Assessing domain adaptation in adverse drug event extraction on real-world breast cancer records.

BACKGROUND: Adverse Drug Events (ADE) are key information present in unstructured portions of Electronic Health Records. These pose a significant challenge in healthcare, ranging from mild discomfort to severe complications, and can impact patient safety and treatment outcomes. METHODS: We explore the influence of domain shift between a set of dummy clinical notes and a real-world hospital corpus of Japanese clinical notes of breast cancer treatment when extracting ADEs from free text. We annotated a subset of the hospital dataset and used it to fine-tune a Named Entity Recognition (NER) model, initially trained with the set of dummy documents. We used increasing amounts of the annotated data and evaluated the impact on the model's performance. Additionally, we examined the extracted information to identify combinations of drugs that are likely to cause ADEs. RESULTS: We show that domain adaptation can significantly improve model performance in the new domain, as by feeding a small subset of 100 documents for the fine-tuning process we saw a 40% improvement in model performance. However, we also noticed diminishing returns when fine-tuning the model with a larger dataset. For instance, by feeding eight times more data, we only saw further 18% improvement in extraction performance. CONCLUSION: While variations in writing style and vocabulary in clinical corpora can significantly impact the quality of NER results. We show that domain adaptation can be of great aid in mitigating these discrepancies and achieving better performance. Yet, while providing in-domain data to a model helps, there are diminishing returns when fine-tuning with large amounts of data.

Travel burdens to access care among children with cancer between 2016 and 2019: Analysis of a national population-based cancer registry in Japan.

BACKGROUND: Centralization of cancer care increases survival but increases the travel burden (i.e., travel durations, distances, and expenditures) in visiting hospitals. This study investigated the travel burdens to access cancer care for children aged 18 years and younger in Japan. METHODS: The study population comprised 10,709 patients diagnosed between 2016 and 2019 obtained from a national population-based cancer registry in Japan. Their residences were classified as urban or rural. We counted the number of patients treated at specialized hospitals and investigated the treatment centralization across diagnostic groups by Pareto plot. Travel burdens to access care were estimated using a route-planner web service and summarized using median values. A multivariable logistic model was performed to investigate factors associated with the events of car travel duration exceeding 1 h. RESULTS: Of the patients, 76.7% lived in urban areas, and 82.5% received treatment in designated hospitals for childhood cancer. The Pareto plot suggested that the top five hospitals treated 63.5% of patients with retinoblastoma. The estimated travel burdens for all patients were 0.62 h (0.57 h in urban areas and 1.00 h in rural areas), 16.9 km, and 0.0 dollars of toll charges. Regarding travel duration, 21.7% of patients had travel exceeding 1 h, and rural areas, retinoblastoma, malignant bone tumors, and childhood cancer-hub hospitals were associated with travel duration exceeding 1 h (adjusted odds ratios of 6.93, 3.59, 1.94, and 1.91, respectively). CONCLUSIONS: Most patients were treated in specialized hospitals and the treatments for specific diseases were centralized. However, most patients were estimated to travel less than 1 h, and the travel burden tended to increase for patients in rural areas, those with specific diseases, and those going to specialized hospitals. Cancer control measures in Japan have steadily improved centralized treatment while keeping the travel burden relatively manageable.

Narrowing the Patient-Physician Gap Based on Self-Reporting and Monthly Hepatologist Feedback for Patients With Alcohol-Related Liver Disease: Interventional Pilot Study Using a Journaling Smartphone App.

