Challenges and pitfalls of Smartphone Contact Tracing

Contact tracing apps for smart phones, combined with testing for the virus and antibodies, help form part of the digital social immune system. Using proximity detection via Bluetooth radio, or else carefully processing location service data, we can infer who was likely to have been co-located with someone within meters and for minutes. When the phone carrier reports systems, or tests positive for infection, the list of recent contacts, together with an interview to check for additional people who may not have been carrying phones, can give a set of people to notify that they may have been infected and to be alert for symptoms, or to get tests as soon as possible if they are not already immune.

Additionally, the distribution of contacts and timing of presentation with symptoms and immunity give public health experts more precise models to predict the progress of the outbreak, to measure the effectiveness of interventions such as social distancing advice or lockdown, by area, demographic, and to model how the relaxing of interventions will play out. All systems are subject to false positives and false negatives, so designs should maximise effectiveness in correct notifi-cations while minimising false contact tracing workload.

Such systems need to be extremely cautious about subject privacy, since this is in the realm of medical confidential data, and since the public need to have very high levels of trust in such systems to be happy to use them in enough numbers for them to be effective. At no point should patient app data be re-linkable, except with explicit (or in the case of hospitalisation, implicit) consent. Notifications to potential contacts should be privacy preserving.

In the absence of universal frequent (e.g. weekly) testing, contact tracing has the potential to complement the coarse grained statistics, processing such fine grain data will allow early lifting of the lockdown with assurance that we can prevent the dead cat bounce of the pandemic.  Ref: https://github.com/DP-3T/documents