This was written in early 2019 and may require some revision for 2020, updates will be noted here as they are made.

Context

Australian hospitals today face a variety of challenges, some obvious, and altruistic, have been addressed in detail in research, for example; the development and advantages of laparoscopy, and then movements to minimal trauma, trocar’s (Tarnay, 1999), and analysis and development of best practice techniques such as pre procedure time outs and sharps tracking (EQUALS, 2013). However, the people who work in hospitals face overlooked challenges daily that impact their ability to effectively treat patients.

Key challenges identified during contextual research include; Hospital struggle with space, not only for patients. In a patient and therefore bed driven economy, stock rooms and warehousing are minimised to add more capacity. Patients are not stationary objects, they move through the hospitals ecosystem to access different services; pathology, radiology, treatment areas, even toilets. At all times these patients need to be accounted for and easily located, in a sea of beds, with minimal staff and shift changes, sometimes they are not. Staff often work across multiple sites utilising a large temporary employment workforce, these operating capacities must be balanced with patient loads and theatre availability, where overestimates increase costs, but shortfalls could leave people in need waiting (Stanton, Willis, & Young, 2005). Surgical site infections and hand hygiene are driving changes in practice across Australian hospitals, outwardly something quite trivial but with far reaching ramifications (Johnson et al., 2005). By addressing these issues hospitals can more effectively treat patients and reduce the pressure on their teams, improving employee well being.

Data and effective interrogation of the data through Data Science can drive innovation in these neglected challenges faced by Australian hospitals; Automated data collection, and tracking can reduce the burden on users and reduce input errors (Hardgrave, Aloysius, & Goyal, 2009). Exploration, Analysis and prediction enable organisations to find obscure issues and successes (LaValle, Lesser, Shockley, Hopkins, & Kruschwitz, 2011), to evaluate processes and services, and see what’s coming next. This can be further developed through machine learning to produce deeper insights and sometimes unexpected ideas. Considered, informed visualisations enable efficient comprehension, providing at a glance business intelligence to drive productive decision making (Popovič, Hackney, Coelho, & Jaklič, 2012). These tools when properly implemented push organisations toward change, in processes and services, empowering individuals and driving innovative ideas.

Space Management

Challenge

The capacity of an Australian hospital is defined by its bed count, where larger bed count hospitals are perceived as more successful, this creates a drive to increase bed capacity for a hospital site. Space is reduced in other departments to increase bed capacity, one affected department and an issue raised by hospital administrators and employees is stock warehousing space. Conventionally stock moves between multiple warehouses within a hospital, a specific example being stock for operating theatres; Stock is unloaded to a dock holding area, moves to the primary hospital in cartons or on pallets stockroom, in cartons to the theatre stockroom, unboxed into theatre stores for picking and finally onto a racking for the day’s surgeries. At each step of the process two conflicting influences interact; A need for adequate stock to perform daily activities, changes in procedure lists, and prepare for emergencies, against minimising space taken up by stock and reducing stock held onsite due to small areas dedicated to stock holding.

This back and forth induces stress on the stakeholders involved; Deliveries take longer as unloading is more complex, warehousing staff struggle to find space to fit all the necessary stock, nurses fear the items needed for a procedure may not be available, surgeons worry they may not have the correct tools for the job, executives see excess stock unused and reducing available area for beds. More effective management can reduce these stresses while improving the efficiency of the hospital as a business.

Opportunities

A comprehensive RFID tracking system with zoning, real-time location systems or ultra-wide band can track the location of items throughout a facility. By having multiple reading stations, with established zones, or through software calculation, location can be established to within centimetres and to 100% quantity accuracy (Roberti, Mark, 2016). This detailed location tracking can be applied not only to stock, but to patients, equipment, staff, and beds.

Visual tracking is an emerging method of providing IoT like data without the necessity of chip and receiver systems like RFID. Visual tracking utilises camera systems and image analysis enabled by deep learning to follow a target through a system (Wang et al., 2017). While this can be somewhat easier to apply than procuring tagged products, drawbacks include reduced accuracy especially among like products, and privacy concerns if facial identification is used. With similar outcomes a mixed approach could be beneficial.

