Loyalty cards - the new early warning system?

Summary: Using loyalty card data on over-the-counter (OTC) medicine purchases could help detect ovarian cancer earlier, according to a study led by Imperial College London. The study of almost 300 women found that women who were subsequently diagnosed with ovarian cancer, made more purchases of indigestion and pain medication, compared to women who did not have the disease. Interestingly this change in purchasing behaviour was seen as much as eight months before diagnosis. Small sample size so more research needed.

Why this is important: Early detection in many diseases can be the difference between a reassuring or devastating prognosis.

The big theme: Healthcare systems are facing ongoing pressure from growing populations, an ever changing landscape of disease and funding pressures. As severity of symptoms or later stages of conditions present themselves, there is an increase in costs and resources required to treat. If we can detect conditions in an earlier stage, then resources required to treat can be reduced and freed up for other cases. Another theme discussed here is the benefits of alternative data in improving decision-making.

Summary of a story from Imperial College London

A study led by Dr James Flanagan from Imperial College's department of Surgery & Cancer, examined up to six years of loyalty card data from two UK based high street retailers for monthly purchases of pain and indigestion medications (such as antacids) in 273 women prior to testing for ovarian cancer.  Just over half were diagnosed with ovarian cancer within a two year period of them being recruited for the study; the remainder showed no signs of ovarian cancer and were thus used as a control group. The results showed a difference in purchases of pain and indigestion medications among women with and without ovarian cancer up to 8 months before diagnosis.

Ovarian cancer is the sixth most common cancer in the UK, with approximately 7,400 women diagnosed each year and over 4,000 deaths. 20% of diagnoses are in emergency departments (ED). The findings from this study suggest a potential early warning system enabling the identification of cases at an earlier stage and therefore improving survival rates.  The small sample size warrants follow up.

Why this is important

93% of women diagnosed with ovarian cancer survive their disease for 5 years or more if diagnosed at the earliest stage. This survival rate drops dramatically to only 13% when diagnosed at the latest stage. The difficulty is that the symptoms can be unclear in those early stages - bloating, eating problems and abdominal pains. As a result, many would tend to buy over-the-counter (OTC) medication from a local pharmacy or other retailer like a supermarket, rather than visit a GP. They don't think their condition is serious. Unfortunately, 20% of sufferers end up being diagnosed in the ED when symptoms have become severe, often the cancer has spread to other areas, they are too weak to receive treatment and chances of survival are sadly massively reduced. Even if treatment options are available they will typically be hard for the patient and costly. So any early warning systems that can lead to diagnosis at the early stage would be greatly beneficial to all concerned.

What is particularly interesting about the results of the study is the potential applicability to other ailments and diseases. There are plenty of serious conditions that present with seemingly innocuous or common symptoms in the early stages. Even a visit to the doctors can have diagnostic pitfalls, exacerbated with the move to online/telephone consultations. There can be reasons why symptoms can be missed when collecting oral history. The patient may not understanding what a symptom is (for example 'jaundiced'), to not spotting a symptom or the severity of it ("how painful is it?"), to 'not wanting to make a fuss.' External data such as OTC medication purchasing, especially changes in the rate of purchasing, can therefore be invaluable input into decision-making, diagnosis and ultimately treatment.

In the Imperial College study, the mean age of the participants was just under 65 years old and almost 60% of them used their loyalty cards all the time providing robust purchasing data. With any data there will always be privacy concerns - In this study, explicit consent was obtained from both the retailers and all of the participants to use their data and verify their identification. However, for a wider scheme and analysis extending to a broader demographic and a wider range of potential diseases, an effective enrolment process would need to be devised. For example, a NHS scheme for extracting data from GP surgeries (General Practice Data for Planning and Research) had attracted criticism from privacy rights campaigners, despite strong support from practice workers for the potential medical benefits.

The use of alternative data to supplement traditional analysis in the investing world really gathered pace in the last 10 years - working on a quantitative desk at the time, the effectiveness of that use varied greatly! The use of loyalty card data is really one other example of that, just applied in the medical and healthcare space. Real-time monitoring of behaviour can provide holistic benefits as well as specific early diagnoses. For example, many of us have gotten used to wearable fitness devices providing real time data on our health and even using gamification to help us improve our health (e.g. hitting that daily steps target). Health insurance companies often provide incentives for buying and using fitness trackers as a micro method for reducing their underwriting risk.

The use of data and digitisation (including the use of electronic medical records) more broadly in healthcare has enhanced the effectiveness and efficiency of it - from the coordination of blood-work and scanning to enabling geographically disparate multi-disciplinary teams to work together. I found this first hand when I recently had an adverse reaction to some anti-fungal medication I was taking which caused my liver to say "adios!" Two observations: firstly, the coordination between different teams because they had my full history including results immediately made for a very efficient trip for me (and for them!); secondly, sitting in the waiting area, whilst triage (the first line assessment of which patients arriving at an ED to prioritise) was very good, it became clear that identifying patient priority real time was hard, i.e. changes in patient condition whilst in the waiting area. This is where a virtual ward approach and associated technology could become game changing. Another area is in early hospital discharge where rather than keeping patients in for observations and taking up expensive resources (bed, staff, equipment etc), they could be sent home and monitored virtually. A number of companies including Doccla are innovating in this area.

Please read: important legal stuff.