The Digital Nudge in Social Security Administration

This is an abridged version of our full research paper, published in the journal International Social Security Review, which can be read in full for free here.

There are significant trends occurring across three major pillars of public administration, namely social investment (policy), nudge (process) and predictive analytics (technology).

The European Commission defines social investment as policies designed to strengthen people’s skills and capacities, supporting them to participate fully in employment and social life (EC, 2015). Key policy areas include education, quality childcare, healthcare, training, job-search assistance and rehabilitation. Active labour market policies to encourage investment in training and re-skilling are other examples of social investment, which is now a feature of most unemployment insurance and benefit schemes around the world.

Within technology, predictive analytics is a specific field of analytics where data is analysed to determine patterns based on pre‐existing data to predict future outcomes and trends. As more data becomes available, the ability to forecast what might happen in the future at the individual level, within an acceptable degree of confidence, provides a new dimension for social investment decision making.

Regarding process, a trend emerging is the widespread adoption of ‘nudges’. Nudge theory recognizes a disconnect between intentions and behaviour as a result of influences from natural human tendencies including biases, heuristics, and emotion (Knoll, 2010). A nudge is a method that leverages these human tendencies to influence individuals to make better decisions that lead to better social outcomes. For example, government agencies have run many successful trials with simple textual nudges designed to positively influence behaviours such as tax compliance, voter registration and student attrition.

At the intersection of these trends is a concept we term digital nudge, which are nudges facilitated by information technology to achieve a social policy outcome. Digital nudges could potentially assist in the nudge development process at many points: for example, in identifying where nudges are most needed or in delivering a nudge through the personalization of a website experience. These digital nudges are proposed as enablers of better outcomes within a social investment model. Importantly, the person receives the nudge in the context of a service plan towards achieving a desired outcome they have agreed to.

People experiencing social risk can fall into a vicious cycle of ongoing disadvantage as one event, such as losing a job, leads to another, the inability to pay their rent and so on. People can become trapped inside the social security system and despite best intentions and effort they can become socially and economically excluded. A digital nudge, leveraging nudge theory and predictive analytics, provides an opportunity to assist people make better life choices leading to a virtuous circle scenario where a good outcome, such as securing stable housing, delivers the capacity to hold onto a job leading to a pathway towards social and economic independence.

When applying a digital nudge, additional consideration of public concerns regarding ethics and privacy are crucial, as implementation occurs in a dynamic manner at an individual level rather than a cohort level. The use of data and personal information to drive the nudge process needs to be managed in a way that ensures individual rights are protected. This requirement has to be reconciled with the broader societal interest towards influencing outcomes within the boundaries of the finance available for social programmes.

Digital nudge cases

Our understanding of digital nudges, predictive analytics and social security is illustrated with two sets of case studies. The two cases aim to reveal how organizations covering different social domains implemented components of digital nudge processes to achieve societal outcomes. The first case concerns reductions in hospital readmission rates, while the second shows the use of advanced analytics to improve insight into the needs of pensioners – allowing targeted marketing to encourage pensioners to switch to lower cost channels and to claim full entitlements.

Reduction in hospital readmissions

The United States (U.S.) Centre for Medicare and Medicaid Services (CMS) defines a readmission as an admission to a hospital within 30 days of a discharge from the same or another hospital. The U.S. Social Security Act requires hospitals to take appropriate steps to reduce readmissions and improve patient safety. Hospital readmissions are common and costly. According to a PwC survey (PwC, 2014), the 30-day readmission rate in the U.S. is at 18 per cent with an estimated cost of USD 17 billion annually among Medicare beneficiaries. While in some cases patients’ conditions may unavoidably get worse, at least one fifth of patients return to the hospital after discharge because of a misunderstanding or overestimation of medical conditions (AHRQ, 2009).

It has been shown that a large number of unnecessary readmissions can be resolved, not with more complex medical procedures, but with simple education and communication. For example, patients who have a clear understanding of their after-hospital care instructions, including how to take medicines and when to take follow-up appointments, are 30 per cent less likely to be readmitted or visit the emergency department than those who lack this information (AHRQ, 2009). Yet, there are still a number of patients who receive no education about how to care for themselves after discharge (Chetty et al., 2011).

Many U.S. hospitals employ a multidisciplinary approach to reduce the readmission rate for high‐risk chronic heart failure patients. The key processes of a multidisciplinary approach include identifying high-risk patients, designing intervention, and implementing them (McAlister et al., 2004; Stewart, Markey and Horowiz, 1999). First, the medical team develop and implement protocols for identifying patients who require intervention (i.e. elderly people, people living alone) based on individual patient’s physical, social, psychological, cultural and spiritual needs. Second, the medical team will prepare intervention services for patients at higher risk at the time of discharge from hospital. Third, the medical team implements the intervention (for example, comprehensive after-discharge education for the patient and family). Although such multidisciplinary strategies have led to positive results, readmission rates still remain high. One reason is that it is hard to precisely identify target patients, and there is a lack of tailored preventive intervention for patient preferences.

Against this background, some hospitals have sought a more advanced solution in predicting readmission risks, with the use of health information technology (HIT). HIT refers to digital technology that facilitates patient diagnosis, healthcare delivery, and treatment based on electronic medical records (EMR).

As one example, Bardhan et al. (2015) built a model using data for chronic heart failure across 18 counties and 67 hospitals in North Texas for patients from 2006–2009. Their study concluded that building an early warning system identifying predictors for likely readmissions is crucial. Hospitals that utilized HIT had lower readmission rates compared with hospitals without HIT.

