In their paper, Holmgren, Patel, Charles and Adler-Milstein describe various odds ratios. The odds ratios can be interpreted by reference to Table 2 in the paper.
What I found interesting about this paper is that Odds Ratios are being used to describe the properties of a system – specifically whether having an electronic health record is related to meeting the four core domains of interoperability.
The Odds Ratio is more commonly used to investigate relationships where causality is suspected but still unclear – for instance the effects of a treatment in a specific disease.
When the Odds Ratio is being used to investigate the properties of a system – the relationships under investigation relate directly to human thought. The properties of a system flow from policies and procedures.
In one sense if there is a policy – then what is the point in investigating that relationship since we ought to understand that from the policy itself?
The answer I think is that policy implementation is most likely probabilistic in the real world. There are many reasons – individual cases for example – where policy cannot be implemented. The Odds Ratio as a balance of clearly stated probabilities clarifies relationships of interest.
To give an example. If there is a policy which should result in specific behaviours then the probability of those behaviours can be clarified under different circumstances. Let us say that there are circumstances A and B. We could then state the probability of a policy driven behaviour in situation A – p(A) and in situation B – p(B).
The Odds Ratio for the policy driven behaviour in situations A compared to situation B would then be p(A)/p(B).
This approach enables a policy to be explored and understood in the real world, characterised in different scenarios. That knowledge can then be utilised to modify policy according to the context.
Let us suppose for example that the Odds Ratio is quite high for p(A)/p(B). It may then be possible to examine situation A in more detail (e.g. A1 and A2). By calculating another Odds Ratio p(A1)/p(A2) it may then be possible to explain the initial finding p(A)/p(B) and so on in an iterative fashion.
This approach can in turn be adapted for use in audit with appropriate sampling (for practical purposes) to improve outcomes.
What Does the Odds Ratio Mean?
The odds ratios tells us about a relationship between two things. In the clinical setting these two things might be expected to have a causal relationship e.g. Blood Pressure and exercise.
How is it Calculated?
Taking the Blood Pressure example. Let’s suppose that a group of people have their Blood Pressure checked before and after an exercise/non-exercise intervention. Let’s also suppose that people are classed as doing regular exercise (intervention) or not doing regular exercise (non-exercise).
We want to see if there is a relationship between the two and it would be great if we could have a number to sum up that relationship. Let’s also suppose that we think the relationship is ‘Doing regular exercise reduces Blood Pressure’. The exposure is doing regular exercise and the outcome is reduced Blood Pressure.
Let’s suppose there are 100 people in the study (These are illustrative numbers only). 60 people do regular exercise and 50 people have low Blood Pressure. Of the 60 people that do regular exercise 40 people have low Blood Pressure. I will write this out in the following statements
60 people do regular exercise and 40 have reduced Blood Pressure
60 people do regular exercise and 20 have high Blood Pressure
40 people do not do regular exercise and 10 have reduced Blood Pressure
40 people do not do regular exercise and 30 have high Blood Pressure
Before we can talk about the odds ratio we need to look at probabilities and also clarify which relationship we are examining.
Firstly let us look at some probabilities.
What is the probability that someone who is in the exercise intervention group will have reduced Blood Pressure?
This would be 40/60 or 0.67 (2 significant figures). Just to explain – if someone is in the exercise group then we know that 40 have reduced Blood Pressure and 20 have high Blood Pressure. The probability is the ratio of the outcome of interest to all of the outcomes in that group. Therefore this is 40/60.
What is the probability that someone who is in the exercise intervention group will have high Blood Pressure?
The reasoning is much the same as the example above except we are substituting 10 for 40 i.e. the probability is 20/60 or 0.33 (2 significant figures).
What is the probability that someone who is in the non-exercise group will have lower Blood Pressure?
This will be 10/40 or 0.25
What is the probability that someone who is in the non-exercise group will have high Blood Pressure?
