Predictors of shorter- and longer-term mortality after COVID-19 presentation among dialysis patients: parallel use of machine learning models in Latin and North American countries.

BACKGROUND: We developed machine learning models to understand the predictors of shorter-, intermediate-, and longer-term mortality among hemodialysis (HD) patients affected by COVID-19 in four countries in the Americas. METHODS: We used data from adult HD patients treated at regional institutions of a global provider in Latin America (LatAm) and North America who contracted COVID-19 in 2020 before SARS-CoV-2 vaccines were available. Using 93 commonly captured variables, we developed machine learning models that predicted the likelihood of death overall, as well as during 0-14, 15-30, > 30 days after COVID-19 presentation and identified the importance of predictors. XGBoost models were built in parallel using the same programming with a 60%:20%:20% random split for training, validation, & testing data for the datasets from LatAm (Argentina, Columbia, Ecuador) and North America (United States) countries. RESULTS: Among HD patients with COVID-19, 28.8% (1,001/3,473) died in LatAm and 20.5% (4,426/21,624) died in North America. Mortality occurred earlier in LatAm versus North America; 15.0% and 7.3% of patients died within 0-14 days, 7.9% and 4.6% of patients died within 15-30 days, and 5.9% and 8.6% of patients died > 30 days after COVID-19 presentation, respectively. Area under curve ranged from 0.73 to 0.83 across prediction models in both regions. Top predictors of death after COVID-19 consistently included older age, longer vintage, markers of poor nutrition and more inflammation in both regions at all timepoints. Unique patient attributes (higher BMI, male sex) were top predictors of mortality during 0-14 and 15-30 days after COVID-19, yet not mortality > 30 days after presentation. CONCLUSIONS: Findings showed distinct profiles of mortality in COVID-19 in LatAm and North America throughout 2020. Mortality rate was higher within 0-14 and 15-30 days after COVID-19 in LatAm, while mortality rate was higher in North America > 30 days after presentation. Nonetheless, a remarkable proportion of HD patients died > 30 days after COVID-19 presentation in both regions. We were able to develop a series of suitable prognostic prediction models and establish the top predictors of death in COVID-19 during shorter-, intermediate-, and longer-term follow up periods.

Modifiable and Non-Modifiable Factors Associated With Increased COVID-19 Related Mortality Among Dialysis Patients

BACKGROUND AND AIMS Patients with end-stage kidney disease (ESKD) face higher risk for severe outcomes from COVID-19 infection. Moreover, it is not well known to which extent potentially modifiable risk factors contribute to mortality risk. In this study, we investigated the incidence and risk factors for 30-day case-fatality of COVID-19 in haemodialysis patients treated in the European Fresenius Medical Care (FMC) Nephrocare network. METHOD In this historical cohort study, we included unvaccinated adult dialysis patients with a first documented SARS-CoV-2 infection between 1 February 2020 and 31 March 2021 (study period) registered in the European Clinical Database (EuCliD®). The first SARS-CoV-2 suspicion date for all documented infections was considered the index date for the analysis. Patients were followed for up to 30 days. Follow-up time was defined from the index date until the date of death, end of follow-up period or lost to follow-up, whichever occurred first. We ascertained patients’ characteristics in the 6-month period prior to index date. We used logistic regression and XGBoost to assess risk factors for 30-day mortality. RESULTS We included 9211 patients meeting the inclusion criteria for the study (Table 1). Age was 65.4 ± 13.7 years, dialysis vintage was 4.2 ± 3.7 years. In the follow up period, 1912 patients died within 30 days (20.8%, 95% confidence interval: 19.9%–21.6%). Correlates of COVID-19 related mortality are summarized in Table 2. Several potentially modifiable factors were associated with increased risk of death: patients on HD compared with online haemodiafiltration had shorter survival after presentation with COVID-19 as well as those who did not achieve the therapeutic targets for serum albumin, erythropoietin resistance index, protein catabolic rate, haemodynamic status, C-reactive protein, single-pool Kt/V, hydration status and serum sodium in the months before infection. The discrimination accuracy of prediction models developed with XGBoost was similar to that observed for main-effect logistic regression (AUC 0.69 and 0.71, respectively) suggesting that no major cross-interaction and non-linear effect could improve prediction accuracy. CONCLUSION We observed high 30-day COVID-19 related mortality among unvaccinated dialysis patients. Older patients, men and those with greater comorbidities had higher risk of death after COVID-19 infection. Derangement in potentially modifiable factors in the 6 months prior to COVID-19 infection was independently associated with increased mortality. Whether achievement of clinical therapeutic targets is associated with improved survival after COVID-19 infection is a matter of further research.

