Elarin is a company in the US formed to commercialise this falls predictor. I developed the algorithm, while other colleagues in Elarin developed the dashboard, coding, working prototype and provide the real time monitor for vital signs. Elarin is seeking funding to implement this applicaion including funding a usability trial.
email me at mike_tremblay@skythunder.net if you are interested in investing in Elarin.
Overview of the Predictor
Falls among older people are rarely caused by a single factor. They usually arise from a changing combination of physiological instability, frailty, medication effects, cognitive or behavioural changes, environmental conditions and immediate circumstances. Conventional falls assessments identify people who are generally at increased risk, the vast majority of current falls detectors on the market detect falls about to happen, but do not provide a prediction with sufficient time to intervene and avoid the fall itself. The most useful feature of the Elarin Falls Predictor is that it gives staff time to react, alter clinical treatment or medicines or other factors that are falls determinants.
The Elarin Falls Predictor is designed to address this temporal problem. Its purpose is to estimate the probability that an individual will fall within the next 24 to 48 hours and to update that estimate as new information becomes available. It is a prediction system rather than a fall-detection device: the intention is to provide clinicians and care teams with advance warning of a developing period of vulnerability, before a fall occurs.
The Predictor brings together several categories of information that are commonly considered separately. These include continuously or regularly measured physiological data, movement and behavioural indicators, clinical characteristics, medication exposure and environmental conditions. The model assesses how these factors interact and change over time rather than treating them as a static checklist.
A central feature is its N-of-1 approach to personalisation. Instead of relying on population averages, the Predictor learns each person’s usual physiological and behavioural pattern. Current observations can then be assessed against that individual’s personalised baseline. A change that appears unremarkable when compared with a general population may be clinically important when it represents a marked departure from the person’s normal state.
Relevant physiological inputs may include heart rate, respiratory rate, blood pressure, oxygen saturation, temperature, activity and sleep-related measures. The model is intended to detect patterns of deviation and instability across several variables, rather than reacting to an isolated measurement. This allows the prediction to reflect the accumulated effect of small changes that may collectively indicate deteriorating balance, reduced physiological reserve, infection, dehydration, fatigue, orthostatic instability or another emerging source of fall vulnerability.
Medication is treated as a dynamic contributor to risk. The Predictor can incorporate the timing, dose and cumulative effect of medicines associated with falls (called Falls Risk Increasing Drugs, FRIDs), including sedatives, psychotropics, antihypertensives and other fall-risk-increasing drugs. It can also account for interactions between medicines and for changes following the introduction, withdrawal or adjustment of treatment. The objective is not merely to identify that a person is prescribed a potentially relevant medicine, but to estimate how current medication exposure may be contributing to risk at a particular time. The predictor enables clinicians to model a person’s falls risk when medications are substituted, started or stopped, enabling the clinician to assess falls risk before prescribing.
The model includes additional functionality for people living with dementia. Dementia can alter fall risk through gait instability, wandering, agitation, disrupted sleep and circadian rhythm, impaired hazard recognition and increased sensitivity to medication. The dementia-aware component incorporates these behavioural and cognitive factors alongside physiological measurements. A dedicated psychotropic-risk module considers the potential effects of antipsychotics, benzodiazepines, antidepressants, mood stabilisers and cognitive-enhancing medicines, including their sedative, motor, extrapyramidal and orthostatic effects.
Environmental conditions are also relevant. Temperature extremes, humidity, wind, atmospheric pressure and precipitation may influence fall risk directly, through slippery surfaces and difficult walking conditions, or indirectly, through changes in physiological strain, mobility, hydration and postural stability. Some weather effects may be immediate, while others may persist or become evident after a delay. The Predictor therefore uses current conditions, recent trends and lagged environmental information rather than treating weather as a simple present-or-absent hazard.
The resulting probability is intended to support, and augment clinical judgement. It provides a structured synthesis of information that would otherwise be dispersed across monitoring systems, medication records, observations and environmental data. Where risk is increasing, the system can identify the principal contributing factors and help the care team consider proportionate preventive measures. These might include closer observation, assistance with mobility, hydration review, medication review, environmental modification or investigation of an emerging clinical problem.
The Elarin Falls Predictor does not assume that every predicted fall can be prevented. Its value lies in identifying periods when vulnerability is materially different from the individual’s usual state and when targeted intervention may be justified. By moving falls prevention from periodic risk classification towards continuous, personalised foresight, the Predictor is intended to help clinicians and care organisations intervene earlier, use resources more selectively and reduce avoidable harm.