Ocuvera’s Technology

At Ocuvera, we make an automated video monitoring system: a video-based monitoring system where computers monitor patient behavior instead of monitoring technicians. The Ocuvera automated video monitoring system is built around a three-dimensional (3D) video stream. Using a 3D camera provides much more information than traditional 2D cameras. If the cameras can see more, then the system can process more information and be more confident when automatically monitoring patients.

Custom algorithms and automated pattern recognition monitor for patient activity that increases their risk of falling, especially if the patient is trying to exit the bed without a nurse helping or watching them. Once the system has enough information to think that the patient is trying to exit the bed, nurses are automatically notified by live video of the patient’s behavior sent to the nurse’s station, phone, or smart watch. This allows nurses to see for themselves what happened to trigger the alert and decide how best to respond, rather than having to simply trust an alarm. Nursing staff set the system to one of four sensitivity levels to control the amount of movement necessary to trigger an alert – depending on their patient’s specific needs, some nurses might only be interested in knowing whether their patient is about to exit the bed, while some might want to be alerted if their patient moves at all. Ocuvera alerts do not sound in the patient room to avoid unnecessarily agitating patients. The system is automated only up to the point of sending the alert – the ultimate decision of when and how to best address patient needs is left up to the nurse.

Ocuvera’s algorithms are designed to predict patient attempts to exit a bed ahead of time. The algorithms underlying the Ocuvera system have been under development for 5 years and are based on over 150,000 hours of patient video through collaboration with over 20 inpatient, critical access, and rehabilitation hospitals. This data contains thousands of real bed exits from patients, including almost two thousand unattended bed exits. As Ocuvera takes patient privacy very seriously, all patient data is securely stored as non-personally identifiable depth images. Analyzing this data from real patients helps our algorithms learn what risky patient behavior looks like and be more likely to send an alert when it’s necessary.

Our algorithms are complex, but simply put; they work by identifying the bed, the patient, and what the patient is doing. In order for our algorithms to work, they must first find the bed in the patient room. They begin by finding the floor, and then decide which pixels in the scene most likely represent the bed. Once the system finds the bed, it is constantly monitored for changes in position and updates automatically so that nurses can adjust the bed as they normally would and be confident the Ocuvera system will still track their patient. The system then works on identifying which pixels represent the patient, and what position that patient is in – whether they are laying down, sitting up, facing the edge of the bed, or about to get out of bed. The system must then decide whether or not to send an alert to nurses. Here is a simplified example of how that decision is made:

  • Has there been movement above the bed for a certain amount of time?
  • If yes, is there a patient in the bed?
  • If yes, has the patient been sitting up for a while?
  • If yes, has the patient been facing the edge of the bed for a while?
  • If yes, has it been a while since a nurse was last in the room?
  • If yes, send an alert.

At every level, our technology is designed to make the nurse’s job easier so they can provide the best care for their patients. We are working closely with medical researchers at the University of Nebraska Omaha to develop an adaptive training curriculum for integration of the Ocuvera system into existing nurse workflows. This will help ensure that the system is as easy for nurses to use as possible, and that it works with existing fall risk reduction strategies rather than replaces them. Through the adaptive training curriculum, we’ll develop straightforward educational materials for nurses on how to best use the system to maximize its effectiveness in reducing patient fall risk.

Nurses and Falls

Patient falls have a substantial impact on the healthcare system. When a patient falls, not only do they suffer the adverse effects, but so do the nurses who care for them. Ocuvera’s focus is on creating a fall risk reduction technology that helps minimize the impact patient falls have on nurses so they can better provide care. We are always listening to nurses who have experience using our system to better understand how they’re making it part of their jobs every day.

No nurse wants their patient to fall. For one, it can cause them guilt and personal anxiety. As one nurse told us, “you can’t help but wonder, ‘does this mean I’m a bad nurse?’ Even if you know it isn’t your fault.” There is also a risk that a nurse could be injured attempting to prevent a patient from falling. In the aftermath of a patient fall, the amount of work that nurses must do is significant. Nurses always stress this to us – they are already extremely busy with their standard workload, and when a patient falls, their list of things to do increases dramatically. Patient assessments, CT scans, X-rays, neuro checks, and seemingly endless paperwork are just some of the things a nurse might have to do following a patient fall. Longer term, a fall can also contribute to patient decline, changing their needs and possibly leading them to require closer monitoring or even a longer stay in the hospital.

Ocuvera understands the burden patient falls place on nurses. Nurses have a bag of many tools to help minimize the risk that their patients will fall – the Ocuvera system is a tool that is uniquely focused on nurses. The nurses that we spoke to pointed out three aspects of Ocuvera alerts that are especially helpful to them: they are proactive, visual, and customizable. Most other fall risk reduction tools are reactive, meaning nurses only become aware of a fall after it has already happened. Because the Ocuvera system is able to predict when a patient is at risk of falling, nurses are given the chance to prevent that potential fall before it occurs.

