Hand-washing seems like a mundane topic until you realise that hundreds of thousands of lives could potentially be saved, just by improving hand-hygiene among healthcare workers.
It is a deceptively complex problem to solve.
Globally, hand hygiene compliance among healthcare workers is estimated to be only 38.7%. Even though there is a WHO protocol for hand-washing, training is often patchy and easily forgotten. The problem is particularly severe in Indian public hospitals as overloaded healthcare workers rush from one patient to the next.
In order to solve this problem, Quicksand, a Delhi-based design studio, partnered with Jhpiego, a non-profit associated with John Hopkins University, and AI/ML startup Datakalp for a project called Safehands, funded by USAID’s MOMENTUM Country and Global Leadership project. The project was conducted in labour rooms at district hospitals and community health centres across India, and aimed to improve hand washing compliance by healthcare workers.
When Tech Needs Support
In June 2021, the COVID-19 pandemic brought hand washing into the public imagination like never before, and made compliance even more critical in healthcare settings. Traditionally, most hospitals use manual training and supervision to improve hand-washing, but this approach is costly, and evaluations are prone to error and bias.
To tackle this problem, Datakalp created an AI-powered device called HAIgenie, (formerly known as Vajra Hands) consisting of a camera to capture hand movements, a processing unit that uses machine learning algorithms to analyse the images, and a screen to provide real-time feedback based on the WHO protocol to the user. The system also provided aggregated insights to hospital management. Despite the promise of the technological solution, however, adoption was lower than expected, and in some hospitals, a few healthcare workers actively avoided using the washbasin with the system.
Human-Centric Design Interventions Identified
Quicksand conducted a design research program involving focus groups, interviews, and observations at labour rooms in six hospitals in Madhya Pradesh, Chhattisgarh and Jharkhand, working on the ground to develop a contextual understanding of hand-washing, and understand barriers to device usage. The team identified the following product design improvement areas as well as supporting interventions.
1. Perception of Technology
During the interview process, several respondents shared concerns about the camera and vaguely expressed fears of being watched by higher authorities, with comments such as “Wo dekh rahe hai“ (They are watching). The Safehands team worked to dispel these concerns through orientation sessions to explain the machine learning technology and its purpose without complex jargon. These sessions helped to improve the healthcare workers relationship with the device, and explore how they could leverage it in their routines instead.
The team also incorporated a friendly mascot called the Safehands Sakhi throughout the system, designed to emulate the spirit of nurses who were often approached by their colleagues with questions. This helped shift the perception of the system from a strict supervisor to a friend.
2. Understanding the Screen and Hand Washing Protocol
The original design used visuals on the screen to show the user what to do next, but the team found that many healthcare workers did not understand the visual instructions shown on the screen, which involved a series of images arranged clockwise, with arrow marks indicating which one is next.
It was also difficult to look at the screen and wash hands at the same time, leading to errors. The team redesigned the interface to be more intuitive and showed simple graphics one at a time while the user washed their hands.
“We often think of literacy as reading text. But visual language matters too. Your users may not be familiar with the graphics and symbols that are otherwise common or popular with an urban, tech savvy audience. ”Niyoshi Shah, Consultant, Quicksand
3. Shifting the Design from a Supervisory to Self-Learning Tool
While the team was figuring out how to address surveillance concerns, they noticed a small subset of workers who had the opposite view. They appreciated how this technology could help them learn independently and make mistakes without fear of being judged or shamed by supervisors.
The new interface was designed to be a friendly tutorial. Earlier, feedback would only appear after a user had completed a step. However, in the new design, each step of the protocol is prominently displayed on the screen, guiding the user along. Users simply need to follow each step for a set duration to earn a green tick. Even if they make a mistake, the system simply shows them a cross mark, allows them to continue through the rest of the process, and presents a score at the end based on their performance in the six core steps.
This change has led to great feedback, making the team explore using the device in skill labs and other educational settings within public health too.
4. Performance Reports and Dashboards
The system included a dashboard to help hospital management monitor handwashing and develop data-informed strategies to improve compliance. However, supervisors did not use the original dashboard as much as intended, as it was number-heavy and hard to read.
The new design provides easy-to-understand visualisations on the dashboard in English, and supplements this with reports shared over WhatsApp in English and Hindi.
The team also introduced a rewards system, providing prizes such as a cooling water dispenser, chest of drawers and certificates to wards that got more points over a two month period. This also helped underline that individuals were not identified by the system.
5. Non-digital Interventions
In addition to design changes in the digital system, the Quicksand team also developed various offline interventions to orient staff and management, designed to be experiential instead of lecture-based.
The team designed card games, with each card depicting one step of the WHO hand-washing protocol. The staff was challenged to arrange these cards in the right order. This game was well received, even by those who hadn’t received formal training on the WHO protocol, but could instantly grasp the sequence to their delight.
Impact and Future of the Safehands Project
The design updates and supporting interventions were a success, leading to a 150% average increase in hand washing compliance according to the WHO protocol across sites. The team now plans to explore the use of the HAIgenie device in educational settings such as medical colleges, which experience high foot traffic and require scalable solutions.
Learnings & Lessons
Although HAIgenie was developed for a healthcare setting, the Quicksand team believes there are learnings for designing tech systems in any sector, particularly if they are to be used by large populations who are often new to digital tools.
1. Taking on responsibility for data privacy and ethics
India does not have any law on data privacy, but systems that require it have developed organically at a rapid pace. Any team that works with new technology should therefore dedicate resources and time to plug these gaps themselves and prioritise data protection.
2. Allocate time for machine learning and set expectations
Most users expect machine learning or AI based tools to provide 100% repeatable results just like a physical machine. AI product designers should work to sensitise customers about the machine learning and training time required before the system reaches an effective level of accuracy and implement a system to quickly learn from initial errors reported by end users. It is also important to budget and plan for a grievance redressal system to build and maintain users’ confidence in the product.
3. Involving end-users from the beginning is key
The Safehands project revealed that certain aspects of the technology could have been improved from the start if end-users were consulted more thoroughly.. For example, staff in many facilities have developed their own hand-washing protocols which have been passed on for a long time. The The team believes this shows a need to build more flexibility in algorithms and tech to accommodate such nuances.
Testing early prototypes with end-users can help incorporate such insights before it is too late in the development cycle.
“Our tech solutions should be responsive. They need to change and evolve as people’s needs change. If it stays static, and doesn’t assimilate into their environment, the solution will fall out of use very soon. We need to pay equal attention to adoption and accuracy in developing AI systems.”Arshmeen Baveja, Senior Associate, Quicksand
The Quicksand team’s parting advice is to focus on developing technology that is more people-centric rather than just problem-centric.
“AI has the potential to solve a lot of long term challenges for the Indian Public Health System. But in the short to medium term, considered and consistent investments need to be made into experiments that make this technology more people centred and resilient in the face of resource constraints.”Rishabh Sachdeva, Principal, Quicksand
The THC Take
During an onslaught of artificially intelligent products, the Safehands project is a great reminder about the importance of human users in every system, and the larger environment every product fits into. The design team’s ability to flip the perception of an AI-powered device – from that of surveillance to a helpful tool – demonstrates the two sides of all new technology, and how design could decide which side the coin lands on.