The implementation of electronic medical records (EMRs) has generally been thought to improve medical efficiency and safety, but consistent evidence of improved healthcare quality due to EMRs in population-based studies is lacking. We assessed the relationship between the degree of EMR adoption and patient outcomes. We performed an observational study using discharge data from Tri-service General Hospital from 2013 to 2018. The levels of EMR utilization were divided into no EMRs, partial EMRs and full EMRs. The primary healthcare quality indicators were inpatient mortality, readmission within 14 days, and 48-hour postoperative mortality. We performed a Cox proportional hazards regression analysis to evaluate the relationship between the EMR utilization level and healthcare quality. In total, 262,569 patients were included in this study. Compared with no EMRs, full EMR implementation led to lower inpatient mortality [adjusted hazard ratio (HR) 0.947, 95% confidence interval (CI): 0.897–0.999, P = ..049] and a lower risk of readmission within 14 days (adjusted HR 0.627, 95% CI: 0.577–0.681, P < .001). Full EMR implementation was associated was a lower risk of 48-hour postoperative mortality (adjusted HR 0.372, 95% CI: 0.208–0.665, P = .001) than no EMRs. Partial EMR implementation was associated with a higher risk of readmission within 14 days than no EMRs (HR 1.387, 95% CI: 1.298–1.485, P < .001). Full EMR adoption improves healthcare quality in medical institutions treating severely ill patients. A prospective study is needed to confirm this finding.

Background Most current state-of-the-art models for searching the International Classification of Diseases, Tenth Revision Clinical Modification (ICD-10-CM) codes use word embedding technology to capture useful semantic properties. However, they are limited by the quality of initial word embeddings. Word embedding trained by electronic health records (EHRs) is considered the best, but the vocabulary diversity is limited by previous medical records. Thus, we require a word embedding model that maintains the vocabulary diversity of open internet databases and the medical terminology understanding of EHRs. Moreover, we need to consider the particularity of the disease classification, wherein discharge notes present only positive disease descriptions. Objective We aimed to propose a projection word2vec model and a hybrid sampling method. In addition, we aimed to conduct a series of experiments to validate the effectiveness of these methods. Methods We compared the projection word2vec model and traditional word2vec model using two corpora sources: English Wikipedia and PubMed journal abstracts. We used seven published datasets to measure the medical semantic understanding of the word2vec models and used these embeddings to identify the three–character-level ICD-10-CM diagnostic codes in a set of discharge notes. On the basis of embedding technology improvement, we also tried to apply the hybrid sampling method to improve accuracy. The 94,483 labeled discharge notes from the Tri-Service General Hospital of Taipei, Taiwan, from June 1, 2015, to June 30, 2017, were used. To evaluate the model performance, 24,762 discharge notes from July 1, 2017, to December 31, 2017, from the same hospital were used. Moreover, 74,324 additional discharge notes collected from seven other hospitals were tested. The F-measure, which is the major global measure of effectiveness, was adopted. Results In medical semantic understanding, the original EHR embeddings and PubMed embeddings exhibited superior performance to the original Wikipedia embeddings. After projection training technology was applied, the projection Wikipedia embeddings exhibited an obvious improvement but did not reach the level of original EHR embeddings or PubMed embeddings. In the subsequent ICD-10-CM coding experiment, the model that used both projection PubMed and Wikipedia embeddings had the highest testing mean F-measure (0.7362 and 0.6693 in Tri-Service General Hospital and the seven other hospitals, respectively). Moreover, the hybrid sampling method was found to improve the model performance (F-measure=0.7371/0.6698). Conclusions The word embeddings trained using EHR and PubMed could understand medical semantics better, and the proposed projection word2vec model improved the ability of medical semantics extraction in Wikipedia embeddings. Although the improvement from the projection word2vec model in the real ICD-10-CM coding task was...

