Longitudinal cohorts for harnessing the electronic health record for disease prediction in a US population

Analysis and prediction of long-term survival using a clinically applicable risk score based on the Electronic Health Record

BACKGROUND

The long-term survival of a population assigned to a hospital can be essential to anticipate, manage, and provide appropriate hospital healthcare resources or lead preventive actions for high-risk mortality individuals. In this study, we discriminate which electronic health record variables are most relevant to predict the long-term survival of a population, and apply the results to identify high-risk mortality groups.

MATERIALS AND METHODS

A prospective cohort study was conducted on a population of 113,403 individuals alive on July 1st, 2018 from the General Hospital of Castellón (Spain). Considering electronic health record patients' variables and survival days from the start date of the study, a Kaplan-Meier analysis and a multivariate Cox regression model were performed, and a risk score based on Cox coefficients was applied to predict survival over 3 years.

RESULTS

All significant covariates from the Cox model (91.5% c-index) were associated with increased mortality risk. Using the proposed risk score, Kaplan-Meier curves show that survival probability in the 3rd year is 99.23% (95% confidence interval (CI) 99.18-99.29) for the low-risk, 91.21% (95% CI 90.67-91.76) for medium-risk, 76.52% (95% CI 75.59-77.46) for the high-risk, and 48.61 % (95% CI 46.85-50.36) for the very high-risk groups.

DISCUSSION

The Cox model obtained is highly predictive, and it has been found that some electronic health record variables little studied to date, such as Clinical Risk Groups, have a strong impact on survival. Regarding clinical application, the proposed risk score is particularly useful for identifying high-risk subpopulations within a large population.

Discordance Between Very Low-Density Lipoprotein Cholesterol and Low-Density Lipoprotein Cholesterol Increases Cardiovascular Disease Risk in a Geographically Defined Cohort

BACKGROUND

Clinical risk scores are used to identify those at high risk of atherosclerotic cardiovascular disease (ASCVD). Despite preventative efforts, residual risk remains for many individuals. Very low-density lipoprotein cholesterol (VLDL-C) and lipid discordance could be contributors to the residual risk of ASCVD.

METHODS AND RESULTS

Cardiovascular disease-free residents, aged ≥40 years, living in Olmsted County, Minnesota, were identified through the Rochester Epidemiology Project. Low-density lipoprotein cholesterol (LDL-C) and VLDL-C were estimated from clinically ordered lipid panels using the Sampson equation. Participants were categorized into concordant and discordant lipid pairings based on clinical cut points. Rates of incident ASCVD, including percutaneous coronary intervention, coronary artery bypass grafting, stroke, or myocardial infarction, were calculated during follow-up. The association of LDL-C and VLDL-C with ASCVD was assessed using Cox proportional hazards regression. Interaction between LDL-C and VLDL-C was assessed. The study population (n=39 098) was primarily White race (94%) and female sex (57%), with a mean age of 54 years. VLDL-C (per 10-mg/dL increase) was significantly associated with an increased risk of incident ASCVD (hazard ratio, 1.07 [95% CI, 1.05-1.09]; P<0.001]) after adjustment for traditional risk factors. The interaction between LDL-C and VLDL-C was not statistically significant (P=0.11). Discordant individuals with high VLDL-C and low LDL-C experienced the highest rate of incident ASCVD events, 16.9 per 1000 person-years, during follow-up.

CONCLUSIONS

VLDL-C and lipid discordance are associated with a greater risk of ASCVD and can be estimated from clinically ordered lipid panels to improve ASCVD risk assessment.