BACKGROUND: Screening and intervention for alcohol use disorders (AUDs) are recommended to improve the prognosis of patients with alcohol-related liver disease (ALD). Most patients' smartphone app diaries record drinking behavior for self-monitoring. A smartphone app can be expected to also be helpful for physicians because it can provide rich patient information to hepatologists, leading to suitable feedback. We conducted this prospective pilot study to assess the use of a smartphone app as a journaling tool and as a self-report-based feedback source for patients with ALD. OBJECTIVE: The aims of this study were assessment of whether journaling (self-report) and self-report-based feedback can help patients maintain abstinence and improve liver function data. METHODS: This pilot study used a newly developed smartphone journaling app for patients, with input data that physicians can review. After patients with ALD were screened for harmful alcohol use, some were invited to use the smartphone journaling app for 8 weeks. Their self-reported alcohol intake, symptoms, and laboratory data were recorded at entry, week 4, and week 8. Biomarkers for alcohol use included gamma glutamyl transferase (GGT), percentage of carbohydrate-deficient transferrin to transferrin (%CDT), and GGT-CDT (GGT-CDT= 0.8 × ln[GGT] + 1.3 × ln[%CDT]). At each visit, their recorded data were reviewed by a hepatologist to evaluate changes in alcohol consumption and laboratory data. The relation between those outcomes and app usage was also investigated. RESULTS: Of 14 patients agreeing to participate, 10 completed an 8-week follow-up, with diary input rates between 44% and 100% of the expected days. Of the 14 patients, 2 withdrew from clinical follow-up, and 2 additional patients never used the smartphone journaling app. Using the physician's view, a treating hepatologist gave feedback via comments to patients at each visit. Mean self-reported alcohol consumption dropped from baseline (100, SD 70 g) to week 4 (13, SD 25 g; P=.002) and remained lower at week 8 (13, SD 23 g; P=.007). During the study, 5 patients reported complete abstinence. No significant changes were found in mean GGT and mean %CDT alone, but the mean GGT-CDT combination dropped significantly from entry (5.2, SD 1.2) to the week 4 visit (4.8, SD 1.1; P=.02) and at week 8 (4.8, SD 1.0; P=.01). During the study period, decreases in mean total bilirubin (3.0, SD 2.4 mg/dL to 2.4, SD 1.9 mg/dL; P=.01) and increases in mean serum albumin (3.0, SD 0.9 g/dL to 3.3, SD 0.8 g/dL; P=.009) were recorded. CONCLUSIONS: These pilot study findings revealed that a short-term intervention with a smartphone journaling app used by both patients and treatment-administering hepatologists was associated with reduced drinking and improved liver function. TRIAL REGISTRATION: UMIN CTR UMIN000045285; http://tinyurl.com/yvvk38tj.

Trends of travel burdens to access cancer care among children with cancer: analysis of a population-based cancer registry data in Aichi, Japan.

Centralization of childhood cancer treatment in specialized hospitals is necessary for quality treatment and care, but imposes a time and cost burden for patients and their families. We investigated the 20-year trend in the patients' car travel burden to reach cancer-care hospitals in Aichi Prefecture, Japan. From the Aichi population-based cancer registry data, 1,741 cases diagnosed in 1998-2017 under 15 years of age were extracted and assigned to three treatment groups: invasive treatment (n = 697), radiotherapy (n = 371), or chemotherapy groups (n = 1,462), allowing for duplicate assignment. Their travels to access each treatment hospital were estimated and summarized as the estimated travel times (ETT), estimated travel distances (ETD), and direct distances (DD). The ETTs were compared using the Brunner-Munzel test. The average cases per year for each hospital were plotted. The annual trends during 1998-2017 on ETT, ETD, and DD were investigated using Joinpoint regression models. The ETTs were 0.38-0.45 hours on median for three periods (1998-2005, 2006-2012, and 2013-2017) in three treatment groups and increased by 0.02-0.07 hours from 2006-2012 to 2013-2017, with a statistically significant difference in the radiotherapy group (0.07 hours, P = 0.037). The average cases per year increased for the top hospital in each group, and regression model analyses showed no joinpoint on the annual median trend. In conclusion, the increases in travel times were small and not considered clinically significant, and treatment centralization was observed from 2006-2012 to 2013-2017.

Major Causes of Death among Older Adults after the Great East Japan Earthquake: A Retrospective Study.

This retrospective study investigated the 3-year impact of the Great East Japan Earthquake (GEJE) of 2011 on deaths due to neoplasm, heart disease, stroke, pneumonia, and senility among older adults in the primarily affected prefectures compared with other prefectures, previous investigations having been more limited as regards mortality causes and geographic areas. Using death certificates issued between 2006 and 2015 (n = 7,383,253), mortality rates (MRs) and risk ratios (RRs) were calculated using a linear mixed model with the log-transformed MR as the response variable. The model included interactions between the area category and each year of death from 2010 to 2013. The RRs in the interaction significantly increased to 1.13, 1.17, and 1.28 for deaths due to stroke, pneumonia, and senility, respectively, in Miyagi Prefecture in 2011, but did not significantly increase for any of the other areas affected by the GEJE. Moreover, increased RRs were not reported for any of the other years. The risk of death increased in 2011; however, this was only significant for single-year impact. In 2013, decreased RRs of pneumonia in the Miyagi and Iwate prefectures and of senility in Fukushima Prefecture were observed. Overall, we did not find evidence of strong associations between the GEJE and mortality.

