Measuring the Unmeasurable
How the heck do you measure health?
At first glance it seems pretty overwhelming. That's because something like "How healthy are we?" Is a big, hazy question that's hard to out your finger on. You might as well ask people what love means, or whether they believe in unicorns.
But if you start narrowing down your definitions a bit, it becomes quite a simple task - who do you mean by "we"? "How healthy" in what terms? Whether we have gum disease? Diabetes? Chicken pox? When we're statistically likely to die? In epidemiology we have a series of basic definitions that help us get to grips with the really big questions.
1) Who are you asking about?
In other words, define your population. Who are you asking this question about? Mr Donald's geography class who just got back from a field trip? The entire female population of Iran? All men over 50 who are listed as patients at a medical centre in Bethnal Green? We need to be clear who we want to find out about.
Some populations, like Mr Donald's geography class, will be small. Some, like the women of Iran, are huge. That's part of the variety and scale of epidemiology, and why it's a big topic. Really big. But each end of this spectrum has it's own hassles.
Random things happen. Pure chance. That's life and we have to deal with it. But chance is about more than meeting your partner or having a car accident, it also affects the results of scientific studies. There are ways to calculate how much chance affects a study or experiment, but the fact remains that you'll never get rid if it completely and, what's more, the smaller your population size, the more likely chance is to affect it (we'll talk more about this another time, but trust me for now). So a small population means that observations you get might be down to chance.
On the other hand, a big population might give a more certain result, but it's also really hard to study from a practical point of view. Say you want to find out about gum disease rates among the people of China. what are you going to do, send everyone a questionnaire asking about their oral hygiene practices? Even looking at dental records would be an enormous task, taking either a lot of epidemiologists, a lot of time, or both. In these cases we take a sample of the population - a reasonable size chunk that we try to make as representative of everyone in the population as possible.
2) What do we mean by "health"?
I think of myself as a reasonably healthy woman. I feel OK, have no major ailments apart from the odd cold and haven't been to visit my GP in a few years. My doctor, on the other hand, might look at me, see I'm about 20lbs over the ideal weight for my height and ask about my woeful lack of regular exercise. My doctor might gently suggest that I could be healthier.
Health is big and vague, and will mean different things depending on the circumstances. It might mean a small child who doesn't have cholera, an elderly person in Ukraine who can still live independently in his own home, a woman in the USA who was just declared "clear" of her breast cancer, or a man in Brazil whose exercise regime has lost him 30lbs in the last 6 months and helped him avoid Type 2 diabetes. It's going back to the whole "what is love" and "do you believe in unicorns" thing.
Epidemiologists need to define what it is they're looking for, and how they decide who has it. They will establish a syndrome for a health condition - a collection of symptoms that, when most are there, means you're probably a case of that condition. The trouble is, it's very rare that either you have something or you don't. Most illnesses exist in a spectrum of severity. Take cholera, which we've talked about before. At one end of the spectrum, certainly, you could die horribly like many of the people in John Snow's neighbourhood. But at the other end of the scale, incredibly you might not even notice you had cholera at all. It could just be a mild dose of the runs which you put down to the week old chicken sandwich you had for lunch yesterday (salmonella, the most likely cause of that, actually takes longer to incubate which is a good clue).
"Has this person got X?" might not be the question to ask. It may be more like "How much has this person got X?". A team of epidemiologists might choose to make their own diagnosis, but it can often be more practical to take a doctors's word for it and categorise a population by previous medical diagnosis. But disease diagnoses are constantly shifting as medical knowledge expands. For instance the threshold of blood sugar levels used to diagnose diabetes has dropped in the last 15 years, because we realised that some levels lower than the previous threshold almost always lead to diabetes anyway.
There are many medical textbooks which give definitions of disease, but the most widely used is the World Health Organisation's International Classification of Diseases and Related Health Problems (ICD) now in it's tenth edition. The ICD gives an alpha numerical code system to classify diseases, and currently has 22 disease categories divided into codes with one letter and two numbers. For instance, E10 - E14 cover various forms of diabetes mellitus, and I21 (with suffixes) covers miocardial infarction. The ICD is used worldwide as a handy way to categorise disease for statistics, death certificates and the like.
3) Rates and Proportions - Measuring Disease
There are big, rough-estimate ways of measuring disease when we're talking global health, and smaller more complex ones that work well on smaller populations. We can sit over coffee in an epidemiology conference somewhere talking morbidity and mortality, life expectancy and potential years of life lost, incidence and prevalence. I'll talk about the Big Guns soon, but for now let's take those last two, incidence and prevalence.
Let's get all hypothetical. Imagine I'm the Manager of a nursery school and have, say, 50 children aged 2-4 in my care. One day, 3 of those children don't turn up to nursery because they have chicken pox, and another child is sent home by lunch time because, when changing her nappy, a carer notices she has a rash. It's pretty clear to me that in the next few days this will spread and I'll end up with even less kids in class, and I decide to track it. For Science!.
So I need to work out something called the prevalence proportion, and track the incidence rate of chicken pox in my nursery school for around a month, by which time I'm sincerely hoping the outbreak will have passed. That first day is quite simple. I can work out the prevalence proportion by dividing the number of kids with chicken pox (4) by the total number of kids in the nursery school (50). This will give me 0.08, or 8% of my class off with chicken pox.
Now that's fine and dandy, but it only gives me a picture of what's happening on that one day. What we really need is some sense of a time scale. In fact, we need three things to measure disease - the number of cases, the number of people in the population and an indication of time.
The incidence rate in my nursery school is a much more useful figure, but also a bit more complicated. It I involves figuring out the number of cases in an observed population, divided by the total amount of time they were observed for.
On the face of it straightforward, huh? The "observed population" is the 50 kids attending my nursery school and the time is a month, or 30 days. To get the total amount of "person-days" for our measurement of total observation time, we multiply the 50 kids by 30 days and get 1500 person-days. Person-days seems a reasonable measure for something quick and infectious like chicken pox, but epidemiologists could use person-months, or even person-years when describing something like heart disease, or the spread of disease within a nation.
Still with me? Good.
But wait. If the population at risk is kids attending my nursery, then if they're off sick with chicken pox they're no longer "at risk". They have it. So we need to take them out of the observation as we go, and when they come back, they'll be unlikely to catch it again. There might be a child who has two days off to attend a wedding, another who has a week's holiday, and another few who are off sick with other ailments. All of those children will have their infection risk reduced because they are exposed to the rest of the class less, and that needs to be accounted for in our person-days figure. In reality, the total number of person-days might be a fair bit less if we use the actual daily attendance register. We can total up the number of kids who are "at risk" who go every day, work out the average and work out our person-days from there.
Calculating the incidence rate can be a bit complicated but it's by far the best way for epidemiologists to judge the increase or decrease of a particular illness in a particular population, and it's used widely. And all part of the maths that is such an essential tool.
So while "how healthy are we?" seems like a huge, unanswerable question epidemiologists actually have some pretty good methods of tackling it by using these practical definitions. It's just a small way of turning populations, with all their differences and soap opera personal lives and personal ambitions, into scientifically measurable data - but it's a start. Next post we'll take it to a global scale.