The visual aspect of Ocuvera’s alerts helps nurses to know exactly what is going on in their patient’s room. With standard, non-visual alarms, such as those triggered by pressure pads, nurses say they typically have to “drop everything and run” to respond, because they don’t know what triggered it, and it could be an emergency. Ocuvera alerts allow nurses to use the Ocuvera device in their pocket to easily see video of the patient and decide for themselves how to respond. This means that Ocuvera alerts don’t interrupt a nurse’s usual workflow to the extent that traditional alarms do, while also being “more sensitive” than existing bed alarm systems. One nurse stated that Ocuvera causes “no alarm fatigue,” which is “a positive influence on our nurses.” Nurses find the ability to have vision and “transparency” into a patient’s room particularly useful for any impulsive or stubborn patients, especially those with brain injuries or balance problems. Aside from sending alerts when a patient might be about to fall, some nurses have found other uses for Ocuvera’s visual component. Ocuvera allows nurses to monitor patients at night, so that there’s no need to disturb them unnecessarily while they’re sleeping. As one nurse stated, “we are able to watch [the patient] from the live video stream to keep him safe.” Nurses have also mentioned to us that the system can be useful for trial one-to-ones or taking patients off one-to-one sitter supervision, stating that it “helps keep the number of one-to-one supervisions down.”

Ocuvera allows nurses to use the system how they see fit, depending on each patient’s individual needs. Not all patients are at the same risk of falling, and nurses can choose when they’d like to receive alerts for a particular patient – when they have already gotten out of bed, when they’re about to get out of bed, or even if they move at all. They can also change this setting as needed as a patient’s condition changes over the course of their stay. For patients that require extra attention, Ocuvera gives nurses “the option of checking on our patients more often.” This gives nurses the autonomy to care for each patient the best they can.

Reducing and managing patient falls is just one of countless things a nurse has to juggle on any given day. In striving to be the best possible fall risk reduction tool, Ocuvera also strives to fit neatly into the workflow that already exists for nurses, helping them care for their patients without interrupting their already full schedules. Nurses using the system tell us that it fits in very smoothly with the rest of their work, that “all the components are in the hands of the nurses,” and that it “feels like part of everything else.” Nurses can set up the system easily before the patient even comes in the room, and it’s also simple to discontinue once their patient is discharged or no longer needs to be monitored. Easy integration means that nurses can spend more time focusing on patient care. One nurse testified that Ocuvera “just makes life easier,” which is exactly our goal.

The Problem of Falls

Patient falls are a common, costly, and serious problem in hospitals and healthcare facilities. The numbers are shocking — every year, over 200,000 people fall.[i] More than 2% of hospital patients fall,[ii] with around 25% of those falls result in injury.[iii]Around 11,000 falls are fatal.[iv] These disheartening numbers are despite decades of effort to better protect patients. Falls are still one of the most common causes of harm to patients in hospitals. For these reasons, serious fall-related injuries are designated preventable Hospital-Acquired Condition by the Centers for Medicare and Medicaid.[v] CMS does not reimburse hospitals for costs related to patient falls with injury.

To us, a fall is any time a patient descends when they did not mean to.[vi] An assisted fall is a fall in which any staff member (whether a nursing service employee or not) was with the patient and attempted to minimize the impact of the fall by slowing the patient’s descent. An assisted fall is much less likely to result in injury than an unassisted fall.[vii] To improve patient recovery and outcomes, best practices today focus on increasing mobility. Nurses might encourage patients to get up and move around with supervision while they’re recovering, which can lead to the patient falling. If the nurse is there supervising the patient, they will usually be able to minimize any injury from the fall while continuing to improve the patient’s mobility. Unassisted falls, on the other hand, occur when a nurse is not there to meet the patient’s needs. Ocuvera focuses on unassisted falls, helping nurses get to patients faster and either prevent the unassisted fall entirely, or turn it into an assisted fall.

One drawback of current fall risk reduction interventions, like pressure pads and bed exit alarms, is that nurses don’t have enough time to get to the patient before they’ve gotten out of bed and possibly fallen. Addressing this limitation could help reduce falls and fall-related injuries, which decrease quality of life and increase health care costs. Nurses aren’t able to see the vast majority of patient falls as they happen, which means they don’t know what happened before the fall. Often, falls happen after a patient attempts to get out of bed on their own, events we refer to as unattended bed exits. About 45% of falls happen after unattended bed exits, 21% after unattended chair exits, and 21% happen in the bathroom, after the patient exits either the bed or the chair.[viii] Preventing the unattended bed exits that often lead to falls could have a significant impact on decreasing the falls themselves. This is what Ocuvera’s technology does, by monitoring patient movements and predicting when they are likely to attempt to get out of bed.

