measuring social class differences in mortality but I won't go into that just now.
3. The sick role. Some groups of people need to be defined as ill in order to legitimately be excused certain kinds of role obligations (my favorite is "Not tonight, Josephine"). An academic (in my younger days) would not have to get a medical diagnosis and sick certificate in order to work at home for a day with a headache, period pains or a hangover.In some studies this results in people in middle class type jobs appearing to have less illness (and a lower consultation rate) than people in more routine types of jobs. When measuring social differences in consultation, you need to keep in mind that doctors are gatekeepers of the sick role, i.e. legitimate non-performance of various obligations. This varies with the type of obligation, and with its observability and the intensity with which it is policed. This kind of thing is pretty variable in place and time. For example, at UCL for some years now, after 12 days' sickness absence you get called for a disciplinary meeting even with a doctors' certificate. The the role of GPs in signing people off for long term sickness has totally changed in the UK in the past 5 years or so.
4. Quality of the Doctor-Patient relationship. GPs who know their patients and communicate will with them will tend to understand about illness behaviour. Although Phil Strong and others wrote about the ways in which many GPs (and other doctors) communicated less well with working class than middle class patients. In fact, looking back at the classical work on illness behaviour, we probably might not agree with all of it today. It was thought that a good GP screened for high blood pressure for example, on an opportunistic basis, and treated it. This would mean more people would tell a survey that they "had an illness". If GPs took more care of their middle class patients on the other hand this would produce a bias in the class specific rates of reported hypertension.I am not sure everyone is so keen on this kind of screening today, as there is little evidence of long term effectiveness. You could say the same of hypercholesterolaemia. I suspect an example everyone would still agree with would be high blood sugar. Although I have not seen much up to date evidence on this, I would not be surprised if screening surveys showed that more working class people had undiagnosed type 2 diabetes than middle class people.
You can see from these very quick examples that illness behaviour -produced bias does not all go in the same direction. Some factors will make inequalities in illness look bigger than they are and some will make them look smaller. When David Blane, Chris Power and I gave this paper to the RSS about 20 years ago, Derek Cook commented that we had "opened a can of worms that no one wants to think about".