SafeHands

Research questions

SafeHands is a comprehensive research initiative dedicated to optimizing hand hygiene within public health environments. By examining the complex interplay of structural barriers, knowledge gaps, and behavioral motivators, the project seeks to understand the factors shaping compliance.

A core component of the study involves evaluating the effectiveness of AI-driven feedback systems designed to monitor and improve adherence to global WHO handwashing protocols. Through a collaborative approach, the project ultimately aims to design and implement a range of digital and analog interventions that are both contextually relevant and deeply rooted in the daily realities of healthcare professionals.

Methodology

Foundational research Quicksand, with support from Jhpeigo and DataKalp, conducted qualitative research with 30+ stakeholders across labour room staff cadres and management levels through remote in-depth interviews, followed by field observations, and focused group discussions. The team approached the development of research tools by relying on storytelling instead of emphasising  on technical terminology. This decision was shaped by the formative research that indicated low data literacy and limited time between different tasks.

To synthesise the findings from primary research, the team applied the Manoff Group's Toolkit for Behaviour Integration. This toolkit became central in understanding the factors impacting two distinct behaviours in the handwashing patterns of staff members: uptake of VajraHands and adherence to WHO protocols. These factors were prioritized based on importance and feasibility, some key barriers included surveillance concerns, confusing interface design and decision logic, as well as technical inconsistencies. However, the research also pointed towards some enabling factors, most importantly—that labour room in-charges could train teams using real-time feedback, and duration information to encourage self-monitoring.

Co-design process

Following research synthesis, brainstorming sessions with stakeholders were organized to ideate on behavior change frameworks. A vital development were the concept cards used for qualitative research to help interviewees visualise potential interventions, based on feedback gathered from the hospital sites.  The design team took the intervention ideas that staff responded positively to during concept testing and checked them against the behavioral factors—the barriers and enablers—identified during research. This ensured that the redesigned interface and complementary activities directly addressed the real obstacles and motivations affecting handwashing compliance.

VajraHands 2.0

Interface

Rather than providing retrospective feedback after handwash completion, the new interface adopted a sequential model where users performed each step as it appeared on screen. Custom animated GIFs were replaced with static graphics, and red crosses appeared instantly for incorrect steps, in contrast with the old interface that indicated errors after completion of handwash.

A "SafeHands Sakhi" mascot also provided encouragement, adapting from observations indicating that health workers rely heavily on peer learning. Visual tools—enlarged step demonstrations, real-time progress feedback along with educational screensavers—and replaced instructions in Hindi were used to create effective communication between staff members and the tech tool.

The launch of VajraHands 2.0 was complemented by an orientation for the staff and management at all sites These complementary activities, on the other hand, became imperative in addressing larger organisational and motivational factors which included adult learning principles, interactive activities, and video resources.

Differentiated intervention groups

Group A: Training of Trainers (ToT) sessions equipped selected staff and management to conduct peer learning on hand hygiene and team cohesion using provided resource kits. Management participated in regular meetings discussing performance reports and establishing targets. Facilities collected patient feedback on ward hand hygiene. A six-week competition offered rewards based on compliance and improvement points.

Group B (Lean): Received orientation and performance reports only. This stratification enabled assessment of inclusion, cohesion, and varied accountability mechanisms such as meetings, self-driven targets, patient feedback, and rewards on hand hygiene outcomes.

Implications

The SafeHands project demonstrates both potential and limitations of AI-enabled behavior change systems in resource-constrained settings. VajraHands proved most effective when integrated with human-mediated supervision rather than operating as standalone solution. The transition from version 1.0 to 2.0 illustrates how interface design and decision logic transparency fundamentally shape user trust and adoption.

Hospital context created differential effects: larger facilities serving broader geographical areas faced greater challenges maintaining protocol adherence under time pressure. While performance visibility generated initial behavior change, sustainability demanded sustained management commitment and integration into institutional quality assurance processes. The research establishes groundwork for scaling AI-enabled systems while highlighting necessary conditions: responsive interface design, integration with existing infrastructure, management capacity building, and attention to power dynamics within hierarchical health systems.