A Natural Language Processing Model for COVID-19 Detection Based on Dutch General Practice Electronic Health Records by Using Bidirectional Encoder Representations From Transformers: Development and Validation Study.

BACKGROUND: Natural language processing (NLP) models such as bidirectional encoder representations from transformers (BERT) hold promise in revolutionizing disease identification from electronic health records (EHRs) by potentially enhancing efficiency and accuracy. However, their practical application in practice settings demands a comprehensive and multidisciplinary approach to development and validation. The COVID-19 pandemic highlighted challenges in disease identification due to limited testing availability and challenges in handling unstructured data. In the Netherlands, where general practitioners (GPs) serve as the first point of contact for health care, EHRs generated by these primary care providers contain a wealth of potentially valuable information. Nonetheless, the unstructured nature of free-text entries in EHRs poses challenges in identifying trends, detecting disease outbreaks, or accurately pinpointing COVID-19 cases. OBJECTIVE: This study aims to develop and validate a BERT model for detecting COVID-19 consultations in general practice EHRs in the Netherlands. METHODS: The BERT model was initially pretrained on Dutch language data and fine-tuned using a comprehensive EHR data set comprising confirmed COVID-19 GP consultations and non-COVID-19-related consultations. The data set was partitioned into a training and development set, and the model's performance was evaluated on an independent test set that served as the primary measure of its effectiveness in COVID-19 detection. To validate the final model, its performance was assessed through 3 approaches. First, external validation was applied on an EHR data set from a different geographic region in the Netherlands. Second, validation was conducted using results of polymerase chain reaction (PCR) test data obtained from municipal health services. Lastly, correlation between predicted outcomes and COVID-19-related hospitalizations in the Netherlands was assessed, encompassing the period around the outbreak of the pandemic in the Netherlands, that is, the period before widespread testing. RESULTS: The model development used 300,359 GP consultations. We developed a highly accurate model for COVID-19 consultations (accuracy 0.97, F1-score 0.90, precision 0.85, recall 0.85, specificity 0.99). External validations showed comparable high performance. Validation on PCR test data showed high recall but low precision and specificity. Validation using hospital data showed significant correlation between COVID-19 predictions of the model and COVID-19-related hospitalizations (F1-score 96.8; P<.001; R2=0.69). Most importantly, the model was able to predict COVID-19 cases weeks before the first confirmed case in the Netherlands. CONCLUSIONS: The developed BERT model was able to accurately identify COVID-19 cases among GP consultations even preceding confirmed cases. The validated efficacy of our BERT model highlights the potential of NLP models to identify disease outbreaks early, exemplifying the power of multidisciplinary efforts in harnessing technology for disease identification. Moreover, the implications of this study extend beyond COVID-19 and offer a blueprint for the early recognition of various illnesses, revealing that such models could revolutionize disease surveillance.

Sex-specific patterns and lifetime risk of multimorbidity in the general population: a 23-year prospective cohort study.

BACKGROUND: Multimorbidity poses a major challenge for care coordination. However, data on what non-communicable diseases lead to multimorbidity, and whether the lifetime risk differs between men and women are lacking. We determined sex-specific differences in multimorbidity patterns and estimated sex-specific lifetime risk of multimorbidity in the general population. METHODS: We followed 6,094 participants from the Rotterdam Study aged 45 years and older for the occurrence of ten diseases (cancer, coronary heart disease, stroke, chronic obstructive pulmonary disease, depression, diabetes, dementia, asthma, heart failure, parkinsonism). We visualised participants' trajectories from a single disease to multimorbidity and the most frequent combinations of diseases. We calculated sex-specific lifetime risk of multimorbidity, considering multimorbidity involving only somatic diseases (1) affecting the same organ system, (2) affecting different organ systems, and (3) multimorbidity involving depression. RESULTS: Over the follow-up period (1993-2016, median years of follow-up 9.2), we observed 6334 disease events. Of the study population, 10.3% had three or more diseases, and 27.9% had two or more diseases. The most frequent pair of co-occurring diseases among men was COPD and cancer (12.5% of participants with multimorbidity), the most frequent pair of diseases among women was depression and dementia (14.9%). The lifetime risk of multimorbidity was similar among men (66.0%, 95% CI: 63.2-68.8%) and women (65.1%, 95% CI: 62.5-67.7%), yet the risk of multimorbidity with depression was higher for women (30.9%, 95% CI: 28.4-33.5%, vs. 17.5%, 95% CI: 15.2-20.1%). The risk of multimorbidity with two diseases affecting the same organ is relatively low for both sexes (4.2% (95% CI: 3.2-5.5%) for men and 4.5% (95% CI: 3.5-5.7%) for women). CONCLUSIONS: Two thirds of people over 45 will develop multimorbidity in their remaining lifetime, with women at nearly double the risk of multimorbidity involving depression than men. These findings call for programmes of integrated care to consider sex-specific differences to ensure men and women are served equally.

Prevalence and determinants of healthcare avoidance during the COVID-19 pandemic: A population-based cross-sectional study.

