Applying a Device for Artificial Intelligence Decision Support in practice during screening for cervical cancer in Bangladesh and Uganda: a CFIR analysis (Preprint)
Bibliographic record
Abstract
Abstract Background Screening is important for early detection of cervical cancer in low- and middle-income countries. Visual inspection with acetic acid (VIA) is usually the method of choice in these settings. However, interpretation of VIA results is subject to interobserver and intraobserver variability. AI decision support systems (AI-DSSs) could contribute to better decisions by health workers. Objective The aim of this study was to analyze the barriers and facilitators of introducing an AI-DSS device under field conditions in the context of VIA screening in rural Bangladesh and Uganda, with the goal of improving the operational systems of applying an AI-DSS device. Methods We operationalized the Consolidated Framework for Implementation Research for this specific study and defined the constructs for analysis. The study was performed in rural Uganda and Bangladesh. We extracted relevant information from routine data, patient surveys, and facility surveys. We also interviewed health workers who used the AI-DSS devices. A panel of experts performed an analysis of the quality of pictures and the performance of health workers. We monitored implementation through quarterly meetings and documented the process. This trial was registered under ClinicalTrials.gov identifier NCT05234112. Results The applied AI-DSS passed tests under laboratory conditions but performed less well under field conditions. The hardware design using an adapted mobile phone was adequate, and the user interface was user-friendly and intuitive. However, operating the device while performing VIA in clinics was challenging. Since the device was not registered as a medical device, it was only used for research purposes. In both countries, there was no official government policy concerning the use of AI in health care. All facilities had gynecological examination rooms, but some facilities did not have permanent electricity or internet connection. The normal procedure for VIA was followed, and AI-DSS did not interfere with routine screening procedures. The AI-DSS was well accepted by women when privacy was guaranteed, but there seemed to be more trust in the judgment of health workers. The pictures helped supervisors to give a second opinion from a distance and were used for training purposes. The technical support team was able to help remotely and improved the performance. Training health workers in taking good pictures was very important. A permanent monitoring process during implementation was established, which led to early detection and correction of shortfalls. Conclusions According to this study, the device has the potential for improving the quality of the assessment by health workers who perform VIA. However, competent staff trained in performing VIA will be indispensable for capturing high-quality pictures. The algorithm requires further fine-tuning. This study showed that the Consolidated Framework for Implementation Research is a proper tool for categorizing barriers and facilitators when introducing an AI-DSS device in practice.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.029 | 0.105 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.004 | 0.006 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.004 | 0.003 |
| Open science | 0.001 | 0.004 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.005 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".