Publications

@article{nokey_67,

title = {Few-Shot Transfer Learning to improve Chest X-Ray pathology detection using limited triplets},

author = {Ananth Reddy Bhimireddy, John Lee Burns, Saptarshi Purkayastha, Judy Wawira Gichoya},

url = {https://arxiv.org/pdf/2204.07824.pdf},

year  = {2023},

date = {2023-04-16},

abstract = {Deep learning approaches applied to medical imaging have reached near-human or better-than-human performance on many diagnostic tasks. For instance, the CheXpert competition on detecting pathologies in chest x-rays has shown excellent multi-class classification performance. However, training and validating deep learning models require extensive collections of images and still produce false inferences, as identified by a human-in-the-loop. In this paper, we introduce a practical approach to improve the predictions of a pre-trained model through Few-Shot Learning (FSL). After training and validating a model, a small number of false inference images are collected to retrain the model using textbf{textit{Image Triplets}} - a false positive or false negative, a true positive, and a true negative. The retrained FSL model produces considerable gains in performance with only a few epochs and few images. In addition, FSL opens rapid retraining opportunities for human-in-the-loop systems, where a radiologist can relabel false inferences, and the model can be quickly retrained. We compare our retrained model performance with existing FSL approaches in medical imaging that train and evaluate models at once.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_60,

title = {Cost-effectiveness of low-dose aspirin for the prevention of preterm birth: a prospective study of the Global Network for Women’s and Children’s Health Research},

author = {Patterson JK, Neuwahl S, Goco N, Moore J, Goudar SS, Derman RJ, Hoffman M, Metgud M, Somannavar M, Kavi A, Okitawutshu J, Lokangaka A, Tshefu A, Bose CL, Mwapule A, Mwenechanya M, Chomba E, Carlo WA, Chicuy J, Figueroa L, Krebs NF, Jessani S, Saleem S, Goldenberg RL, Kurhe K, Das P, Patel A, Hibberd PL, Achieng E, Nyongesa P, Esamai F, Bucher S, Liechty EA, Bresnahan BW, Koso-Thomas M, McClure EM},

url = {https://www.thelancet.com/pdfs/journals/langlo/PIIS2214-109X(22)00548-4.pdf},

year  = {2023},

date = {2023-03-15},

abstract = {Background

Premature birth is associated with an increased risk of mortality and morbidity, and strategies to prevent preterm birth are few in number and resource intensive. In 2020, the ASPIRIN trial showed the efficacy of low-dose aspirin (LDA) in nulliparous, singleton pregnancies for the prevention of preterm birth. We sought to investigate the cost-effectiveness of this therapy in low-income and middle-income countries.



Methods 

In this post-hoc, prospective, cost-effectiveness study, we constructed a probabilistic decision tree model to compare the benefits and costs of LDA treatment compared with standard care using primary data and published results from the ASPIRIN trial. In this analysis from a health-care sector perspective, we considered the costs and effects of LDA treatment, pregnancy outcomes, and neonatal health-care use. We did sensitivity analyses to understand the effect of the price of the LDA regimen, and the effectiveness of LDA in reducing both preterm birth and perinatal death.



Findings 

In model simulations, LDA was associated with 141 averted preterm births, 74 averted perinatal deaths, and 31 averted hospitalisations per 10 000 pregnancies. The reduction in hospitalisation resulted in a cost of US$248 per averted preterm birth, $471 per averted perinatal death, and $15·95 per disability-adjusted life year.



Interpretation 

LDA treatment in nulliparous, singleton pregnancies is a low-cost, effective treatment to reduce preterm birth and perinatal death. The low cost per disability-adjusted life year averted strengthens the evidence in support of prioritising the implementation of LDA in publicly funded health care in low-income and middle-income countries.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_87,

title = {Birth weight and gestational age distributions in a rural Kenyan population},

author = {Sherri Bucher, Kayla Nowak, Kevin Otieno, Constance Tenge, Irene Marete, Faith Rutto, Millsort Kemboi, Emmah Achieng, Osayame A. Ekhaguere, Paul Nyongesa, Fabian O. Esamai & Edward A. Liechty },

url = {https://bmcpediatr.biomedcentral.com/articles/10.1186/s12887-023-03925-2},

year  = {2023},

date = {2023-03-08},

abstract = {Background

With the increased availability of access to prenatal ultrasound in low/middle-income countries, there is opportunity to better characterize the association between fetal growth and birth weight across global settings. This is important, as fetal growth curves and birthweight charts are often used as proxy health indicators. As part of a randomized control trial, in which ultrasonography was utilized to establish accurate gestational age of pregnancies, we explored the association between gestational age and birthweight among a cohort in Western Kenya, then compared our results to data reported by the INTERGROWTH-21st study.



Methods

This study was conducted in 8 geographical clusters across 3 counties in Western Kenya. Eligible subjects were nulliparous women carrying singleton pregnancies. An early ultrasound was performed between 6 + 0/7 and 13 + 6/7 weeks gestational age. At birth, infants were weighed on platform scales provided either by the study team (community births), or the Government of Kenya (public health facilities). The 10th, 25th, median, 75th, and 90th BW percentiles for 36 to 42 weeks gestation were determined; resulting percentile points were plotted, and curves determined using a cubic spline technique. A signed rank test was used to quantify the comparison of the percentiles generated in the rural Kenyan sample with those of the INTERGROWTH-21st study.



Results

A total of 1291 infants (of 1408 pregnant women randomized) were included. Ninety-three infants did not have a measured birth weight. The majority of these were due to miscarriage (n = 49) or stillbirth (n = 27). No significant differences were found between subjects who were lost to follow-up. Signed rank comparisons of the observed median of the Western Kenya data at 10th, 50th, and 90th birthweight percentiles, as compared to medians reported in the INTERGROWTH-21st distributions, revealed close alignment between the two datasets, with significant differences at 36 and 37 weeks. Limitations of the current study include small sample size, and detection of potential digit preference bias.



Conclusions

A comparison of birthweight percentiles by gestational age estimation, among a sample of infants from rural Kenya, revealed slight differences as compared to those from the global population (INTERGROWTH-21st).



},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_43,

title = {Azithromycin to Prevent Sepsis or Death in Women Planning a Vaginal Birth},

author = {Tita ATN, Carlo WA, McClure EM, Mwenechanya M, Chomba E, Hemingway-Foday JJ, Kavi A, Metgud MC, Goudar SS, Derman R, Lokangaka A, Tshefu A, Bauserman M, Bose C, Shivkumar P, Waikar M, Patel A, Hibberd PL, Nyongesa P, Esamai F, Ekhaguere OA, Bucher S, Jessani S, Tikmani SS, Saleem S, Goldenberg RL, Billah SM, Lennox R, Haque R, Petri W, Figueroa L, Mazariegos M, Krebs NF, Moore JL, Nolen TL, Koso-Thomas M; A-PLUS Trial Group

},

url = {https://www.nejm.org/doi/10.1056/NEJMoa2212111},

year  = {2023},

date = {2023-02-09},

abstract = {BACKGROUND

The use of azithromycin reduces maternal infection in women during planned cesarean delivery, but its effect on those with planned vaginal delivery is unknown. Data are needed on whether an intrapartum oral dose of azithromycin would reduce maternal and offspring sepsis or death.



METHODS

In this multicountry, placebo-controlled, randomized trial, we assigned women who were in labor at 28 weeks’ gestation or more and who were planning a vaginal delivery to receive a single 2-g oral dose of azithromycin or placebo. The two primary outcomes were a composite of maternal sepsis or death and a composite of stillbirth or neonatal death or sepsis. During an interim analysis, the data and safety monitoring committee recommended stopping the trial for maternal benefit.



RESULTS

A total of 29,278 women underwent randomization. The incidence of maternal sepsis or death was lower in the azithromycin group than in the placebo group (1.6% vs. 2.4%), with a relative risk of 0.67 (95% confidence interval [CI], 0.56 to 0.79; P<0.001), but the incidence of stillbirth or neonatal death or sepsis was similar (10.5% vs. 10.3%), with a relative risk of 1.02 (95% CI, 0.95 to 1.09; P=0.56). The difference in the maternal primary outcome appeared to be driven mainly by the incidence of sepsis (1.5% in the azithromycin group and 2.3% in the placebo group), with a relative risk of 0.65 (95% CI, 0.55 to 0.77); the incidence of death from any cause was 0.1% in the two groups (relative risk, 1.23; 95% CI, 0.51 to 2.97). Neonatal sepsis occurred in 9.8% and 9.6% of the infants, respectively (relative risk, 1.03; 95% CI, 0.96 to 1.10). The incidence of stillbirth was 0.4% in the two groups (relative risk, 1.06; 95% CI, 0.74 to 1.53); neonatal death within 4 weeks after birth occurred in 1.5% in both groups (relative risk, 1.03; 95% CI, 0.86 to 1.24). Azithromycin was not associated with a higher incidence in adverse events.



CONCLUSIONS

Among women planning a vaginal delivery, a single oral dose of azithromycin resulted in a significantly lower risk of maternal sepsis or death than placebo but had little effect on newborn sepsis or death.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_61,

title = {NTD Health: An electronic medical record system for neglected tropical diseases},

author = {Rodrigo Ochoa, Alessa Álvarez, Jordan Freitas, Saptarshi Purkayastha, Iván D Vélez},

url = {http://www.scielo.org.co/scielo.php?script=sci_arttext&pid=S0120-41572022000400602},

year  = {2022},

date = {2022-12-01},

abstract = {Introduction:



The use of technological resources to support processes in health systems has generated robust, interoperable, and dynamic platforms. In the case of institutions working with neglected tropical diseases, there is a need for specific customizations of these diseases.



Objectives:



To establish a medical record platform specialized in neglected tropical diseases which could facilitate the analysis of treatment evolution in patients, as well as generate more accurate data about various clinical aspects.



Materials and methods:



A set of requirements to develop state of the art forms, concepts, and functionalities to include neglected tropical diseases were compiled. An OpenMRS distribution (version 2.3) was used as reference to build the platform, following the recommended guidelines and shared-community modules.



Results:



All the customized information was developed in a platform called NTD Health, which is web-based and can be upgraded and improved by users without technological barriers.