Once implemented both suppliers and hospital staff gain access to highly accurate tracking information of both stock and patients, with potentially greater detail than previously experienced, while decreasing workload manually accounting for patients, stock and transfers. Decreasing the time spent on easily automated tasks, in a change empowered organisation gives team members the time and opportunity to seek out improvements and potential innovation in their department (Kotter & Cohen, 2002).

Stage 1 — IoT, RFID and Visual Tracking

A comprehensive RFID tracking system with zoning, real-time location systems or ultra-wide band can track the location of items throughout a facility. By having multiple reading stations, with established zones, or through software calculation, location can be established to within centimetres and to 100% quantity accuracy (Roberti, Mark, 2016). This detailed location tracking can be applied not only to stock, but to patients, equipment, staff, and beds.

Visual tracking is an emerging method of providing IoT like data without the necessity of chip and receiver systems like RFID. Visual tracking utilises camera systems and image analysis enabled by deep learning to follow a target through a system (Wang et al., 2017). While this can be somewhat easier to apply than procuring tagged products, drawbacks include reduced accuracy especially among like products, and privacy concerns if facial identification is used. With similar outcomes a mixed approach could be beneficial.

Once implemented both suppliers and hospital staff gain access to highly accurate tracking information of both stock and patients, with potentially greater detail than previously experienced, while decreasing workload manually accounting for patients, stock and transfers. Decreasing the time spent on easily automated tasks, in a change empowered organisation gives team members the time and opportunity to seek out improvements and potential innovation in their department (Kotter & Cohen, 2002).

Stage 2 — Just in Time Inventory

Just-In-Time (JIT) inventory is the receipt of stock on, or as near to the date it is needed as possible. This method of inventory management can reduce the inventory footprint by reducing stock on hand at the risk of not having items in times of high demand. The concept of JIT Inventory developed out of JIT manufacturing as part of the Toyota Production System (TPS) a process that was later developed into LEAN manufacturing. TPS and LEAN have proven to be effective strategies for increasing business efficiencies (Bagchi, Raghunathan, & Bardi, 1987). However, one of the core concerns around JIT inventory is not having the available resources when needed, due to; unexpected demand, supplier issues, stock recalls and other factors. Data Science can reduce these concerns by delivering accurate forecasting and predicting upcoming issues.

Though it is possible to produce a forecast through simple analysis such as linear regression, by combining more information and variables into the model, it is possible to more accurately predict stock needs. A robust, optimised forecasting model enables the most effective use of space through JIT inventory. One option is to use multivariate time series analysis to develop a forecast based on a weighting of multiple variables and historical data (Chakraborty, Mehrotra, Mohan, & Ranka, 1992). This model will be more complex than a liner model by considering external factors that are affecting the system; events, weather, etc. Ideally this higher level of complexity produces a more accurate day to day forecast than the simpler alternative. Another more contemporary option is to utilise neural networks and machine learning to optimise the business forecast. A neural network can be set up with rules that govern failure and success criteria (Chakraborty et al., 1992). By tasking the network with maximising success, in our case, accurate stock provisioning, the network can then be trained to forecast, order, predict and adapt to the current stock conditions, including unexpected events. These data science methods provide a more robust system for automating stock management and thus reducing fear, offsetting some of the drawbacks and increasing the efficiency and therefore space saving using JIT Inventory.

Patient Tracking

Challenge

The hospital staff need to be able to account for all patients under their care at a given time. Patients move about a hospital throughout their stay; transferring between wards, utilising services, undertaking procedures, receiving treatment, and just because that is what people do. In order to track patients, duty of care is placed on the nurses in the patient’s ward, and digital or manual paperwork tracks each movement around the hospital. This paperwork is time consuming and not always accurately filled out either by intention or when external factors complicate this further; emergencies, changing nurses, shift changes, patient handovers, changes in schedules and treatment plans.

Opportunities

Enabling simpler more effective patient tracking through more comprehensive data collection, information sharing and modelling. Will first take stress off the caretakers and reduce paperwork, and potentially enable further improvements (Sangwan, Qiu, & Jessen, 2005).

