Abstract
Objectives
Evaluate whether online search patterns are different in malignant and benign gynecological diagnoses. Determine whether online search data (OSD) can enable the earlier detection of gynecological cancer.
Methods
This is a prospective cohort study, evaluating OSD in symptomatic individuals (Google users). They were referred with suspected cancer to Imperial College Healthcare NHS Trust, between December 2020-June 2022. OSD (24 months) was extracted via a Google Takeout (GT) file and pseudo-anonymised. A health-filter was applied to extract relevant data. Clinical data including age and clinical/histological diagnosis were extracted from clinical records. A focused symptom questionnaire was completed. OSD from various time intervals (630-to-0 days) before GP referral were utilised to build vector-space models to predict outcome (malignant). Area under the ROC curve (AUC) was used to evaluate model performance.
Results
255 patients were enrolled, and 20 were excluded due to empty GT files, resulting in a cohort of 235 patients with a median age of 53 (range 20-81). The rate of malignancy was 26.0%, with 42 ovarian (68.9%) and 15 endometrial cancers (24.6%) respectively. The OSD-based model had a predictive signal (AUC 0.64) for malignancy 360 days before GP referral. The best performing OSD-based model, (630-to-60 days), reached an AUC of 0.82 at 60 days before GP referral, in individuals who searched for medical conditions (n = 153, 65.1%). The questionnaire-based model comparatively had a lower predictive performance (AUC 0.62).
Conclusions
This study indicates that OSD appears to be different between individuals with a benign and malignant gynecological diagnosis. Furthermore, there appears to be a predictive signal in advance of GP referral date, which could be utilised to enable the earlier detection of gynecological cancer. An OSD-based model could provide real-time, individualised gynecological cancer risk profiles and has potential as an accessible disease screening tool.
Evaluate whether online search patterns are different in malignant and benign gynecological diagnoses. Determine whether online search data (OSD) can enable the earlier detection of gynecological cancer.
Methods
This is a prospective cohort study, evaluating OSD in symptomatic individuals (Google users). They were referred with suspected cancer to Imperial College Healthcare NHS Trust, between December 2020-June 2022. OSD (24 months) was extracted via a Google Takeout (GT) file and pseudo-anonymised. A health-filter was applied to extract relevant data. Clinical data including age and clinical/histological diagnosis were extracted from clinical records. A focused symptom questionnaire was completed. OSD from various time intervals (630-to-0 days) before GP referral were utilised to build vector-space models to predict outcome (malignant). Area under the ROC curve (AUC) was used to evaluate model performance.
Results
255 patients were enrolled, and 20 were excluded due to empty GT files, resulting in a cohort of 235 patients with a median age of 53 (range 20-81). The rate of malignancy was 26.0%, with 42 ovarian (68.9%) and 15 endometrial cancers (24.6%) respectively. The OSD-based model had a predictive signal (AUC 0.64) for malignancy 360 days before GP referral. The best performing OSD-based model, (630-to-60 days), reached an AUC of 0.82 at 60 days before GP referral, in individuals who searched for medical conditions (n = 153, 65.1%). The questionnaire-based model comparatively had a lower predictive performance (AUC 0.62).
Conclusions
This study indicates that OSD appears to be different between individuals with a benign and malignant gynecological diagnosis. Furthermore, there appears to be a predictive signal in advance of GP referral date, which could be utilised to enable the earlier detection of gynecological cancer. An OSD-based model could provide real-time, individualised gynecological cancer risk profiles and has potential as an accessible disease screening tool.
| Originalsprog | Engelsk |
|---|---|
| Artikelnummer | OC12.06 |
| Tidsskrift | Ultrasound in Obstetrics & Gynecology |
| Vol/bind | 62 |
| Udgave nummer | S1 |
| Sider (fra-til) | 29-29 |
| ISSN | 0960-7692 |
| DOI | |
| Status | Udgivet - 2023 |