Search

Traumatic brain injury (TBI) disproportionately affects low- and middle-income countries (LMICs). In these settings, accurate patient prognostication is both difficult and essential for high-quality patient care. With the ultimate goal of enhancing TBI triage in LMICs, we aim to develop the first deep learning model to predict outcomes after TBI and compare its performance with that of less complex algorithms.TBI patients' data were prospectively collected in Kampala, Uganda, from 2016 to 2020. To predict good versus poor outcome at hospital discharge, we created deep neural network, shallow neural network, and elastic-net regularized logistic regression models. Predictors included 13 easily acquirable clinical variables. We assessed model performance with 5-fold cross-validation to calculate areas under both the receiver operating characteristic curve and precision-recall curve (AUPRC), in addition to standardized partial AUPRC to focus on comparisons at clinically relevant operating points.We included 2164 patients for model training, of which 12% had poor outcomes. The deep neural network performed best as measured by the area under the receiver operating characteristic curve (0.941) and standardized partial AUPRC in region maximizing recall (0.291), whereas the shallow neural network was best by the area under the precision-recall curve (0.770). In several other comparisons, the elastic-net regularized logistic regression was noninferior to the neural networks.We present the first use of deep learning for TBI prognostication, with an emphasis on LMICs, where there is great need for decision support to allocate limited resources. Optimal algorithm selection depends on the specific clinical setting; deep learning is not a panacea, though it may have a role in these efforts.

Hospitals in low- and middle-income countries (LMICs) could benefit from decision support technologies to reduce time to triage, diagnosis, and surgery for patients with traumatic brain injury (TBI). Corticosteroid Randomization after Significant Head Injury (CRASH) and International Mission for Prognosis and Clinical Trials in Traumatic Brain Injury (IMPACT) are robust examples of TBI prognostic models, although they have yet to be validated in Sub-Saharan Africa (SSA). Moreover, machine learning and improved data quality in LMICs provide an opportunity to develop context-specific, and potentially more accurate, prognostic models. We aim to externally validate CRASH and IMPACT on our TBI registry and compare their performances to that of the locally derived model (from the Kilimanjaro Christian Medical Center [KCMC]). We developed a machine learning-based prognostic model from a TBI registry collected at a regional referral hospital in Moshi, Tanzania. We also used the core CRASH and IMPACT online risk calculators to generate risk scores for each patient. We compared the discrimination (area under the curve [AUC]) and calibration before and after Platt scaling (Brier, Hosmer-Lemeshow Test, and calibration plots) for CRASH, IMPACT, and the KCMC model. The outcome of interest was unfavorable in-hospital outcome defined as a Glasgow Outcome Scale score of 1-3. There were 2972 patients included in the TBI registry, of whom 11% had an unfavorable outcome. The AUCs for the KCMC model, CRASH, and IMPACT were 0.919, 0.876, and 0.821, respectively. Prior to Platt scaling, CRASH was the best calibrated model (χ

OBJECTIVE Traumatic brain injury (TBI), a burgeoning global health concern, is one condition that could benefit from prognostic modeling. Risk stratification of TBI patients on presentation to a health facility can support the prudent use of limited resources. The CRASH (Corticosteroid Randomisation After Significant Head Injury) model is a well-established prognostic model developed to augment complex decision-making. The authors’ current study objective was to better understand in-hospital decision-making for TBI patients and determine whether data from the CRASH risk calculator influenced provider assessment of prognosis. METHODS The authors performed a choice experiment using a simulated TBI case. All participant doctors received the same case, which included a patient history, vitals, and physical examination findings. Half the participants also received the CRASH risk score. Participants were asked to estimate the patient prognosis and decide the best next treatment step. The authors recruited a convenience sample of 28 doctors involved in TBI care at both a regional and a national referral hospital in Uganda. RESULTS For the simulated case, the CRASH risk scores for 14-day mortality and an unfavorable outcome at 6 months were 51.4% (95% CI 42.8%, 59.8%) and 89.8% (95% CI 86.0%, 92.6%), respectively. Overall, participants were overoptimistic when estimating the patient prognosis. Risk estimates by doctors provided with the CRASH risk score were closer to that score than estimates made by doctors in the control group; this effect was more pronounced for inexperienced doctors. Surgery was selected as the best next step by 86% of respondents. CONCLUSIONS This study was a novel assessment of a TBI prognostic model’s influence on provider estimation of risk in a low-resource setting. Exposure to CRASH risk score data reduced overoptimistic prognostication by doctors, particularly among inexperienced providers.

