Modelo predictivo para turnos de salud

The City of Mendoza is piloting a predictive model to reduce missed appointments in its municipal primary healthcare network. The challenge is significant: analysis of more than 193,000 appointments between 2021 and 2025 found a no-show rate of 21.2%. At the same time, approximately 49% of appointment requests were rejected because no slots were available. Missed appointments therefore represent not only an efficiency problem, but also a barrier to making scarce healthcare capacity available to other residents.

Developed through the LAC AI Accelerator in collaboration with the World Bank and NTT DATA, the solution uses a calibrated logistic regression model running on the City’s Microsoft Fabric environment. It combines historical attendance patterns with characteristics of each appointment to estimate the risk of non-attendance. Results are displayed through an operational Power BI dashboard so municipal staff can use the information to support preventive follow-up and communication.

The pilot combines predictive analytics with an experimental evaluation. It begins in one Primary Health Care Center and progressively expands to the four-center municipal network. Randomized control trials will compare the usual process with different reminder strategies, including messages that facilitate early cancellation or rescheduling. Importantly, risk scores do not determine experimental assignment, allowing the City to separately evaluate both the predictive performance of the model and the causal impact of the interventions.

As this is an active pilot, impact is being measured rather than claimed in advance. The main target is to reduce the no-show rate by five percentage points; based on historical volumes, this could recover approximately 20–25 appointments per week across the municipal network. Additional targets include timely preventive contact and earlier release of appointments that patients cannot use.

A distinctive element of the project is its governance framework. In September 2026, the City formally adopted a Responsible Use and Transparency Protocol and a Public Algorithmic Transparency Sheet, establishing human oversight, traceability, performance and equity monitoring, risk management, security safeguards and public transparency. The model is explicitly limited to decision support and cannot autonomously cancel appointments or restrict access to healthcare.

The pilot is implemented under the City of Mendoza’s responsible AI and data governance framework. Ordinance No. 4221/2025 establishes principles of transparency, explainability, security and human oversight for municipal AI. Resolution No. 1/2026 introduced a specific governance framework for this model, including human oversight, traceability, risk management, performance and equity monitoring, security safeguards and public algorithmic transparency. The model is explicitly limited to decision support and cannot autonomously affect access to healthcare.