The Importance of Estimating Excess Deaths Regionally During the COVID-19 Pandemic

National or statewide estimates of excess deaths have limited value to understanding the impact of the COVID-19 pandemic regionally. We assessed excess deaths in a 9-county geographically defined population that had low rates of COVID-19 and widescale availability of testing early in the pandemic, well-annotated clinical data, and coverage by 2 medical examiner's offices. We compared mortality rates (MRs) per 100,000 person-years in 2020 and 2021 with those in the 2019 reference period and MR ratios (MRRs). In 2020 and 2021, 177 and 219 deaths, respectively, were attributed to COVID-19 (MR = 52 and 66 per 100,000 person-years, respectively). COVID-19 MRs were highest in males, older persons, those living in rural areas, and those with 7 or more chronic conditions. Compared with 2019, we observed a 10% excess death rate in 2020 (MRR = 1.10 [95% CI, 1.04 to 1.15]), with excess deaths in females, older adults, and those with 7 or more chronic conditions. In contrast, we did not observe excess deaths overall in 2021 compared with 2019 (MRR = 1.04 [95% CI, 0.99 to 1.10]). However, those aged 18 to 39 years (MRR = 1.36 [95% CI, 1.03 to 1.80) and those with 0 or 1 chronic condition (MRR = 1.28 [95% CI, 1.05 to 1.56]) or 7 or more chronic conditions (MRR = 1.09 [95% CI, 1.03 to 1.15]) had increased mortality compared with 2019. This work highlights the value of leveraging regional populations that experienced a similar pandemic wave timeline, mitigation strategies, testing availability, and data quality.

Association Between Fluctuations in Blood Lipid Levels Over Time With Incident Alzheimer Disease and Alzheimer Disease-Related Dementias

BACKGROUND AND OBJECTIVES

Prevention strategies for Alzheimer disease and Alzheimer disease-related dementias (AD/ADRDs) are urgently needed. Lipid variability, or fluctuations in blood lipid levels at different points in time, has not been examined extensively and may contribute to the risk of AD/ADRD. Lipid panels are a part of routine screening in clinical practice and routinely available in electronic health records (EHR). Thus, in a large geographically defined population-based cohort, we investigated the variation of multiple lipid types and their association to the development of AD/ADRD.

METHODS

All residents living in Olmsted County, Minnesota on the index date January 1, 2006, aged 60 years or older without an AD/ADRD diagnosis were identified. Persons with ≥3 lipid measurements including total cholesterol, triglycerides, low-density lipoprotein cholesterol (LDL-C), or high-density lipoprotein cholesterol (HDL-C) in the 5 years before index date were included. Lipid variation was defined as any change in individual's lipid levels over time regardless of direction and was measured using variability independent of the mean (VIM). Associations between lipid variation quintiles and incident AD/ADRD were assessed using Cox proportional hazards regression. Participants were followed through 2018 for incident AD/ADRD.

RESULTS

The final analysis included 11,571 participants (mean age 71 years; 54% female). Median follow-up was 12.9 years with 2,473 incident AD/ADRD cases. After adjustment for confounding variables including sex, race, baseline lipid measurements, education, BMI, and lipid-lowering treatment, participants in the highest quintile of total cholesterol variability had a 19% increased risk of incident AD/ADRD, and those in highest quintile of triglycerides, variability had a 23% increased risk.

DISCUSSION

In a large EHR derived cohort, those in the highest quintile of variability for total cholesterol and triglyceride levels had an increased risk of incident AD/ADRD. Further studies to identify the mechanisms behind this association are needed.

Knowledgebase strategies to aid interpretation of clinical correlation research

OBJECTIVE

Knowledgebases are needed to clarify correlations observed in real-world electronic health record (EHR) data. We posit design principles, present a unifying framework, and report a test of concept.

MATERIALS AND METHODS

We structured a knowledge framework along 3 axes: condition of interest, knowledge source, and taxonomy. In our test of concept, we used hypertension as our condition of interest, literature and VanderbiltDDx knowledgebase as sources, and phecodes as our taxonomy. In a cohort of 832 566 deidentified EHRs, we modeled blood pressure and heart rate by sex and age, classified individuals by hypertensive status, and ran a Phenome-wide Association Study (PheWAS) for hypertension. We compared the correlations from PheWAS to the associations in our knowledgebase.