The rapid emergence of machine learning in the form of large-scale computational statistics and accumulation of data offers global health implementing partners an opportunity to adopt, adapt, and apply these techniques and technologies to low- and middle-income country (LMIC) contexts where we work. These benefits reside just out of the reach of many implementing partners because they lack the experience and specific skills to use them. Yet the growth of available analytical systems and exponential growth of data require the global digital health community to become conversant in this technology to continue to make contributions to help fulfill our missions. In this community case study, we describe the approach we took at IntraHealth International to inform the use case for machine learning in global health and development. We found that the data needed to take advantage of machine learning were plentiful and that an international, interdisciplinary team can be formed to collect, clean, and analyze the data at hand using cloud-based (e.g., Dropbox, Google Drive) and open source tools (e.g., R). We organized our work as a “sprint” lasting roughly 10 weeks in length so that we could rapidly prototype these approaches in order to achieve institutional buy in. Our initial sprint resulted in two requests in subsequent workplans for analytics using the data we compiled and directly impacted program implementation.

OBJECTIVETraumatic brain injury (TBI) is a leading cause of death and disability worldwide, with a disproportionate burden of this injury on low- and middle-income countries (LMICs). Limited access to diagnostic technologies and highly skilled providers combined with high patient volumes contributes to poor outcomes in LMICs. Prognostic modeling as a clinical decision support tool, in theory, could optimize the use of existing resources and support timely treatment decisions in LMICs. The objective of this study was to develop a machine learning–based prognostic model using data from Kilimanjaro Christian Medical Centre in Moshi, Tanzania.METHODSThis study is a secondary analysis of a TBI data registry including 3138 patients. The authors tested nine different machine learning techniques to identify the prognostic model with the greatest area under the receiver operating characteristic curve (AUC). Input data included demographics, vital signs, injury type, and treatment received. The outcome variable was the discharge score on the Glasgow Outcome Scale–Extended.RESULTSThe AUC for the prognostic models varied from 66.2% (k-nearest neighbors) to 86.5% (Bayesian generalized linear model). An increasing Glasgow Coma Scale score, increasing pulse oximetry values, and undergoing TBI surgery were predictive of a good recovery, while injuries suffered from a motor vehicle crash and increasing age were predictive of a poor recovery.CONCLUSIONSThe authors developed a TBI prognostic model with a substantial level of accuracy in a low-resource setting. Further research is needed to externally validate the model and test the algorithm as a clinical decision support tool.

Traumatic brain injury (TBI) disproportionately affects low- and middle-income countries (LMICs). In these low-resource settings, effective triage of patients with TBI-including the decision of whether or not to perform neurosurgery-is critical in optimizing patient outcomes and healthcare resource utilization. Machine learning may allow for effective predictions of patient outcomes both with and without surgery. Data from patients with TBI was collected prospectively at Mulago National Referral Hospital in Kampala, Uganda, from 2016 to 2019. One linear and six non-linear machine learning models were designed to predict good versus poor outcome near hospital discharge and internally validated using nested five-fold cross-validation. The 13 predictors included clinical variables easily acquired on admission and whether or not the patient received surgery. Using an elastic-net regularized logistic regression model (GLMnet), with predictions calibrated using Platt scaling, the probability of poor outcome was calculated for each patient both with and without surgery (with the difference quantifying the "individual treatment effect," ITE). Relative ITE represents the percent reduction in chance of poor outcome, equaling this ITE divided by the probability of poor outcome with no surgery. Ultimately, 1766 patients were included. Areas under the receiver operating characteristic curve (AUROCs) ranged from 83.1% (single C5.0 ruleset) to 88.5% (random forest), with the GLMnet at 87.5%. The two variables promoting good outcomes in the GLMnet model were high Glasgow Coma Scale score and receiving surgery. For the subgroup not receiving surgery, the median relative ITE was 42.9% (interquartile range [IQR], 32.7% to 53.5%); similarly, in those receiving surgery, it was 43.2% (IQR, 32.9% to 54.3%). We provide the first machine learning-based model to predict TBI outcomes with and without surgery in LMICs, thus enabling more effective surgical decision making in the resource-limited setting. Predicted ITE similarity between surgical and non-surgical groups suggests that, currently, patients are not being chosen optimally for neurosurgical intervention. Our clinical decision aid has the potential to improve outcomes.

