Learning debiased graph representations from the OMOP common data model for synthetic data generation.

BACKGROUND: Generating synthetic patient data is crucial for medical research, but common approaches build up on black-box models which do not allow for expert verification or intervention. We propose a highly available method which enables synthetic data generation from real patient records in a privacy preserving and compliant fashion, is interpretable and allows for expert intervention. METHODS: Our approach ties together two established tools in medical informatics, namely OMOP as a data standard for electronic health records and Synthea as a data synthetization method. For this study, data pipelines were built which extract data from OMOP, convert them into time series format, learn temporal rules by 2 statistical algorithms (Markov chain, TARM) and 3 algorithms of causal discovery (DYNOTEARS, J-PCMCI+, LiNGAM) and map the outputs into Synthea graphs. The graphs are evaluated quantitatively by their individual and relative complexity and qualitatively by medical experts. RESULTS: The algorithms were found to learn qualitatively and quantitatively different graph representations. Whereas the Markov chain results in extremely large graphs, TARM, DYNOTEARS, and J-PCMCI+ were found to reduce the data dimension during learning. The MultiGroupDirect LiNGAM algorithm was found to not be applicable to the problem statement at hand. CONCLUSION: Only TARM and DYNOTEARS are practical algorithms for real-world data in this use case. As causal discovery is a method to debias purely statistical relationships, the gradient-based causal discovery algorithm DYNOTEARS was found to be most suitable.

The Relationship Between Hospital Procedure Volume and Outcomes After Endovascular or Open Surgical Revascularisation for Peripheral Arterial Disease: An Analysis of Health Insurance Claims Data.

OBJECTIVE: There is a paucity of data on the relationship between hospital procedure volume and outcomes after inpatient treatment of symptomatic peripheral arterial disease (PAD). This study aimed to generate meaningful hypotheses to support the ongoing discussion. METHODS: Data derived from BARMER, Germany's second largest insurance provider, were linked with nationwide hospital procedure volumes from mandatory hospital quality reports. All endovascular (EVR) and open surgical revascularisations (OSR) provided to patients (≥ 40 years) with symptomatic PAD between 1 January 2013 and 31 December 2018 were included. Hospital volume was defined as the number of procedures performed by a hospital in the previous calendar year (in quartiles). Freedom from re-intervention, amputation, and overall mortality rate within 12 months after discharge were analysed using multivariable Cox proportional hazards models. In hospital mortality was determined by generalised estimating equations logistic regression models. RESULTS: There were 88 187 revascularisations (72.4% EVR; EVR: 72.7 years and 45.2% females; OSR: 71.9 years and 41.9% females) registered by 668 hospitals. No statistically significant association was found between 12 month freedom from re-intervention and hospital volume (EVR: 4; quartile HR 1.05; 95% CI 0.94 - 1.16. OSR: 4; quartile HR 1.05; 95% CI 0.92 - 1.21). Patients with OSR had a decreased hazard of 12 month mortality in a high volume hospital compared with a low volume hospital (HR 0.85; 95% CI 0.73 - 0.98), but not with EVR (HR 1.03; 95% CI 0.91 - 1.16). Patients who were treated in hospitals with highest volumes showed decreased hazards of 12 month freedom from amputation when compared with low volume hospitals (EVR: HR 0.72; 95% CI 0.52 - 0.99. OSR: HR 0.61; 95% CI 0.44 - 0.85). CONCLUSION: This large retrospective analysis of insurance claims suggests that higher procedure volume is associated with lower major amputation rates, although there is a need for standardisation of the definition of volume stratification. Future studies should address the impact of subsequent outpatient care and surveillance to further examine the complex interaction between treatment and outcomes.

A health insurance claims analysis on the effect of female sex on long-term outcomes after peripheral endovascular interventions for symptomatic peripheral arterial occlusive disease.

