Sophisticated nudging in HIV: combining predictive analytics and behavioural insights

Many of the challenges with HIV care such as low uptake, loss to follow up (LTFU) and poor adherence can be addressed by adapting HIV care and service delivery to what the user needs. As efforts to reach the UNAIDS 95-95-95 goals increase, differentiated models of care have taken centre stage for their potential to improve quality of care and reduce patient costs. HIV services are tailored to the needs of specific populations with the understanding that people are different, and so are their needs and preferences. In addition to tailoring services to specific groups, should we also tailor how we “nudge” individuals to use these services? A recent paper by Stuart Mills details the concept of personalising nudges by personalising choices, or the delivery of a nudge. How do we apply this to HIV service delivery?

 

Combining predictive analytics and behavioural economics can help tailor services and interventions to patient needs (personalise choice). It can also facilitate a more targeted approach to nudging individuals (personalise delivery). These two disciplines, if used well, can improve patient outcomes among PLHIV (people living with HIV). Let’s unpack what this means.

 

Figure 1: Nudging via tailoring of both choices and delivery of nudges. Predictive analytics is used for segmentation and behavioural economics principles are used to design tailored nudges for the different segments/groups

 

Predictive analytics

Predictive analytics uses data from current and past events to make predictions about the future. In HIV care it can be used to identify individuals at risk for specific outcomes like LTFU and failure to be virally suppressed. Identifying these individuals who could benefit from different interventions allows us to match them to models of care that are likely to meet their needs and therefore reduce this risk. While predictive analytics can help us understand who is likely to have undesired outcomes, it is unlikely to give us any understanding of behaviour and how to influence this.

 

Behavioural economics and nudging

Behavioural economics, on the other hand, offers insights into people’s decision making and these insights can be used to influence behaviour through nudges.

Thaler and Sunstein defined a nudge as “any aspect of the choice architecture that alters people’s behaviour in a predictable way without forbidding any options or significantly changing their economic incentives”. The term ‘choice architecture’ refers to the environments within which people make choices; importantly people and the environment in which they make decisions are different. These differences can influence how individuals respond to certain nudges. To achieve a certain behaviour, not all people need nudging and not all people will respond to the same nudge. Whilst it is important to provide the right behavioural interventions, it is not always possible to know which individuals or groups will benefit from a nudge and which strategy will have the largest positive impact.

 

“it is not always possible to know which individuals or group will benefit from a nudge and which strategy will have the largest positive impact”

 

Personalised/tailored nudging

In an ideal world, with no resource constraints, a personalised approach to both HIV care and nudging can be used. A different approach is identifying similar sub-groups and tailoring models of care and the delivery of nudges to these groups. This approach retains the main advantages of personalisation at a lower cost. Combining predictive analytics and behavioural economics insights allows us to use this approach.

 

The benefit of intersecting these two disciplines goes beyond what we have described above. Insights from behavioural economics may be leveraged to nudge providers to efficiently use information from predictive analytics. The full potential of combining these disciplines in the delivery of more customized, patient-centred care is likely much bigger than what we currently know. This approach is however not without challenges and implementers must think of creative ways to overcome them.

 

“A different approach is identifying similar sub-groups and tailoring models of care and the delivery of nudges to these groups.”

 

Challenges of this approach

Large volumes of quality data containing personal information may be required to accurately predict outcomes and to tailor nudges.  In 2020, Protection of Personal Information Act (or POPIA) came into effect in South Africa to protect the personal information of individuals. Whilst usable data for analytics exists, electronic health information systems in South Africa are not universal. Considerations of data availability need to be made and where data is available we must address issues of quality, access and privacy.

 

Cultural differences make it difficult to generalise behavioural insights and this is a concern that behavioural scientists share. Tailored nudging requires a context specific evidence base that explores variations in behavioural tendencies across different cultures and groups. Creating this evidence base remains one of Indlela’s goals.

 

We explore this sophisticated approach to nudging in one of our BITs. Our partners, Right to Care, Palindrome Data and HE2RO have designed a predictive tool to help healthcare providers triage patients according to risk of future LTFU. The BIT leverages insights from behavioural economics to design a tool that nudges recipients of care and providers to translate the information from the predictive tool into actions that can improve outcomes.

 

As HIV care moves towards an emphasis on person-centred care, resource limited settings need to explore options to facilitate “personalisation”. Segmentation using data analytics in combination with behavioural economics offers a way to do this.

 By: Caroline Govathson  Indlela Senior Nudge Associate |09.09.2021


This disclaimer informs readers that the views, thoughts, and opinions expressed in the text belong solely to the author, and not necessarily to the author’s employer, organization, committee or other group or individual.


References

  1. Ford N, Geng E, Ellman T, Orrell C, Ehrenkranz P, Sikazwe I, Jahn A, Rabkin M, Ayisi Addo S, Grimsrud A, Rosen S. Emerging priorities for HIV service delivery. PLoS medicine. 2020 Feb 14;17(2):e1003028.
  2. Available from: https://differentiatedservicedelivery.org/about
  3. MILLS S. Personalized nudging. Behavioural Public Policy. 2020:1-0.
  4. Samson A. The behavioral economics guide 2016 (with an introduction by Gerd Gigerenzer).
  5. Thaler RH, Cass SR. Nudge: Improving Decisions about Health, Wealth and Happiness. 
  6. Sunstein CR. Impersonal default rules vs. active choices vs. personalized default rules: A triptych. Active Choices vs. Personalized Default Rules: A Triptych (May 19, 2013). 2013 May 19.
  7. Maskew M, Sharpey-Schafer K, De Voux L, Bor J, Rennick M, Crompton T, Majuba P, Sanne I, Pisa P, Miot J. Machine learning to predict retention and viral suppression in South African HIV treatment cohorts. medRxiv. 2021 Jan 1.

One thought on “Sophisticated nudging in HIV: combining predictive analytics and behavioural insights

  1. Shawn Malone says:

    We heard about the Palindrome data analytics a few years ago, but at that point the most salient point seemed to be “Someone is at higher risk of treatment interruption if they have previously had a treatment interruption”, which seemed a bit obvious. Will be eager to see what can be done when the analytics are combined with behavioural nudges.

Leave a Reply

Your email address will not be published.