Purpose The study's purpose was to assess population demographics and resource utilization of the Medical Student Run Clinic, which provides primary care services to patients in El Paso, Texas along the Texas-Mexico border. Methods A retrospective cross-sectional chart review was performed on 760 patients evaluated at the medical student-run clinic between 2013 and 2016. Data included demographic characteristics, chief complaints, diagnoses, and interventions, which were analyzed with calculations of means, standard deviations, and percentages. Results Most (79.7%) patients were female; average age was 38.43 years; 91% of patients were Hispanic, and 66.8% spoke Spanish. Average BMI was 30.9 kg/m2. Less than 1% of patients presented with a psychiatric complaint; however, 17.9% screened positive for anxiety, and 16.5% screened positive for depression. Conclusions This study shows that diabetes, hypertension, obesity, anxiety, and depression represent avenues for future patient-centered interventions and provide insight into challenges patients face along the border.

Background Traumatic brain injury (TBI) prognostic models are potential solutions to severe human and technical shortages. Although numerous TBI prognostic models have been developed, none are widely used in clinical practice, largely because of a lack of feasibility research to inform implementation. We previously developed a prognostic model and Web-based application for in-hospital TBI care in low-resource settings. In this study, we tested the feasibility, acceptability, and usability of the application with potential end-users. Methods We performed our feasibility assessment with providers involved in TBI care at both a regional and national referral hospital in Uganda. We collected qualitative and quantitative data on decision support needs, application ease of use, and implementation design. Results We completed 25 questionnaires on potential uses of the app and 11 semistructured feasibility interviews. Top-cited uses were informing the decision to operate, informing the decision to send the patient to intensive care, and counseling patients and relatives. Participants affirmed the potential of the application to support difficult triage situations, particularly in the setting of limited access to diagnostics and interventions, but were hesitant to use this technology with end-of-life decisions. Although all participants were satisfied with the application and agreed that it was easy to use, several expressed a need for this technology to be accessible by smartphone and offline. Conclusions We elucidated several potential uses for our app and important contextual factors that will support future implementation. This investigation helps address an unmet need to determine the feasibility of TBI clinical decision support systems in low-resource settings.

OBJECTIVEGlobal neurosurgery is a rapidly emerging field that aims to address the worldwide shortages in neurosurgical care. Many published outreach efforts and initiatives exist to address the global disparity in neurosurgical care; however, there is no centralized report detailing these efforts. This scoping review aims to characterize the field of global neurosurgery by identifying partnerships between high-income countries (HICs) and low- and/or middle-income countries (LMICs) that seek to increase neurosurgical capacity.METHODSA scoping review was conducted using the Arksey and O’Malley framework. A search was conducted in five electronic databases and the gray literature, defined as literature not published through traditional commercial or academic means, to identify studies describing global neurosurgery partnerships. Study selection and data extraction were performed by four independent reviewers, and any disagreements were settled by the team and ultimately the team lead.RESULTSThe original database search produced 2221 articles, which was reduced to 183 final articles after applying inclusion and exclusion criteria. These final articles, along with 9 additional gray literature references, captured 169 unique global neurosurgery collaborations between HICs and LMICs. Of this total, 103 (61%) collaborations involved surgical intervention, while local training of medical personnel, research, and education were done in 48%, 38%, and 30% of efforts, respectively. Many of the collaborations (100 [59%]) are ongoing, and 93 (55%) of them resulted in an increase in capacity within the LMIC involved. The largest proportion of efforts began between 2005–2009 (28%) and 2010–2014 (17%). The most frequently involved HICs were the United States, Canada, and France, whereas the most frequently involved LMICs were Uganda, Tanzania, and Kenya.CONCLUSIONSThis review provides a detailed overview of current global neurosurgery efforts, elucidates gaps in the existing literature, and identifies the LMICs that may benefit from further efforts to improve accessibility to essential neurosurgical care worldwide.