Impact of the Great East Japan Earthquake on spontaneous abortion and induced abortion: A population-based cross-sectional and longitudinal study in the Fukushima Prefecture based on the census survey of the Fukushima maternity care facility and vital statistics.

AIM: The 2011 Great East Japan Earthquake (GEJE) was a disaster leading to radiation exposure and psychological distress, particularly among pregnant women. However, it is not known how this affected the seasonal changes of pregnancy and childbirth. Therefore, this study investigated the impact of the GEJE in the Fukushima Prefecture on spontaneous and induced abortions with regard to seasonal variability. METHODS: We used the data of vital statistics of live birth and stillbirth registry and the census survey of the Fukushima Maternity Care Facility. We calculated the spontaneous and induced abortion rate for 2011-2016 using two different methods (cross-sectional and longitudinal). We calculated the quartiles and outliers to determine the impact and duration of the GEJE. Periodicity was investigated using spectral density analysis. The data were analyzed for the entire Fukushima Prefecture and by region. RESULTS: The spontaneous abortion rate did not show specific changes after the GEJE. Contrarily, the monthly analysis in the cross-sectional method, revealed specific increases in induced abortion rate during the year after the GEJE; in the longitudinal method, induced abortions increased among women who became pregnant within 1 year after the GEJE. Spontaneous abortion showed no specific periodicity, while induced abortion showed cycles of 6 and 12 months, with a particular increase in May each year. CONCLUSIONS: The spontaneous abortion rate was not affected by the GEJE. The changes in the induced abortion rate after the disaster may have overlapped with the timing of the increased periodicity, and cannot be attributed solely to the GEJE.

Pregnant women's migration patterns before childbirth after large-scale earthquakes and the added impact of concerns regarding radiation exposure in Fukushima and five prefectures.

The 2011 Great East Japan Earthquake (within Fukushima, Iwate, and Miyagi prefectures) was a complex disaster; it caused a tsunami and the Fukushima Daiichi Nuclear Power Plant accident, resulting in radiation exposure. This study investigated the earthquake's effects on the migration patterns of pregnant women and their concerns regarding radiation exposure. We also considered the following large-scale earthquakes without radiation exposure: Great Hanshin-Awaji (Hyogo prefecture), Niigata-Chuetsu, and Kumamoto. Pregnant women were categorized as outflow and inflow pregnant women. Data on the annual number of births three years before and after the earthquake were used as a denominator to calculate the outflow and inflow rates per 100 births. The odds ratios of annual outflow and inflow rates after the earthquake, using three years before the earthquake as the baseline, were calculated. The odds-ratio for outflow significantly increased for Hyogo, Fukushima, Miyagi, and Kumamoto prefectures after the earthquake, particularly for Fukushima, showing a significant increase until three years post the Great East Japan Earthquake (disaster year: odds-ratio: 2.66 [95% confidence interval: 2.44-2.90], 1 year post: 1.37 [1.23-1.52], 2 years post: 1.13 [1.00-1.26], 3 years post: 1.18 [1.05-1.31]), while the remaining three prefectures reported limited increases post one year. The inflow decreased after the earthquake, particularly in Fukushima, showing a significant decrease until 2 years post the Great East Japan Earthquake (disaster year: 0.58 [0.53-0.63], 1 year post: 0.76 [0.71-0.82], 2 years post: 0.83 [0.77-0.89]). Thus, pregnant women's migration patterns changed after large-scale earthquakes, suggesting radiation exposure concerns possibly have a significant effects. These results suggested that plans for receiving assistance and support that considers the peculiarities of disaster related damage and pregnant women's migration patterns are needed in both the affected and non-affected areas.