Conclusions:



The electronic medical record system can become a useful tool for other institutions to improve their health practices as well as the quality of life for neglected tropical disease patients, simplifying the customization of healthcare systems able to interoperate with other platforms.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_62,

title = {CVAD: An Anomaly Detector for Medical Images Based on Cascade VAE},

author = {Xiaoyuan Guo, Judy Wawira Gichoya, Saptarshi Purkayastha, Imon Banerjee},

url = {https://link.springer.com/chapter/10.1007/978-3-031-16760-7_18},

year  = {2022},

date = {2022-09-15},

abstract = {Anomaly detection in medical imaging plays an important role to ensure AI generalization. However, existing out-of-distribution (OOD) detection approaches fail to account for OOD data granularity in medical images, where identifying both intra-class and inter-class OOD data is essential to the generalizability in the medical domain. We focus on the generalizability of outlier detection for medical images and propose a generic Cascade Variational autoencoder-based Anomaly Detector (CVAD). We use variational autoencoders’ cascade architecture, which combines latent representation at multiple scales, before being fed to a discriminator to distinguish the OOD data from the in-distribution data. Finally, both the reconstruction error and the OOD probability predicted by the binary discriminator are used to determine the anomalies. We compare the performance with the state-of-the-art deep learning models to demonstrate our model’s efficacy on various open-access natural and medical imaging datasets for intra- and inter-class OOD. Extensive experimental results on multiple datasets show our model’s effectiveness and generalizability. The code will be publicly available.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_39,

title = {Using Mobile Virtual Reality Simulation to Prepare for In-Person Helping Babies Breathe Training: Secondary Analysis of a Randomized Controlled Trial (the eHBB/mHBS Trial)},

author = {Ezenwa BN, Umoren R, Fajolu IB, Hippe DS, Bucher S, Purkayastha S, Okwako F, Esamai F, Feltner JB, Olawuyi O, Mmboga A, Nafula MC, Paton C, Ezeaka VC},

url = {https://mededu.jmir.org/2022/3/e37297/},

year  = {2022},

date = {2022-09-12},

urldate = {2022-09-12},

abstract = {Background:

Neonatal mortality accounts for approximately 46% of global under-5 child mortality. The widespread access to mobile devices in low- and middle-income countries has enabled innovations, such as mobile virtual reality (VR), to be leveraged in simulation education for health care workers.



Objective:

This study explores the feasibility and educational efficacy of using mobile VR for the precourse preparation of health care professionals in neonatal resuscitation training.



Methods:

Health care professionals in obstetrics and newborn care units at 20 secondary and tertiary health care facilities in Lagos, Nigeria, and Busia, Western Kenya, who had not received training in Helping Babies Breathe (HBB) within the past 1 year were randomized to access the electronic HBB VR simulation and digitized HBB Provider’s Guide (VR group) or the digitized HBB Provider’s Guide only (control group). A sample size of 91 participants per group was calculated based on the main study protocol that was previously published. Participants were directed to use the electronic HBB VR simulation and digitized HBB Provider’s Guide or the digitized HBB Provider’s Guide alone for a minimum of 20 minutes. HBB knowledge and skills assessments were then conducted, which were immediately followed by a standard, in-person HBB training course that was led by study staff and used standard HBB evaluation tools and the Neonatalie Live manikin (Laerdal Medical).



Results:

A total of 179 nurses and midwives participated (VR group: n=91; control group: n=88). The overall performance scores on the knowledge check (P=.29), bag and mask ventilation skills check (P=.34), and Objective Structured Clinical Examination A checklist (P=.43) were similar between groups, with low overall pass rates (6/178, 3.4% of participants). During the Objective Structured Clinical Examination A test, participants in the VR group performed better on the critical step of positioning the head and clearing the airway (VR group: 77/90, 86%; control group: 57/88, 65%; P=.002). The median percentage of ventilations that were performed via head tilt, as recorded by the Neonatalie Live manikin, was also numerically higher in the VR group (75%, IQR 9%-98%) than in the control group (62%, IQR 13%-97%), though not statistically significantly different (P=.35). Participants in the control group performed better on the identifying a helper and reviewing the emergency plan step (VR group: 7/90, 8%; control group: 16/88, 18%; P=.045) and the washing hands step (VR group: 20/90, 22%; control group: 32/88, 36%; P=.048).



Conclusions:

The use of digital interventions, such as mobile VR simulations, may be a viable approach to precourse preparation in neonatal resuscitation training for health care professionals in low- and middle-income countries.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_49,

title = {Examining maternal morbidity  across a spectrum of delivery locations: An analysis of the Global Network's Maternal and Neonatal Health Registry},

author = {Oberlin A, Wallace J, Moore JL, Saleem S, Lokangaka A, Tshefu A, Bauserman M, Figueroa L, Krebs NF, Esamai F, Liechty E, Bucher S, Patel AB, Hibberd PL, Chomba E, Carlo WA, Goudar S, Derman RJ, Koso-Thomas M, McClure EM, Goldenberg RL},

url = {https://doi.org/10.1002/ijgo.14391},

year  = {2022},

date = {2022-08-18},

abstract = {Objective

To better understand maternal morbidity, using quality data from low- and middle-income countries (LMICs), including out-of-hospital deliveries. Additionally, to compare to the WHO estimate that maternal morbidity occurs in 15% of pregnancies, which is based largely on hospital-level data.



Methods

The Global Network for Women's and Children's Health Research Maternal Newborn Health Registry collected data on all pregnancies from seven sites in six LMICs between 2015 and 2020. Rates of maternal mortality and morbidity and the differences in morbidity across delivery location and birth attendant type were evaluated.



Results

Among the 280 584 deliveries included in the present analysis, the overall maternal mortality ratio was 138 per 100 000, while 11.7% of women experienced at least one morbidity. Rates of morbidity were generally higher for deliveries occurring within hospitals (19.8%) and by physicians (23.6%). The lowest rates of morbidity were noted among women delivering in non-hospital healthcare facilities (5.6%) or with non-physician clinicians (e.g. nurses, midwives [5.4%]).



Conclusion

The present study shows important differences in reported maternal morbidity across delivery sites, with a trend towards lower morbidity in non-hospital healthcare facilities and among non-physician clinicians.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_59,

title = {Examining maternal morbidity across a spectrum of delivery locations: An analysis of the Global Network's Maternal and Neonatal Health Registry},

author = {Oberlin A, Wallace J, Moore JL, Saleem S, Lokangaka A, Tshefu A, Bauserman M, Figueroa L, Krebs NF, Esamai F, Liechty E, Bucher S, Patel AB, Hibberd PL, Chomba E, Carlo WA, Goudar S, Derman RJ, Koso-Thomas M, McClure EM, Goldenberg RL},

url = {https://doi.org/10.1002/ijgo.14391},

year  = {2022},

date = {2022-08-18},

abstract = {Objective

To better understand maternal morbidity, using quality data from low- and middle-income countries (LMICs), including out-of-hospital deliveries. Additionally, to compare to the WHO estimate that maternal morbidity occurs in 15% of pregnancies, which is based largely on hospital-level data.



Methods

The Global Network for Women's and Children's Health Research Maternal Newborn Health Registry collected data on all pregnancies from seven sites in six LMICs between 2015 and 2020. Rates of maternal mortality and morbidity and the differences in morbidity across delivery location and birth attendant type were evaluated.



Results

Among the 280 584 deliveries included in the present analysis, the overall maternal mortality ratio was 138 per 100 000, while 11.7% of women experienced at least one morbidity. Rates of morbidity were generally higher for deliveries occurring within hospitals (19.8%) and by physicians (23.6%). The lowest rates of morbidity were noted among women delivering in non-hospital healthcare facilities (5.6%) or with non-physician clinicians (e.g. nurses, midwives [5.4%]).



Conclusion

The present study shows important differences in reported maternal morbidity across delivery sites, with a trend towards lower morbidity in non-hospital healthcare facilities and among non-physician clinicians.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_38,

title = {The Global Network Socioeconomic Status Index as a predictor of stillbirths, perinatal mortality, and neonatal mortality in rural communities in low and lower middle income country sites of the Global Network for Women's and Children's Health Research},

author = {Patel AB, Bann CM, Kolhe CS, Lokangaka A, Tshefu A, Bauserman M, Figueroa L, Krebs NF, Esamai F, Bucher S, Saleem S, Goldenberg RL, Chomba E, Carlo WA, Goudar S, Derman RJ, Koso-Thomas M, McClure EM, Hibberd PL

},

url = {https://doi.org/10.1371/journal.pone.0272712},

year  = {2022},

date = {2022-08-16},

abstract = {Background

Globally, socioeconomic status (SES) is an important health determinant across a range of health conditions and diseases. However, measuring SES within low- and middle-income countries (LMICs) can be particularly challenging given the variation and diversity of LMIC populations.



Objective

The current study investigates whether maternal SES as assessed by the newly developed Global Network-SES Index is associated with pregnancy outcomes (stillbirths, perinatal mortality, and neonatal mortality) in six LMICs: Democratic Republic of the Congo, Guatemala, India, Kenya, Pakistan, and Zambia.



Methods

The analysis included data from 87,923 women enrolled in the Maternal and Newborn Health Registry of the NICHD-funded Global Network for Women’s and Children’s Health Research. Generalized estimating equations models were computed for each outcome by SES level (high, moderate, or low) and controlling for site, maternal age, parity, years of schooling, body mass index, and facility birth, including sampling cluster as a random effect.



Results

Women with low SES had significantly higher risks for stillbirth (p < 0.001), perinatal mortality (p = 0.001), and neonatal mortality (p = 0.005) than women with high SES. In addition, those with moderate SES had significantly higher risks of stillbirth (p = 0.003) and perinatal mortality (p = 0.008) in comparison to those with high SES.



Conclusion

The SES categories were associated with pregnancy outcomes, supporting the validity of the index as a non–income-based measure of SES for use in studies of pregnancy outcomes in LMICs.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_53,

title = {Providing Emotional Support and Physical Comfort During a Time of Social Distanci ng: A Thematic Analysis of Doulas’ Experiences During the Coronavirus Pandemic},

author = {Jewett CG, Sobiech KL, Donahue MC, Alexandrova M, Bucher S},

url = {https://doi.org/10.1177/0272684X221094172},

year  = {2022},

date = {2022-06-30},

abstract = {Doulas are trained, non-clinical professionals that provide a continuum of support for mothers. An interpretive phenomenological approach was used to explore the professional experiences of doulas (n = 17) during the COVID-19 pandemic in the US. Data were collected using brief intake surveys, in-depth semi-structured interviews, and an online discussion group. After a list of significant statements was created and grouped during emergent themes analysis, the reflections were summarized into three themes, (1) Doula Resilience, (2) Experiencing Vulnerability, and (3) Concern for Client Vulnerability that encapsulate the experiences of doulas during the COVID-19 pandemic. We conclude that as part of the COVID-19 recovery process, policy makers should look to non-clinical interventions for improving maternal health, such as promoting and supporting synergy between doulas and other maternal health service providers.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_58,

title = {Impacts of biomedical hashtag-based Twitter campaign: #DHPSP utilization for promotion of open innovation in digital health, patient safety, and personalized medicine},