Coronavirus disease-2019 (COVID-19) is a global pandemic affecting numerous countries around the world. This study elaborates Taiwan's epidemiological characteristics from the 2020 to 2021 COVID-19 pandemic from human, temporal, and geographical dimensions. Big data for cases were obtained from a public database from the Taiwan Centers for Disease Control (CDC) in April 2021. The data were analyzed and used to compare differences, correlations, and trends for human, temporal, and geographical characteristics for imported and domestic COVID-19 cases. During the study period, 1030 cases were confirmed and the mortality rate of 1.0%. The epidemiological features indicated that most cases (953/1030, 92.5%) were imported. A comparison of the domestic confirmed and imported cases revealed the following findings: No significant difference of COVID-19 between males and females for sex was observed; For age, the risk of domestic transmission was significantly lower for 20 to 29 years old, higher for 50 to 59 years old, and >60 years old with odds ratios (ORs) ( P value < .05) of 0.36, 3.37, and 2.50, respectively; For the month of infection, the ORs ( P value < .05) of domestic confirmed cases during January and February 2020 were 22.428; and in terms of area of residence, the ORs ( P value < .05) for domestic confirmed cases in northern and southern Taiwan were 4.473 and 0.033, respectively. Thus, the increase in domestic cases may have been caused by international travelers transmitting the virus in March 2020 and December 2020, respectively. Taiwan has been implementing effective screening and quarantine measures at airports. Moreover, Taiwan has implemented and maintained stringent interventions such as large-scale epidemiological investigation, rapid diagnosis, wearing masks, washing hands frequently, safe social distancing, and prompt clinical classifications for severe patients who were given appropriate medical measures. This is the first report comparing imported and domestic cases of COVID-19 from surveillance data from the Taiwan Centers for Disease Control during January 2020 and March 2021. It illustrates that individuals infected during overseas travel are the main risk factors for the spread of COVID-19 in Taiwan. The study also highlights the importance of longitudinal and geographically extended studies in understanding the implications of COVID-19 transmission for Taiwan's population.

Background: Hemorrhagic fever with renal syndrome (HFRS) is an illness caused by hantaviruses. Numerous factors modify the risk of hantavirus transmission. This study explored the epidemiological characteristics, differences, and trends in terms of gender, age, season, and living areas of those diagnosed with domestically acquired HFRS in Taiwan from 2001 to 2019. Methods: We examined publicly available annual summary data on the domestic cases with HFRS from 2001 to 2019; these data were obtained from the web database of Taiwan’s Centers for Disease Control (CDC). Results: This study analyzed 21 domestic cases with HFRS from Taiwan’s CDC databases. In this study of the cases of HFRS in Taiwan, a gradual increase in the cases of those aged ≥40 years acquiring the disease was noted, and a distinct pattern of seasonal variation (spring) was observed. Furthermore, more men had domestically acquired HFRS, and living in Taipei metropolitan area (6 cases [28.6%]) and the rural areas (Gao-Ping region, 9 cases [42.9%]) was identified as a potential risk factor. This study represents the first report of confirmed cases of domestically acquired HFRS from surveillance data from Taiwan’s CDC, 2001–2019. Conclusion: This study highlights the importance of longitudinal studies covering a wide geographical area, particularly for highly fluctuating pathogens, to understanding the implications of the transmission of zoonotic diseases in human populations. Important data were identified to inform future surveillance and research efforts in Taiwan.

Background: Chikungunya fever is caused by the chikungunya virus. Numerous factors affect the risk of chikungunya transmission. This study explored the epidemiological characteristics, differences, and trends in domestic and imported cases of chikungunya fever in Taiwan in terms of patient sex, age, month of confirmation, and area of residence from 2007 to 2019. Methods: Public annual chikungunya data from Taiwan’s Centers for Disease Control (CDC) were analyzed. In total, 21 confirmed domestic and 198 imported cases of chikungunya were reported. Of the domestic cases, one was sporadic and reported in July 2019, and 20 were attributed to a cluster event during August and September 2019. Results: In a comparison between domestic and imported cases reported from July to October 2019, differences in sex were nonsignificant (p = 0.555), whereas significant differences were observed for age (p < 0.001), month of confirmation (p = 0.005), and place of residence (p = 0.001). An age of 69–69 years (odds ratio (OR) = 6.66, 95% confidence interval (95%CI) = 2.15–20.65), month of confirmation of September (OR = 5.25, 95%CI = 1.89–14.61) and place of residence of New Taipei City (OR = 48.70, 95%CI = 6.17–384.44) were identified as potential risk factors. Additionally, domestic cases in August and September 2019 increased in proportion to the increase in imported cases during July and August 2019. Increased domestic patients may have been caused by the domestic mosquitoes that transmitted the virus by biting the imported patients to Taiwan. This is the first report comparing domestic and imported cases of chikungunya from surveillance data from the Taiwan CDC from 2007 to 2019. Conclusion: This study highlights the importance of longitudinal and geographically extended studies to understand the implications of zoonotic disease transmission on Taiwan’s population. Critical data were identified to inform future surveillance and research efforts in Taiwan.

