Clinical Epidemiology最新文献

Purpose: To estimate the positive predictive value (PPV) of case ascertainment algorithm for hypocalcemia leading to hospitalization or emergency visit in the Swedish National Patient Register among women with postmenopausal osteoporosis (PMO) treated with antiresorptive agents. This was a regulator-requested validation study to support a multidatabase postauthorisation safety study (PASS) of antiresorptive treatment.

Methods: The Swedish part of the PASS was based on data from Swedish population registries. Potential cases of hypocalcemia, identified among women with PMO, included in the PASS in 2010-2016, were defined based on non-specific International Classification of Diseases, 10th Revision (ICD-10) codes for disorders of calcium metabolism at hospitalization or emergency visit, as recorded in the Swedish Patient Register through 2018. Presence of hypocalcemia among the potential cases was confirmed using a standardized abstraction of medical charts. PPV was estimated as a measure of validity.

Results: There were 164 potential cases of hypocalcemia, of which 121 had medical charts with sufficient information available. Among these 121 cases, 19 had confirmed hypocalcemia, PPV 15.7% (95% confidence interval: 10.0 to 23.0).

Conclusion: The case-defining algorithm based on the non-specific ICD-10 codes had a low PPV. Reliance on the algorithm may bias results of epidemiologic studies relying it. Limitations include non-response and low precision of some PPV estimates.

Background: The Danish Prehospital Medical Record (DPMR) represents a pioneering nationwide electronic prehospital medical record system. While routinely collected data from the DPMR are increasingly used for research, a comprehensive description of its system and content is needed.

Objective: To provide an overview of the DPMR as a tool for research, including its structure, variables, and current volume of records.

Methods: We examined the DPMR's history, data structure, content, and data usage. We also analyzed aggregated DPMR data from 2016 to 2023 for selected key variables. Further, we searched MEDLINE to identify studies utilizing this data source in the past decade.

Results: Since 2016, the DPMR system has grown to include 1.8 million unique prehospital patients with over 6 million associated patient contacts. For each patient contact, the DPMR compiles information on the emergency medical call (dispatch criteria, level of urgency, and pre-arrival treatment), characteristics of the incident (patient examination, treatment, response time, on-scene time, and transport time), emergency medical services units (ambulances, rapid response vehicles with paramedics, anesthesiologists in ground-based mobile emergency care units and/or helicopters, or patient transports without treatment capability), and extensive patient-related data. The system currently encompasses 528 variables, standardized across all emergency medical services units. There are a limited number of studies on the data quality of the system and the proportion of patients with missing civil registration numbers has varied between approximately 5% and 9%, which should be taken into account when using it for research.

Conclusion: The DPMR is growing in importance as a research tool in Denmark. It provides nationwide patient-related and logistical prehospital data going back to 2016, enabling linkage with national registries for outcome research.

Background: Gilbert's syndrome (GS) is a common genetic disorder marked by elevated bilirubin levels due to UGT1A1 enzyme deficiency. While jaundice and some adverse drug reactions are the primary recognised clinical features, individuals with GS frequently report non-specific symptoms like fatigue, brain fog, and abdominal pain. This study investigates the symptoms and diagnostic triggers of GS using UK primary care electronic health records.

Methods: We analysed data from the IQVIA Medical Research Database, covering over 11 million active UK patients. Individuals with a recorded GS diagnosis were identified and their sociodemographic profiles described. Using a nested case-control design, we applied machine learning-based feature selection to pinpoint key clinical features recorded up to five years before diagnosis. These features were then examined longitudinally by sex to distinguish persistent symptoms from short-term diagnostic triggers.

Results: The estimated UK prevalence of GS was 180.4 per 100,000 (95% CI: 174.4-186.6), with diagnoses more common in men, peaking around age 35, and more frequent in areas of least social deprivation. Among 9,240 GS cases and 150,846 controls, machine learning identified key diagnostic themes including jaundice, abnormal liver function tests, abdominal pain, fatigue, bowel changes, and sleep disturbances. While most of these features appeared primarily in the year prior to diagnosis, only abdominal pain and fatigue were consistently more common in GS cases up to five years before diagnosis.