BACKGROUND: During the Coronavirus Disease 2019 (COVID-19) pandemic, the number of consultations and diagnoses in primary care and referrals to specialist care declined substantially compared to prepandemic levels. Beyond deferral of elective non-COVID-19 care by healthcare providers, it is unclear to what extent healthcare avoidance by community-dwelling individuals contributed to this decline in routine healthcare utilisation. Moreover, it is uncertain which specific symptoms were left unheeded by patients and which determinants predispose to healthcare avoidance in the general population. In this cross-sectional study, we assessed prevalence of healthcare avoidance during the pandemic from a patient perspective, including symptoms that were left unheeded, as well as determinants of healthcare avoidance. METHODS AND FINDINGS: On April 20, 2020, a paper COVID-19 survey addressing healthcare utilisation, socioeconomic factors, mental and physical health, medication use, and COVID-19-specific symptoms was sent out to 8,732 participants from the population-based Rotterdam Study (response rate 73%). All questionnaires were returned before July 10, 2020. By hand, prevalence of healthcare avoidance was subsequently verified through free text analysis of medical records of general practitioners. Odds ratios (ORs) for avoidance were determined using logistic regression models, adjusted for age, sex, and history of chronic diseases. We found that 1,142 of 5,656 included participants (20.2%) reported having avoided healthcare. Of those, 414 participants (36.3%) reported symptoms that potentially warranted urgent evaluation, including limb weakness (13.6%), palpitations (10.8%), and chest pain (10.2%). Determinants related to avoidance were older age (adjusted OR 1.14 [95% confidence interval (CI) 1.08 to 1.21]), female sex (1.58 [1.38 to 1.82]), low educational level (primary education versus higher vocational/university 1.21 [1.01 to 1.46), poor self-appreciated health (per level decrease 2.00 [1.80 to 2.22]), unemployment (versus employed 2.29 [1.54 to 3.39]), smoking (1.34 [1.08 to 1.65]), concern about contracting COVID-19 (per level increase 1.28 [1.19 to 1.38]) and symptoms of depression (per point increase 1.13 [1.11 to 1.14]) and anxiety (per point increase 1.16 [1.14 to 1.18]). Study limitations included uncertainty about (perceived) severity of the reported symptoms and potentially limited generalisability given the ethnically homogeneous study population. CONCLUSIONS: In this population-based cross-sectional study, 1 in 5 individuals avoided healthcare during lockdown in the COVID-19 pandemic, often for potentially urgent symptoms. Healthcare avoidance was strongly associated with female sex, fragile self-appreciated health, and high levels of depression and anxiety. These results emphasise the need for targeted public education urging these vulnerable patients to timely seek medical care for their symptoms to mitigate major health consequences.

Design, implementation and initial findings of COVID-19 research in the Rotterdam Study: leveraging existing infrastructure for population-based investigations on an emerging disease.

The Rotterdam Study is an ongoing prospective, population-based cohort study that started in 1989 in the city of Rotterdam, the Netherlands. The study aims to unravel etiology, preclinical course, natural history and potential targets for intervention for chronic diseases in mid-life and late-life. It focuses on cardiovascular, endocrine, hepatic, neurological, ophthalmic, psychiatric, dermatological, otolaryngological, locomotor, and respiratory diseases. In response to the COVID-19 pandemic, a substudy was designed and embedded within the Rotterdam Study. On the 20th of April, 2020, all living non-institutionalized participants of the Rotterdam Study (n = 8732) were invited to participate in this sub-study by filling out a series of questionnaires administered over a period of 8 months. These questionnaires included questions on COVID-19 related symptoms and risk factors, characterization of lifestyle and mental health changes, and determination of health care seeking and health care avoiding behavior during the pandemic. As of May 2021, the questionnaire had been sent out repeatedly for a total of six times with an overall response rate of 76%. This article provides an overview of the rationale, design, and implementation of this sub-study nested within the Rotterdam Study. Finally, initial results on participant characteristics and prevalence of COVID-19 in this community-dwelling population are shown.

Children and young people's consultation rates for psychosocial problems between 2016 and 2021 in the Netherlands.

BACKGROUND: Worldwide, there are concerns about declining mental health of children and young people (CYP). OBJECTIVES: To examine trends in GP consultation rates for psychosocial problems and the impact of the COVID-19 pandemic. METHODS: We performed a population-based cohort study using electronic GP records of CYP (0-24 years) living in the Rotterdam metropolitan area between 2016 and 2021. We calculated monthly consultation rates for psychosocial problems, stratified by age group and sex. We used negative binomial models to model the pre-COVID-19 trend, and estimate expected rates post-COVID-19 onset. We modelled the effect of COVID-19 infection rate and school closure on consultation rates per sex and age group. RESULTS: The cohort increased from 64801 to 92093 CYP between January 2016 and December 2021. Median age was 12.5 years and 49.3% was female. Monthly consultation rates increased from 2,443 to 4,542 consultations per 100,000 patient months over the six years. This trend (RR 1.009, 95%CI 1.008-1.011) started well before the COVID-19 pandemic. Consultation rates of adolescent girls and young women increased most strongly. Between March and May 2020, there was a temporary reduction in consultation rates, whereupon these returned to expected levels. COVID-19 infection rate and school closures showed small but significant associations with consultation rates for psychosocial problems but this did not affect the overall trend. Although consultation rates for psychosocial problems increased, this increment was stable over the entire study period. CONCLUSION: The COVID-19 pandemic did not significantly increase consultation rates for psychosocial problems in CYP. The consultation rates increased, especially in adolescent girls and young women.

