Accounting for missing data caused by drug cessation in observational comparative effectiveness research: a simulation study

ObjectivesTo assess the performance of statistical methods used to compare the effectiveness between drugs in an observational setting in the presence of attrition.MethodsIn this simulation study, we compared the estimations of low disease activity (LDA) at 1 year produced by complete case analysis (CC), last observation carried forward (LOCF), LUNDEX, non-responder imputation (NRI), inverse probability weighting (IPW) and multiple imputations of the outcome. All methods were adjusted for confounders. The reasons to stop the treatments were included in the multiple imputation method (confounder-adjusted response rate with attrition correction, CARRAC) and were either included (IPW2) or not (IPW1) in the IPW method. A realistic simulation data set was generated from a real-world data collection. The amount of missing data caused by attrition and its dependence on the ‘true’ value of the data missing were varied to assess the robustness of each method to these changes.ResultsLUNDEX and NRI strongly underestimated the absolute LDA difference between two treatments, and their estimates were highly sensitive to the amount of attrition. IPW1 and CC overestimated the absolute LDA difference between the two treatments and the overestimation increased with increasing attrition or when missingness depended on disease activity at 1 year. IPW2 and CARRAC produced unbiased estimations, but IPW2 had a greater sensitivity to the missing pattern of data and the amount of attrition than CARRAC.ConclusionsOnly multiple imputation and IPW2, which considered both confounding and treatment cessation reasons, produced accurate comparative effectiveness estimates.

TB prevalence in Zimbabwe: a national cross-sectional survey, 2014

BACKGROUND: We conducted the first national TB prevalence survey to provide accurate estimates of bacteriologically confirmed pulmonary TB disease among adults aged ≥15 years in 2014.METHODS: A TB symptoms screen and chest X-ray (CXR) were used to identify presumptive TB cases who submitted two sputum samples for smear microscopy, liquid and solid culture. Bacteriological confirmation included acid-fast bacilli smear positivity confirmed using Xpert® MTB/RIF and/or culture. Prevalence estimates were calculated using random effects logistic regression with multiple imputations and inverse probability weighting.RESULTS: Of 43,478 eligible participants, 33,736 (78%) were screened; of these 5,820 (17%) presumptive cases were identified. There were 107 (1.9%) bacteriologically confirmed TB cases, of which 23 (21%) were smear-positive. The adjusted prevalences of smear-positive and bacteriologically confirmed TB disease were respectively 82/100,000 population (95% CI 47–118/100,000) and 344/100,000 (95% CI 268–420/100,000), with an overall all-ages, all-forms TB prevalence of 275/100,000 population (95% CI 217–334/100,000). TB prevalence was higher in males, and age groups 35–44 and ≥65 years. CXR identified 93/107 (87%) cases vs. 39/107 (36%) using the symptom screen.CONCLUSION: Zimbabwe TB disease prevalence has decreased relative to prior estimates, possibly due to increased antiretroviral therapy coverage and successful national TB control strategies. Continued investments in TB diagnostics for improved case detection are required.

Association between Nafamostat Mesylate and In-Hospital Mortality in Patients with Coronavirus Disease 2019: A Multicenter Observational Study

Nafamostat mesylate may be effective against coronavirus disease 2019 (COVID-19). However, it is not known whether its use is associated with reduced in-hospital mortality in clinical practice. We conducted a retrospective observational study to evaluate the effect of nafamostat mesylate in patients with COVID-19 using the Medical Data Vision Co. Ltd. hospital-based database in Japan. We compared patients with COVID-19 who were (n = 121) and were not (n = 15,738) administered nafamostat mesylate within 2 days of admission between January and December 2020. We conducted a 1:4 propensity score matching with multiple imputations for smoking status and body mass index and combined the 20 imputed propensity score-matched datasets to obtain the adjusted odds ratio for in-hospital mortality. Crude in-hospital mortality was 13.2% (16/121) and 5.0% (790/15,738), respectively. In the propensity score-matched analysis with multiple imputations, the adjusted odds ratio (use vs. no use of nafamostat mesylate) for in-hospital mortality was 1.27 (95% confidence interval: 0.61–2.64; p = 0.52). Sensitivity analyses showed similar results. The results of this retrospective observational study did not support an association between nafamostat mesylate and improved in-hospital outcomes in patients with COVID-19, although further studies with larger sample sizes are warranted to assess the generalizability of our findings.