Conclusion: Our findings highlight both expected and novel GS diagnostic triggers. While many features likely reflect known symptomology or incidental detection via routine testing, the persistent presence of fatigue and abdominal pain suggests they may be under-recognised symptoms of GS. These findings warrant further investigation, and the data-driven approach used here may help uncover early signs of other underdiagnosed genetic conditions.

Background: Chronic obstructive pulmonary disease (COPD) incurs significant healthcare costs, often accompanied by multimorbidity. Advanced patients may need short-term stays for rehabilitation, treatment, or respite to maintain home living.

Aim: To identify predictors for a first short-term stay and study the healthcare utilization and costs compared with similar patients without a short-term stay.

Patients and methods: Data on COPD patients in the cities Oslo and Trondheim 2010-2019 and including information on specialist, primary and long-term care, diagnoses, sociodemographics and -economics were collected from national and municipal registries, resulting in a sample of 24,613 patients. Using discrete time survival models, we identified predictors for a short-term stay. We described the costs before and after admission, and the duration of living at home, compared to non-recipients matched on age, comorbidities and healthcare use.

Results: Depression, anxiety, mental disorders, alcoholism, prior hospitalization and reception of home care were associated with higher odds of short-term stays. One to two GP visits for respiratory diseases, being in the top quartile for GP visits for non-respiratory diseases, visits to specialists, and physiotherapist visits for non-respiratory issues were significantly associated with lower odds of short-term institutional stay. Patients admitted to short-term stays incurred markedly higher costs both in the year before admission and during subsequent years compared to matched non-recipients, primarily due to increased use of inpatient and home care services.

Conclusion: Prior receipt of home care, unlike standard outpatient services, was linked to a higher likelihood of short-term stays. This suggests that some outpatient services may delay the need for such stays, or that patients already in municipal services are more readily admitted. Additionally, patients with psychosocial issues may have greater care needs, indicating that resource allocation aligns with these needs. The findings suggest that by the time short-term stays are required, health deterioration has already become considerable.

Background: In the presence of competing risks, when the baseline risk is unclear, if only the sub-distribution hazard ratio (SHR) is reported in the results, which is related to the cumulative incidence function, the survival disparity of events of interest between groups cannot be clarified. In contrast, the difference in restricted mean time lost (RMTLd), which is the difference in the areas under the cumulative incidence between two groups, can well compensate for the deficiencies of SHR and explain the effects on a time scale, facilitating clinical interpretation and communication.

Methods: The Surveillance, Epidemiology, and End Results (SEER) database was used to collect information on female patients with locally advanced breast cancer diagnosed between 2010 and 2015. The prognostic factors of breast cancer death were evaluated considering competing risk. Univariable and multivariable analyses were conducted to get SHR and RMTLd.

Results: SHR can indicate the direction of prognostic factors, while RMTLd can quantify prognostic effects and provide time-scale interpretation. For instance, in adjuvant radiotherapy, the SHR showed a protective effect, which can be quantified as an average increase of 4.15 months in survival time.

Discussion: In the presence of competing risks, the combined use of absolute measure RMTLd can more intuitively explain the prognostic effect, which is convenient for clinical practice and communication.

Introduction: The impact of type 2 diabetes (T2DM) on mortality following lung cancer diagnosis remains unclear, with conflicting evidence across studies. We aimed to assess differences in all-cause and cause-specific mortality between people with lung cancer with and without T2DM within a primary care population in England.

Methods: The study population was 69,674 people with incident lung cancer within the Clinical Practice Research Datalink (CPRD) Aurum primary care database (2010-2022). The study exposure was T2DM at cancer diagnosis, and the outcomes were all-cause and cause-specific mortality (cancer, cardio-vascular, respiratory). Cox models were fitted for each outcome adjusting for age, gender, smoking status, body mass index, calendar year and socioeconomic status (Index of Multiple Deprivation).