author = {Maria Kletecka-Pulker, Himel Mondal, Dongdong Wang, R. Gonzalo Parra, Abdulkadir Yusif Maigoro, Soojin Lee, Tushar Garg, Eoghan J. Mulholland, Hari Prasad Devkota, Bikramjit Konwar, Sourav S. Patnaik, Ronan Lordan, Faisal A. Nawaz, Christos Tsagkaris, Rehab A. Rayan, Anna Maria Louka, Ronita De, Pravin Badhe, Eva Schaden, Harald Willschke, Mathias Maleczek, Hemanth Kumar Boyina, Garba M. Khalid, Md. Sahab Uddin, Sanusi, Johra Khan, Joy I. Odimegwu, Andy Wai Kan Yeung, Faizan Akram, Chandragiri Siva Sai, Sherri Bucher, Shravan Kumar Paswan, Rajeev K. Singla, Bairong Shen, Sara Di Lonardo, Anela Tosevska, Jesus Simal-Gandara, Manja Zec, Elena González-Burgos, Marija Habijan, Maurizio Battino, Francesca Giampieri, Aleksei Tikhonov, Danila Cianciosi, Tamara Y. Forbes-Hernandez, José L. Quiles, Bruno Mezzetti, Smith B. Babiaka, Mosa E.O. Ahmed, Paula Piccard, Mágali S. Urquiza, Jennifer R. Depew, Fabien Schultz, Daniel Sur, Sandeep R. Pai, Mihnea-Alexandru Găman, Merisa Cenanovic, Nikolay T. Tzvetkov, Surya Kant Tripathi, Kiran R. Kharat, Alfonso T. Garcia-Sosa, Simon Sieber, Atanas G. Atanasov},

url = {https://doi.org/10.1016/j.crbiot.2021.04.004},

year  = {2022},

date = {2022-06-30},

abstract = {The open innovation hub Digital Health and Patient Safety Platform (DHPSP) was recently established with the purpose to invigorate collaborative scientific research and the development of new digital products and personalized solutions aiming to improve human health and patient safety. In this study, we evaluated the effectiveness of a Twitter-based campaign centered on using the hashtag #DHPSP to promote the visibility of the DHPSP initiative. Thus, tweets containing #DHPSP were monitored for five weeks for the period 20.10.2020–24.11.2020 and were analyzed with Symplur Signals (social media analytics tool). In the study period, a total of 11,005 tweets containing #DHPSP were posted by 3020 Twitter users, generating 151,984,378 impressions. Analysis of the healthcare stakeholder-identity of the Twitter users who used #DHPSP revealed that the most of participating user accounts belonged to individuals or doctors, with the top three user locations being the United States (501 users), the United Kingdom (155 users), and India (121 users). Analysis of co-occurring hashtags and the full text of the posted tweets further revealed that the major themes of attention in the #DHPSP Twitter-community were related to the coronavirus disease 2019 (COVID-19), medicine and health, digital health technologies, and science communication in general. Overall, these results indicate that the #DHPSP initiative achieved high visibility and engaged a large body of Twitter users interested in the DHPSP focus area. Moreover, the conducted campaign resulted in an increase of DHPSP member enrollments and website visitors, and new scientific collaborations were formed. Thus, Twitter campaigns centered on a dedicated hashtag prove to be a highly efficient tool for visibility-promotion, which could be successfully utilized by healthcare-related open innovation platforms or initiatives.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_72,

title = {Multi-label Medical Image Retrieval via Learning Multi-class Similarity},

author = {Xiaoyuan Guo, Jiali Duan, Judy Gichoya, Hari Trivedi, Saptarshi Purkayastha, Ashish Sharma, Imon Banerjee},

url = {https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4149616},

year  = {2022},

date = {2022-06-29},

abstract = {Introduction: 

Multi-label image retrieval is a challenging problem in the medical area. First, compared to natural images, labels in the medical domain exhibit higher class-imbalance and much nuanced variations. Second, pair-based sampling for positives and negatives during similarity optimization are ambiguous in the multi-label setting, as samples with the same set of labels are limited. 

Methods: 

To address the aforementioned challenges, we propose a proxy-based multi-class similarity (PMS) framework, which compares and contrasts samples by comparing their similarities with the discovered proxies. In this way, samples of different sets of label attributes can be utilized and compared indirectly, without the need for complicated sampling. PMS learns a class-wise feature decomposition and maintains a memory bank for positive features from each class. The memory bank keeps track of the latest features, used to compute the class proxies. We compare samples based on their similarity distributions against the proxies, which provide a more stable mean against noise.

Results: 

We benchmark over 10 popular metric learning baselines on two public chest X-ray datasets and experiments show consistent stability of our approach under both exact and non-exact match settings.

Conclusions: 

We proposed a methodology for multi-label medical image retrieval and design a proxy-based multi-class similarity metric, which compares and contrasts samples based on their similarity distributions with respect to the class proxies. With no perquisites, the metrics can be applied to various multi-label medical image applications. The implementation code repository will be publicly available after acceptance.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_63,

title = {OSCARS: An Outlier-Sensitive Content-Based Radiography Retrieval System},

author = {Xiaoyuan Guo, Jiali Duan, Saptarshi Purkayastha, Hari Trivedi, Judy Wawira Gichoya, Imon Banerjee},

url = {https://dl.acm.org/doi/abs/10.1145/3512527.3531425},

year  = {2022},

date = {2022-06-27},

abstract = {Improving the retrieval relevance on noisy datasets is an emerging need for the curation of a large-scale clean dataset in the medical domain. While existing methods can be applied for class-wise retrieval (aka. inter-class), they cannot distinguish the granularity of likeness within the same class (aka. intra-class). The problem is exacerbated on medical external datasets, where noisy samples of the same class are treated equally during training. Our goal is to identify both intra/inter-class similarities for fine-grained retrieval. To achieve this, we propose an Outlier-Sensitive Content-based rAdiologhy Retrieval System (OSCARS), consisting of two steps. First, we train an outlier detector on a clean internal dataset in an unsupervised manner. Then we use the trained detector to generate the anomaly scores on the external dataset, whose distribution will be used to bin intra-class variations. Second, we propose a quadruplet (a, p, nintra, ninter) sampling strategy, where intra-class negatives nintra are sampled from bins of the same class other than the bin anchor a belongs to, while n_inter are randomly sampled from inter-classes. We suggest a weighted metric learning objective to balance the intra and inter-class feature learning. We experimented on two representative public radiography datasets. Experiments show the effectiveness of our approach. The training and evaluation code can be found in https://github.com/XiaoyuanGuo/oscars.



},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_64,

title = {Risks and Benefits of AI-generated Text Summarization for Expert Level Content in Graduate Health Informatics},

author = {Regina Merine, Saptarshi Purkayastha},

url = {https://ieeexplore.ieee.org/abstract/document/9874678},

year  = {2022},

date = {2022-06-11},

abstract = {AI-generated text summarization (AI-GTS) is now a popular topic in applied computer science education. It has proven helpful in various sectors, but its benefits and risks in education have not been thoroughly investigated. Few researchers have demonstrated the benefits of employing AI-generated text summaries in learning to generate ideas swiftly and to explore insights and hidden knowledge. AI-GTS has made it easier for students to understand electronically-available critical information. On the other hand, the risks linked with its implementation in education are understudied. Some anticipated risks include harming pupils' writing skills, overdependence, reduced critical thinking capacity, and increased plagiarism. This paper presents the application of AI-generated text summarization in a graduate health informatics course and discusses the risks and benefits to students. Furthermore, utilizing the Bidirectional Encoder Representations from Transformers (BERT) model, we demonstrate that the current state-of-the-art AI-generated text summarization has the potential to create expert knowledge content. We conducted a study with 58 health informatics graduate students in the Fall of 2019 to write annotated bibliography for 25 articles each, to which we also added the AI-generated article summaries. We then asked the students to peer grade and distinguish the AI-generated annotations from the student-written summary. Using the Kruskal-Wallis test, we found no significant difference in the peer grades between the two. The robustness of such AI-generated text summarization raises important questions for educators teaching in health informatics.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_37,

title = {Predictors of Plasmodium falciparum Infection in the First Trimester Among Nulliparous Women From Kenya, Zambia, and the Democratic Republic of the Congo},

author = {Leuba SI, Westreich D, Bose CL, Powers KA, Olshan A, Taylor SM, Tshefu A, Lokangaka A, Carlo WA, Chomba E, Liechty EA, Bucher SL, Esamai F, Jessani S, Saleem S, Goldenberg RL, Moore J, Nolen T, Hemingway-Foday J, McClure EM, Koso-Thomas M, Derman RJ, Hoffman M, Bauserman M

},

url = {https://europepmc.org/article/med/34888658

},

year  = {2022},

date = {2022-06-01},

urldate = {2022-06-01},

abstract = {Background

Malaria can have deleterious effects early in pregnancy, during placentation. However, malaria testing and treatment are rarely initiated until the second trimester, leaving pregnancies unprotected in the first trimester. To inform potential early intervention approaches, we sought to identify clinical and demographic predictors of first-trimester malaria.

Methods

We prospectively recruited women from sites in the Democratic Republic of the Congo (DRC), Kenya, and Zambia who participated in the ASPIRIN (Aspirin Supplementation for Pregnancy Indicated risk Reduction In Nulliparas) trial. Nulliparous women were tested for first-trimester Plasmodium falciparum infection by quantitative polymerase chain reaction. We evaluated predictors using descriptive statistics.

Results

First-trimester malaria prevalence among 1513 nulliparous pregnant women was 6.3% (95% confidence interval [CI], 3.7%-8.8%] in the Zambian site, 37.8% (95% CI, 34.2%-41.5%) in the Kenyan site, and 62.9% (95% CI, 58.6%-67.2%) in the DRC site. First-trimester malaria was associated with shorter height and younger age in Kenyan women in site-stratified analyses, and with lower educational attainment in analyses combining all 3 sites. No other predictors were identified.

Conclusions

First-trimester malaria prevalence varied by study site in sub-Saharan Africa. The absence of consistent predictors suggests that routine parasite screening in early pregnancy may be needed to mitigate first-trimester malaria in high-prevalence settings.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_65,

title = {AI recognition of patient race in medical imaging: a modelling study},

author = {Judy Wawira Gichoya, Imon Banerjee, Ananth Reddy Bhimireddy, John L Burns, Leo Anthony Celi, Li-Ching Chen, Ramon Correa, Natalie Dullerud, Marzyeh Ghassemi, Shih-Cheng Huang, Po-Chih Kuo, Matthew P Lungren, Lyle J Palmer, Brandon J Price, Saptarshi Purkayastha, Ayis T Pyrros, Lauren Oakden-Rayner, Chima Okechukwu, Laleh Seyyed-Kalantari, Hari Trivedi, Ryan Wang, Zachary Zaiman, Haoran Zhang},

url = {https://www.sciencedirect.com/science/article/pii/S2589750022000632},

year  = {2022},

date = {2022-06-01},

abstract = {Background

Previous studies in medical imaging have shown disparate abilities of artificial intelligence (AI) to detect a person's race, yet there is no known correlation for race on medical imaging that would be obvious to human experts when interpreting the images. We aimed to conduct a comprehensive evaluation of the ability of AI to recognise a patient's racial identity from medical images.