The problem of patient falls is complex, but it does not necessarily need a complex solution. As part of our research while developing Ocuvera, we’re pushing to have better data on patient falls to improve overall knowledge. Video is an immensely powerful tool to give insight into questions that have not had answers before – why does a patient get out of bed? What happens when they do? How often do they do it? Our technology provides nurses with literal vision into these otherwise unobserved situations, helping them better understand their patients and provide better care.

Throughout our research into falls, we have been driven by measurement, reproducible methodology, and verifiable numbers and evidence. In 2018, Ocuvera completed two state and federal grant funded studies to test the feasibility of our system to reduce unattended bed exits in rural Critical Access Hospitals in Nebraska. Figure 1 shows that unattended bed exits decreased by 89% when the system was in use. Other results from these studies showed that the Ocuvera system was able to detect unattended bed exits 96% of the time, and that 56% of the alerts sent by the system were in response to behaviors that warranted intervention from nursing staff. We found that the system alerted nurses of risky patient behavior with approximately 20 seconds of lead time.

Ocuvera began testing its field testing its system in 2017 at a neuro med-surg unit. At this site, Ocuvera has been used for over 150 of the unit’s highest-fall-risk patients and 3,000 total patient days. The unit sees approximately 7,800 patient days a year. Before Ocuvera was introduced, the unit had a high fall rate: 7.17 falls per 1,000 patient days. Since Ocuvera was introduced, the unit has seen a nearly 40% decline in falls, with a fall rate of 4.36 through July 2018. In Q4 2017, Ocuvera began another field test at one of the leading academic medical centers. Twelve cameras were used in a 50-bed med-surg unit. After 4 months, there was a 64% reduction in the number of unassisted and unobserved falls among patients where the Ocuvera system was used compared to other patients on the unit. Ocuvera was used for these patients because they were at very high risk of falling, so results could be more pronounced with broader patient selection criteria.

Our results so far are very promising. We believe they show our technology really can have an impact on the problem of patient falls by preventing unattended bed exits. We hope to conduct more studies and deploy our system to more hospitals to broaden our understanding of patient falls and how our technology can help nurses reduce them.

[i] https://www.ahrq.gov/professionals/quality-patient-safety/pfp/index.html

[ii] (Bouldin, et al., 2013)

[iii] (Oliver, Healey, & Haines, 2010)

[iv] (Agency for Healthcare Research and Quality, 2016)

[v] (Centers for Medicare and Medicaid Services)

[vi] For research purposes, we use the definition of falls from The Nati­­­onal Database of Nursing Quality Indicators® (NDNQI), which is: an unintentional descent, with or without injury to the patient, that results in the patient coming to rest on the floor, or against some other surface, on another person, or an object.

[vii] (Jones, Skinner, Kennel, & High, October 15, 2016)

[viii] (Jones, Origin of Falls, 2017)

Ethical Artificial Intelligence

Though Artificial Intelligence (AI) is a fascinating and useful application of technology, it is also fraught with ethical questions. Nurses train for years to gain the experience and expertise necessary to give their patients the best possible care – why, then, should they trust their patients’ safety and wellbeing to a computer? This is a legitimate question, and one that Ocuvera has gone to great lengths to address at every level of our project.

First, it is worth noting that AI is not entirely new within the healthcare space. Nurses have long trusted software to monitor other aspects of the patient experience, such as vitals. AI has proved valuable in this role because of its ability to automate otherwise menial tasks and offload them to a computer, freeing up time in nurses’ busy schedules to focus on more complex and demanding duties. This is what the Ocuvera system does. Watching video of patients diligently for hours on end is not only not feasible, but also dangerous: even the most attentive and dedicated person could not watch a motionless patient for hours on end with the level of concentration necessary to ensure fall risk is minimized. Ocuvera takes the most tedious aspects of fall risk reduction out of people’s hands, but leaves nurses at the center of the solution when human attention is most needed. The automation ends when the alert is sent, and the decisions about when and how to intervene to best help the patient are left up to the nurse.

The question still remains why nurses should be confident that the Ocuvera system will alert them when their patient is behaving in a way that increases their risk of falling. Our confidence in our algorithms is based on years of research and hundreds of thousands of hours of video. Computer vision is data hungry, and the more data we use to train our predictive algorithms the more accurate the system’s predictions are.