RESULTS

We produced PhecodeKbHtn: a knowledgebase comprising 167 hypertension-associated diseases, 15 of which were also negatively associated with blood pressure (pos+neg). Our hypertension PheWAS included 1914 phecodes, 129 of which were in the PhecodeKbHtn. Among the PheWAS association results, phecodes that were in PhecodeKbHtn had larger effect sizes compared with those phecodes not in the knowledgebase.

DISCUSSION

Each source contributed unique and additive associations. Models of blood pressure and heart rate by age and sex were consistent with prior cohort studies. All but 4 PheWAS positive and negative correlations for phecodes in PhecodeKbHtn may be explained by knowledgebase associations, hypertensive cardiac complications, or causes of hypertension independently associated with hypotension.

CONCLUSION

It is feasible to assemble a knowledgebase that is compatible with EHR data to aid interpretation of clinical correlation research.

Comprehensive Assessment of 16S rRNA Gene Amplicon Sequencing for Microbiome Profiling across Multiple Habitats

The 16S rRNA gene works as a rapid and effective marker for the identification of microorganisms in complex communities; hence, a huge number of microbiomes have been surveyed by 16S amplicon-based sequencing. The resolution of the 16S rRNA gene is always considered only at the genus level; however, it has not been verified on a wide range of microbes yet. To fully explore the ability and potential of the 16S rRNA gene in microbial profiling, here, we propose Qscore, a comprehensive method to evaluate the performance of amplicons by integrating the amplification rate, multitier taxonomic annotation, sequence type, and length. Our in silico assessment by a "global view" of 35,889 microbe species across multiple reference databases summarizes the optimal sequencing strategy for 16S short reads. On the other hand, since microbes are unevenly distributed according to their habitats, we also provide the recommended configuration for 16 typical ecosystems based on the Qscores of 157,390 microbiomes in the Microbiome Search Engine (MSE). Detailed data simulation further proves that the 16S amplicons produced with Qscore-suggested parameters exhibit high precision in microbiome profiling, which is close to that of shotgun metagenomes under CAMI metrics. Therefore, by reconsidering the precision of 16S-based microbiome profiling, our work not only enables the high-quality reusability of massive sequence legacy that has already been produced but is also significant for guiding microbiome studies in the future. We have implemented the Qscore as an online service at http://qscore.single-cell.cn to parse the recommended sequencing strategy for specific habitats or expected microbial structures. IMPORTANCE 16S rRNA has long been used as a biomarker to identify distinct microbes from complex communities. However, due to the influence of the amplification region, sequencing type, sequence processing, and reference database, the accuracy of 16S rRNA has not been fully verified on a global range. More importantly, the microbial composition of different habitats varies greatly, and it is necessary to adopt different strategies according to the corresponding target microbes to achieve optimal analytical performance. Here, we developed Qscore, which evaluates the comprehensive performance of 16S amplicons from multiple perspectives, thus providing the best sequencing strategies for common ecological environments by using big data.

Health Care Utilization and Death in Patients With Heart Failure During the COVID-19 Pandemic

Objective

To compare the 1-year health care utilization and mortality in persons living with heart failure (HF) before and during the coronavirus disease 2019 (COVID-19) pandemic.

Patients and Methods

Residents of a 9-county area in southeastern Minnesota aged 18 years or older with a HF diagnosis on January 1, 2019; January 1, 2020; and January 1, 2021, were identified and followed up for 1-year for vital status, emergency department (ED) visits, and hospitalizations.