Methods

Using private (Emory CXR, Emory Chest CT, Emory Cervical Spine, and Emory Mammogram) and public (MIMIC-CXR, CheXpert, National Lung Cancer Screening Trial, RSNA Pulmonary Embolism CT, and Digital Hand Atlas) datasets, we evaluated, first, performance quantification of deep learning models in detecting race from medical images, including the ability of these models to generalise to external environments and across multiple imaging modalities. Second, we assessed possible confounding of anatomic and phenotypic population features by assessing the ability of these hypothesised confounders to detect race in isolation using regression models, and by re-evaluating the deep learning models by testing them on datasets stratified by these hypothesised confounding variables. Last, by exploring the effect of image corruptions on model performance, we investigated the underlying mechanism by which AI models can recognise race.



Findings

In our study, we show that standard AI deep learning models can be trained to predict race from medical images with high performance across multiple imaging modalities, which was sustained under external validation conditions (x-ray imaging [area under the receiver operating characteristics curve (AUC) range 0·91–0·99], CT chest imaging [0·87–0·96], and mammography [0·81]). We also showed that this detection is not due to proxies or imaging-related surrogate covariates for race (eg, performance of possible confounders: body-mass index [AUC 0·55], disease distribution [0·61], and breast density [0·61]). Finally, we provide evidence to show that the ability of AI deep learning models persisted over all anatomical regions and frequency spectrums of the images, suggesting the efforts to control this behaviour when it is undesirable will be challenging and demand further study.



Interpretation

The results from our study emphasise that the ability of AI deep learning models to predict self-reported race is itself not the issue of importance. However, our finding that AI can accurately predict self-reported race, even from corrupted, cropped, and noised medical images, often when clinical experts cannot, creates an enormous risk for all model deployments in medical imaging.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_66,

title = {Development and Evaluation of a Natural Language Conversational Bot for Identifying Appropriate Clinician Referral from Patient Narratives},

author = {Sriharsha Tummala, Saptarshi Purkayastha, Josette Jones},

url = {https://scholarworks.iupui.edu/bitstream/handle/1805/28780/BISP-Tummala%26Purkayastha-OCR.pdf?sequence=1&isAllowed=y},

year  = {2022},

date = {2022-04-26},

abstract = {Recent years have seen a significant increase in automated conversational agent chatbots. Conversational agents like chatbots for health may provide timely and cost-effective support in clinical care. Some studies show that chatbots could have an impact on patient engagement. Additionally, health systems are attempting to connect with patients over social networks, mainly where specialists are limited. By 2025, the Association of American Medical Colleges estimates that the United States will have a shortfall of 61,700-94,700 physicians and critical shortage in many specialties, delaying available appointments by months in many cases. Thus, we need innovative solutions that can manage the time of limited specialists appropriately. Recent research has demonstrated that deep learning methods are superior for natural language classification tasks compared to other machine learning methods. The primary objective of this study was to develop a telegram chatbot which reads patient narratives and acts as a conversational agent by redirecting the case to the appropriate specialist. Besides simply working on improving conversational capabilities of chatbots, we developed a novel method for referring the cases to specialists based on their responses to previous cases on a social network group. As far as we know, no other chatbot has the level of accuracy or referral system like our developed chatbot.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_73,

title = {Energy Efficiency of Quantized Neural Networks in Medical Imaging},

author = {Priyanshu Sinha, Sai Sreya Tummala, Saptarshi Purkayastha, Judy W Gichoya},

url = {https://scholarworks.iupui.edu/bitstream/handle/1805/30206/Sinha2022Energy-preprint.pdf?sequence=1&isAllowed=y},

year  = {2022},

date = {2022-04-05},

abstract = {The main goal of this paper is to compare the energy efficiency of quantized neural networks to perform medical image analysis on different processors and neural network architectures. Deep neural networks have demonstrated outstanding performance in medical image analysis but require high computation and power usage. In our work, we review the power usage and temperature of processors when running Resnet and UNet architectures to perform image classification and segmentation respectively. We compare Edge TPU, Jetson Nano, Apple M1, Nvidia Quadro P6000 and Nvidia A6000 to infer using full-precision FP32 and quantized INT8 models. The results will be useful for designers and implementers of medical imaging AI on hand-held or edge computing devices.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_68,

title = {Using Machine Learning Approaches to Identify Exercise Activities from a Triple-Synchronous Biomedical Sensor},

author = {Yohan Mahajan, Jahnavi Pinnamraju, John L Burns, Judy W Gichoya, Saptarshi Purkayastha},

url = {https://link.springer.com/chapter/10.1007/978-3-030-96308-8_113},

year  = {2022},

date = {2022-03-27},

abstract = {Human activity recognition (HAR) is the method for identifying a person’s activity using sensors that sense and capture movements during activity. HAR is an evolving area of research in computing, applied to sensors found in smartphones, smartwatches, and cameras. This paper illustrates how to identify some types of human activities using a BioStamp sensor that consists of an inertial sensors-based smart patch. Our dataset has five activities: stair climbing, jogging, skipping, kettlebell lifting, and passing a basketball. We employed both supervised and unsupervised learning approaches to perform HAR. For supervised learning, the dataset was divided into 67% training data and 33% testing data to train and evaluate the classification models. We evaluated the models on the accuracy, F1-score, precision, and recall. Random forest (F1-score = 97%) performed best followed by artificial neural networks (F1-score = 94%) and lastly, Gradient boosting (F1-score = 93%). We performed feature extraction on the original dataset and then performed unsupervised learning on the extracted features to obtain a Sørensen index value of 0.74, which shows good cluster separation. Geriatric or motor disease patients often use the BioStamp nPoint sensor and we demonstrate that their activities can be identified fairly accurately using machine learning approaches.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_69,

title = {LibreHealth Cost-of-Care Explorer: Mobile Application for Patient-friendly Access to Hospital Chargemasters},

author = {Regina Merine, Jahnavi Pinnamraju, Darshpreet Singh, Judy W Gichoya, Saptarshi Purkayastha},

url = {https://ieeexplore.ieee.org/abstract/document/9744078},

year  = {2022},

date = {2022-03-10},

abstract = {The Centers for Medicare & Medicaid Services (CMS) mandated that Charge Description Master (CDM) be available online in a machine-readable format by January 1, 2019. These changes were made to increase availability of hospital price information in order to provide patients with more access to their health information and allow physicians to spend more time with their patients. However, the mandated machine-readable formats make it impossible for a patient to interpret the pricing and hence limit price transparency. To bridge this gap, we developed a mobile application that can be used by patients to reveal the cost of a medical procedure or service before receiving it. Patients are able to compare the costs of medical procedures available in surrounding hospitals by viewing the hospital’s CDM and determine which hospital delivers the best care/medical procedures at the best price.Clinical relevance—The LibreHealth cost-of-care explorer app presents the cost of medical procedures in a consumer-friendly format. It provides patient-friendly cost estimates for medical procedures performed in US hospitals and promotes price transparency to the end users.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_70,

title = {Leapfrogging medical ai in low-resource contexts using edge tensor processing unit},

author = {Priyanshu Sinha, Judy W Gichoya, Saptarshi Purkayastha},

url = {https://ieeexplore.ieee.org/abstract/document/9744071},

year  = {2022},

date = {2022-03-10},

abstract = {With each passing year, the state-of-the-art deep learning neural networks grow larger in size, requiring larger computing and power resources. The high compute resources required by these large networks are alienating the majority of the world population that lives in low-resource settings and lacks the infrastructure to benefit from these advancements in medical AI. Current state-of-the-art medical AI, even with cloud resources, is a bit difficult to deploy in remote areas where we don’t have good internet connectivity. We demonstrate a cost-effective approach to deploying medical AI that could be used in limited resource settings using Edge Tensor Processing Unit (TPU). We trained and optimized a classification model on the Chest X-ray 14 dataset and a segmentation model on the Nerve ultrasound dataset using INT8 Quantization Aware Training. Thereafter, we compiled the optimized models for Edge TPU execution. We find that the inference performance on edge TPUs is 10x faster compared to other embedded devices. The optimized model is 3x and 12x smaller for the classification and segmentation respectively, compared to the full precision model. In summary, we show the potential of Edge TPUs for two medical AI tasks with faster inference times, which could potentially be used in low-resource settings for medical AI-based diagnostics. We finally discuss some potential challenges and limitations of our approach for real-world deployments.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_75,

title = {Margin-aware intraclass novelty identification for medical images},

url = {https://www.spiedigitallibrary.org/journals/journal-of-medical-imaging/volume-9/issue-1/014004/Margin-aware-intraclass-novelty-identification-for-medical-images/10.1117/1.JMI.9.1.014004.short?SSO=1},

year  = {2022},

date = {2022-02-03},

abstract = {Purpose: Existing anomaly detection methods focus on detecting interclass variations while medical image novelty identification is more challenging in the presence of intraclass variations. For example, a model trained with normal chest x-ray and common lung abnormalities is expected to discover and flag idiopathic pulmonary fibrosis, which is a rare lung disease and unseen during training. The nuances of intraclass variations and lack of relevant training data in medical image analysis pose great challenges for existing anomaly detection methods.



Approach: We address the above challenges by proposing a hybrid model—transformation-based embedding learning for novelty detection (TEND), which combines the merits of classifier-based approach and AutoEncoder (AE)-based approach. Training TEND consists of two stages. In the first stage, we learn in-distribution embeddings with an AE via the unsupervised reconstruction. In the second stage, we learn a discriminative classifier to distinguish in-distribution data and the transformed counterparts. Additionally, we propose a margin-aware objective to pull in-distribution data in a hypersphere while pushing away the transformed data. Eventually, the weighted sum of class probability and the distance to margin constitutes the anomaly score.



Results: Extensive experiments are performed on three public medical image datasets with the one-vs-rest setup (namely one class as in-distribution data and the left as intraclass out-of-distribution data) and the rest-vs-one setup. Additional experiments on generated intraclass out-of-distribution data with unused transformations are implemented on the datasets. The quantitative results show competitive performance as compared to the state-of-the-art approaches. Provided qualitative examples further demonstrate the effectiveness of TEND.



Conclusion: Our anomaly detection model TEND can effectively identify the challenging intraclass out-of-distribution medical images in an unsupervised fashion. It can be applied to discover unseen medical image classes and serve as the abnormal data screening for downstream medical tasks. The corresponding code is available at https://github.com/XiaoyuanGuo/TEND_MedicalNoveltyDetection.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_71,

title = {CVAD - An unsupervised image anomaly detector},

author = {Xiaoyuan Guo, Judy Wawira Gichoya, Saptarshi Purkayastha, Imon Banerjee},

url = {https://www.sciencedirect.com/science/article/pii/S2665963821000853},

year  = {2022},

date = {2022-02-01},

abstract = {Detecting out-of-distribution samples for image applications plays an important role in safeguarding the reliability of machine learning model deployment. In this article, we developed a software tool to support our OOD detector CVAD - a self-supervised Cascade Variational autoencoder-based Anomaly Detector , which can be easily applied to various image applications without any assumptions. The corresponding open-source software is published for better public research and tool usage.