ObjectiveThe objective of this study was to better understand how the lack of emergency child and obstetric care can be related to maternal and neonatal mortality levels.MethodsWe performed spatiotemporal geospatial analyses using data from Brazilian municipalities. An emergency service accessibility index was derived using the two-step floating catchment area (2SFCA) for 951 hospitals. Mortality data from 2000 to 2015 was used to characterize space-time trends. The data was overlapped using a spatial clusters analysis to identify regions with lack of emergency access and high mortality trends.ResultsFrom 2000 to 2015 Brazil the overall neonatal mortality rate varied from 11,42 to 11,71 by 1000 live births. The maternal mortality presented a slightly decrease from 2,98 to 2,88 by 100 thousand inhabitants. For neonatal mortality the Northeast and North regions presented the highest percentage of up trending. For maternal mortality the North region exhibited the higher volume of up trending. The accessibility index obtained highlighted large portions of the rural areas of the country without any coverage of obstetric or neonatal beds.ConclusionsThe analyses highlighted regions with problems of mortality and access to maternal and newborn emergency services. This sequence of steps can be applied to other low and medium income countries as health situation analysis tool.Significance statementLow and middle income countries have greater disparities in access to emergency child and obstetric care. There is a lack of approaches capable to support analysis considering a spatiotemporal perspective for emergency care. Studies using Geographic Information System analysis for maternal and child care, are increasing in frequency. This approach can identify emergency child and obstetric care saturated or deprived regions. The sequence of steps designed here can help researchers, and policy makers to better design strategies aiming to improve emergency child and obstetric care.

The development of smart wearable devices has strengthened a new area: mHealth. This is characterized by the use of mobile technologies for the diagnosis, monitoring, and intervention of health conditions. When inserted into the wider universe comprising the Internet of Things, mHealth solutions emerge with real-world applications. The myriad of mHealth-based applications already draws a scenario in which these new technologies will be responsible for changing the current logic of healthcare delivery. Nevertheless, there are numerous challenges for mHealth before its full potential can be realized. The Brazilian context is marked by peculiarities regarding the challenges for the dissemination of mHealth. There are few professionals, resistance to adopting new health technologies, an inexistent regulatory framework, regional inequalities, the absence of guidelines for protecting personal data, and difficulty associated with communication networks. Such aspects are discussed throughout this work together with components relating to the current status of development linked to mHealth.

OBJECTIVEIn addition to the rising burden of surgical disease globally, infrastructure and human resources for health remain a great challenge for low- and middle-income countries, especially in Uganda. In this study, the authors aim to explore the trends of neurosurgical care at a regional referral hospital in Uganda and assess the long-term impact of the institutional collaboration between Mulago National Referral Hospital and Duke University.METHODSAn interrupted time series is a quasi-experimental design used to evaluate the effects of an intervention on longitudinal data. The authors applied this design to evaluate the trends in monthly mortality rates for neurosurgery patients at Mbarara Regional Referral Hospital (MRRH) from March 2013 to October 2015. They used segmented regression and autoregressive integrated moving average models for the analysis.RESULTSOver the study timeframe, MRRH experienced significant increases in referrals received (from 117 in 2013 to 211 in 2015), neurosurgery patients treated (from 337 in 2013 to 625 in 2015), and operations performed (from 61 in 2013 to 173 in 2015). Despite increasing patient volumes, the hospital achieved a significant reduction in hospital mortality during 2015 compared to prior years (p value = 0.0039).CONCLUSIONSThis interrupted time series analysis study showed improving trends of neurosurgical care in Western Uganda. There is a steady increase in volume accompanied by a sharp decrease in mortality through the years. Multiple factors are implicated in the significant increase in volume and decrease in mortality, including the addition of a part-time neurosurgeon, improvement in infrastructure, and increased experience. Further in-depth prospective studies exploring seasonality and long-term outcomes are warranted.