This will be 30/40 or 0.75
Now we can turn to the odds ratio which is the ratio of probabilities.
What is the odds of having high Blood Pressure with exercise compared to non-exercise.
The probability of high Blood Pressure given exercise is 0.33.
The probability of high Blood Pressure given non-exercise is 0.75
The odds ratio of having high Blood Pressure with exercise compared to non-exercise is 0.33/0.75 = 0.44
We can have several odds ratios in the same sample – it depends on what question we are asking.
What is the odds of having lower Blood Pressure with exercise compared to non-exercise.
The probability of lower Blood Pressure given exercise is 0.67
The probability of lower Blood Pressure given exercise is 0.25
The odds ratio of having lower Blood Pressure with exercise compared to non-exercise is 0.67/0.25 = 2.68.
So just to summarise the odds ratio is a ratio of probabilities. You need to work out the probabilities first. Then you need to clearly state the relationship you are interested in and then calculate the ratio of the probabilities for that relationship.
Finally just a few key points.
- The odds ratio can be between 0 and infinity. So for example if all the people in the non-exercise intervention group had high Blood Pressure then the odds ratio of having lower Blood Pressure with exercise compared to non-exercise would be infinity (assuming all other figures remained the same).
- The odds ratio can be affected by sample bias. So if the sample is not characteristic of the general population and is small then the odds ratio could be abnormally high or low. The confidence interval is usually calculated for the odds ratio although this can be a little complicated.
There is a good explanation of the odds ratio in this paper.
What are the Four Domains of Interoperability?
The ONC defined the four core domains of interoperability thus:
The ONC is short for the Office of the National Coordinator for Health Information Technology. This has emerged from the American Health Information Technology for Economic and Clinical Health Act (HITECH) and there is a good write-up on this here.
Interoperability is relevant for a Health Information Exchange.
Patient records are central to the delivery of healthcare and serve a number of functions including the recording of clinical assessments and interventions. Aggregated data is also utilised at a local and national level to inform commissioning.
Electronic Patient Records
The digitisation of patient records offers a number of advantages over paper based records. These advantages include automated backup of records, reduced use of physical storage space (since paper based notes are switched to servers), off-site access to records using mobile devices and the potential to develop analytical clinical support tools which use computers to process clinical data to help improve clinical decisions. Not all healthcare services have electronic patient records but most providers are moving in this direction.
Getting Electronic Patient Records to Talk to Each Other
When patients move between healthcare providers – for instance between primary care and the hospital – they may find that one provider does not have information that the other provider has. There are many providers and many electronic paper record systems. For two systems to talk to each other they have to solve a number of problems. When these problems are solved a patient can move between providers and healthcare information can be accessed by the different providers. A key solution to this problem of health information gaps is the Health Information Exchange (HIE).
The Health Information Exchange
There are many definitions of what a Health Information Exchange is. (Hersh et al, 2015) define a HIE as follows:
‘Health information exchange (HIE), the electronic sharing of clinical information across the boundaries of health care organizations’
Whilst this definition is simple, the process of sharing clinical information between healthcare organisations is technically complex and encompasses a range of software, hardware and governance issues. The process of helping systems to talk to each other is helped by the development of standards. A set of standards is outlined in the NHS interoperability framework.
Links to Other Posts in the Health Information Exchange Series
Twitter: You can follow ‘The Amazing World of Psychiatry’ Twitter by clicking on this link.
TAWOP Channel: You can follow the TAWOP Channel on YouTube by clicking on this link.
Responses: If you have any comments, you can leave them below or alternatively e-mail firstname.lastname@example.org.
Disclaimer: The comments made here represent the opinions of the author and do not represent the profession or any body/organisation. The comments made here are not meant as a source of medical advice and those seeking medical advice are advised to consult with their own doctor. The author is not responsible for the contents of any external sites that are linked to in this blog.
Conflicts of Interest: *For potential conflicts of interest please see the About section