},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_76,

title = {MedShift: identifying shift data for medical dataset curation},

author = {Xiaoyuan Guo, Judy Wawira Gichoya, Hari Trivedi, Saptarshi Purkayastha, Imon Banerjee},

url = {https://arxiv.org/pdf/2112.13885.pdf},

year  = {2021},

date = {2021-12-27},

abstract = {Automated dataset curation in the medical domain has long been demanding as AI technologies are often hungry for annotated data. To curate a high-quality dataset, identifying data variance between the internal and external sources is a fundamental and crucial step as the data distributions from different sources can vary significantly and thus affect the performance of the AI models. However, methods to detect shift or variance in data have not been significantly researched. Challenges to this are the lack of effective approaches to learn dense representation of a dataset by capturing its semantics and difficulties of sharing private data across medical institutions. To overcome the problems, we propose a unified pipeline called MedShift to detect the top-level shift samples and thus facilitate the medical curation. Given an internal dataset A as the base source, we first train anomaly detectors for each class of dataset A to learn internal distributions in an unsupervised way. Second, without exchanging data across sources, we run the trained anomaly detectors on an external dataset B for each class. The data samples with high anomaly scores are identified as shift data. To quantify the shiftness of the external dataset, we cluster B’s data into groups classwise based on the obtained scores. We then train a multi-class classifier on A and measure the shiftness with the classifier’s performance variance on B by gradually dropping the group with the largest anomaly score for each class. Additionally, we adapt a dataset quality metric to help inspect the distribution differences for multiple medical sources. We verify the efficacy of MedShift with musculoskeletal radiographs (MURA) and chest X-rays datasets from more than one external source. Experiments show our proposed shift data detection pipeline can be beneficial for medical centers to curate high-quality datasets more efficiently. An interface introduction video to visualize our results is available at https://youtu.be/V3BF0P1sxQE.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_77,

title = {CVAD: A generic medical anomaly detector based on Cascade VAE},

author = {Xiaoyuan Guo, Judy Wawira Gichoya, Saptarshi Purkayastha, Imon Banerjee},

url = {https://arxiv.org/pdf/2110.15811.pdf},

year  = {2021},

date = {2021-10-29},

abstract = {Detecting out-of-distribution (OOD) samples in medical imaging plays an important role for downstream medical diagnosis. However, existing OOD detectors are demonstrated on natural images composed of classes with clear inter-class variations and have difficulty generalizing to medical images. The key issue is the granularity of OOD data in the medical domain, where intra-class OOD samples are predominant. We focus on the generalizability of OOD detection for medical images and propose a self-supervised Cascade Variational autoencoderbased Anomaly Detector (CVAD). We use a cascaded variational autoencoder architecture, which combines latent representation at multiple scales, before being fed to a discriminator to distinguish the OOD data from the in-distribution (ID) data. Finally, both the reconstruction error and the OOD probability

predicted by the binary discriminator are used to determine the

anomalies. We compare the performance with the state-of-theart deep learning models to demonstrate our model’s efficacy on

various open-access medical imaging datasets for both intra- and

inter-class OOD. Further extensive results on datasets including

common natural datasets show our model’s effectiveness and

generalizability},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_78,

title = {PLHI-MC10: A dataset of exercise activities captured through a triple synchronous medically-approved sensor},

author = {Yohan Mahajan, Ananth Bhimireddy, Areeba Abid, Judy W Gichoya, Saptarshi Purkayastha},

url = {https://www.sciencedirect.com/science/article/pii/S2352340921005710},

year  = {2021},

date = {2021-10-01},

abstract = {Most human activity recognition datasets that are publicly available have data captured by using either smartphones or smartwatches, which are usually placed on the waist or the wrist, respectively. These devices obtain one set of acceleration and angular velocity in the x-, y-, and z-axis from the accelerometer and the gyroscope planted in these devices. The PLHI-MC10 dataset contains data obtained by using 3 BioStamp nPoint® sensors from 7 physically healthy adult test subjects performing different exercise activities. These sensors are the state-of-the-art biomedical sensors manufactured by MC10. Each of the three sensors was attached to the subject externally on three muscles-Extensor Digitorum (Posterior Forearm), Gastrocnemius (Calf), and Pectoralis (Chest)-giving us three sets of 3 axial acceleration, two sets of 3 axial angular velocities, and 1 set of voltage values from the heart. Using three different sensors instead of a single sensor improves precision. It helps distinguish between human activities as it simultaneously captures the movement and contractions of various muscles from separate parts of the human body. Each test subject performed five activities (stairs, jogging, skipping, lifting kettlebell, basketball throws) in a supervised environment. The data is cleaned, filtered, and synced.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_74,

title = {Optimizing medical image classification models for edge devices},

author = {Areeba Abid, Priyanshu Sinha, Aishwarya Harpale, Judy Gichoya, Saptarshi Purkayastha},

url = {https://link.springer.com/chapter/10.1007/978-3-030-86261-9_8},

year  = {2021},

date = {2021-09-02},

abstract = {Machine learning algorithms for medical diagnostics often require resource-intensive environments to run, such as expensive cloud servers or high-end GPUs, making these models impractical for use in the field. We investigate the use of model quantization and GPU-acceleration for chest X-ray classification on edge devices. We employ 3 types of quantization (dynamic range, float-16, and full int8) which we tested on models trained on the Chest-XRay14 Dataset. We achieved a 2–4x reduction in model size, offset by small decreases in the mean AUC-ROC score of 0.0%–0.9%. On ARM architectures, integer quantization was shown to improve inference latency by up to 57%. However, we also observe significant increases in latency on x86 processors. GPU acceleration also improved inference latency, but this was outweighed by kernel launch overhead. We show that optimization of diagnostic models has the potential to expand their utility to day-to-day devices used by patients and healthcare workers; however, these improvements are context- and architecture-dependent and should be tested on the relevant devices before deployment in low-resource environments.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_44,

title = {eHBB: a randomised controlled trial of virtual reality or video for neonatal resuscitation refresher training in healthcare workers in resource-scarce settings},

author = {Umoren R, Bucher S, Hippe DS, Ezenwa BN, Fajolu IB, Okwako FM, Feltner J, Nafula M, Musale A, Olawuyi OA, Adeboboye CO, Asangansi I, Paton C, Purkayastha S, Ezeaka CV, Esamai F},

url = {https://bmjopen.bmj.com/content/11/8/e048506},

year  = {2021},

date = {2021-08-25},

abstract = {Abstract

Objective To assess the impact of mobile virtual reality (VR) simulations using electronic Helping Babies Breathe (eHBB) or video for the maintenance of neonatal resuscitation skills in healthcare workers in resource-scarce settings.



Design Randomised controlled trial with 6-month follow-up (2018–2020).



Setting Secondary and tertiary healthcare facilities.



Participants 274 nurses and midwives assigned to labour and delivery, operating room and newborn care units were recruited from 20 healthcare facilities in Nigeria and Kenya and randomised to one of three groups: VR (eHBB+digital guide), video (video+digital guide) or control (digital guide only) groups before an in-person HBB course.



Intervention(s) eHBB VR simulation or neonatal resuscitation video.



Main outcome(s) Healthcare worker neonatal resuscitation skills using standardised checklists in a simulated setting at 1 month, 3 months and 6 months.



Results Neonatal resuscitation skills pass rates were similar among the groups at 6-month follow-up for bag-and-mask ventilation (BMV) skills check (VR 28%, video 25%, control 22%, p=0.71), objective structured clinical examination (OSCE) A (VR 76%, video 76%, control 72%, p=0.78) and OSCE B (VR 62%, video 60%, control 49%, p=0.18). Relative to the immediate postcourse assessments, there was greater retention of BMV skills at 6 months in the VR group (−15% VR, p=0.10; −21% video, p<0.01, –27% control, p=0.001). OSCE B pass rates in the VR group were numerically higher at 3 months (+4%, p=0.64) and 6 months (+3%, p=0.74) and lower in the video (−21% at 3 months, p<0.001; −14% at 6 months, p=0.066) and control groups (−7% at 3 months, p=0.43; −14% at 6 months, p=0.10). On follow-up survey, 95% (n=65) of respondents in the VR group and 98% (n=82) in the video group would use their assigned intervention again.



Conclusion eHBB VR training was highly acceptable to healthcare workers in low-income to middle-income countries and may provide additional support for neonatal resuscitation skills retention compared with other digital interventions.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_80,

title = {A DICOM Framework for Machine Learning and Processing Pipelines Against Real-time Radiology Images},

author = {Pradeeban Kathiravelu, Puneet Sharma, Ashish Sharma, Imon Banerjee, Hari Trivedi, Saptarshi Purkayastha, Priyanshu Sinha, Alexandre Cadrin-Chenevert, Nabile Safdar, Judy Wawira Gichoya},

url = {https://link.springer.com/article/10.1007/s10278-021-00491-w},

year  = {2021},

date = {2021-08-17},

abstract = {Real-time execution of machine learning (ML) pipelines on radiology images is difficult due to limited computing resources in clinical environments, whereas running them in research clusters requires efficient data transfer capabilities. We developed Niffler, an open-source Digital Imaging and Communications in Medicine (DICOM) framework that enables ML and processing pipelines in research clusters by efficiently retrieving images from the hospitals’ PACS and extracting the metadata from the images. We deployed Niffler at our institution (Emory Healthcare, the largest healthcare network in the state of Georgia) and retrieved data from 715 scanners spanning 12 sites, up to 350 GB/day continuously in real-time as a DICOM data stream over the past 2 years. We also used Niffler to retrieve images bulk on-demand based on user-provided filters to facilitate several research projects. This paper presents the architecture and three such use cases of Niffler. First, we executed an IVC filter detection and segmentation pipeline on abdominal radiographs in real-time, which was able to classify 989 test images with an accuracy of 96.0%. Second, we applied the Niffler Metadata Extractor to understand the operational efficiency of individual MRI systems based on calculated metrics. We benchmarked the accuracy of the calculated exam time windows by comparing Niffler against the Clinical Data Warehouse (CDW). Niffler accurately identified the scanners’ examination timeframes and idling times, whereas CDW falsely depicted several exam overlaps due to human errors. Third, with metadata extracted from the images by Niffler, we identified scanners with misconfigured time and reconfigured five scanners. Our evaluations highlight how Niffler enables real-time ML and processing pipelines in a research cluster.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_79,

title = {Electronic Patient Records as a Substrate for Collaboration for Distributed Care in Low-Resource Contexts},

author = {Saptarshi Purkayastha},

url = {https://ieeexplore.ieee.org/abstract/document/9565720},

year  = {2021},

date = {2021-08-09},

abstract = {Collaboration to provide patient care in low-resource contexts has been a challenge due to heavy patient load, limited connectivity, and knowledge-gap between primary and tertiary care. Through the design, development, and implementation of a private social network-connected, large-scale hospital information system, which has scaled to several zonal and district hospitals in a small hilly country in South East Asia, we present the case study of a system that has enabled collaboration. Using coordination mechanisms as a theoretical framework, we discuss some methods of collaboration. In the paper, we present electronic patient records (EPR) as the substrate that enables collaboration between providers, departments, developers throughout the health systems. In our analysis, we present useful learnings of collaboration between provider-provider, developer-developer, provider-patient, implementer-provider, and how the balance of these is a necessary condition to create a useful substrate for collaboration.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_35,

title = {Critical Interpretive Synthesis of Qualitative Data on the Health Care Ecosystem for Vulnerable Newborns in Low- to Middle-Income Countries},

author = {Ray H, Sobiech KL, Alexandrova M, Songok JJ, Rukunga J, Bucher S},

url = {https://doi.org/10.1016/j.jogn.2021.05.001},

year  = {2021},

date = {2021-07-21},

urldate = {2021-07-21},

abstract = {Objective: To critically assess and synthesize qualitative findings regarding the health care ecosystem for vulnerable (low-birth-weight or sick) neonates in low- to middle-income countries (LMICs).