Results: After adjusting for age and gender, there was no evidence for a difference in all-cause mortality in people with T2DM compared with people without T2DM (IRR 0.98 95% CI 0.96, 1.01). After fully-adjusting for measured confounders, there was a small positive effect (IRR 1.07 95% CI 1.04, 1.09). After adjusting for age and gender, people with T2DM had lower rates of cancer-specific mortality compared to people without T2DM (IRR 0.96 95% CI 0.94, 0.98). However, after adjustment for all measured confounders there was a small positive association (IRR 1.05 95% CI 1.02, 1.07). In both age and gender adjusted and fully adjusted models people with T2DM had higher cardiovascular (fully adjusted HR 1.30 95% CI 1.15, 1.47) and respiratory disease mortality (fully adjusted HR 1.30 95% CI 1.15, 1.47).

Conclusion: There was robust evidence that people with T2DM had higher cardiovascular and respiratory disease mortality following lung cancer diagnosis. The relationships between T2DM and all-cause and cancer-specific mortality were highly sensitive to adjustment for confounding. Differences in studies on approaches to confounding and levels of missing data may contribute to the mixed findings on this association in the literature.

Purpose: COVID-19 has been extensively researched; however, the lack of standardized COVID-19 severity categorization in register-based research complicates comparison of studies. The WHO COVID-19 Clinical Progression Scale is a standardized disease severity tool for clinical data, though not adapted to data available in health registries. We aimed to develop and validate such a novel categorization with international applicability.

Methods: The WHO Clinical Progression Scale was translated to a severity index utilizing ICD- and procedure-codes from outpatient, inpatient, intensive care, and mortality registries using the adult Swedish population and SARS-CoV-2 positive-test data (January 2020 - July 2022). Cox proportional hazards were applied to determine whether increasing severity correlates with mortality in COVID-19 patients compared to the population.

Results: The WHO-Scale was translated to ten categories reflecting the increasing need for advanced care, encompassing 8,245,474 individuals including 1,981,946 SARS-CoV-2 infections. Fatal COVID-19 cases were older with more comorbidities. Those receiving mechanical ventilation and ECMO were younger with fewer comorbidities. Among survivors beyond 30 days, 90-day all-cause mortality increased with severity using category zero (no laboratory-verified SARS-CoV-2) as reference. Mortality was lowest for patients without health care adjusted for age, sex, comorbidities and socio-economic variables (adjusted hazard ratio (aHR) 1.18, 95% confidence interval (CI) 1.13-1.22). Those hospitalized >5 days had higher mortality (aHR 5.83, 5.5-6.17). Those requiring ECMO/ECLS had the highest mortality (aHR 593.54, 317.77-1108.65).

Conclusion: The novel COVID-19 severity index associated with all-cause 90-day mortality and aligned with previous literature. This index will enable comparative studies of COVID-19, which is important for public health policies and development of clinical guidelines. This is an innovative epidemiologic tool with potential applicability in all countries with centralised health registers. The index also has the potential to be used for other infectious diseases and in real-time data for modelling predictions.

Clinical algorithms are widely used tools for predicting, diagnosing, and managing diseases. However, race correction in these algorithms has faced increasing scrutiny for potentially perpetuating health disparities and reinforcing harmful stereotypes. This narrative review synthesizes historical, clinical, and methodological literature to examine the origins and consequences of race correction in clinical algorithms. We focus primarily on developments in the United States and the United Kingdom, where many race-based algorithms originated. Drawing on interdisciplinary sources, we discuss the persistence of race-based adjustments, the implications of their removal, and emerging strategies for bias mitigation and fairness in algorithm development. The practice began in the mid-19th century with the spirometer, which measured lung capacity and was used to reinforce racial hierarchies by characterizing lower lung capacity for Black people. Despite critiques that these differences reflect environmental exposure rather than inherited traits, the belief in race-based biological differences in lung capacity and other physiological functions, including cardiac, renal, and obstetric processes, persists in contemporary clinical algorithms. Concerns about race correction compounding health inequities have led many medical organizations to re-evaluate their algorithms, with some removing race entirely. Transitioning to race-neutral equations in areas like pulmonary function testing and obstetrics has shown promise in enhancing fairness without compromising accuracy. However, the impact of these changes varies across clinical contexts, highlighting the need for careful bias identification and mitigation. Future efforts should focus on incorporating diverse data sources, capturing true social and biological health determinants, implementing bias detection and fairness strategies, ensuring transparent reporting, and engaging with diverse communities. Educating students and trainees on race as a sociopolitical construct is also important for raising awareness and achieving health equity. Moving forward, regular monitoring, evaluation, and refinement of approaches in real-world settings are needed for clinical algorithms serve all patients equitably and effectively.

