SafeVoice

1. Problem

Diabetics get routinely admitted to the hospital with dangerously low sugar levels, manifesting a range of symptoms, from anxiety and mild confusion to seizures, coma, and irreversible brain damage. Many of these instances of low sugar levels or “hypoglycemia” are a result of accidental overdose of self-injected insulin. Such cases are particularly exacerbated in ‘brittle diabetics’ who have widely swinging and difficult to control blood sugar levels and also in elderly chronic diabetics on multiple medications who happen to have other comorbid illnesses like stroke, dementia, kidney and liver disease. These diabetics are not only more vulnerable to the devastating complications of hypoglycemia but are also more likely to inadvertently inject the inappropriate dose at the wrong time. This is because diabetics with initial stages of hypoglycemia are very often anxious, confused and may have slurred speech. Patients in such a state of disorientation may inadvertently continue to inject the next scheduled dose of insulin because of the confusion and poor judgment brought on by the initial low blood sugar levels. This vicious cycle can be disrupted by someone who can recognize these symptoms and intervene in real time, by preventing the insulin overdose and summoning medical help. Unfortunately, many such patients may have inadequate supervision or may be alone at home at the time of injecting insulin.

Impact:

• A prospective study of patients aged >80 years with well-controlled type 2 diabetes reported that hypoglycemia was responsible for 25 % of hospitalizations associated with diabetes and a mortality rate estimated at between 4.9 and 9 %.

• Hypoglycemia represents a significant economic burden on healthcare systems. Reported costs of a severe hypoglycemic episode varied from approximately $80 to $5,000, depending on the requirement for resources, including hospitalization, emergency services, healthcare professionals, and diagnostic tests. This multiplied with the number of actual instances of hypoglycemic episodes is sufficient to highlight the burden of hypoglycemia given a study demonstrated that about 30% of all insulin users experience hypoglycemia.

2. Solution

To prevent such medication errors, we propose a voice recognition system that can interact intelligently with the uniquely matched patient-voice and the insulin-dispensing device (either an injectable pen or a pump) in real time, prior to administration. The purpose of this interaction would be to confirm that the patient is alert, oriented and able to understand the dose being given. The interaction will require the system to engage and recognize the patient's unique speech patterns, test the patient's cognition based on predetermined questions and password type answers and then decide if it is safe to allow the patient to self-inject. In the absence of such confirmation, either because of patient's slurred speech or confusion, the system will lock the insulin-dispensing device, preventing the patient from continuing to inject and/or trigger an alarm for help. The system will unlock only after confirmation of patient safety either by a caregiver or a first responder.

3. How it Works

The system comprises of an IoT / internet-enabled smart insulin-dispensing device that can be managed remotely by SafeVoice.

SafeVoice is a cloud-based voice recognition software running in AWS, that recognizes the unique speech pattern of the user and can also assess the user’s cognition level based on prior input.

Users can access SafeVoice by either downloading an Alexa skill on their Amazon Echo/Dot devices or installing an app from the Google or Apple app store on their mobile devices.

SafeVoice uses a machine learning model that is trained to increasingly improve its recognition of patient’s speech patterns, pitch, intonation, accent, and variable usage in different moods so as to differentiate an alert & oriented user from a confused and disoriented hypoglycemic patient. SafeVoice also remotely controls the dispensing of insulin by the smart insulin device.

SafeVoice will engage and assess the patient prior to insulin self- administration by the patient.

The smart insulin-dispensing device will be remotely locked or permitted to proceed with dispensing insulin based on SafeVoice’s assessment of the patient’s mental status.

In case SafeVoice recognizes an alteration in patient’s mental status based on alteration in speech patterns or inadequate cognition it will notify previously designated caregivers or EMS to alert them of a possible episode of ongoing hypoglycemia.

4. Customers and Acquisition Strategy

Given Medicare’s focus on value-based purchasing and incentives for providing quality care to patients, any effort toward preventing hospitalization of chronically ill type 2 diabetics is a highly sought after endeavor. Third-party payers and Insurance companies are extremely interested in any interventions that will potentially avoid expensive hospitalizations and treatment of complications related to hypoglycemia. Primary care offices and home health care services will also save on resource utilization in terms of personnel and time spent on follow up for such patients.

Instead of making complex plans that are based on a lot of assumptions, our strategy is to make constant adjustments with a steering wheel called the "Build-Measure-Learn" feedback loop. Our plan is to roll out a MVP (Minimum viable product) to a limited number of patients who have had prior episodes of hypoglycemia and based on the feedback we will either make a sharp "pivot" to adjust the solution or we will "persevere" along the current path.

5. Monetization

The software can be downloaded by individual users on a monthly subscription based model or by hospitals and care provider offices for the entire practice for a ‘wholesale’ fee. As patient data is aggregated, data analytics and predictive analytic services for care providers and savvy patients can be another source of revenue.

References:

1. Greco D, Angileri G, Drug-induced severe hypoglycaemia in Type 2 diabetic patients aged 80 years or older, Diabetes Nutr Metab, 2004;17:23–6.
2. Marrett E, Radican L, Davies MJ, Zhang Q, Assessment of severity and frequency of self-reported hypoglycemia on quality of life in patients with type 2 diabetes treated with oral antihyperglycemic agents: A survey study, BMC Res Notes, 2011;4:251.
3. Amiel SA, Dixon T, Mann R, Jameson K, Hypoglycaemia in Type 2 diabetes, Diabet Med, 2008;25:245–54.
4. Hypoglycemia in Type 2 Diabetes—Consequences and Risk Assessment Ildiko Lingvay