Data sources: Between May 4 and June 2, 2020, we searched four databases (Medline [PubMed], SCOPUS, PsycINFO, and Web of Science) for articles published from 2010 to 2020. Inclusion criteria were peer-reviewed reports of original studies focused on the health care ecosystem for vulnerable neonates in LMICs. We also searched the websites of several international development agencies and included findings from primary data collected between May and July 2019 at a tertiary hospital in Kenya. We excluded studies and reports if the focus was on healthy neonates or high-income countries and if they contained only quantitative data, were written in a language other than English, or were published before 2010.



Study selection: One of the primary authors conducted an initial review of titles and abstracts (n = 102) and excluded studies that were not consistent with the purpose of the review (n = 60). The two primary authors used a qualitative appraisal checklist to assess the validity of the remaining studies (n = 42) and reached agreement on the final 13 articles.



Data extraction: The two primary authors independently conducted open and axial coding of the data. We incorporated data from studies with different units of analysis, types of methodology, research topics, participant types, and analytical frameworks in an emergent conceptual development process according to the critical interpretive synthesis methodology.



Data synthesis: We synthesized our findings into one overarching theme, Pervasive Turbulence Is a Defining Characteristic of the Health Care Ecosystem in LMICs, and two subthemes: Pervasive Turbulence May Cause Tension Between the Setting and the Caregiver and Pervasive Turbulence May Result in a Loss of Synergy in the Caregiver-Parent Relationship.



Conclusion: Because pervasive turbulence characterizes the health care ecosystems in LMICs, interventions are needed to support the caregiver-parent interaction to mitigate the effects of tension in the setting.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_81,

title = {Reading Race: AI Recognises Patient's Racial Identity In Medical Images},

author = {Imon Banerjee, Ananth Reddy Bhimireddy, John L Burns, Leo Anthony Celi, Li-Ching Chen, Ramon Correa, Natalie Dullerud, Marzyeh Ghassemi, Shih-Cheng Huang, Po-Chih Kuo, Matthew P Lungren, Lyle Palmer, Brandon J Price, Saptarshi Purkayastha, Ayis Pyrros, Luke Oakden-Rayner, Chima Okechukwu, Laleh Seyyed-Kalantari, Hari Trivedi, Ryan Wang, Zachary Zaiman, Haoran Zhang, Judy W Gichoya},

url = {https://arxiv.org/ftp/arxiv/papers/2107/2107.10356.pdf},

year  = {2021},

date = {2021-07-21},

abstract = {Background: 

In medical imaging, prior studies have demonstrated disparate AI performance by race, yet there is no known correlation for race on medical imaging that would be obvious to the human expert interpreting the images.



Methods: 

Using private and public datasets we evaluate: A) performance quantification of deep learning models to detect race from medical images, including the ability of these models to generalize to external environments and across multiple imaging modalities, B) assessment of possible confounding anatomic and phenotype population features, such as disease distribution and body habitus as predictors of race, and C) investigation into the underlying mechanism by which AI models can recognize race.



Findings: 

Standard deep learning models can be trained to predict race from medical images with high performance across multiple imaging modalities. Our findings hold under external validation conditions, as well as when models are optimized to perform clinically motivated tasks. We demonstrate this detection is not due to trivial proxies or imaging-related surrogate covariates for race, such as underlying disease distribution. Finally, we show that performance persists over all anatomical regions and frequency spectrum of the images suggesting that mitigation efforts will be challenging and demand further study.



Interpretation:

We emphasize that model ability to predict self-reported race is itself not the issue of importance. However, our findings that AI can trivially predict self-reported race -- even from corrupted, cropped, and noised medical images -- in a setting where clinical experts cannot, creates an enormous risk for all model deployments in medical imaging: if an AI model secretly used its knowledge of self-reported race to misclassify all Black patients, radiologists would not be able to tell using the same data the model has access to.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_82,

title = {Predicting Opioid Prescriptions based on Patient Demographics in MIMIC-IV},

author = {Snigdha Kodela, Jahnavi Pinnamraju, Judy W Gichoya, Saptarshi Purkayastha},

url = {https://ieeexplore.ieee.org/abstract/document/9474745},

year  = {2021},

date = {2021-06-07},

abstract = {Opioids are widely used analgesics because of their efficacy, mild sedative and anxiolytic properties, and flexibility to administer through multiple routes. Understanding the demographics of the patients receiving these medications helps provide customized care for the susceptible group of people. We conducted a demographic evaluation of the frequently prescribed opioid drug prescriptions from the MIMIC IV database. We analyzed prescribing patterns of six commonly used opioids with demographics such as age, gender, ethnicity, marital status, and year predominantly. After conducting exploratory data analysis, we built models using Logistic Regression, Random Forest, and XGBoost to predict opioid prescriptions and demographics for those. We also analyzed the association between demographics and the frequency of prescribed medications for pain management. We found statistically significant differences in opioid prescriptions among the male and female population, married and unmarried, various ages, ethnic groups, and an association with in-hospital deaths.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_83,

title = {Patient-specific COVID-19 resource utilization prediction using fusion AI model},

author = {Amara Tariq, Leo Anthony Celi, Janice M Newsome, Saptarshi Purkayastha, Neal Kumar Bhatia, Hari Trivedi, Judy Wawira Gichoya, Imon Banerjee},

url = {https://www.nature.com/articles/s41746-021-00461-0},

year  = {2021},

date = {2021-06-03},

abstract = {The strain on healthcare resources brought forth by the recent COVID-19 pandemic has highlighted the need for efficient resource planning and allocation through the prediction of future consumption. Machine learning can predict resource utilization such as the need for hospitalization based on past medical data stored in electronic medical records (EMR). We conducted this study on 3194 patients (46% male with mean age 56.7 (±16.8), 56% African American, 7% Hispanic) flagged as COVID-19 positive cases in 12 centers under Emory Healthcare network from February 2020 to September 2020, to assess whether a COVID-19 positive patient’s need for hospitalization can be predicted at the time of RT-PCR test using the EMR data prior to the test. Five main modalities of EMR, i.e., demographics, medication, past medical procedures, comorbidities, and laboratory results, were used as features for predictive modeling, both individually and fused together using late, middle, and early fusion. Models were evaluated in terms of precision, recall, F1-score (within 95% confidence interval). The early fusion model is the most effective predictor with 84% overall F1-score [CI 82.1–86.1]. The predictive performance of the model drops by 6 % when using recent clinical data while omitting the long-term medical history. Feature importance analysis indicates that history of cardiovascular disease, emergency room visits in the past year prior to testing, and demographic factors are predictive of the disease trajectory. We conclude that fusion modeling using medical history and current treatment data can forecast the need for hospitalization for patients infected with COVID-19 at the time of the RT-PCR test.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_51,

title = {Cost estimation alongside a multiregional, multi-country randomized trial of antenatal ultrasound in five low-and-middle-income countries},

author = {Bresnahan BW, Vodicka E, Babigumira JB, Malik AM, Yego F, Lokangaka A, Chitah BM, Bauer Z, Chavez H, Moore JL, Garrison LP, Swanson JO, Swanson D, McClure EM, Goldenberg RL, Esamai F, Garces AL, Chomba E, Saleem S, Tshefu A, Bose CL, Bauserman M, Carlo W, Bucher S, Liechty EA, Nathan RO},

url = {https://bmcpublichealth.biomedcentral.com/articles/10.1186/s12889-021-10750-8},

year  = {2021},

date = {2021-05-20},

abstract = {Background

Improving maternal health has been a primary goal of international health agencies for many years, with the aim of reducing maternal and child deaths and improving access to antenatal care (ANC) services, particularly in low-and-middle-income countries (LMICs). Health interventions with these aims have received more attention from a clinical effectiveness perspective than for cost impact and economic efficiency.



Methods

We collected data on resource use and costs as part of a large, multi-country study assessing the use of routine antenatal screening ultrasound (US) with the aim of considering the implications for economic efficiency. We assessed typical antenatal outpatient and hospital-based (facility) care for pregnant women, in general, with selective complication-related data collection in women participating in a large maternal health registry and clinical trial in five LMICs. We estimated average costs from a facility/health system perspective for outpatient and inpatient services. We converted all country-level currency cost estimates to 2015 United States dollars (USD). We compared average costs across countries for ANC visits, deliveries, higher-risk pregnancies, and complications, and conducted sensitivity analyses.



Results

Our study included sites in five countries representing different regions. Overall, the relative cost of individual ANC and delivery-related healthcare use was consistent among countries, generally corresponding to country-specific income levels. ANC outpatient visit cost estimates per patient among countries ranged from 15 to 30 USD, based on average counts for visits with and without US. Estimates for antenatal screening US visits were more costly than non-US visits. Costs associated with higher-risk pregnancies were influenced by rates of hospital delivery by cesarean section (mean per person delivery cost estimate range: 25–65 USD).