Results

We identified 5631 patients with HF (mean age, 76 years; 53% men) on January 1, 2019, 5996 patients (mean age, 76 years; 52% men) on January 1, 2020, and 6162 patients (mean age, 75 years; 54% men) on January 1, 2021. After adjustment for comorbidities and risk factors, patients with HF in 2020 and patients with HF in 2021 experienced similar risks of mortality compared with those in 2019. After adjustment, patients with HF in 2020 and 2021 were less likely to experience all-cause hospitalizations (2020: rate ratio [RR], 0.88; 95% CI, 0.81-0.95; 2021: RR, 0.90; 95% CI, 0.83-0.97) compared with patients in 2019. Patients with HF in 2020 were also less likely to experience ED visits (RR, 0.85; 95% CI, 0.80-0.92).

Conclusion

In this large population-based study in southeastern Minnesota, we observed an approximately 10% decrease in hospitalizations among patients with HF in 2020 and 2021 and a 15% decrease in ED visits in 2020 compared with those in 2019. Despite the change in health care utilization, we found no difference in the 1-year mortality between patients with HF in 2020 and those in 2021 compared with those in 2019. It is unknown whether any longer-term consequences will be observed.

Variability in Lipid Levels and Risk for Cardiovascular Disease: An Electronic Health Record-Based Population Cohort Study

Background Larger within-patient variability of lipid levels has been associated with increased risk of cardiovascular disease (CVD); however, measures of lipid variability require ≥3 measurements and are not currently used clinically. We investigated the feasibility of calculating lipid variability within a large electronic health record-based population cohort and assessed associations with incident CVD. Methods and Results We identified all individuals ≥40 years of age who resided in Olmsted County, MN, on January 1, 2006 (index date), without prior CVD, defined as myocardial infarction, coronary artery bypass graft surgery, percutaneous coronary intervention, or CVD death. Patients with ≥3 measurements of total cholesterol, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, or triglycerides during the 5 years before the index date were retained. Lipid variability was calculated using variability independent of the mean. Patients were followed through December 31, 2020 for incident CVD. We identified 19 652 individuals (mean age 61 years; 55% female), who were CVD-free and had variability independent of the mean calculated for at least 1 lipid type. After adjustment, those with highest total cholesterol variability had a 20% increased risk of CVD (Q5 versus Q1 hazard ratio, 1.20 [95% CI, 1.06-1.37]). Results were similar for low-density lipoprotein cholesterol and high-density lipoprotein cholesterol. Conclusions In a large electronic health record-based population cohort, high variability in total cholesterol, high-density lipoprotein cholesterol, and low-density lipoprotein cholesterol was associated with an increased risk of CVD, independent of traditional risk factors, suggesting it may be a possible risk marker and target for intervention. Lipid variability can be calculated in the electronic health record environment, but more research is needed to determine its clinical utility.

Predicting Alzheimer's Disease and Related Dementias in Heart Failure and Atrial Fibrillation

BACKGROUND

The Framingham Heart Study Dementia Risk Score (FDRS) was developed in a general population of older persons. It is unknown how the FDRS variables predict Alzheimer's disease and Alzheimer's disease-related dementias (AD/ADRD) in heart failure and atrial fibrillation populations. We aimed to evaluate the predictive ability of the FDRS variables in population-based cohorts of heart failure and atrial fibrillation and to determine whether the addition of other comorbidities and risk factors improves risk prediction for AD/ADRD.

METHODS

Residents aged ≥50 years from 7 southeastern Minnesota counties with a first diagnosis of heart failure or atrial fibrillation between January 1, 2013, and December 31, 2017, were identified. Patients with AD/ADRD before or within 6 months after index atrial fibrillation or heart failure and patients who died within 6 months after index were excluded. For both cohorts, models were constructed to predict AD/ADRD after index including the variables in the FDRS. Additional comorbidities and risk factors were added to the models. For all models, c-statistics using 5-fold cross-validation were calculated.

