Nermina Ferizovic | Health Economics Expert

In a previous blog post, I argued that it would not be appropriate to include value of hope as a value attribute in value assessment frameworks as it can be considered a special case of clinical benefits which accounts for an individual’s risk preferences. In this blog post, I discuss ‘uncertainty’ as a value attribute and argue that ‘uncertainty’ should not be considered a value attribute as it is an inherent component of other attributes.

Defined as “a situation in which something is not known” by the Cambridge English Dictionary, uncertainty features regularly in healthcare decision making discussions. It can refer to multiple elements including, but not limited to, uncertainty on the number of patients to be treated, uncertainty from the chosen model structure, uncertainty around the service capacity to provide the treatment, and uncertainty surrounding the clinical data. There are ways of handling certain uncertainties such as those employed by health economic modellers which include cost-effectiveness planes and cost-effectiveness acceptability curves. The uncertainty that is the focus of this blog post is the uncertainty that is associated with the value of a health intervention.  

Why is this type of uncertainty the focus of this blog post? Because this uncertainty is the only one which could potentially be argued as a value attribute of health interventions due to its potential impact on the value of the intervention. The question here is whether uncertainty is a value attribute of a health intervention in its own right or an inherent component of other value attributes. I will argue that uncertainty is an inherent component of other value attributes meaning that it should not be treated as a separate value attribute with the premise of the argument built on the idea of association. 

Uncertainty is commonly associated with the attributes of clinical benefits (from current treatment and any subsequent treatment), risk of adverse events, and impacts of current treatment on any future treatment options (also known as real option value). The reason why they are associated with uncertainty is that there is ‘reasonable doubt’ around the impact of a health intervention on these attributes. As an example of this, imagine an intervention which has a high level of uncertainty around its expected clinical benefits and now the same intervention which has much lower uncertainty in expected clinical benefits. The latter intervention has higher value as it is deemed likelier of providing a desired health outcome. Other types of uncertainty as mentioned above do not impact the value of a health intervention. 

The uncertainty associated with these attributes can be due to many factors including lack of (robust) studies demonstrating the impact of interventions on these attributes, lack of methods to measure the impact of interventions on these attributes, and the lack of data including long-term real-world data demonstrating (temporal) value. This leads to the first point: as these three factors generate the uncertainty around the value attributes they are thus associated with the respective value attributes resulting in a strong correlation between uncertainty and the respective value attributes. This means that the quantification of uncertainty without consideration of the associated value attribute is meaningless given it is the uncertainty that lends itself to the quantification of the respective attribute. It could be argued that uncertainty should be treated as a special case of value attribute but then how would the contributing value that uncertainty brings to a health intervention be quantified given its association with (other) value attributes?  

To continue this point, uncertainty can be due to a lack of studies, including long-term data, and lack of methods. This means that uncertainty primarily relates to the data on the value attribute and how it is generated and measured with reduced data quality leading to increased uncertainty. That is, while it may be clear that these attributes give the intervention its value, the uncertainty is specifically related to the measurement of these attributes and thereby the quality of the data. If the data were perfect, beyond any reasonable doubt, there would be no uncertainty. This further makes the point that uncertainty is associated with the respective value attributes. 

Not only is uncertainty associated with the respective value attributes, but it is also associated with individual risk. Individuals who are risk averse would not accept high uncertainty and vice versa. Despite risk being individualistic meaning that it cannot be a value attribute, this point reinforces that uncertainty is not only associated with value attributes but with other concepts too. 

In summary, uncertainty is strongly associated with other value attributes due to lack of information and/or an inability to measure the information. This association precludes uncertainty being an independent and quantifiable value attribute. Therefore, uncertainty contributes to the value of an intervention, as part of other attributes, but is not independent of them and should not be included separately in value assessment frameworks.