Conclusions

Despite substantial differences among countries in infrastructures and health system capacity, there were similarities in resource allocation, delivery location, and country-level challenges. Overall, there was no clear suggestion that adding antenatal screening US would result in either major cost savings or major cost increases. However, antenatal screening US would have higher training and maintenance costs. Given the lack of clinical effectiveness evidence and greater resource constraints of LMICs, it is unlikely that introducing antenatal screening US would be economically efficient in these settings--on the demand side (i.e., patients) or supply side (i.e., healthcare providers).},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_50,

title = {Cost estimation alongside a multiregional, multi-country randomized trial of antenatal ultrasound in five low-and-middle-income countries},

author = {Bresnahan BW, Vodicka E, Babigumira JB, Malik AM, Yego F, Lokangaka A, Chitah BM, Bauer Z, Chavez H, Moore JL, Garrison LP, Swanson JO, Swanson D, McClure EM, Goldenberg RL, Esamai F, Garces AL, Chomba E, Saleem S, Tshefu A, Bose CL, Bauserman M, Carlo W, Bucher S, Liechty EA, Nathan RO},

url = {https://bmcpublichealth.biomedcentral.com/articles/10.1186/s12889-021-10750-8},

year  = {2021},

date = {2021-05-20},

abstract = {Background

Improving maternal health has been a primary goal of international health agencies for many years, with the aim of reducing maternal and child deaths and improving access to antenatal care (ANC) services, particularly in low-and-middle-income countries (LMICs). Health interventions with these aims have received more attention from a clinical effectiveness perspective than for cost impact and economic efficiency.



Methods

We collected data on resource use and costs as part of a large, multi-country study assessing the use of routine antenatal screening ultrasound (US) with the aim of considering the implications for economic efficiency. We assessed typical antenatal outpatient and hospital-based (facility) care for pregnant women, in general, with selective complication-related data collection in women participating in a large maternal health registry and clinical trial in five LMICs. We estimated average costs from a facility/health system perspective for outpatient and inpatient services. We converted all country-level currency cost estimates to 2015 United States dollars (USD). We compared average costs across countries for ANC visits, deliveries, higher-risk pregnancies, and complications, and conducted sensitivity analyses.



Results

Our study included sites in five countries representing different regions. Overall, the relative cost of individual ANC and delivery-related healthcare use was consistent among countries, generally corresponding to country-specific income levels. ANC outpatient visit cost estimates per patient among countries ranged from 15 to 30 USD, based on average counts for visits with and without US. Estimates for antenatal screening US visits were more costly than non-US visits. Costs associated with higher-risk pregnancies were influenced by rates of hospital delivery by cesarean section (mean per person delivery cost estimate range: 25–65 USD).



Conclusions

Despite substantial differences among countries in infrastructures and health system capacity, there were similarities in resource allocation, delivery location, and country-level challenges. Overall, there was no clear suggestion that adding antenatal screening US would result in either major cost savings or major cost increases. However, antenatal screening US would have higher training and maintenance costs. Given the lack of clinical effectiveness evidence and greater resource constraints of LMICs, it is unlikely that introducing antenatal screening US would be economically efficient in these settings--on the demand side (i.e., patients) or supply side (i.e., healthcare providers).},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_34,

title = {Digital ≠ paperless: novel interfaces needed to address global health challenges},

author = {Kumar P, Sammut SM, Madan JJ, Bucher S, Kumar MB},

url = {http://dx.doi.org/10.1136/bmjgh-2021-005780},

year  = {2021},

date = {2021-04-06},

urldate = {2021-04-06},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_84,

title = {Failures hiding in success for artificial intelligence in radiology},

author = {Saptarshi Purkayastha, Hari Trivedi, Judy Wawira Gichoya},

url = {https://www.jacr.org/article/S1546-1440(20)31227-8/fulltext},

year  = {2021},

date = {2021-03-01},

abstract = {Reports of computer algorithms outperforming radiologists have persisted over the last 15 years, starting with the 2005 publication by Rubin et al on detecting pulmonary nodules from CT scans [ 1 ]. Back then, these technologies were referred to as computer-aided diagnosis, which could be considered as a precursor, of sorts, to what is now referred to broadly as artificial intelligence (AI). Technology gains in hardware over the past 5 years have facilitated the training of deep neural networks with millions of parameters, exponentially accelerating the pace of AI publications. However, like every other scientific field, successes of AI in radiology are published and publicized with much fanfare, and failures are not discussed or made public. In fact, most AI failures are discovered anecdotally from personal experience or when shared in social media as tweets or blog posts. In this article, we discuss some pitfalls frequently encountered in reporting the success of AI in radiology, which might be considered failures when considered differently.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_86,

title = {Human Activity Recognition using Deep Learning Models on Smartphones and Smartwatches Sensor Data},

author = {Bolu Oluwalade, Sunil Neela, Judy Wawira, Tobiloba Adejumo, Saptarshi Purkayastha},

url = {https://arxiv.org/ftp/arxiv/papers/2103/2103.03836.pdf},

year  = {2021},

date = {2021-02-28},

abstract = {In recent years, human activity recognition has garnered considerable attention both in industrial and academic research because of the wide deployment of sensors, such as accelerometers and gyroscopes, in products such as smartphones and smartwatches. Activity recognition is currently applied in various fields where valuable information about an individual's functional ability and lifestyle is needed. In this study, we used the popular WISDM dataset for activity recognition. Using multivariate analysis of covariance (MANCOVA), we established a statistically significant difference (p<0.05) between the data generated from the sensors embedded in smartphones and smartwatches. By doing this, we show that smartphones and smartwatches don't capture data in the same way due to the location where they are worn. We deployed several neural network architectures to classify 15 different hand and non-hand-oriented activities. These models include Long short-term memory (LSTM), Bi-directional Long short-term memory (BiLSTM), Convolutional Neural Network (CNN), and Convolutional LSTM (ConvLSTM). The developed models performed best with watch accelerometer data. Also, we saw that the classification precision obtained with the convolutional input classifiers (CNN and ConvLSTM) was higher than the end-to-end LSTM classifier in 12 of the 15 activities. Additionally, the CNN model for the watch accelerometer was better able to classify non-hand oriented activities when compared to hand-oriented activities.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_52,

title = {Association of hemoglobin levels in the first trimester and at 26 to 30 weeks with fetal and neonatal outcomes: A secondary analyses of the Global Network for Women's and Children's Health's ASPIRIN Trial},

author = {Jessani S, Saleem S, Hoffman MK, Goudar SS, Derman RJ, Moore JL, Garces A, Figueroa L, Krebs NF, Okitawutshu J, Tshefu A, Bose CL, Mwenechanya M, Chomba E, Carlo WA, Das PK, Patel A, Hibberd PL, Esamai F, Liechty EA, Bucher S, Nolen TL, Koso-Thomas M, Miodovnik M, McClure EM, Goldenberg RL},

url = {https://doi.org/10.1111/1471-0528.16676},

year  = {2021},

date = {2021-02-24},

abstract = {Objective

Limited data are available from low- and middle-income countries (LMICs) on the relationship of haemoglobin levels to adverse outcomes at different times during pregnancy. We evaluated the association of haemoglobin levels in nulliparous women at two times in pregnancy with pregnancy outcomes.



Design

ASPIRIN Trial data were used to study the association between haemoglobin levels measured at 6+0–13+6 weeks and 26+0–30+0 weeks of gestation with fetal and neonatal outcomes.



Setting

Obstetric care facilities in Pakistan, India, Kenya, Zambia, The Democratic Republic of the Congo and Guatemala.



Population

A total of 11 976 pregnant women.



Methods

Generalised linear models were used to obtain adjusted relative risks and 95% CI for adverse outcomes.



Main outcome measures

Preterm birth, stillbirth, neonatal death, small for gestational age (SGA) and birthweight <2500 g.



Results

The mean haemoglobin levels at 6+0–13+6 weeks and at 26–30 weeks of gestation were 116 g/l (SD 17) and 107 g/l (SD 15), respectively. In general, pregnancy outcomes were better with increasing haemoglobin. At 6+0–13+6 weeks of gestation, stillbirth, SGA and birthweight <2500 g, were significantly associated with haemoglobin of 70–89 g/l compared with haemoglobin of 110–129 g/l The relationships of adverse pregnancy outcomes with various haemoglobin levels were more marked at 26–30 weeks of gestation.



Conclusions

Both lower and some higher haemoglobin concentrations are associated with adverse fetal and neonatal outcomes at 6+0–13+6 weeks and at 26–30 weeks of gestation, although the relationship with low haemoglobin levels appears more consistent and generally stronger.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_85,

title = {Usability and Security of Different Authentication Methods for an Electronic Health Records System},

author = {Saptarshi Purkayastha, Shreya Goyal, Bolu Oluwalade, Tyler Phillips, Huanmei Wu, Xukai Zou},

url = {https://arxiv.org/ftp/arxiv/papers/2102/2102.11849.pdf},

year  = {2021},

date = {2021-02-23},

abstract = {We conducted a survey of 67 graduate students enrolled in the Privacy and Security in Healthcare course at Indiana University Purdue University Indianapolis. This was done to measure user preference and their understanding of usability and security of three different Electronic Health Records authentication methods: single authentication method (username and password), Single sign-on with Central Authentication Service (CAS) authentication method, and a bio-capsule facial authentication method. This research aims to explore the relationship between security and usability, and measure the effect of perceived security on usability in these three aforementioned authentication methods. We developed a formative-formative Partial Least Square Structural Equation Modeling (PLS-SEM) model to measure the relationship between the latent variables of Usability, and Security. The measurement model was developed using five observed variables (measures). - Efficiency and Effectiveness, Satisfaction, Preference, Concerns, and Confidence. The results obtained highlight the importance and impact of these measures on the latent variables and the relationship among the latent variables. From the PLS-SEM analysis, it was found that security has a positive impact on usability for Single sign-on and bio-capsule facial authentication methods. We conclude that the facial authentication method was the most secure and usable among the three authentication methods. Further, descriptive analysis was done to draw out the interesting findings from the survey regarding the observed variables.

},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_56,

title = {Development of a Mobile, Self-Sovereign Identity Approach for Facility Birth Registration in Kenya},

author = {Freytsis Maria, Barclay Iain, Radha Swapna Krishnakumar, Czajka Adam, Siwo Geoffery H., Taylor Ian, Bucher Sherri

},

url = {https://www.frontiersin.org/articles/10.3389/fbloc.2021.631341/full},

year  = {2021},

date = {2021-02-15},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_57,

title = {Safety of daily low-dose aspirin use during pregnancy in low-income and middle-income countries},

author = {Short VL, Hoffman M, Metgud M, Kavi A, Goudar SS, Okitawutshu J, Tshefu A, Bose CL, Mwenechanya M, Chomba E, Carlo WA, Figueroa L, Garces A, Krebs NF, Jessani S, Saleem S, Goldenberg RL, Das PK, Patel A, Hibberd PL, Achieng E, Nyongesa P, Esamai F, Bucher S, Nowak KJ, Goco N, Nolen TL, McClure EM, Koso-Thomas M, Miodovnik M, Derman RJ},

url = {https://doi.org/10.1016/j.xagr.2021.100003},

year  = {2021},

date = {2021-01-27},

abstract = {BACKGROUND

The daily use of low-dose aspirin may be a safe, widely available, and inexpensive intervention for reducing the risk of preterm birth. Data on the potential side effects of low-dose aspirin use during pregnancy in low- and middle-income countries are needed.