Prognostic models to support clinical decisions is an innovative solution to mitigate the increasing burden of traumatic brain injury (TBI) in low and medium income countries (LMIC). Hospitals, equipped with predictive models, can make data driven decisions based on patient prognosis to improve efficiency, productivity and outcomes. Furthermore, attempts to implement TBI predictive models are limited in both high and LMIC. The main objective of this study was to develop a predictive model for TBI outcomes. Our study is categorized as retrospective, with diagnostic temporal perspective, using the classification approach. Data comprising 3138 patients admitted with TBI were collected from 2014 to 2017. Data about demographics, injury type, vital signs, Glasgow Coma Score, surgeries and outcome observed were analyzed for each patient. Our model considered the good recovery obtained through the Glasgow outcome score extended scale as a result measure. Data were analyzed using Bayesian Generalized Linear Model. The average age of patients were 31.18% and 82.02% were men. Of all TBI patients 67.87% were victims of road traffic injury. Considering the good recovery as outcome our model achieved an accuracy of 0.865 (95% CI 0.859, 0.870), with a sensitivity level of 0.890 and specificity of 0.713. our detection rate of a good recovery due TBI interventions based on admission data were of 0.762. The main predictors of a good recovery were: Glasgow coma level score, pulse oximetry, availability of intensive care unit bed. A predictive model with high accuracy and kappa level can be generated from data available at patient admission in a low resource setting such as Tanzania, Kilimanjaro Christian Medical Centre hospital. To our knowledge, this is the first prognostic model for severe TBI using machine learning techniques in Sub-Saharan Africa.

Background Castleman disease (CD) is an uncommon disorder of deregulated lymphoproliferation with unicentric (UCD) and multicentric forms based on extent of nodal involvement. Gross resection with histopathologic analysis remains the gold standard for diagnosis of UCD and is curative in most cases. Symptomatic paraspinal UCD is a rare presentation with potentially dangerous complications, and its tendency to mimic more common spinal tumors presents a significant diagnostic challenge. Case Presentation A 25-year-old Hispanic man with no past medical history was evaluated for a known left-sided paraspinal mass that was incidentally discovered during an emergency department work-up for hematuria. Computed tomography on initial presentation revealed a 5.3 cm × 3.3 cm × 4.8 cm heterogeneously enhancing left paraspinal mass adjacent to the T11 vertebral body with tonguelike extension into the T11-T12 neural foramen. Although he remained neurologically intact throughout most of the diagnostic work-up, an inconclusive biopsy, worsening hematuria, and late-onset radiculopathy with severe back pain prompted surgical intervention. Microscopic histomorphology was consistent with CD. He continued to have intermittent hematuria and dysuria postoperatively, but repeat computed tomography at 7 months confirmed no recurrence of the mass. Conclusions Compared with previous reports, our case of postcoital hematuria and radiculopathy accompanying a paraspinal thoracic mass in a young Mexican-American man is a unique presentation. Awareness and early consideration of UCD in the work-up of a paraspinal mass may spare affected patients adverse and dangerous sequelae, such as spinal cord compression and excessive intraoperative hemorrhage.

OBJECTIVEThe purpose of this study was to determine if patients with traumatic brain injury (TBI) in low- and middle-income countries who receive surgery have better outcomes than patients with TBI who do not receive surgery, and whether this differs with severity of injury.METHODSThe authors generated a series of Kaplan-Meier plots and performed multiple Cox proportional hazard models to assess the relationship between TBI surgery and TBI severity. The TBI severity was categorized using admission Glasgow Coma Scale scores: mild (14, 15), moderate (9–13), or severe (3–8). The authors investigated outcomes from admission to hospital day 14. The outcome considered was the Glasgow Outcome Scale–Extended, categorized as poor outcome (1–4) and good outcome (5–8). The authors used TBI registry data collected from 2013 to 2017 at a regional referral hospital in Tanzania.RESULTSOf the final 2502 patients, 609 (24%) received surgery and 1893 (76%) did not receive surgery. There were significantly fewer road traffic injuries and more violent causes of injury in those receiving surgery. Those receiving surgery were also more likely to receive care in the ICU, to have a poor outcome, to have a moderate or severe TBI, and to stay in the hospital longer. The hazard ratio for patients with TBI who underwent operation versus those who did not was 0.17 (95% CI 0.06–0.49; p < 0.001) in patients with moderate TBI; 0.2 (95% CI 0.06–0.64; p = 0.01) for those with mild TBI, and 0.47 (95% CI 0.24–0.89; p = 0.02) for those with severe TBI.CONCLUSIONSThose who received surgery for their TBI had a lower hazard for poor outcome than those who did not. Surgical intervention was associated with the greatest improvement in outcomes for moderate head injuries, followed by mild and severe injuries. The findings suggest a reprioritization of patients with moderate TBI—a drastic change to the traditional practice within low- and middle-income countries in which the most severely injured patients are prioritized for care.