This major Malaysian hospital group operated nine private hospitals across Peninsular Malaysia and East Malaysia, collectively handling over 620,000 emergency department visits per year. Triage was performed by nursing staff using a modified version of the Canadian Triage and Acuity Scale, but inter-rater reliability assessments revealed a 26% disagreement rate on triage classifications between nurses across different facilities and shifts.
Average emergency department wait times had climbed to 52 minutes for non-critical presentations, contributing to a patient satisfaction score of only 58% — well below the group's target of 80%. More alarmingly, a clinical governance review identified seven cases of under-triage over the previous 12 months where patients with time-sensitive conditions such as acute coronary syndrome or sepsis were initially classified as lower priority. Two of these cases had resulted in adverse patient outcomes that triggered Ministry of Health investigations.
The hospital group's medical director had explored computerized triage support tools from international vendors, but these systems were trained predominantly on Western patient populations and did not account for the different disease prevalence patterns, patient presentation norms, and cultural communication styles common in Malaysia's diverse patient demographic.
Pertama Partners began with an AI Readiness Audit that included clinical observation sessions across four of the group's busiest emergency departments. We shadowed triage nurses through over 300 patient encounters to document the clinical reasoning process, identify common decision challenges, and understand the contextual factors that influenced triage accuracy — including language barriers, cultural reluctance to express pain, and the high prevalence of tropical diseases uncommon in Western training datasets.
Our AI Pilot Program developed a clinical decision support system trained on the hospital group's own electronic health records, which contained structured data on over 2.4 million historical emergency visits. The model incorporated presenting complaints, vital signs, patient demographics, medical history, medication lists, and Malaysia-specific epidemiological patterns. It was designed to present a recommended triage level alongside the three most clinically similar historical cases with their outcomes, giving nurses both a recommendation and context for their decision.
We deployed the pilot at three facilities with extensive Team Training for triage nurses. Training emphasized that the AI was a safety net — not a replacement for clinical judgment — and included scenario-based exercises where nurses practiced integrating AI recommendations with their own assessment. Weekly calibration sessions allowed nurses to discuss cases where they agreed or disagreed with the AI, building both confidence and refinement of the system.
"In emergency medicine, minutes matter and consistency saves lives. Pertama Partners built a system that respects our clinical expertise while catching the cases that might slip through during a chaotic night shift."— Dr. Aisha Binti Rahman, Group Chief Medical Officer
Every transformation starts with a conversation. Let's discuss your challenges and opportunities.
Discuss Your Challenge