Stage 1 — Tracking

Utilising previously described IoT technologies patient location data can easily be collected and provided to the use. RFID gowns have begun to be developed and active RFID is easily applied to capital equipment such as patient beds. Alternatively, visual tracking either through labels or facial identification could be implemented to follow patients’ movements throughout the hospital. This data ensures accurate tracking of patients during the time in a facility (Sangwan et al., 2005), and ease of location for team members.

Stage 2 — Visualisation

The location data itself can be provided the staff through in multiple ways. A simple form displaying the ward or room the patient is currently located in by name can enable quick reference. More complex visualisations could improve understanding and empower the user to engage with he information more thoughtfully. By giving the user a map with multiple patient’s current locations a Nurse can quickly asses that multiple patients are where they should be, rather than having to review multiple individual’s files. Further the patients next known movements could be plotted for the Nurse, based on upcoming tests and procedures. Providing this information to the Nursing team may empower them to improve the organisation of patients, for example, placing beds in order of next procedures.

Step 3 — Predictive Modelling

A predictive model for patient movements could provide increased efficiencies within the hospital. After sufficient patient location data has been collected, this can be combined with other data such as reason for stay, length of stay, and contraindications to develop a predictive model of patient movements (Choudhary, Bafna, Heo, Hendrich, & Chow, 2010). This model could the be interrogated to predict upcoming loads on hospital services based on bookings and external factors. Patients could be moved closer to services they are likely to need and reduce the incidence of patients receiving incorrect care due to errors or mix-ups. For example, high risk patients closer to emergent care, or simple changes such as placing patients with easy access to bathrooms if they are likely to need them and/or have low mobility. A neural network may provide real time assistance and pre-warning of upcoming issues, or potentially increase bed capacity through increased space efficiency, while predicting the need for services to improve scheduling. Patient tracking and prediction, in these cases simplifying daily duties and reducing decisions made throughout the day.

These benefits must be weighed against potential ethical concerns, especially when involving vulnerable individuals such patients. The chances for data misuse in healthcare are high both intentionally and unintentionally. A system that tracks patients so accurately reduces individual privacy and may highlight activity the patient would not normally have disclosed. Predictive analysis can cause both excessive care under-care if users are too reliant on predictive models over interacting and assessing current needs, in such a high-risk area, consequences could be severe. These concerns need to be evaluated and addressed if a system is to be beneficial.

Horizon 1 — Dual Purpose Spaces

A potential innovative solution to hospital space issues is the development of dual-purpose spaces, split between stock holding and patient care. By utilising real time stock and patient tracking, a dual-purpose area with a moveable division, could be actively resized to meet demand for either stock or patients. This enables the placement of stock nearer to areas of use during times of high demand, and the movement of patients to ideal locations based on services needed. Effective utilisation of dual-purpose areas could improve; patient well being, staff stress levels, and potentially hospital bed capacity.

Data science, through data tracking, analysis, automation, machine learning and other tools, enable innovative solutions like these to business issues. The best people to develop innovative solutions are often the people on the ground (Kotter & Cohen, 2002), and effective data utilisation and communication enables team members to take an introspective look at their day to day duties with new information and insights. This drives thought and ideas, these ideas need to be combined with a culture that empowers its people (Kotter & Cohen, 2002), enabling them to create innovative change.

Challenges, Ethics and Privacy

Data Collection and Integration

Data produced in the healthcare environment though now increasingly recorded digitally is often unstructured, in the form of notes, comments and recommendations. This data while sometimes easier to produce places additional burden on data processing to provide accurate interpretation. Ensuring inputs are structured in a format that is more easily interpreted programmatically reduces the chance for error, however attention must be paid to the user experience in the development and rollout of any changes. Additionally, systems that are used to track data within hospitals are often hard to integrate with and have stringent privacy protections. Aggregating data from these systems requires understanding and collaboration between additional stakeholders and parter companies to ensure positive outcomes.

Data Sharing and Sensitive Information

To fully realise some of the opportunities identified within this article, information would need to be shared between external partners such as, suppliers, data service providers, data warehousing and team members. Ensuring that the data is secure, only the information required and does not expose individuals to potential privacy issues is paramount. An ongoing project and ethics review process should be implemented to ensure these values are maintained.

