In the current era of technology, the use of data processing has become an undeniable necessity, especially in the health context. One of the uses in managing health data is the classification of disease text data, such as cardiovascular disease, therefore this research aims to evaluate the performance of the KNN method using 3 types of distance measurements, namely Euclidean, Manhattan/City Block, and Mahalanobis. Although simple, this algorithm has succeeded in producing high performance in some cases. This KNN approach classifies an object based on its similarity or closest distance to the objects in the training data. The research produces accuracy values obtained from variations in the k value from the numbers 1 to 31 with odd multiples. Results of calculating Euclidien, Manhattan and Mahalanobis distances on K-NN accuracy. The more the k value increases the accuracy, even though there are conditions where the accuracy decreases for a certain K value, although it is not significant. The lowest total accuracy was obtained from the Euclidien distance, while the highest total accuracy was achieved by the Manhattan distance, followed by the Mahalanobis distance. The Manhattan and Mahalanobis distances produce the best total accuracy and produce the best accuracy in most K sizes, while Euclidien is the calculation with the best total accuracy. Increasing the number of K in each distance calculation can increase classification accuracy. The optimal number of K in this experiment was 29, indicating the highest effectiveness. It is important to note that selecting an appropriate number of K can have a significant impact on classifier performance, and experimental results support the conclusion that by using K29 values, we can achieve the highest accuracy in fitting the model to the given data.