Some hospitals in North Texas used algorithms with EMR systems to identify patients in need of follow‐up tests and to schedule preventive care visits, thereby influencing patients’ decision making. The medical team could move quickly back and forth between details of the patients’ electronic profiles from different perspectives. Such flexibility allowed the medical team to identify high‐risk patients more accurately and to predict the incidence and timing of future readmissions. It was concluded that some patients had an increased sense of medical safety and thus were less likely to seek readmission. For example, the Parkland Hospital in Dallas calculates a risk score for each heart failure patient from data in EMR systems and a high‐risk score triggers an alert for special follow‐up care from a heart failure SWAT team (Hagland, 2011).

United Kingdom Department for Work and Pensions

The Department for Work and Pensions (DWP) is the largest public service delivery department in the United Kingdom with over 20 million customers. The major mission for the DWP is to help individuals become financially independent and to avoid poverty. In general, this mission is achieved by modelling the relationship between economic variables such as pensions, savings, incomes and related parameters, thereby providing important information to support policy‐making (Emmerson, Reed and Shephard, 2004). However, it is a challenge to project everything from pensioner income distribution to benefit expenditure in an effective and legal manner.

The DWP used predictive analytics business software to create a number of robust segmentation options for consideration, drawing upon well-established principles (DWP, 2010). Predictive analytics are derived from an intelligent business analytic software suite that helps organizations analyse, model, and forecast “what if” scenarios, plus data access and data management capabilities to provide new insights for decision making. The business software suite chosen by the DWP fulfilled a range of key functions such as long-run pensioner income distributions and forecasting benefit expenditure, for example, the likely changes if the government raised the state pension age by a given amount. The key function the DWP’s business software suite offers is modelling the predicted relationship between parameters. This enables the life events of individuals to be simulated and predicted in sequence (Emmerson, Reed and Shephard, 2004). A report on the project states how the DWP developed “Genesis”:

“… an engine that combines internal and external data to create an accurate picture of what will happen – and what could happen – in the future. Faced with an ageing society and reduced funding, the DWP relies on Genesis to reliably forecast pensioner incomes and shortfalls, and plan benefit expenditure fairly and accurately … and … to reliably predict the impact of proposed policy changes on individuals, families and communities …” (Mohamed, 2012, p. 4).

Intelligent use of digital-enabled predictive analysis has delivered considerable benefits to the DWP. The result is a better understanding of pensioner and pre-pensioner behaviour, which informs interventions regarding policy-making. Since predictive analytics were introduced, the DWP has improved insight into the needs of some 15 million pensioners and pre-pensioners (SAS, 2011). The DWP was able to identify short-term demands on resources and target marketing campaigns to encourage pensioners to switch to lower cost channels. As a result, the DWP has reduced the costs of marketing campaigns, while tripling the rate of uptake of key social security benefits (SAS, 2011).

Conclusions

In bothcases, integrated data bases and predictive analytics were being used to design nudges, primarily in terms of market segmentation. Both positive and negative nudges were identified, and these appeared to be mainly of the “mindful” type.

Ethical and privacy considerations regarding the use and analysis of personal data to influence social investment decision making will loom large over social security administrations as they examine further the digital nudge concept. A focus on ethics however will provide a legal and procedural framework to guide policy-makers while also setting boundaries for technologists. This is important for ensuring that value and benefit for the socially disadvantaged come from the digital nudge without further disempowering them.

A future research question is measuring the impact and return on investment from a digital nudge programme in the context of the achievement of better social outcomes. Many social security administrators fall back on measuring and reporting on outputs rather than using information at their disposal to understand the efficacy of the social programmes they manage in addressing social disadvantage. The digital nudge will add to the options available for social security administrators when working with people to achieve a better outcome within a social investment model. Administrators will need to exercise judgment on when and where to use the digital nudge and seek to evaluate its contribution to the overall measurement and evaluation of social impact.

References

• AHRQ. 2009. National healthcare quality report. Rockville, MD, Agency for Healthcare Research and Quality
• Bardhan, I. et al. 2015. “New analytics model predicts readmission of congestive heart failure patients”, in Information Systems Research, Vol. 26, No. 1
• Chetty, V. K. et al. 2011. “Lower hospital readmission rates and costs”, in American Family Physician, Vol. 83, No. 9
• DWP. 2010. Analysis of the DWP working age customer base. London, Customer segmentation, Department for Work and Pensions
• EC. 2015. Social investment. Brussels, European Commission – Employment, Social Affairs and Inclusion
• Emmerson, C.; Reed, H.; Shephard, A.2004. An Assessment of PENSIM2. London, IFS
• Hagland, M. 2011. “Mastering readmissions: Laying the foundation for change”, in Healthcare Informatics, Vol. 28, No. 4
• Knoll, M. A. 2010. “The role of behavioural economics and behavioural decision making in American’s retirement saving decisions”, in Social Security Bulletin, Vol. 70, No. 4
• PwC. 2014. 30 day readmissions: How do you avoid them?PricewaterhouseCoopers LLP (powerpoint).
• McAlister, F. et al. 2004. “Multidisciplinary strategies for the management of heart failure patients at high risk for admission: A systematic review of randomized trials”, in Journal of the American College of Cardiology, Vol. 44, No. 4
• Mohamed, S. 2012. Harnessing big data to mend our broken economy: High‐performance analytics can help boost public sector efficiency (SAS White paper). Marlow, Buckinghamshire, SAS
• SAS. 2011. The benefits of analytics in the public sector: The smart way to cut costs, optimise performance and deliver reform in the public sector (JMP Statistical Discovery – White paper). Marlow, Buckinghamshire, SAS