Purpose: No studies have validated psychiatric diseases diagnoses in Taiwan's National Health Insurance Research Database (NHIRD). We aimed to assess the interrater reliability of chart-review among psychiatrists, examine the validity of the diagnostic codes for psychotic disorders and affective diseases in the NHIRD against review-based diagnoses, and examine whether the change in the coding system from the ICD-9-CM to the ICD-10-CM affected the validity of the diagnostic codes.

Patients and methods: The study participants were psychiatric inpatients aged 18 to 65 years who were admitted in 2015 and 2017, respectively, to the main and three branch hospitals of National Taiwan University Hospital. A chart review was conducted among 48 purposively selected inpatients with discharge diagnoses in five core categories to assess interrater reliability. This chart-review procedure was then used to generate diagnostic codes for a stratified sampling of 727 inpatients with discharge diagnoses in 12 diagnostic categories of psychotic disorders and affective disorders to examine the validity of the diagnostic codes.

Results: The intraclass correlation coefficient reliability of schizophrenia and three broad categories of diagnoses indicated good interrater reliability. The positive predictive value and sensitivity of common diagnoses in the narrow category (eg, schizophrenia) or the broad category (eg, psychotic disorders, bipolar disorders, and major depressive disorders) were high-performing (≥ 0.70), whereas those of the diagnoses of low prevalence were modest. The validity indices of claims-based diagnoses using the ICD-10-CM tended to be better than those using the ICD-9-CM.

Conclusion: This first-ever study validating psychiatric diagnoses in Taiwan's NHIRD using a structured chart review suggests that the diagnostic codes of narrow categories of schizophrenia or other broad categories are recommended for high-performing validity indices. Intensive training for the coding plus the specific details requested by the ICD-10 may increase the validity of the claims-based databases for psychotic and affective disorders.

Background: Twin pregnancies, accounting for a rising proportion of births globally, present significant public health challenges in China. Birthweight discordance (BWD), a critical complication, remains understudied in its epidemiological context, particularly regarding its population-level associations with adverse neonatal outcomes.

Methods: This multi-center, retrospective cohort study leveraged data from 21 hospitals across 18 Chinese cities (2018-2020) to assess BWD and its epidemiological implications. Ordinal logistic regression with random effects was used to explore their association. BWD was defined as: [(larger birthweight - smaller birthweight) / larger birthweight] × 100% and categorized into four grades: I (≤15%), II (>15% to 20%), III (>20% to 25%), and IV (>25%).

Results: Among 6437 twin pairs, 73.6% were classified as Grade I (no BWD), while 10.7%, 7.1%, and 8.6% constituted Grades II, III, and IV discordance, respectively. Dose-response relationships emerged: each incremental BWD elevated risks of small vulnerable newborns (aOR = 1.83, 95% CI 1.76-1.90), small for gestational age (aOR = 1.23, 95% CI 1.18-1.29), low birthweight (LBW, aOR = 1.16, 95% CI 1.13-1.20), very LBW (aOR = 1.63, 95% CI 1.53-1.73) and extreme LBW (aOR = 1.82, 95% CI 1.61-2.05). Smaller twins exhibited disproportionately higher adverse outcome rates than larger twins. Sensitivity analyses confirmed robustness across specific subgroups.

Conclusion: BWD exceeding 20% affects 15.7% of live-born twins in China, mirroring rates in high-income settings. BWD demonstrates strong dose-response relationships with adverse outcomes, validating its utility for twin health stratification. These findings call for integrating BWD assessment into prenatal surveillance and risk-adapted care to reduce neonatal morbidity/mortality, urging clinicians and policymakers to prioritize perinatal outcome equity.