Modern patient segmentation: How to make sure patient interventions succeed

Modern patient segmentation: How to make sure patient interventions succeed

As humans we often think other people have the same beliefs we do. This is why we can be completely baffled when other people – especially those we like – have completely different ways of thinking and make completely different decisions, from ice cream choices (mint choc chip for me, I know it’s controversial) to political views (don’t worry, I won’t get into those here). You only have to look at the clothes we wear to be reminded that humans are all quite different to one another.

When it comes to behavioral interventions, a ‘one size fits all’ approach goes some way in trying to change behavior. Behavioral teams in governments throughout the world aim to nudge whole populations towards a particular desired behavior. This idea is based on Richard Thaler and Cass Sunstein’s popular book Nudge. For example, a government decision to change UK work pension schemes from opt-in to optout increased participation from 61% to 83% according to one study1, showing how this small change effectively nudged a decent proportion of the population towards the desired behavior.

Given that people differ from one another, the ‘one size fits all’ approach isn’t the most effective way to change behavior. As programming and machine learning become more accessible, we can harness these tools to do more than just give each individual the same nudge. The most successful way to change behavior would be to have a different approach for each individual, so that behavior change techniques are designed specifically to the needs of that per son. Of course, this wouldn’t be cost-effective and wouldn’t work in practical terms. But balance the needs for a targeted and cost-effective approach and we land on segmentation – dividing the market into groups, where people in each group are similar to one another in some way, but are distinct from the other groups. The idea is “going with the grain of human nature, rather than rubbing us up the wrong way”2, so working with what there is to get the best effect.

Balance the needs for a targeted and cost-effective approach and we land on segmentation

Grayscale photo of female

In healthcare, segmenting patients can help us create patient support programmes (PSPs) that drive adherence to medication, which can lead to better patient outcomes and increased pharmaceutical sales3. Segmentations of HCPs can also lead to these outcomes, but we’ll focus on patients here. There are different approaches to segmentations, where patients can be grouped based on different criteria, such as:

  • Clinical – splitting patients into groups based on clinically-relevant characteristics such as blood pressure, age, comorbidities etc.
  • Sociodemographic or socioeconomic – splitting patients into groups based on characteristics, such as economic status, education background, gender, ethnicity, and so on.
  • Behavioral – splitting patients into groups based on behaviors relevant to the one we want to change.

Which is the best approach for patient segmentations?

If you’re planning a patient intervention of some kind and want to make sure it succeeds, which of these three approaches can help us the most?

Some studies have shown that sociodemographic factors, such as age and gender, affect healthcare behaviors. For example, adherence to heart medication after a heart attack is lower in women than in men and in elderly compared to middle-aged patients4. But we can’t design interventions to change someone from female to male or elderly to middle-aged to make them more likely to take their heart medication. For this reason, clinical, sociodemographic and socioeconomic segmentations are useful for adding context to segments, but beyond this they’re intrinsically limited.

Other researchers used segmentation to understand why some people delay going to their doctor when they have symptoms, while others go straight away5. Age was the biggest factor in explaining delaying going to the doctor – older people were less likely to delay – and higher income, higher education and being female was also associated with less delay. But behavioral and attitudinal factors added even more to this picture, where positive experience with healthcare providers and feeling in control of a situation was also associated with less delay in going to the doctor.

Now we know that these attitudes and behaviors affect health behaviors, we can tailor interventions so that those who feel like they can’t control their situation can receive training to increase their sense of control, so the next time they experience symptoms they feel they can change their situation and they go to the doctor sooner. This is why behavioral segmentation is so important in healthcare.

The beauty of behavioral profiles is that they aren’t just descriptive. We can actually predict future behaviors with these profiles and make actionable recommendations on how to change behaviors. Understanding why, when and how people do things (i.e. collecting enough data) means we can design tailored interventions that are more likely to be effective because we understand the underlying reasons for their behavior.

Data-driven segmentation

In our work on people with autoimmune conditions, our statistical algorithms have shown that there tend to be four profiles of patients. By using data, we can have confidence that these profiles are really there in the real world. There might be slight differences and nuances depending on the particular autoimmune condition, but these four profiles explain a lot of how people with different autoimmune conditions feel and act.