Change in medical and health care decision-making patterns at the End-of-Life: A cohort of the very old people

How does medical and healthcare decision-making among the very old people change in their last year before death? We explored patterns of decision-making in the Health ABC cohort study in 2011-14 (years 15-17), which involved 12 waves of quarterly phone interviews. When the participant was unable to do the interview, a proxy completed it instead. We identified a sample of 291 decedents (aged 90.0±2.9 at death, 35.7% Black, 52.6% female) with at least 1-year follow-up before death. Percentages of decedents who have made medical or healthcare decisions in the last four quarters before death were 32.0%, 31.2%, 32.6%, 41.9%, respectively. Decedents made more healthcare decisions in the last quarter before death (P<0.01), compared to the baseline. Across all quarters, decision-making is most in need for medications (17.6%), hospital admission (13.2%), and ER/urgent care visit (13.2%). We matched a 1:1 sample of survivors at the time of death by race, sex, and age (within ±3 years). In random effects models with multiple imputations of missing data, we found that decedents were more likely to make healthcare decisions than survivors across all four quarters before death or censor (Odds ratio=1.99, 95%CI: 1.49-2.65, P<0.01). The significance still held after adjusting for age, female, race, education, and interview methods. Overall, compared to matched survivors, the frequency of making medical and healthcare decisions among the very old decedents has already been high in the last year before death. This frequency rose sharply in the last quarter before death.

Computing Bayes factors from data with missing values

The Bayes factor is increasingly used for the evaluation of hypotheses. These may betraditional hypotheses specified using equality constraints among the parameters of thestatistical model of interest or informative hypotheses specified using equality andinequality constraints. So far no attention has been given to the computation of Bayesfactors from data with missing values. A key property of such a Bayes factor should bethat it is only based on the information in the observed values. This paper will show thatsuch a Bayes factor can be obtained using multiple imputations of the missing values.

Comparison of three cohorts of COVID-19 patients with different severity: exposure to antihypertensive, antidiabetic and lipid-lowering agents

Aim: The association between COVID-19 disease severity and certain medicines for the treatment of chronic diseases is currently under discussion. We herein evaluated if previous exposure to antihypertensive, hypoglycaemic, and lipid-lowering drugs increases the risk of poorer COVID-19 outcomes. Methods: We performed a retrospective study on three cohorts of COVID-19 adult patients between March 2020 and May 2020 at the Vall d’Hebron University Hospital. Information relating to the patient lifestyle, comorbidities, and chronic exposures was retrieved from primary healthcare electronic records. Three cohorts were examined, namely patients who had died or required intensive care treatment (ICU/Death [ICU-D] Cohort), patients who required hospitalisation (Hospitalisation [HOSP] Cohort), and patients who only attended the emergency department (Emergencies [EM] Cohort). Descriptive statistics and a multivariate logistic regression model were used to investigate associations with drug exposure, where EM was employed as the reference cohort. Results: We included 1,778 patients: 417 (23.5%) from the ICU-D Cohort, 1,052 (59.2%) from the HOSP Cohort, and 309 (17.4%) from the EM Cohort. After multiple imputations and data adjustment by potential confounders, no statistically significant association was observed between the COVID-19 severity and the use of antihypertensives, hypoglycaemic agents, or lipid-lowering agents, with the exception of calcium channel blockers (CCB) (ICU-D Cohort: OR 2.23; CI 95% [1.03–4.83]; P = 0.042). Conclusions: Most results on lifestyle characteristics and comorbidities related to COVID-19 severity were in agreement with current knowledge, although some associated factors are nowadays a matter of controversy and further investigation is required.

