We are using the hospital data to conduct a case-control study where our case will be children visited hospital due to respiratory diseases and control will be children visited hospital due to gastrointestinal diseases or injury. However, as respiratory diseases were accounted for the majority of the outpatient visits among children, our case # is 5 times higher than our control #.
As there are quite a lot of articles talking about selecting multiple control for cases, there is no detailed discussion about biases that may raise when # of control is less than cases.
Will there be any bias or is it just a power issue? Thank you very much and I am looking forward to any suggestion or comments.
provided the controls are selected independently of the exposure of interest and the diseases in the controls are not caused by the exposure of interest, a smaller number of controls than cases should only lead to loss of efficiency (power),not necessarily selection bias. the controls should also come from the same hospital referral region that appled for the cases..
provided the controls are selected independently of the exposure of interest and the diseases in the controls are not caused by the exposure of interest, a smaller number of controls than cases should only lead to loss of efficiency (power),not necessarily selection bias. the controls should also come from the same hospital referral region that appled for the cases..
I agree that smaller # of controls than cases is more of a power issue. Based on your question, "case" definition seems to be any admission related to any respiratory disease. It is important that controls' conditions are not related to your primary exposure of interest. Do you mean "indoor environment of hospital" is your exposure? If so, you would be better of selecting controls admitted with trauma/injuries that occurred prior to hospital admission.
Yi Lu
Please be very careful in selecting your cases and controls. Your cases should include children with certain respiratory diseases like wheezy chest, pneumonia and other conditions which could assumed to be related to air pollution. Hospital cases are convenient although they may represent the severe ones and the heavily exposed to risk factors. The controls must be children matched for age and sex and free from respiratory diseases and any other disease which is related to air pollution also. In this case the controls suggested by you and by Hlaing ( traumatic cases and appendicitis cases) are good enough but they may be small in number so look for other categories of controls. One suggestion is to use neighborhood controls.
Hi
The role of control group is to estimate the proportion of exposed and unexposed individuals in the base population that gives rise to the cases. Usually, the most appropriate control group is the subpopulation of individuals who are at risk of developing the event under study and from which it can be assured, with relative certainty, that in the theoretical case that they will develop the event or disease under study they would be included in the case group of study population.
There is no problem per se with having less controls than cases. We usually chooses an equal or larger group of controls to increase power to be able to detect even a small odds ratio when exposure is relatively rare. A smaller number of controls than cases suggests that the investigators had more cases than needed given an expected a priori a large relative risk and a high prevalence of exposure among controls for exemple.
My suggestion is that if you have enough number of cases ("statistical power") you can match “the best” control for each case:
- Controls should be representative of the source population that produced the cases.
- Controls must be sampled in a way that is independent of the exposure, meaning that their selection should not be more (or less) likely if they have the exposure of interest
The comments made by Pere are quite relevant with two points:
1. Controls need to be representative of the universe from which cases were drawn
2. There should be no attempt to introduce bias in selection of controls. At the stage of selection of cases and controls, there should be no knowledge of past exposure to proposed risk factors.
Many such situation arises when number of controls can be less than no. of cases. As already suggested their may be small comprise in the power. Nevertheless there is no rule stating that number of controls should always be 1:1. As long as controls come from the same population which has given cases it does not matter.
Every body answered nicely what i suggest is that dont take sick controls as you mentioned gastrointestinal problems, appendicitis,trauma such controls are not reliable because there is some underlying known cause that can not be compared .So cause and effect relationship will not be established and conclusion drawn will be based on bias .As for as number of cases and controls is concerned , Number of the two may vary but percentage will conclude the results
I agree it's mostly a statistical power issue. You should also be aware of something called Berkson's Bias, which is a form of selection bias that can arise when the hospitalized case patients have a higher exposure rate to the respiratory disease-causing agent than controls do, which results in a higher risk of being hospitalized. The idea is that many other subjects in the community, who are not hospitalized, could also have some exposure but not become sick.
- Badges
- Science topic
- Similar topics
- Medicine
- Epi
- Age Groups
- Children