Related Experiment Video
Updated: Aug 7, 2026
A Machine Learning Approach to Design an Efficient Selective Screening of Mild Cognitive Impairment
Published on: January 11, 2020
[Validation of a screening questionnaire for stroke detection in Spanish-speaking communities]
O H Del Brutto1, L Idrovo, A Mosquera
1Departamento de Ciencias Neurológicas, Hospital-Clínica Kennedy, Guayaquil, Ecuador. odb@gye.satnet.net
Objective:
To evaluate the accuracy of an screening questionnaire for stroke detection in speaking-speaking communities.
Patients And Methods:
We performed a door-to-door survey in Atahualpa (a rural community in coastal Ecuador) using a modified version of the Spanish translation of the OMS questionnaire for stroke detection. Subjects suspected of having a stroke as well as a 2% sample of negative subjects were evaluated by neurologists to evaluate sensitivity, specificity, and predictive value of the questionnaire and that of each of its questions.
Results:
We found 18 possible cases among 1,568 individuals around 15 years old. Of these, 10 were confirmed stroke patients and 8 were false-positive (crude stroke prevalence of 6.38 per 1,000). We did not find false-negative cases. Sensitivity of the questionnaire was 100%, specificity was 99.5%, positive predictive value was 0.55 and negative predictive value was 1. The accuracy of each question as well as the number of questions answered as affirmative were different between patients and false-positive cases.
Conclusions:
The current questionnaire is highly sensitive but its positive predictive value is poor. This causes problems in large-scale studies, as the detection of many false-positives may compromise its viability. We propose a modification of the questionnaire that will turn it more accurate.