RESULTS

Among 3052 patients with heart failure (mean age 75 years, 53% male), 626 developed AD/ADRD; among 4107 patients with atrial fibrillation (mean age 74 years, 57% male), 736 developed AD/ADRD. Among patients with heart failure, the FDRS variables predicted AD/ADRD with c-statistic = 0.69. Adding comorbidities and risk factors improved the c-statistic slightly to 0.70. The FDRS variables also performed well (c-statistic = 0.73) in patients with atrial fibrillation; adding comorbidities and risk factors slightly improved performance (c-statistic = 0.75).

CONCLUSIONS

The variables from the FDRS predict AD/ADRD well in both heart failure and atrial fibrillation populations. The addition of comorbidities and risk factors only modestly improved prediction, indicating that the FDRS variables are appropriate to predict AD/ADRD in patients with heart failure and atrial fibrillation.

Rethinking blood eosinophil counts: Epidemiology, associated chronic diseases, and increased risks of cardiovascular disease

Background

The distribution and determinants of blood eosinophil counts in the general population are unclear. Furthermore, whether elevated blood eosinophil counts increase risk for cardiovascular disease (CVD) and other chronic diseases, other than atopic conditions, remains uncertain.

Objective

We sought to describe the distribution of eosinophil counts in the general population and determine the association of eosinophil count with prevalent chronic disease and incident CVD.

Methods

A population-based adult cohort was followed from January 1, 2006, to December 31, 2020. Electronic health record data regarding demographic characteristics, prevalent clinical characteristics, and incident CVD were extracted. Associations between blood eosinophil counts and demographic characteristics, chronic diseases, laboratory values, and risks of incident CVD were assessed using chi-square test, ANOVA, and Cox proportional hazards regression.

Results

Blood eosinophil counts increased with age, body mass index, and reported smoking and tobacco use. The prevalence of chronic obstructive pulmonary disease, hypertension, cardiac arrhythmias, hyperlipidemia, diabetes mellitus, chronic kidney disease, and cancer increased as eosinophil counts increased. Eosinophil counts were significantly associated with coronary heart disease (hazard ratio [HR], 1.44; 95% CI, 1.12-1.84) and heart failure (HR, 1.62; 95% CI, 1.30-2.01) in fully adjusted models and with stroke/transient ischemic attack (HR, 1.37; 95% CI, 1.16-1.61) and CVD death (HR, 1.49; 95% CI, 1.10-2.00) in a model adjusting for age, sex, race, and ethnicity.

Conclusions

Blood eosinophil counts differ by demographic and clinical characteristics as well as by prevalent chronic disease. Moreover, elevated eosinophil counts are associated with risk of CVD. Further prospective investigations are needed to determine the utility of eosinophil counts as a biomarker for CVD risk.

Research on Medical Big Data Analysis and Disease Prediction Method Based on Artificial Intelligence

In recent years, the continuous development of big data, cloud services, Internet+, artificial intelligence, and other technologies has accelerated the improvement of data communication services in the traditional pharmaceutical industry. It plays a leading role in the development of my country’s pharmaceutical industry, deepening the reform of the health system, improving the efficiency and quality of medical services, and developing new technologies. In this context, we make the following research and draw the following conclusions: (1) the scale of my country’s medical big data market is constantly increasing, and the global medical big data market is also increasing. Compared with the global medical big data market, China’s medical big data has grown at a faster rate. From the initial 10.33% in 2015, the proportion has reached 38.7% after 7 years, and the proportion has increased by 28.37%. (2) Generally speaking, urine is mainly slightly acidic, that is, the pH is around 6.0, the normal range is 5.0 to 7.0, and there are also neutral or slightly alkaline. 8 and 7.5 are generally people with some physical problems. In recent years, the pharmaceutical industry has continuously developed technologies such as big data, cloud computing, Internet+, and artificial intelligence by improving data transmission services. As an important strategic resource of the country, the generation of great medical skills and great information is of great significance to the development of my country’s pharmaceutical industry and the deepening of the reform of the national medical system. Improve the efficiency and level of medical services, and establish forms and services. Accelerate economic growth. In this sense, we set out to explore.