Understanding the Difference Between Indication and Diagnosis in Medicines Data

Understanding and using NHS data effectively often requires grappling with technical details that are not widely documented. At the Bennett Institute, we occasionally publish technical explainers to help fill this “missing manual” of NHS data, clarifying concepts that underpin safe, data-driven healthcare improvement. This piece explores the distinction between “indication” and “diagnosis” in health data, terms that are frequently used interchangeably, but which carry important differences in some contexts.

In healthcare, especially within NHS systems, we encounter vast data around medicines, often with terms like “indication” and “diagnosis” used interchangeably when talking about medicines data. While these terms might seem similar, understanding their differences – and when they truly matter – is important for anyone working with medicines data. Currently, there is rarely a choice at the hand of researchers, between indication and diagnosis, but understanding the distinction is important for acknowledging the limitations of the data available. Indeed a lot of the time, neither indication nor diagnosis will be available but that doesn’t stop good work like OpenPrescribing where we only have the medication prescribed.

What are “Indication” and “Diagnosis”?

In simple terms, a diagnosis refers to identifying a condition based on symptoms, tests, and/or clinical expertise. For instance, a diagnosis might label a patient with asthma or diabetes.

Indication, on the other hand, is more specific to the treatment being used. It’s the reason a particular medicine is prescribed. For example, inhalers are prescribed with the indication of treating asthma symptoms, not to “cure” asthma itself.

In the NHS, indications and diagnosis can be recorded in patients’ clinical records. In general practice this is done using a code terminology system SNOMED CT, that represents a clinician’s intended meaning as a code, for example 195967001 asthma (disorder).

How available is Diagnosis data?

When considering Electronic Health Records (EHRs), diagnosis data is often available for research purposes. Although we believe diagnosis is generally well recorded in general practice, some of our recent work identified that up to one-third of patients prescribed an antidepressant do not have a diagnosis of depression or anxiety formally coded in their notes, and this gap is especially prevalent among patients with learning disabilities (manuscript: BMJ Mental Health). It is likely that other indications were recorded in free text, which is often less accessible for data analysis. However, even if this is true, particularly with the passage of time, it can get harder for clinicians to link the medication back to the diagnosis it was intended for.

What about coding of Indication, is it complete?

The National Overprescribing Review highlighted that recording “indication” is not a routine part of NHS clinicians’ workflows, which means that critical details about why a medication was prescribed often go unrecorded. In most cases, this practice does not matter since diagnosis is frequently documented. However, the Review also recommended that national NHS organisations should work closely with software suppliers to embed indication recording more seamlessly into clinical care. This is a move that could improve data accuracy, support safer prescribing​, and prompt deprescribing or prompting a conversation to stop a medication.

Why the distinction doesn’t always matter

For most GP data, the difference between indication and diagnosis isn’t a huge concern. When we prescribe aspirin for heart disease or antibiotics for an infection, the reasons behind these choices are well understood by healthcare professionals. In these cases, the medication’s indication aligns closely with the patient’s diagnosis.

Although it is possible in some EHR systems, diagnosis and indication aren’t always linked very explicitly to individual prescriptions. Therefore from an analytic perspective, it’s often easier to aggregate and analyse medicines data without splitting hairs between indication and diagnosis, as these align in predictable, straightforward ways for many common conditions.

When the distinction between Indication and Diagnosis might matter

There are, however, situations where the distinction between indication and diagnosis might matter. This is not an exhaustive list but some examples include:

1. Polypharmacy¹, co-morbidities and overprescribing risks
   As the National Overprescribing Review noted, a patient having many health conditions can result in overprescribing. This can lead to complex cases where medications are used off-label, or multiple drugs are given to treat overlapping symptoms rather than a clear diagnosis. Here, tracking the specific indications can help us identify and address potential overprescribing.

2. Multi-indication treatments
   Increasingly, medicines are being developed for more than one indication or diagnosis. An example of this are anticoagulants which can be used to prevent strokes in people with an abnormal heart rhythm (atrial fibrillation) or used to treat blood clots like pulmonary embolism. In addition, different indications may require different doses of the same medication. In such cases knowing the indication could be useful in safety audits like we did with PrescQIPP.

3. Off-Label Prescribing
   Some medications are prescribed for indications outside their initial approvals. If a patient has multiple diagnoses and the medicine is being used for something it is not usually used for, then it is helpful to have the indication recorded explicitly. Knowing when and why a clinician has prescribed a medication off-label helps ensure that practices align with both patient needs and regulatory standards.

4. Indication without a diagnosis
   Sometimes medications are prescribed in relation to a ‘situation’ rather than a diagnosis. For example, a elderly patient taking an NSAID long-term might be prescribed a PPI for prophylaxis of NSAID-associated gastric ulcers. In this case an indication exists, but a directly related diagnosis does not. Similar is true for most medications used for primary prevention, such as prescribing of folic acid in women planning pregnancy.

It might be the case that the distinction matters for your analysis but then again it might not - having a diagnosis or other clinical data might be sufficient to work out what indication a patient is receiving the treatment for.

Implications for NHS data and prescribing practice

Understanding the finer points of why we would like to track indication and diagnosis separately can improve how we analyse GP data and make data-driven improvements in prescribing practices. This document is intended as an introduction into quite a technical field. If you have done more work in this space or you think we have missed any big picture details, we would love to hear from you! brian.mackenna@phc.ox.ac.uk