CYP' consultation rates for psychosocial problems within general practice almost doubled between 2016 and 2021 in the Rotterdam metropolitan area, the Netherlands.This increase was steady over time and was not affected by the COVID-19 pandemic.Consultation rates in general practice for psychosocial problems increased most strongly in adolescent girls and young women.

Healthcare avoidance during the early stages of the COVID-19 pandemic and all-cause mortality: a longitudinal community-based study.

BACKGROUND: During the COVID-19 pandemic, global trends of reduced healthcare-seeking behaviour were observed. This raises concerns about the consequences of healthcare avoidance for population health. AIM: To determine the association between healthcare avoidance during the early stages of the COVID-19 pandemic and all-cause mortality. DESIGN AND SETTING: This was a 32-month follow-up within the population-based Rotterdam Study, after sending a COVID-19 questionnaire at the onset of the pandemic in April 2020 to all communty dwelling participants (n = 6241/8732, response rate 71.5%). METHOD: Cox proportional hazards models assessed the risk of all-cause mortality among respondents who avoided health care because of the COVID-19 pandemic. Mortality status was collected through municipality registries and medical records. RESULTS: Of 5656 respondents, one-fifth avoided health care because of the COVID-19 pandemic (n = 1143). Compared with non-avoiders, those who avoided health care more often reported symptoms of depression (n = 357, 31.2% versus n = 554, 12.3%) and anxiety (n = 340, 29.7% versus n = 549, 12.2%), and more often rated their health as poor to fair (n = 336, 29.4% versus n = 457, 10.1%) . Those who avoided health care had an increased adjusted risk of all-cause mortality (hazard ratio [HR] 1.30, 95% confidence interval [CI] = 1.01 to 1.67), which remained nearly identical after adjustment for history of any non-communicable disease (HR 1.20, 95% CI = 0.93 to 1.54). However, this association attenuated after additional adjustment for mental and physical self-perceived health factors (HR 0.93, 95% CI = 0.71 to 1.20). CONCLUSION: This study found an increased risk of all-cause mortality among individuals who avoided health care during COVID-19. These individuals were characterised by poor mental and physical self-perceived health. Therefore, interventions should be targeted to these vulnerable individuals to safeguard their access to primary and specialist care to limit health disparities, inside and beyond healthcare crises.

Changes in the Diagnosis of Stroke and Cardiovascular Conditions in Primary Care During the First 2 COVID-19 Waves in the Netherlands.

BACKGROUND AND OBJECTIVES: Although there is evidence of disruption in acute cerebrovascular and cardiovascular care during the coronavirus disease 2019 (COVID-19) pandemic, its downstream effect in primary care is less clear. We investigated how the pandemic affected utilization of cerebrovascular and cardiovascular care in general practices (GPs) and determined changes in GP-recorded diagnoses of selected cerebrovascular and cardiovascular outcomes. METHODS: From electronic health records of 166,929 primary care patients aged 30 or over within the Rotterdam region, the Netherlands, we extracted the number of consultations related to cerebrovascular and cardiovascular care, and first diagnoses of selected cerebrovascular and cardiovascular risk factors (hypertension, diabetes, lipid disorders), conditions, and events (angina, atrial fibrillation, TIA, myocardial infarction, stroke). We quantified changes in those outcomes during the first COVID-19 wave (March-May 2020) and thereafter (June-December 2020) by comparing them to the same period in 2016-2019. We also estimated the number of potentially missed diagnoses for each outcome. RESULTS: The number of GP consultations related to cerebrovascular and cardiovascular care declined by 38% (0.62, 95% confidence interval 0.56-0.68) during the first wave, as compared to expected counts based on prepandemic levels. Substantial declines in the number of new diagnoses were observed for cerebrovascular events: 37% for TIA (0.63, 0.41-0.96) and 29% for stroke (0.71, 0.59-0.84), while no significant changes were observed for cardiovascular events (myocardial infarction [0.91, 0.74-1.14], angina [0.77, 0.48-1.25]). The counts across individual diagnoses recovered following June 2020, but the number of GP consultations related to cerebrovascular and cardiovascular care remained lower than expected throughout the June to December period (0.93, 0.88-0.98). DISCUSSION: While new diagnoses of acute cardiovascular events remained stable during the COVID-19 pandemic, diagnoses of cerebrovascular events declined substantially compared to prepandemic levels, possibly due to incorrect perception of risk by patients. These findings emphasize the need to improve symptom recognition of cerebrovascular events among the general public and to encourage urgent presentation despite any physical distancing measures.