Seroprevalence of Human T-Cell Lymphotropic Virus-1 in a Jamaican Antenatal Population and Assessment of Pooled Testing as a Cost Reduction Strategy for Implementation of Routine Antenatal Screening.

Mother to child transmission (MTCT) of human T-cell lymphotropic virus (HTLV)-1 is associated with increased risk of adult T-cell leukemia and can be unrecognized without routine antenatal screening. We assessed the seroprevalence of HTLV-1/2 among pregnant women attending The University Hospital of the West Indies Antenatal Clinic, 2019, and validated a cost-effective strategy to screen antenatal clinic attendees for HTLV-1/2. Residual antenatal samples from 370 women were tested for HTLV-1/2 by chemiluminescence microparticle immunoassay (CMIA). Six samples were confirmed HTLV-1 positive by Western blot (none for HTLV-2) for a prevalence of 1.62%. Four mother-child pairs were able to be recruited for HTLV testing of children, with two children testing HTLV-1/2 positive. Medical records of HTLV-1-infected women revealed that all women breastfed, indicating an unrecognized risk for HTLV MTCT. To assess whether pooling of samples as a cost-reduction strategy could be introduced, we pooled all antenatal samples received between November and December 2021 into 12 pools of eight samples/pool. Two pools were CMIA positive, and de-pooling of samples identified two CMIA-positive samples (one per pool), both confirmed as HTLV-1 by Western blot. These results indicate that HTLV-1 remains prevalent in pregnant Jamaican women and that sample pooling can be a cost-effective strategy to limit MTCT in Jamaica.

AI pitfalls and what not to do: mitigating bias in AI.

Various forms of artificial intelligence (AI) applications are being deployed and used in many healthcare systems. As the use of these applications increases, we are learning the failures of these models and how they can perpetuate bias. With these new lessons, we need to prioritize bias evaluation and mitigation for radiology applications; all the while not ignoring the impact of changes in the larger enterprise AI deployment which may have downstream impact on performance of AI models. In this paper, we provide an updated review of known pitfalls causing AI bias and discuss strategies for mitigating these biases within the context of AI deployment in the larger healthcare enterprise. We describe these pitfalls by framing them in the larger AI lifecycle from problem definition, data set selection and curation, model training and deployment emphasizing that bias exists across a spectrum and is a sequela of a combination of both human and machine factors.

Opportunistic detection of type 2 diabetes using deep learning from frontal chest radiographs.

Deep learning (DL) models can harness electronic health records (EHRs) to predict diseases and extract radiologic findings for diagnosis. With ambulatory chest radiographs (CXRs) frequently ordered, we investigated detecting type 2 diabetes (T2D) by combining radiographic and EHR data using a DL model. Our model, developed from 271,065 CXRs and 160,244 patients, was tested on a prospective dataset of 9,943 CXRs. Here we show the model effectively detected T2D with a ROC AUC of 0.84 and a 16% prevalence. The algorithm flagged 1,381 cases (14%) as suspicious for T2D. External validation at a distinct institution yielded a ROC AUC of 0.77, with 5% of patients subsequently diagnosed with T2D. Explainable AI techniques revealed correlations between specific adiposity measures and high predictivity, suggesting CXRs' potential for enhanced T2D screening.

Adaptation and psychometric testing of the end-of-life professional caregiver survey in Jamaica.

BACKGROUND: Using a validated instrument to measure palliative care (PC) educational needs of health professionals is an important step in understanding how best to educate a well-versed PC workforce within a national health system. The End-of-life Professional Caregiver Survey (EPCS) was developed to measure U.S. interprofessional PC educational needs and has been validated for use in Brazil and China. As part of a larger research project, this study aimed to culturally adapt and psychometrically test the EPCS among physicians, nurses, and social workers practicing in Jamaica. METHODS: Face validation involved expert review of the EPCS with recommendations for linguistic item modifications. Content validation was carried out by six Jamaica-based experts who completed a formal content validity index (CVI) for each EPCS item to ascertain relevancy. Health professionals practicing in Jamaica (n = 180) were recruited using convenience and snowball sampling to complete the updated 25-item EPCS (EPCS-J). Internal consistency reliability was assessed using Cronbach's [Formula: see text] coefficient and McDonald's [Formula: see text]. Construct validity was examined through confirmatory factor analysis (CFA) and exploratory factor analysis (EFA). RESULTS: Content validation led to elimination of three EPCS items based on a CVI < 0.78. Cronbach's [Formula: see text] ranged from 0.83 to 0.91 and McDonald's [Formula: see text] ranged from 0.73 to 0.85 across EPCS-J subscales indicating good internal consistency reliability. The corrected item-total correlation for each EPCS-J item was > 0.30 suggesting good reliability. The CFA demonstrated a three-factor model with acceptable fit indices (RMSEA = 0.08, CFI = 0.88, SRMR = 0.06). The EFA determined a three-factor model had the best model fit, with four items moved into the effective patient care subscale from the other two EPCS-J subscales based on factor loading. CONCLUSIONS: The psychometric properties of the EPCS-J resulted in acceptable levels of reliability and validity indicating that this instrument is suitable for use in measuring interprofessional PC educational needs in Jamaica.