The optimists – believe they will be able to achieve their goals for treatment and motivated to work hard to make this happen, feel in control, positive about their relationship with their HCP. Often referred to as active patients. Highly conscientious, extraverted, agreeable.

The independents – able to manage their health independently of their HCP but this is likely to come from dissatisfaction with their care rather than desire to self-care, may struggle to seek help if needed. Conscientious, not open.

The pessimists – struggling to manage their health condition, feel that treatment isn’t working for them but lack the confidence and motivation to make Modern patient segmentation: How to make sure patient interventions succeed a change. Often referred to as passive patients.Extraverted, neurotic, less conscientious.

The doctor-dependents – reliant on their HCP, feel like they don’t have control over their health and might lack the knowledge or skills to become more involved in their healthcare. Often referred to as avoidant patients. Agreeable, not open.

Research shows that conscientiousness is linked with healthy behaviors such as exercise and eating a healthy diet6, better perceived health7 and reduced mortality8. So if we know the optimists and the independents are conscientious patient groups, they are likely to already be engaging in healthy behaviours. Our task here would be to maintain their conscientiousness through interventions, and to boost the independents slightly to make them highly conscientious.

On the other hand, neuroticism is linked with more impaired functioning9 and having multiple health conditions10, so if we know the pessimists are likely to have these issues, we can build interventions that reduce their neuroticism and encourage them to become more conscientious.

See Rochefort11 for more on how personality and behaviors affect health.

Behavioural segmentation is vital to improve health engagement

Patient health engagement can bring a huge benefit to pharmaceutical companies – increased sales and better patient outcomes. Here we explain why.

Why health engagement?
“As 60 to 70 percent of premature deaths are caused by behaviours that could be changed12, it is essential that patients and the general public become more engaged with adopting positive health behaviours”13. People who are engaged and active in their healthcare understand their role alongside their HCPs, are capable of fulfilling that role and, for those with long-term health conditions, are more likely to manage their health effectively, which leads to better quality outcomes and lower costs for healthcare systems and providers13. This is good news for everyone – the general public, HCPs and pharmaceutical companies.

People become healthier and adhere to medication and HCPs have fewer cost and time restraints. This means that HCPs keep prescribing medication to patients because (1) patients are adhering to them and therefore benefitting
from them and (2) HCPs have more money to spend on medication through healthcare system budgets.

On the flip side, those who have low levels of engagement and activation in their healthcare (which is 25–40% of people14) are likely to feel overwhelmed, have little confidence in their ability to affect their health and often avoid dealing with health concerns completely13. Together, these behaviors are likely to lead to poor outcomes and higher costs, which is bad news for everyone – people become less healthy, HCPs have less time and higher costs and this leaves less money for buying medications from pharmaceutical companies. So thinking about what we can do to shift people from low engagement to high engagement is well worth doing, and we can do this with segmentation.

How does behavioral segmentation improve health engagement?

It can help us increase trust in HCPs
Patients who trust their primary care physician, such as their GP, are more likely to be engaged and active in their own healthcare15. So increasing patient trust in HCPs and strengthening the patient-HCP relationship for people with low engagement could help increase engagement and ultimately improve health.

Patient trust in HCPs is crucial for a positive patient-HCP relationship as it reduces the power imbalance between the two parties and helps patients play a more active role in their healthcare15. Trust in HCPs has also been linked with better self-management16, adherence to treatment17 and better clinical outcomes18. Those who don’t trust their HCP or healthcare are more likely
to suffer from psychological distress19 and self-report poorer health20.

Grayscale Photo of Male

Of course, from these studies we don’t know whether low trust causes poorer outcomes, or poorer outcomes causes low trust, or whether something else causes both. But it seems logical that a stronger and more trusting patient-HCP relationship could foster better health engagement and outcomes for at least some patients. So designing interventions to increase the trust between patients and their HCPs could prove helpful in improving patient outcomes. But how would we go about it? We first need to understand what trustworthiness means to different people, so we need to segment based on attitudes and behaviors related to trust. A study based on 600 people in India did just that21.