Effectiveness of Covid-19 Vaccines in a Large European Haemodialysis Cohort

BACKGROUND AND AIMS To date, no large-scale study has evaluated the effectiveness of COVID-19 vaccines in hemodialysis patients. We sought to evaluate the effectiveness of vaccines against SARS-CoV-2 infections and death in haemodialysis patients registered in the Fresenius Medical Care (FMC) Nephrocare network. METHOD In this historical, 1:1 matched cohort study, we analysed electronic health records (EHR) of individuals receiving in-center haemodialysis therapy in FMC European dialysis clinics from 1 December 2020, to 31 May 2021 (study period). For each vaccinated patient, an unvaccinated patient was selected among patients registered in the same country and attending a dialysis session within +/–3 days from the vaccination date. Matching without replacement was based on demographics, clinical characteristics, past COVID-19 infections and a risk score representing the local (dialysis centre) background risk of infection at each vaccination date. The infection risk score was calculated from an artificial Intelligence model predicting the risk of COVID-19 outbreak in each clinic over a 2-week prediction horizon. The infection risk score was based on trends in regional COVID-19 epidemic metrics, FMC COVID-19 reporting system and clinical practice patterns. The index date was the date of the first vaccination for the vaccinated and the matching treatment date for the unvaccinated controls. To overcome violation of the proportional hazard assumption, we estimated the effectiveness of the COVID-19 vaccines in preventing infection and mortality rates as 1—hazard ratio estimated from a time-dependent extended Cox regression stratified by country and vaccine type. RESULTS We included 44 458 patients, 22 229 vaccinated and matched 22 229 unvaccinated. Distribution of covariates was balanced across study arms after matching (Figure 1A). In the effectiveness analysis on mRNA vaccines, we observed 850 SARS-CoV-2 infections and 201 COVID19-related deaths among the 28 110 patients (14 055 vaccinated and 14 055 unvaccinated) during a mean follow up time of 44 ± 40 days. In the effectiveness analysis of viral-vector vaccines, we observed 297 SARS-CoV-2 infections and 64 COVID19-related deaths among 12 888 patients (6444 vaccinated and 6444 unvaccinated) during a mean a follow-up time of 48 ± 32 days (Figure 1B). We observed 18.5/100 patient-year and 8.5/100 patient-year fewer infections and 5.4/100 patient-year and 5.2/100 patient-year fewer COVID-19-related deaths among patients vaccinated with mRNA and viral-vector vaccines respectively, as compared to matched unvaccinated controls. The effectiveness of COVID-19 vaccines concerning both symptomatic infections and COVID-related death along the follow up period is shown in Figure 2. CONCLUSION In this matched, historical cohort study, we observed a strong reduction in both SARS-CoV-2 symptomatic infection and COVID-19-related death among dialysis patients receiving an mRNA vaccine. Despite seemingly less protective against symptomatic infections, we observed similar reduction in COVID-19 mortality rate among patients receiving a viral-carrier vaccine.FIGURE 1A: Forest Plot demonstrating covariate distribution balance between exposure groups. Effect Sizes calculated as Cohen's d or Cromer's Negative coefficient indicates that mean or relative frequency was greater among vaccinated patients. Effect Size 0.12 negligible Effect Size-0.1-0.2: small.FIGURE 1B: Absolute frequency and incidence density (95% confidence intervall of events across exposure groups.FIGURE 2: Effectiveness (1-HR) estimates by vaccine type concerning symptomatic, documented infection and COVID-19 related death. Estimates were obtained from extended, cox regression with time-varying covariate.

Machine Learning for Prediction of Patients on Hemodialysis with an Undetected SARS-CoV-2 Infection.

Background: We developed a machine learning (ML) model that predicts the risk of a patient on hemodialysis (HD) having an undetected SARS-CoV-2 infection that is identified after the following ≥3 days. Methods: As part of a healthcare operations effort, we used patient data from a national network of dialysis clinics (February-September 2020) to develop an ML model (XGBoost) that uses 81 variables to predict the likelihood of an adult patient on HD having an undetected SARS-CoV-2 infection that is identified in the subsequent ≥3 days. We used a 60%:20%:20% randomized split of COVID-19-positive samples for the training, validation, and testing datasets. Results: We used a select cohort of 40,490 patients on HD to build the ML model (11,166 patients who were COVID-19 positive and 29,324 patients who were unaffected controls). The prevalence of COVID-19 in the cohort (28% COVID-19 positive) was by design higher than the HD population. The prevalence of COVID-19 was set to 10% in the testing dataset to estimate the prevalence observed in the national HD population. The threshold for classifying observations as positive or negative was set at 0.80 to minimize false positives. Precision for the model was 0.52, the recall was 0.07, and the lift was 5.3 in the testing dataset. Area under the receiver operating characteristic curve (AUROC) and area under the precision-recall curve (AUPRC) for the model was 0.68 and 0.24 in the testing dataset, respectively. Top predictors of a patient on HD having a SARS-CoV-2 infection were the change in interdialytic weight gain from the previous month, mean pre-HD body temperature in the prior week, and the change in post-HD heart rate from the previous month. Conclusions: The developed ML model appears suitable for predicting patients on HD at risk of having COVID-19 at least 3 days before there would be a clinical suspicion of the disease.