Today’s society is awash in masses of readily available information or data. This is a result of the ability to digitalize data, manipulate it through computer computations, and make it universally available via the internet. The vast amount of information that exists can be used to increase knowledge. In turn, that knowledge can help improve the ability to increase the likelihood of predicting a future event, even when that ‘event’ is concerning health, one of the more complicated areas to ‘predict’ and yet, probably the most important.
Despite all of our current methods of measuring, the old cliché that “life is uncertain” is as true today as it was in the past. But, knowledge can be used in the form of probability to reduce the “uncertainty about the uncertainty”. Considering the numerous choices people have today to live their lives, the knowledge about the chances (probability) of something happening can help a person have the best chance of getting what they want. At the same time, knowledge can help predict the cost of a decision so that people can determine if the cost is worth the price to get what they want. A gross example of this could be the probability that links cigarette smoking with a poorer quality of health and/or life expectancy. Knowing this can allow an individual to make an informed decision on whether to pick up, or put down their next smoke.
Learning the Language of Statistics
Strangely, in a world where data rules and statistics are used to manipulate people through data, very little attention is given the understanding basic ideas about statistics. Education demands an understanding of English and math but not statistics. Therefore, people suffer from what some authors call a “statistical illiteracy.” Considering how many sources are trying to influence an individual’s decision by using statistical arguments, it seems dangerous not to pay more attention to understanding simple, basic ideas about statistics and what they actually mean. If statistics are used to make a prediction just exactly what does that prediction mean?
Statistical literacy does not require a PhD in statistics. Rather statistical literacy requires a person to acquire a limited ability to understand what statistical statements about healthcare mean. For example, suppose a person has been trying to achieve a pregnancy for over a year, understandably has become frustrated and worried about ever being able to get pregnant, goes online searching for infertility and sees the following headline: Using treatment A increases the chance of getting pregnant by 25%. That seems like a very encouraging treatment so the person asks her physician if she can try the treatment. However, before the physician can answer the question accurately there is much information that is needed.
The first question is whether or not the person is actually infertile: what is the average time to conception for couples just starting to have a family (85% of women who conceive and have a child, i.e. are fertile, do so within the first 6 months of attempting to conceive). A second question assumes that the person is infertile. Does the treatment actually treat the person’s cause of the infertility? To answer that, diagnostic testing must be done. If the diagnostic testing indicates that the treatment might treat the person’s cause of the infertility then a number of questions arise. What type of study was this? How accurate are the predictions. Do the people in the study resemble this person? What are the risks of the treatment? What are the costs of the treatment? Is there a better treatment? Is there an easier, safer treatment? What about the children created through this treatment? What about the person being treated…are there any long-term risks? How often is this treatment used and how old or new is this treatment? Has the person’s doctor used this treatment before and what is that doctor’s personal experience with the treatment?
Statistics are used at each stage of the process to help answer these questions. But, there is no certainty in any of these answers and no such thing as zero risk. Using statistical methods to estimate probability only indicates what might happen and tries to quantitate that. Probability is not; “if this then that.” Rather it is: “if this, then that will happen X% of the time.” The issue of uncertainty arises constantly in medicine in the form of informed consent. Informed consent uses probabilities to educate a person about the risks and benefits of a particular intervention. The rational for this is that people have the right to determine their own healthcare choices. The goal of informed consent is to educate a person about their choices. However, nothing in the informed consent process guarantees a person a certain outcome. For example, suppose a person is considering undergoing a cycle of IVF. The physician has said that there is a 0.3% chance for a complication of either bleeding or infection from the retrieval process. This information only has value prior to the procedure if it helps the person decides to take the risk for the benefit of becoming pregnant. Suppose a person does have one of the complications; for example she develops an infection after the retrieval which requires hospitalization and extended IV antibiotic treatment? Had she known this was going to happen to her, she might not have chosen to do the treatment. Frequently people will say; “if only I had known that, then I would have done something different.” So, while a 0.3% risk of bleeding or infection is reassuring, it does not guarantee that a person will not have the complication. Someone will end up in that 0.3%.
Conceiving your risks
There are certain basic ideas about statistics that are used frequently but which are poorly understood. That being said, everyone can master certain skills in evaluating statistics that are commonly used in today’s medical world. Accepting that all medical issues contain uncertainty and thus risk, what information is important to evaluate the significance of a particular risk? Perhaps a starting point is to define what risk actually is. While it is not customary to talk about pregnancy as a risk, for the consideration here, risk will mean chance of being pregnant and pregnant will mean the birth of a normal healthy child. For example, in IVF, suppose a person is told they have a 20% chance of becoming pregnant. What does pregnant means? Does it mean that the first pregnancy test has a beta-HCG of >10? 25? >100? Does it mean that the second beta-HCG was rising? By how much? Does it mean that there was a gestational sac on the US? Was there a fetus in the gestational sac? Does it mean the pregnancy made it past the first trimester? Or finally does it mean that a child was born? At term? Healthy? Pregnancy rates have used many of these different endpoints as a definition of pregnancy. Most people doing IVF are not interested in anything but the last definition: delivery at term of a healthy child, yet many studies continue to publish pregnancy rates based upon there being a gestational sac in the uterus, a long way from the desired outcome.
The time frame for assessing risk is very important in infertility treatment. Suppose a person is told that they have a 30% chance of becoming pregnant if they do nothing. Does that mean that next month when they have intercourse they have a 30% chance that they will achieve a pregnancy? Or does that mean that they have a 30% chance of achieving a pregnancy if they have well- timed intercourse every month for the next 2 years? So how can this information be useful? Suppose this person is told that she can achieve a 30% pregnancy rate by doing 3 cycles of clomiphene citrate and IUI. Now the person is in the position to determine if the hassle and cost of doing 3 cycles of C-IUI are worth it or would she rather just wait and try to conceive spontaneously? For a woman 20 years old both options are reasonable but for a woman 39 years old, knowing the time frame may allow her to be more aggressive by using the C-IUI and then if not pregnant the window of opportunity for her to use her own eggs may still be open. Because of the effect of age upon fertility, waiting 2 years starting at age 39 may result in missing the chance to use that last good egg.
A very important consideration of risk (pregnancy) is the size of the risk. Suppose a person is told that if they do 3 cycles of C-IUI they have a 25% chance of pregnancy but if they do IVF they have a 65% chance of pregnancy? That might justify doing IVF sooner rather than later. But what if the chances for C-IUI were 25% and the chances using IVF were 30%. That changes significantly the complexion of the decision.
That last consideration about risk is whether the particular risk applies to the person asking the questions. For example, suppose a woman age 40 who has been trying to get pregnant for over a year reads, or is told by her physician, that women have a 20% chance of conceiving every month they try. This is a common fallacy in the field of infertility. The chance of achieving a pregnancy on the next month of trying is related to a number of factors and varies significantly based upon those factors. For example, a young couple just starting to try to conceive may have as high as a 40% chance of conception on their first try. But after trying for six months with no pregnancy, the chance of conceiving on the seventh month may be closer to 2-3%. For the 40 year old who has been trying to conceive for over a year without success, the chance of her conceiving a pregnancy on her next month of trying is <1%.
So applying the concept of risk, the chance of achieving a full-term, healthy, live birth, depends upon many factors. Knowing the limitations and significance of those factors helps a person make decisions that give that person he best chance of achieving the goals they have set for themselves.