People were asked questions about different aspects of trust with their HCPs, eg I trust a doctor if… ‘the doctor listens to me patiently’, ‘the doctor understands my beliefs and practices’. The researchers used hierarchical cluster analysis and factor analysis (if you’re not a statistics person, these are just fancy algorithms) that helped them split the 600 respondents into segments based on how and why they would trust an HCP. The four segments were: comfort-based trust who valued the approachability of their doctors, personal trust who valued a family or community connection with their doctor, emotionally-assessed trust who valued emotional connectedness and HCP similarities to themselves and objectively-assessed trust who valued their HCPs communicating clearly and kindly.

Now we know there are four types of people based on how they trust, we can design four interventions to effectively build more trust between these groups of patients and their HCPs. For example, for the segment that values comfort-based trust, we can coach HCPs on how to come across as more approachable, and for the group that values objectively-assessed trust, we can coach HCPs to explain the illness or treatment clearly and kindly. If these changes increase how trustworthy an HCP seems, then patients might feel they can be more engaged with their health, potentially improving self-management, adherence to treatment and, ultimately, clinical outcomes and pharmaceutical sales.

It can help us improve patient knowledge
ValCronic is a programme in Spain (Health Agency of Valencia and Telefonica) developed to improve integrated care for people with long-term health conditions. Through education and telemonitoring, the aim of the pilot was to reduce complications from long-term conditions22. Patients were segmented and given interventions tailored to their specific needs: those at highest risk of complications were given biometric devices, telemonitoring, education and support, while those at lowest risk were given education alone. At a two-year follow-up on 322 patients on the programme, 86% said the interventions helped them better understand their disease and 84% said it improved their adherence to medication.

It can help us provide the right support for multimorbid patients
A team of researchers segmented people with multiple long-term health conditions (multimorbid patients) based on their health needs and resources23. They measured things like the impact of the health conditions on their daily roles, their health literacy, and their ability to manage their own health. Using similar statistical methods to the HCP trust example earlier, the researchers found that the patients fell into three groups based on their health-related beliefs and behaviors that each had different needs for care and support. Now we know this, we could design three different types of care and support packages (ie three interventions) that meet the needs of each of those three patient segments.

These studies show how segmenting and tailoring interventions can improve health engagement in three different areas: trust, knowledge and support needs, and these are just three examples. Tailoring interventions means the patients get the care, education and support they need most but in a way that is efficient and straightforward for stakeholders in the healthcare system, and beneficial for
pharmaceutical companies.

Tailoring interventions means the patients get the care, education and support they need most but in a way that is efficient and straightforward for stakeholders in the healthcare system, and beneficial for pharmaceutical companies.

Behavioral segmentation maximizes impact

Pharmaceutical companies already do so much to help people cure or manage their diseases, and most have a mandate to provide additional patient support programmes on top of this to both improve patient outcomes and drive sales. In the real world, there isn’t an infinite amount of funding for these patient support programmes, so we have to be strategic to get maximal impact from a fixed budget.

Behavioral segmentation offers a solution to this; segmenting patients based on attitudes and behaviors means we can identify what each segment of patients needs the most in terms of care, education and support and design interventions that provide them with those things. This approach doesn’t waste resources on things that aren’t needed, but focuses on the areas of biggest unmet need, so it uses the budget as effectively as possible. This is how behavioral segmentation allows us to design interventions with real, positive impact for both patients and pharmaceutical companies.

