It is quite often to see data from cross-sectional studies using the words determinants or predictors when describing relationship between their variables. "Predictor"- implies a time-factor, hence longitudinal, whilst "determinant." implies causation and hence cohort. For cross-sectional studies, the correct descriptor of relationships would be factors "associated with" or "relationship between A nd B". Why has the misuse of these words continued over the last 30 years?
You are spot on and have identified an important terminological flaw in many published epidemiological research publications.
There is need for a guideline about when to these terms in the research.
The reason is that there are 2 different kinds of models, predictive and explanatory which require different kinds of methods. See Galit Shuemele To explain or to predict, Stat science, 2010 available onRG. D Booth
"Predictor" comes from the latin word "prae" - before/in front of and "dicere" - saying/speaking and a "Praedictor" in a narrow sense is a "prophet", therefore, telling the future, allright.
One might say that a predictor tells the future and therefore should only be used in longitudinal designs. BUT In a cross sectional design, in a typical, simple model Y = b0 + b1x1, we would call x1 a predictor, since I can predict which score Y will have, if I insert any value for x1. Isn't that predicting too, if I can tell with some margin of error which value Y will have (or in which range Y is likely to be) if I know something about x1? In my understanding of the word prediction or "Vorhersage" in German, this fits very well although not strictly applied to temporal predictions.
I am not a native speaker, so please correct me if I am wrong or have the wrong impression of the usage in English.
what is the difference between research question and hypothesis
Determinants VS Predictor
I would say that if something is a determinant of a dependent (or outcome) variable, you are making the assumption that the relationship is causal. If you are saying that it is a predictor of a dependent variable, you are not saying that the relationship is causal, just predictive. All causal relationships are predictive. Not all predictive relationships are causal.
For example, height is predicted by gender
People who attempt suicide tend to have recently visited their doctor. Visiting the doctor is therefore predictive of suicide attempts. Suicide attempts are not determined by visiting the doctor - it's not that you go to the doctor and feel so bad that you attempt suicide.
Determinants VS Risk Factors
- Effective disease management requires management of the immediate clinical problem, then of the patient’s risk factors and lastly of the underlying determinants.
- For example, a patient presenting with acute angina requires management of the immediate problem: the pain and the risk of complications, then the risk factors such as smoking and lack of exercise, and then the underlying determinants: poverty, environment, social norms.
- Risk factors are of various types. Some have a direct and probabilistic effect on the likelihood of an adverse outcome, as with not wearing a seatbelt in the event of a motor vehicle collision. Others form part of a complex causal pattern, as with the effect of diet on coronary disease, in which the factor interacts with many others in a complex way we often summarize in a statement of probability. While the notion of risk factors implies a causal relation, some are actually just markers for the true risk factor. Age sometimes works like this: it may not be age per se, but other changes that arise with age, that increase risk. In such instances it may be more appropriate to use the term ‘risk marker’ or ‘risk indicator’.
- Health determinants refer to underlying characteristics of society that ultimately shape the health of individuals and communities. They can be thought of as the causes of the causes of ill health, or as ‘upstream factors’. Be careful, however: the term ‘determinant’ is not intended to imply inevitability or determinism; it comes from the Latin de termine, or “from the end” (i.e., the origin) of the causal chain.
Factors associated: better if used used with exposure and and outcome variables and study designs which could not make clear temporal sequence. When one uses predictors, there is a conviction that exposure variable related with outcome came before that outcome. Determinant should be used with study designs which enable researchers identify cause and effect (analytical).
Sure, we have seen and had been using these terms inappropriately. I think one can use determinants and predictors for analytical studies while factors associated with for descriptive/cross-sectional studies.
When we say that the "x" variable is a determinant of the "y" outcome variable, we are assuming that the relationship between "x" and "y" is causal. If we say that the "x" variable is a predictor of the "y" outcome variable, the relationship between the two variables is not necessarily causal. Just, "x" is the predictor of the "y" variable occurrence. Not all predictors have a causal relationship with the outcome variable, but all independent variables which have a causal relationship with the outcome variable are also the predictor of the outcome variable. Therefore, both the "determinant" and "predictor" words should be used for analytic study, while "factors associated with" should be used for cross-sectional studies(descriptive).
Thank you Doctor Lemessa Oljira for raising this interesting topic; it support us to mind our term in scientific writing. These terms are an issue of debate for several years as the terms are synonyms and can be used interchangeably in some disciplines (e.g. Mathematics /Statistics to describe the independent variables in regression), but have distinct definitions in Epidemiological application. I get significant discussion in this paper "Article Association is not prediction: A landscape of confused repor...
"
In my experience people choose a shared vocabulary and its ease of quickly communicating an idea over the precision of its definition. I also experience words being used differently in a variety of cultures and areas of studies. With due respect to what you're asking; being understood is valued more than my own correctness.
Great learning here, Thank you Abe for raising such issues in discussion for scientific communities for better use of evidence
Maybe, in several cross-sectional studies, involved independent variables (predictor candidates) that were assumed would not change within one or two years: residence, gender, education, work type, etc.
Thank you so much comrades for the wonderful lectures. These terms have been confusing me in deciding which one to use when.
Thank you for bringing such issues to the attention of scientific communities in order for them to make better use of evidence.
- Badges
- Science topic