Can the Net Promoter Score predict the Primary Care Assessment Tool? A transfer learning approach

Area of Science:

• Healthcare Management
• Health Services Research
• Predictive Analytics

Background:

• Assessing primary care quality is crucial for effective healthcare delivery.
• The Primary Care Assessment Tool (PCAT) is a comprehensive measure, but its length can be a limitation.
• Exploring parsimonious predictors like the Net Promoter Score (NPS) can streamline quality assessments.

Purpose of the Study:

• To evaluate the Net Promoter Score (NPS) as a concise predictor for the Primary Care Assessment Tool (PCAT).
• To identify a minimal set of variables that accurately estimate PCAT scores for adult and pediatric populations.

Main Methods:

• A cross-sectional survey data from Rio de Janeiro (2024) was used, with samples for adults and children.
• Machine learning algorithms were employed to build predictive models using NPS and other variables.
• Data was split into training (70%) and testing (30%) sets for model validation.

Main Results:

• A parsimonious model using NPS and few additional variables (e.g., age, education, consistent doctor) accurately predicted PCAT scores.
• Predictive model performance was superior in the children population (AUC=0.82) compared to adults (AUC=0.76).
• Transfer learning demonstrated robust adaptation of adult models to pediatric data.

Conclusions:

• The NPS, combined with a small set of demographic and service-related variables, is a viable and parsimonious predictor of PCAT scores.
• This simplified approach can enhance the efficiency of primary care quality assessment.
• The findings support the use of NPS-based models for routine healthcare quality monitoring.