Revisiting the Pancreatic Fistula Risk Score: Clinical Nomogram Accurately Assesses Risk

Objectives Surgeons have created numerous iterations of the pancreatic fistula risk score (FRS) to predict risk for clinically relevant postoperative pancreatic fistula (CR-POPF). The multitude of often conflicting models makes it difficult for surgeons to apply data in clinical practice. Methods We conducted a retrospective cohort study utilizing National Surgical Quality Improvement Program data from 2015 to 2018. The study included patients undergoing pancreaticoduodenectomy. Missing data were resolved with multiple imputations. Results The study included 5975 patients; 1018 (17%) had a CR-POPF. On multivariate analysis, male sex (odds ratio (OR) 1.60 CI: 1.29-1.98 P < .001), obesity (OR 1.65 CI: 1.31-2.08 P < .001), and soft gland texture (OR 3.21 CI: 2.45-4.23 P < .001) were all associated with increased odds of a CR-POPF. Variables not associated with CR-POPF included diabetes, preoperative bilirubin, preoperative albumin, and American Society of Anesthesiologists (ASA) classification. On multivariate analysis, duct diameter >6 mm (OR .52 CI: .34-.77 P = .001), pancreatic adenocarcinoma pathology (OR .67 CI: .53-.84 P < .001), and neoadjuvant treatment (OR .71 CI: .51-.98 P = .042) were all associated with decreased odds of a CR-POPF. We constructed a clinically relevant nomogram from this model known as the Portland FRS. Model characteristics were superior to previously published FRS models. The area under the curve (AUC) for the Portland FRS was .72 (CI: .704-.737). In comparison, AUCs for the Alternative and Seoul FRS were .70 and .64, respectively. Conclusion Utilizing readily available clinical data, the Portland FRS can accurately predict the risk for pancreatic fistula. The nomogram may assist surgeons in patient counseling and perioperative management.

Comparison of Imputation Methods on Retrospective Breast Cancer Data in Tanzania: A Case Study of Muhimbili and Ocean Road Hospitals

Background Clinical data are at risk of having missing or incomplete values for several reasons including patients’ failure to attend clinical measurements, wrong interpretations of measurements, and measurement recorder’s defects. Missing data can significantly affect the analysis and results might be doubtful due to bias caused by omission of missed observation during statistical analysis especially if a dataset is considerably small. The objective of this study is to compare several imputation methods in terms of efficiency in filling-in the missing data so as to increase the prediction and classification accuracy in breast cancer dataset. Methods Five imputation methods namely series mean, k-nearest neighbour, hot deck, predictive mean matching, and multiple imputations were applied to replace the missing values to the real breast cancer dataset. The efficiency of imputation methods was compared by using the Root Mean Square Errors and Mean Absolute Errors to obtain a suitable complete dataset. Binary logistic regression and linear discrimination classifiers were applied to the imputed dataset to compare their efficacy on classification and discrimination. Results The evaluation of imputation methods revealed that the predictive mean matching method was better off compared to other imputation methods. In addition, the binary logistic regression and linear discriminant analyses yield almost similar values on overall classification rates, sensitivity and specificity. Conclusion The predictive mean matching imputation showed higher accuracy in estimating and replacing missing/incomplete data values in a real breast cancer dataset under the study. It is a more effective and good method to handle missing data in this scenario. We recommend to replace missing data by using predictive mean matching since it is a plausible approach toward multiple imputations for numerical variables, as it improves estimation and prediction accuracy over the use complete-case analysis especially when percentage of missing data is not very small.

Association between self-reported binge drinking and absenteeism in the Baltic countries

Objective: Substantial loss of productivity due to absenteeism is associated with alcohol use. This study examined the associations between absenteeism in the workplace and in schools and binge drinking across various beverage types in the Baltic countries. Methods: We utilised a dataset of 3,778 individuals compiled from 2015 to 2016 and performed multiple negative binomial regression analysis with multiple imputations to deal with missing data. Self-reported measures were used for both absenteeism and binge drinking. Results: We found evidence to support the claim that absenteeism, in terms of self-reported absence days, is positively associated with self-reported binge drinking, specifically with beer bingeing. On average, beer bingers reported 49% ( p < .05) more absences than people who drink alcohol but do not binge on beer. For wine and spirits variables, the estimates indicated positive but statistically insignificant associations. No group differences were identified across gender and education. Conclusions: A considerable proportion of days absent from work and from school can be associated with beer bingeing. Therefore, it should be acknowledged that beverage-specific alcohol policies that are more lenient toward beer than other types of alcohol can inadvertently increase absenteeism and decrease workplace productivity.