Our video is of real patients in real care settings. While recorded video is necessary for us to design the Ocuvera system to be effective in true hospital settings, we also understand that our patients are trusting us to treat their recordings with the respect and privacy it deserves. Being recorded is an imposition on the lives of patients and their families. That’s why, as part of our commitment to ethical AI, we take patient privacy and dignity very seriously throughout the data collection process. Most of our data collection has taken place in the context of Institutional Review Board (IRB)-approved studies, several of them grant funded. Part of our IRB agreements consists of detailed informed consent procedures, where patients and their families are briefed on what Ocuvera does, why they are being recorded, and that participation in the study is completely optional and they can stop at any time. A “privacy button” is built into the system to allow patients, family, or nursing staff to temporarily disable the system during sensitive moments. To ensure privacy, all patient video is recorded in depth images only. These are images where featureless shapes represent the activity in a patient room, but where personal identifiers such as skin tone and facial features are not visible. Ocuvera’s use of the depth image to protect patient privacy has been approved by four separate IRBs as containing no personally identifiable images.

Patient privacy remains critically important even after data collection ends. No personally identifiable information is stored at Ocuvera: all information linking patients to their identity is kept exclusively at the hospital. All patient video gathered is encrypted and stored on secure hard drives at Ocuvera. Only Ocuvera employees who have undergone human subjects research training through the Collaborative Institutional Training Initiative (CITI) have access to the data. These employees understand the risks and issues associated with video of patients, and always approach patient video with respect.

All of Ocuvera’s methods to ensure our ethical implementation of our AI stem from the understanding that people are the center of healthcare technology. Nurses and patients are not just the end recipients of our product, but central to our mission of empowering nurses to improve patient lives through technology.

Ocuvera’s Philosophy

Ocuvera was founded on the belief that technology can help people. At our core, we are a group of developers with extensive experience in the fields of engineering, computer vision, and mathematics. We believe that with the help of nurses, our unique perspective and expertise can improve healthcare. In fall risk reduction, Ocuvera has found a space where technology has immense potential to improve people’s lives, by empowering every nurse with tools to provide safer patient care.

Ocuvera was created by the technology startup incubator Nebraska Global Investment Company (NGIC). Ocuvera and Nebraska Global specialize in building complex software using agile methodologies to solve difficult problems. NGIC shares Ocuvera’s philosophy on the power of technology to improve lives. Through NGIC, the same set of engineers who would go on to create Ocuvera developed EliteForm, LLC, a computer vision company in the sports performance industry. EliteForm measures athlete performance during weightlifting by using depth cameras to count repetitions and to measure lift velocity and power generation. Based on the success of EliteForm, Ocuvera was started in 2012 to apply this computer vision expertise to help hospitals prevent inpatient falls.

Fall risk reduction is a field ripe for innovation, where technology has real potential to help improve the lives of both patients suffering falls and the nurses who treat them. Patient falls are a common, costly, and serious problem in hospitals and healthcare facilities. Almost everyone we talk to has a personal story of a friend or loved one who was injured by falling. Nurses often tell us how stressful patient falls can be for them, both because of the workload they add and the negative impact they can have psychologically. There are almost one million patient falls each year, with many resulting in injury, and some even proving fatal. Despite the severity of the problem, little progress has been made in recent years in reliably reducing patient falls. Nurses rely on largely ineffective interventions, such as bed alarms and in-room sitters, which are still standard in most healthcare facilities. Although technologies like centralized video monitoring (CVM) have made some progress in updating these methods, each has its limitations and drawbacks. The continued prevalence of falls is evidence that additional approaches are needed.

Innovative ideas for reducing fall risk are limited because of a lack of information. Although we have numbers about how many falls occur, many details about their surrounding circumstances remain a mystery. When we began research into how Ocuvera could best help the problem of patient falls, we realized it was often not clear why a patient fell, where they were before they fell, or what they did before they fell that increased their fall risk. One of our goals is to gather this information ourselves to better understand falls and make sure that our solution is based on the best possible data in order to provide the best possible care.

Ocuvera’s technology combines real-time, automated, predictive, video-based, and 3-dimensional algorithms to reduce fall risk. This combination is unique among existing technologies that reduce fall risk. Our system uses these Automated Video Monitoring (AVM) algorithms to detect unattended bed exits before they occur. Our early warning system monitors for changes in patient behavior, predicts when they’re about to get out of bed, and alerts a nurse before the patient gets up. This gives the nurse time to help their patients and stop a potential fall, putting the power in the hands of the nurses when it’s most needed. We see our technology as a tool that empowers nurses to better care for their patients: we are there to help nurses make a difference in a patient’s life.

Now is the time to apply computer vision and artificial intelligence to healthcare. Five years ago, the technology did not exist to create an effective automated video monitoring system. In ten years, nurses will be able to use a camera in every hospital room. Ocuvera wants to be the company that delivers this innovation to nurses. Ocuvera believes in a future where artificial intelligence is used to its fullest potential in the healthcare space, and our technology is a step in this direction. We plan to turn our system into a platform for nurses to know everything they need to provide the best patient care.