Dr Hannah Betts
Research & Insights Lead

Craig Mills
Group Managing Director

References

  1. Campbell, N. (2013). Automatic enrolment opt out rates: findings from research with large employers. Available online at https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/227039/opt-out-researchlarge-employers-ad_hoc.pdf [accessed 29 Mar 2022].
  2. Dolan, P., Hallsworth, M., Halpern, D., King, D., Vlaev, I. (2010). MINDSPACE: Influencing behaviour through public policy. Institute for Government. Available online at https://www.instituteforgovernment.org.uk/sites/default/files/publications/MINDSPACE.pdf [accessed 29 Mar 2022].
  3. Wild, H. (2012). The economic rationale for adherence in the treatment of type 2 diabetes mellitus. The American journal of managed care, 18:43-48.
  4. Eindhoven, D.C., Hilt, A.D., Zwaan, T.C., Schalij, M.J. and Borleffs, C.J.W. (2018). Age and gender differences in medical adherence after myocardial infarction: Women do not receive optimal treatment–The Netherlands claims database. European journal of preventive cardiology, 25:181-189.
  5. Kannan, V.D. and Veazie, P.J., 2015. Who avoids going to the doctor and why? Audience segmentation analysis for application of message development. Health communication, 30:635-645.
  6. Bogg, T. and Roberts, B.W. (2004). Conscientiousness and health-related behaviors: A meta-analysis of the leading behavioral contributors to mortality. Psychological Bulletin, 130:887–919.
  7. Jokela, M., Batty, G.D., Nyberg, S.T. et al. (2013). Personality and all-cause mortality: Individual-participant meta-analysis of 3,947 deaths in 76,150 adults. American Journal of Epidemiology, 178:667–675.
  8. Luo, J., Roberts, B.W. (2015) Concurrent and longitudinal relations among conscientiousness, stress, and self-perceived physical health. Journal of Research in Personality 59:93–103.
  9. Chapman B, Duberstein P, Lyness JM. (2007a) Personality traits, education, and health-related quality of life among older adult primary care patients. The Journals of Gerontology: Series B, Psychological Sciences and Social Sciences 62:343–352.
  10. Chapman BP, Lyness JM, Duberstein P. (2007b) Personality and medical illness burden among older adults in primary care. Psychosomatic Medicine 69:277–282.
  11. Rochefort, C., Hoerger, M., Turiano, N.A. and Duberstein, P. (019). Big Five personality and health in adults with and without cancer. Journal of Health Psychology, 24:1494-1504.
  12. Schroeder SA (2007). ‘We can do better – improving the health of the American people’. New England Journal of Medicine, vol 357:1221–8.
  13. Hibbard, J. and Gilburt, H. (2014). Supporting people to manage their health: An introduction to patient activation. Available at https://www.kingsfund.org.uk/sites/default/files/field/field_publication_file/supporting-people-manage-healthpatient-activation-may14.pdf [accessed 29 Mar 2022].
  14. Hibbard J.H. and Cunningham P.J. (2008). How engaged are consumers in their health and health care,and why does it matter?. Health System Change Research Briefs, 8:1–9.
  15. Becker E.R. and Roblin D.W. (2008). Translating primary care practice climate into patient activation: the role of patient trust in physician. Med Care, 46:795-805.
  16. Bonds D., Camacho F., Bell R., et al. (2004). The association of patient trust and self-care among patients with diabetes mellitus. BMC Family Practice 5:1–7.
  17. Trachtenberg F., Dugan E., Hall M.A. (2005). How patients trust relates to their involvement in medical care. Journal Family Practice, 54:344–52.
  18. Lee Y.Y., Lin J.L. (2009). The effects of trust in physician on self-efficacy, adherence and diabetes outcomes. Social Science and Medicine, 68:1060–8.
  19. Ahnquist J., Wamala S.P. and Lindstrom M. (2010). What has trust in the health-care system got to do with psychological distress? Analyses from the national Swedish survey of public health. International Journal for Quality in Health Care, 22:250–8.
  20. Mohseni M. and Lindstrom M. (2007). Social capital, trust in the health-care system and self-rated health: the role of access to health care in a population-based study. Social Science and Medicine, 64:1373–83.
  21. Gopichandran V., Chetlapalli S.K. (2013). Factors influencing trust in doctors: a community segmentation strategy for quality improvement in healthcare. BMJ Open, 3. doi: 10.1136/bmjopen-2013-004115
  22. Mira-Solvesa, J.J., Orozco-Beltrána, D., Sánchez-Mollad, M. and Sánchez Garcíae, J.J. (2014). Evaluación de la satisfacción de los pacientes crónicos con los dispositivos de telemedicina y con el resultado de la atención recibida. Programa ValCrònic. Aten Primaria, 46:16-23
  23. Rijken, M. and van der Heide, I. (2019). Identifying subgroups of persons with multimorbidity based on their needs for care and support. BMC Family Practice, 20:1-10.