OBJECTIVE

This study aimed to assess differences in unexpected emergency medical visits and potential maternal side effects from a randomized, double-blind, multicountry, placebo-controlled trial of low-dose aspirin use (81 mg daily, from 6 to 36 weeks’ gestation).



STUDY DESIGN

This study was a secondary analysis of data from the Aspirin Supplementation for Pregnancy Indicated Risk Reduction In Nulliparas trial, a trial of the Global Network for Women's and Children's Health conducted in India (2 sites), Pakistan, Guatemala, Democratic Republic of the Congo, Kenya, and Zambia. The outcomes for this analysis were unexpected emergency medical visits and the occurrence of the following potential side effects—overall and separately—nausea, vomiting, rash or hives, diarrhea, gastritis, vaginal bleeding, allergic reaction, and any other potential side effects. Analyses were performed overall and by geographic region.



RESULTS

Between the aspirin (n=5943) and placebo (n=5936) study groups, there was no statistically significant difference in the risk of unexpected emergency medical visits or the risk of any potential side effect (overall). Of the 8 potential side effects assessed, only 1 (rash or hives) presented a different risk by treatment group (4.2% in the aspirin group vs 3.5% in the placebo group; relative risk, 1.20; 95% confidence interval, 1.01–1.43; P=.042).



CONCLUSION

The daily use of low-dose aspirin seems to be a safe intervention for reducing the risk of preterm birth and well tolerated by nulliparous pregnant women between 6 and 36 weeks’ gestation in low- and middle-income countries.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_36,

title = {Caesarean birth by maternal request: a poorly understood phenomenon in low- and middle-income countries. Int Health},

author = {Harrison MS, Garces A, Figueroa L, Esamai F, Bucher S, Bose C, Goudar S, Derman R, Patel A, Hibberd PL, Chomba E, Mwenechanya M, Hambidge M, Krebs NF

},

url = {https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7807237/},

year  = {2021},

date = {2021-01-14},

urldate = {2021-01-14},

abstract = {Background

While trends in caesarean birth by maternal request in low- and middle-income countries are unclear, age, education, multiple gestation and hypertensive disease appear associated with the indication when compared with caesarean birth performed for medical indications.



Methods

We performed a secondary analysis of a prospectively collected population-based study of home and facility births using descriptive statistics, bivariate comparisons and multilevel mixed-effects logistic regression.



Results

Of 28 751 patients who underwent caesarean birth and had a documented primary indication for the surgery, 655 (2%) were attributed to caesarean birth by maternal request. The remaining 98% were attributed to maternal and foetal indications and prior caesarean birth. In a multilevel mixed effects logistic regression adjusted for site and cluster of birth, when compared with caesareans performed for medical indications, caesarean birth performed for maternal request had a higher odds of being performed among women ≥35 y of age, with a university or higher level of education, with multiple gestations and with pregnancies complicated by hypertension (P < 0.01). Caesarean birth by maternal request was associated with a two-times increased odds of breastfeeding within 1 h of delivery, but no adverse outcomes (when compared with women who underwent caesarean birth for medical indications; P < 0.01).



Conclusion

Caesarean performed by maternal request is more common in older and more educated women and those with multifoetal gestation or hypertensive disease. It is also associated with higher rates of breastfeeding within 1 h of delivery.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_42,

title = {Regional trends in birth weight in low- and middle-income countries 2013-2018},

author = {Marete I, Ekhaguere O, Bann CM, Bucher SL, Nyongesa P, Patel AB, Hibberd PL, Saleem S, Goldenberg RL, Goudar SS, Derman RJ, Garces AL, Krebs NF, Chomba E, Carlo WA, Lokangaka A, Bauserman M, Koso-Thomas M, Moore JL, McClure EM, Esamai F

},

url = {https://reproductive-health-journal.biomedcentral.com/articles/10.1186/s12978-020-01026-2},

year  = {2020},

date = {2020-12-17},

abstract = {Background

Birth weight (BW) is a strong predictor of neonatal outcomes. The purpose of this study was to compare BWs between global regions (south Asia, sub-Saharan Africa, Central America) prospectively and to determine if trends exist in BW over time using the population-based maternal and newborn registry (MNHR) of the Global Network for Women's and Children's Health Research (Global Network).



Methods

The MNHR is a prospective observational population-based registry of six research sites participating in the Global Network (2013–2018), within five low- and middle-income countries (Kenya, Zambia, India, Pakistan, and Guatemala) in three global regions (sub-Saharan Africa, south Asia, Central America). The birth weights were obtained for all infants born during the study period. This was done either by abstracting from the infants' health facility records or from direct measurement by the registry staff for infants born at home. After controlling for demographic characteristics, mixed-effect regression models were utilized to examine regional differences in birth weights over time.



Results

The overall BW means were higher for the African sites (Zambia and Kenya), 3186 g (SD 463 g) in 2013 and 3149 g (SD 449 g) in 2018, as compared to Asian sites (Belagavi and Nagpur, India and Pakistan), 2717 g (SD450 g) in 2013 and 2713 g (SD 452 g) in 2018. The Central American site (Guatemala) had a mean BW intermediate between the African and south Asian sites, 2928 g (SD 452) in 2013, and 2874 g (SD 448) in 2018. The low birth weight (LBW) incidence was highest in the south Asian sites (India and Pakistan) and lowest in the African sites (Kenya and Zambia). The size of regional differences varied somewhat over time with slight decreases in the gap in birth weights between the African and Asian sites and slight increases in the gap between the African and Central American sites.



Conclusions

Overall, BW means by global region did not change significantly over the 5-year study period. From 2013 to 2018, infants enrolled at the African sites demonstrated the highest BW means overall across the entire study period, particularly as compared to Asian sites. The incidence of LBW was highest in the Asian sites (India and Pakistan) compared to the African and Central American sites.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_41,

title = {Stillbirth 2010-2018: a prospective, population- based, multi-country study from the Global Network},

author = {McClure EM, Saleem S, Goudar SS, Garces A, Whitworth R, Esamai F, Patel AB, Tikmani SS, Mwenechanya M, Chomba E, Lokangaka A, Bose CL, Bucher S, Liechty EA, Krebs NF, Yogesh Kumar S, Derman RJ, Hibberd PL, Carlo WA, Moore JL, Nolen TL, Koso-Thomas M, Goldenberg RL

},

url = {https://reproductive-health-journal.biomedcentral.com/articles/10.1186/s12978-020-00991-y},

year  = {2020},

date = {2020-11-30},

abstract = {Background

Stillbirth rates are high and represent a substantial proportion of the under-5 mortality in low and middle-income countries (LMIC). In LMIC, where nearly 98% of stillbirths worldwide occur, few population-based studies have documented cause of stillbirths or the trends in rate of stillbirth over time.



Methods

We undertook a prospective, population-based multi-country research study of all pregnant women in defined geographic areas across 7 sites in low-resource settings (Kenya, Zambia, Democratic Republic of Congo, India, Pakistan, and Guatemala). Staff collected demographic and health care characteristics with outcomes obtained at delivery. Cause of stillbirth was assigned by algorithm.



Results

From 2010 through 2018, 573,148 women were enrolled with delivery data obtained. Of the 552,547 births that reached 500 g or 20 weeks gestation, 15,604 were stillbirths; a rate of 28.2 stillbirths per 1000 births. The stillbirth rates were 19.3 in the Guatemala site, 23.8 in the African sites, and 33.3 in the Asian sites. Specifically, stillbirth rates were highest in the Pakistan site, which also documented a substantial decrease in stillbirth rates over the study period, from 56.0 per 1000 (95% CI 51.0, 61.0) in 2010 to 44.4 per 1000 (95% CI 39.1, 49.7) in 2018. The Nagpur, India site also documented a substantial decrease in stillbirths from 32.5 (95% CI 29.0, 36.1) to 16.9 (95% CI 13.9, 19.9) per 1000 in 2018; however, other sites had only small declines in stillbirth over the same period. Women who were less educated and older as well as those with less access to antenatal care and with vaginal assisted delivery were at increased risk of stillbirth. The major fetal causes of stillbirth were birth asphyxia (44.0% of stillbirths) and infectious causes (22.2%). The maternal conditions that were observed among those with stillbirth were obstructed or prolonged labor, antepartum hemorrhage and maternal infections.



Conclusions

Over the study period, stillbirth rates have remained relatively high across all sites. With the exceptions of the Pakistan and Nagpur sites, Global Network sites did not observe substantial changes in their stillbirth rates. Women who were less educated and had less access to antenatal and obstetric care remained at the highest burden of stillbirth.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey_45,

title = {Evaluating the effect of care around labor and delivery practices on early neonatal mortality in the Global Network's Maternal and Newborn Health Registry},

author = {Patel AB, Simmons EM, Rao SR, Moore J, Nolen TL, Goldenberg RL, Goudar SS,

Somannavar MS, Esamai F, Nyongesa P, Garces AL, Chomba E, Mwenechanya M, Saleem S, Naqvi F, Bauserman M, Bucher S, Krebs NF, Derman RJ, Carlo WA, Koso-ThomasMcClure MEM, Hibberd PL},

url = {https://reproductive-health-journal.biomedcentral.com/articles/10.1186/s12978-020-01010-w},

year  = {2020},

date = {2020-11-30},

abstract = {Background

Neonatal deaths in first 28-days of life represent 47% of all deaths under the age of five years globally and are a focus of the United Nation’s (UN’s) Sustainable Development Goals. Pregnant women are delivering in facilities but that does not indicate quality of care during delivery and the postpartum period. The World Health Organization’s Essential Newborn Care (ENC) package reduces neonatal mortality, but lacks a simple and valid composite index that measures its effectiveness.



Methods

Data on 5 intra-partum and 3 post-partum practices (indicators) recommended as part of ENC, routinely collected in NICHD’s Global Network’s (GN) Maternal Newborn Health Registry (MNHR) between 2010 and 2013, were included. We evaluated if all 8 practices (Care around Delivery – CAD), combined as an index was associated with reduced early neonatal mortality rates (days 0–6 of life).



Results

A total of 150,848 live births were included in the analysis. The individual indicators varied across sites. All components were present in 19.9% births (range 0.4 to 31% across sites). Present indicators (8 components) were associated with reduced early neonatal mortality [adjusted RR (95% CI):0.81 (0.77, 0.85); p < 0.0001]. Despite an overall association between CAD and early neonatal mortality (RR < 1.0 for all early mortality): delivery by skilled birth attendant; presence of fetal heart and delayed bathing were associated with increased early neonatal mortality.



Conclusions

Present indicators (8 practices) of CAD were associated with a 19% reduction in the risk of neonatal death in the diverse health facilities where delivery occurred within the GN MNHR. These indicators could be monitored to identify facilities that need to improve compliance with ENC practices to reduce preventable neonatal deaths. Three of the 8 indicators were associated with increased neonatal mortality, due to baby being sick at birth. Although promising, this composite index needs refinement before use to monitor facility-based quality of care in association with early neonatal mortality.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}