Understanding clinical risk scores in 4 days. Day 4: Treatment
Clinical trials tell us about the effects of certain drugs in certain patients in certain clinical situations. Arguably the most practical use of risk scores is to help in deciding which patients are most at risk of disease or have disease requiring most urgent treatment. In other words, risk scores can help answer the question, “Who needs treatment?”
An example we have used before is the CHADS2 score
for risk of stroke in patients with atrial fibrillation. Patients with a CHADS2 score of 2 or more should get warfarin therapy. Risk scores used in this way are easy to incorporate into treatment guidelines and are therefore translated into clinical practice fairly well.
However, we have already seen that risk scores are underused in clinical practice. In cardiovascular disease, to some extent, there are too many risk scores. These scores differ in terms of population, predictors and outcomes, which may not fit with those used by clinicians. Ideally, risk scores should be formulated, tried and tested in populations which have not received any treatment, but often patients in these populations have been treated, partially or fully and this is poorly accounted for by most rules.
Refining and re-defining of risk scores can lead to profound impact on who gets treatment, which can lead to a bigger impact on healthcare resources. For example, the CHADS2 system may not fully exclude risk of stroke in patients with a CHADS2 score of <2. Therefore, the CHA(2)DS(2)-VASc score was formulated to better define the patients with atrial fibrillation who truly are at low risk of stroke and therefore do not need warfarin. However, use of the CHA(2)DS(2)-VASc system does mean that more patients will go on warfarin treatment than with the CHADS2 system.
Another example is the impact of adding social deprivation to cardiovascular risk scores. One such score which incorporates the social gradient in cardiovascular disease is the ASSIGN score. Patients with an ASSIGN risk of 20% or more receive statin treatment. One study calculated that if the ASSIGN risk score was implemented, where a person lives would affect the decision to initiate statin treatment in the case of 15.7% of the population (aged 30-74 years), corresponding to 4.6 million in the UK.
The upshot is that risk scores are very versatile tools which can be used to guide diagnosis, prediction, prognosis and treatment in clinical practice, if the populations in which the scores were formulated and tested match your patient population, and you recognise their limitations.
Understanding clinical risk scores in 4 days. Day 3: Prognosis
Closely related and interlinked to prediction is the idea of prognosis. Literally meaning “foreseeing” in Greek, prognosis attempts to quantify the likely outcome over time for a particular patient in a particular clinical setting. For example, in terms of risk prediction, a CHADS2 score of 2 or more in a patient with AF tells us that he or she is at a large enough risk of stroke to warrant treatment with warfarin to thin the blood. 18.2% of patients with a maximum of CHADS2 score of 6 will have had a stroke at one year: this is prognostic information. Therefore, the same score can be used as both a predictive and a prognostic tool.
One of the most enduring and widely used prognostic scores in clinical practice, is the Apgar score, used to predict outcomes in newborns by a score of 1 to 10. As noted in Apgar’s seminal paper, scores of 0-3 indicate poor prognosis for newborn babies.
The prognostic information doctors and patients are most commonly interested in is survival, especially in the case of cancer. The challenge is individualising risk from epidemiologic data to a particular patient.
As in the previous 2 blogs about diagnosis and prognosis, scores are more relevant and more likely to be used in clinical practice when they are tied in a decision-making strategy, i.e. there is no point in prognostic information (other than telling the patient what their outlook is likely to be), unless the management of the patient is altered. This is nicely summarised in this 10-year old JAMA editorial:
“Medicine is an action-oriented profession in which clinicians want to relieve suffering, rather than just watch its course. Regrettably, most prognostic indices are not accompanied by decision thresholds that convert level of risk into degree of action.”
Understanding clinical risk scores in 4 days. Day 1: Diagnosis
Risk is arguably one of the most central concepts in clinical medicine. Patients want to know their risk of a particular disease or perhaps even death. Clinicians are interested in which patients are at highest risk of disease and complications. Researchers want to establish the way in which new treatments change the risk of disease. Clinical risk scores use available information to predict the risk of a particular clinical scenario. Risk scores cannot replace clinical judgment but they can guide it. There are hundreds of risk scores available to clinicians and many are available online as risk calculators and are widely used by patients. In this 4-day series, I will cover the 4 types of clinical risk score.
LESSON 1: DIAGNOSIS: IS IT DISEASE?
Some diseases are easy to diagnose because they are well-defined. For example, infection with HIV is diagnosed by a positive HIV test. However, not all diseases are so easy to diagnose or have such clearly defined criteria. Lack of diagnostic certainty can lead to inappropriate treatment and can even lead to abuse of the social benefits system by some unscrupulous parents in the case of attention deficit hyperactivity disorder(ADHD)
Diagnostic scores can be helpful in identifying disease, but also in describing severity of disease. In depression, the Beck’s depression inventory (BDI) is a score which describes the severity of depression on a continuous scale, with 0–9 indicating minimal depression, 10–18: mild depression, 19–29: moderate depression and 30–63 indicating severe depression. A score of at least 10 on the Beck Depression Inventory is the generally accepted threshold for the indication of possible depression.
The mini-mental test score (MMTS) is the most commonly used diagnostic test to assess cognitive function. Scores of 25-30 out of 30 are considered normal. NICE classify 21-24 as mild, 10-20 as moderate and <10 as severe impairment.
Diagnostic scores must be tested and validated in well-defined populations because the scores may perform differently in different patient groups. For example, the BDI above is good at diagnosing depression in post-stroke patients.
The term “clinical prediction rules” is often used to describe the use of clinical findings (history, physical examination, and test results) to make a diagnosis, OR to predict an outcome. This article from Annals in Internal Medicine sums up what you need to know about diagnostic scores before you use them on your patients.