CSympData: Expert Annotated Patient Symptoms Data

Access to timely and accurate healthcare guidance remains a challenge, particularly in Low- and Middle-Income Countries (LMICs). To help bridge this gap, AI-powered symptom checkers can provide preliminary guidance, helping individuals make a critical decision: whether to seek professional medical care or if self-management with over-the-counter (OTC) medication is appropriate.To support the development and validation of these AI-driven clinical triage tools, we introduce CSympData, a large-scale, expert-annotated dataset. Crucially, this dataset does not contain real patient records. Instead, it consists of 130,637 synthetically generated, clinically plausible patient profiles. Each profile was meticulously crafted by combining symptoms, demographic attributes (age group, gender), symptom duration, and overall severity to represent a realistic medical scenario. These profiles were then validated and labeled by teams of medical experts, who provided a final recommendation: either OTC Drug or Doctor Consultation. The process was informed by disease prevalence data and symptom knowledge from sources like the International Classification of Diseases, 11th Revision (ICD-11), ensuring the scenarios are relevant and aligned with global healthcare standards.<br><br>List of Dataset Attributes:<br>Symptoms: (String) A list of symptoms presented in the profile.Gender: (Categorical) The patient's gender (Male, Female).Age_Group: (Categorical) The patient's binned age group, categorized based on clinical and public health relevance (Below 5 years, 6-15 years, 16-45 years (Female), 16-60 years (Male), Above 45 years (Female), Above 60 years (Male)).Duration: (Categorical) Binned duration of symptoms (&lt;= 3 days, &gt; 3 days).Severity: (Categorical) The holistic severity of the case as assessed by an expert (Mild, Moderate, Severe).Final_Recommendation (Target Label): (Categorical) The expert-validated triage recommendation (OTC Drug or Doctor Consultation).