Foodborne pathogens cause diseases and death, increasing the economic burden. It needs to identify incident places, media food and pathgens. Our aim is to survey empirical data that provide a retrospective historical perspective on foodborne diseases and explore the causes and trends of outbreaks. We examined publicly available annual summary data on reported foodborne disease outbreaks in Taiwan from 2014 to 2018. We calculated the percentage of places, media food, bacteria and natural toxin sources in foodborne diseases and performed a chi-square test for difference evaluation. The higher risk of places and causes in 2018 compared with 2014 was empolyzed with univariate logistic regression. There were 26847 patients with foodborne diseases during the period from 2014 to 2018. The top 2 primary source locations of the foodborne diseases were schools and restaurants. The top 2 primary food media classifications of the foodborne diseases were boxed meals and compounded foods. The top 2 primary incident bacterial classifications of the observed foodborne diseases were Bacillus cereus and Staphylococcus aureus . The top 2 primary natural toxin classifications of the foodborne diseases were plants and histamines. The incidence of foodborne disease in military facilities, fruits and vegetables, and Staphylococcus aureus was increased in our study. Our study confirmed the high risk and increased incidence of foodborne diseases, food media classifications, bacterial classifications, and natural toxins in Taiwan. It is worthy of attention for the government health department-designed policy to promote disease prevention.

A mass casualty incident (MCI) can have an enormous impact on an already crowded emergency department (ED), affecting the quality of health care provided to non-MCI ED patients. On June 26, 2015, a burn MCI (BMCI) occurred due to a cornstarch explosion at a party at a water park. The competing needs of the BMCI patients might have crowded out the needs of the non-BMCI patients. Although crowd-out effects have been previously documented in a variety of health care situations, they have not been extensively evaluated during MCIs. We aimed to determine whether the outcomes of the non-MCI patients were compromised during this incident. We conducted a retrospective observational study comparing several health care parameters and outcomes between non-BMCI patients and historical controls during the designated period using institutional electronic records and the National Health Insurance Research Database. On the night of the incident, 53 patients were sent to our ED; most of them arrived within 3 hours after the BMCI. There was a significant increase in the wait time for ICU beds among non-BMCI patients compared to the wait times during the corresponding week of the previous year (8.09 ± 4.21 hours vs 3.77 ± 2.15 hours, P = .008). At the hospital level, there was a significantly increased length of hospital stay (LOS) in the ICU after the MCI compared with the LOS in the ICU in the same week of the preceding year (median days: 15 vs 8, P ≤ .001). At the regional level, there were no significant differences between the 2 periods in the LOS in acute care, LOS in the ICU or mortality rates at the involved medical centers. Crowd-out effects from the MCI occurred in the ED and at the institutional level. Although there was an increased wait time for admission to the ICU and a longer LOS in the ICU, the LOS in acute care beds, treatment of time-sensitive diseases, and mortality rates were not compromised by the current MCI protocol at either the institutional or regional levels.

The copy-and-paste feature is commonly used for clinical documentation, and a policy is needed to reduce overdocumentation. We aimed to determine if the restricted use of copy and paste by doctors could improve inpatient healthcare quality. Clinical documentation in an inpatient dataset compiled from 2016 to 2018 was used. Copied-and-pasted text was detected in word templates using natural language programming with a threshold of 70%. The prevalence of copying and pasting after the policy introduction was accessed by segmented regression for trend analysis. The rate of readmission for the same disease within 14 days was assessed to evaluate inpatient healthcare quality, and the completion of discharge summary notes within 3 days was assessed to determine the timeliness of note completion. The relationships between these factors were used cross-correlation to detect lag effect. Poisson regression was performed to identify the relative effect of the copy and paste restriction policy on the 14-day readmission rate or the discharge note completion rate within 3 days. The prevalence of copying and pasting initially decreased, then increased, and then flatly decreased. The cross-correlation results showed a significant correlation between the prevalence of copied-and-pasted text and the 14-day readmission rate ( P < .001) and a relative risk of 1.105 ( P < .005), with a one-month lag. The discharge note completion rate initially decreased and not affected long term after restriction policy. Appropriate policies to restrict the use of copying and pasting can lead to improvements in inpatient healthcare quality. Prospective research with cost analysis is needed.