Final Thoughts

Moving forward with increased data utilisation in healthcare is likely inevitable, ensuring the data is utilised to benefit patients and healthcare workers first requires a considered approach. A combination of Data Science and Human Centred Design frameworks could be applied to ensure that not only are data projects effective but that they empower the people at the core of healthcare and further drive innovation in the sector. However, as always in the healthcare environment careful attention must be paid to the ethical and privacy challenges presented by these projects.

References

Bagchi, P. K., Raghunathan, T., & Bardi, E. J. (1987). The Implications of Just-in-Time Inventory Policies on Carr. Logistics and Transportation Review, 23(4), 373.

Chakraborty, K., Mehrotra, K., Mohan, C. K., & Ranka, S. (1992). Forecasting the behavior of multivariate time series using neural networks. Neural Networks, 5(6), 961–970.

Choudhary, R., Bafna, S., Heo, Y., Hendrich, A., & Chow, M. (2010). A predictive model for computing the influence of space layouts on nurses’ movement in hospital units. Journal of Building Performance Simulation, 3(3), 171–184. https://doi.org/10.1080/19401490903174280

EQUALS, P. S. (2013). Aseptic Technique, Surgical Conscience and Time Out. Surgical Technologist.

Hardgrave, B. C., Aloysius, J., & Goyal, S. (2009). Does RFID improve inventory accuracy? A preliminary analysis. International Journal of RF Technologies: Research and Applications, 1(1), 44–56. https://doi.org/10.1080/17545730802338333

Johnson, P. D., Martin, R., Burrell, L. J., Grabsch, E. A., Kirsa, S. W., O’Keeffe, J., … Bolger, C. (2005). Efficacy of an alcohol/chlorhexidine hand hygiene program in a hospital with high rates of nosocomial methicillin‐resistant Staphylococcus aureus (MRSA) infection. Medical Journal of Australia, 183(10), 509–514.

Kotter, J. P., & Cohen, D. S. (2002). The heart of change: real-life stories of how people change their organizations. Boston, Mass: Harvard Business School Press.

LaValle, S., Lesser, E., Shockley, R., Hopkins, M. S., & Kruschwitz, N. (2011). Big data, analytics and the path from insights to value. MIT Sloan Management Review, 52(2), 21.

Popovič, A., Hackney, R., Coelho, P. S., & Jaklič, J. (2012). Towards business intelligence systems success: Effects of maturity and culture on analytical decision making. Decision Support Systems, 54(1), 729–739. https://doi.org/10.1016/j.dss.2012.08.017

Roberti, Mark. (2016, October 31). How Accurate Can RFID Tracking Be? [RFID Journal]. Retrieved 15 April 2019, from https://www.rfidjournal.com/blogs/experts/entry?11454

Sangwan, R. S., Qiu, R. G., & Jessen, D. (2005). Using RFID tags for tracking patients, charts and medical equipment within an integrated health delivery network. Proceedings. 2005 IEEE Networking, Sensing and Control, 2005., 1070–1074. https://doi.org/10.1109/ICNSC.2005.1461346

Saygin, C. (2007). Adaptive inventory management using RFID data. The International Journal of Advanced Manufacturing Technology, 32(9–10), 1045–1051.

Stanton, P., Willis, E., & Young, S. (Eds.). (2005). Workplace reform in the healthcare industry: the Australian experience. New York: Palgrave Macmillan.

Tarnay, C. (1999). Incision characteristics associated with six laparoscopic trocar-cannula systems: a randomized, observer-blinded comparison. Obstetrics & Gynecology, 94(1), 89–93. https://doi.org/10.1016/S0029-7844(99)00287-2

Wang, J., Amos, B., Das, A., Pillai, P., Sadeh, N., & Satyanarayanan, M. (2017). A Scalable and Privacy-Aware IoT Service for Live Video Analytics. Proceedings of the 8th ACM on Multimedia Systems Conference — MMSys’17, 38–49. https://doi.org/10.1145/3083187.3083192