OBJECTIVE: Several reports have addressed sex disparities in peripheral arterial occlusive disease (PAOD) treatment with inconclusive or even conflicting results. However, most previous studies have neither been sufficiently stratified nor used matching or weighting methods to address severe confounding. In the present study, we aimed to determine the disparities between sexes after percutaneous endovascular revascularization (ER) for symptomatic PAOD. METHODS: Health insurance claims data from the second-largest insurance fund in Germany, BARMER, were used. A large cohort of patients who had undergone index percutaneous ER of symptomatic PAOD from January 1, 2010 to December 31, 2018 were included in the present study. The study cohort was stratified by the presence of intermittent claudication, ischemic rest pain, and wound healing disorders. Propensity score matching was used to adjust for confounding through differences in age, treated vessel region, comorbidities, and pharmacologic treatment. Sex-related differences regarding cardiovascular event-free survival, amputation-free survival, and overall survival within 5 years of surgery were determined using Kaplan-Meier time-to-event curves, log-rank test, and Cox regression analysis. RESULTS: In the present study, 50,051 patients (47.2% women) were identified and used to compose a matched cohort of 35,232 patients. Among all strata, female patients exhibited lower mortality (hazard ratio [HR], 0.69-0.90), fewer amputations or death (HR, 0.70-0.89), and fewer cardiovascular events or death (HR, 0.78-0.91). The association between female sex and improved long-term outcomes was most pronounced for the patients with intermittent claudication. CONCLUSIONS: In the present propensity score-matched analysis of health insurance claims, we observed superior cardiovascular event-free survival, amputation-free survival, and overall survival during 5 years of follow-up after percutaneous ER in women with symptomatic PAOD. Future studies should address sex disparities in the open surgical treatment of PAOD to illuminate whether the conflicting data from previous reports might have resulted from insufficient stratification of the studies.

Sex Disparities in Long Term Outcomes After Open Surgery for Chronic Limb Threatening Ischaemia: A Propensity Score Matched Analysis of Health Insurance Claims.

OBJECTIVE: Several studies suggest a disadvantage for women in peri-operative morbidity and mortality after open surgery in peripheral arterial occlusive disease. In addition to their heterogeneity regarding design and analysed cohorts, long term data are mostly missing. This study aimed to determine sex disparities in outcomes after open revascularisation in chronic limb threatening ischaemia (CLTI). METHODS: Using health insurance claims data of the second largest insurance fund in Germany, BARMER, a large cohort of patients was sampled consecutively for analysis including index open surgical revascularisations of CLTI performed between 1 January 2010, and 31 December 2018. Propensity score matching was used to adjust for confounding. Sex related differences regarding overall survival, amputation free survival (AFS), and cardiovascular event free survival (CVEFS) during the five years after surgery were determined using Kaplan-Meier time to event curves, log rank test, logistic, and Cox regression. RESULTS: Among 9 526 patients (49.5% women) in the entire cohort, 6 502 patients were matched. Before matching, women were older at presentation (78.0 vs. 71.8 years, p < .001) and suffered more often from multiple comorbidities (van Walraven score > 9, 55.5% vs. 50.6%, p < .001). During the hospital stay, there were 692 (7.3%) deaths, while 4 631 deaths (48.6%) occurred during the follow up. In the matched cohort, the median follow up was 746 days for women and 871 days for men. In the matched analyses, female sex was significantly associated with better overall survival (hazard ratio, HR, 0.80, log rank p < .001), AFS (HR 0.81, log rank p < .0001), and CVEFS (HR 0.84, log rank p < .001) five years after the index treatment. CONCLUSION: In this largest propensity score matched analysis of health insurance claims to date from Germany, evidence was found for better long term outcomes in women after open surgical revascularisations for chronic limb threatening ischaemia. Future guidelines and studies should address the impact of sex on patient selection practice and outcomes to determine the underlying reasons for existing disparities.

Long-term incidence of cancer after index treatment for symptomatic peripheral arterial disease - a health insurance claims data analysis.

Background: Cancer as a concomitant condition in symptomatic peripheral arterial disease (PAD) patients could have an impact on further therapy and the long-term prognosis of these patients. Aim of this study was to investigate whether there is an increased incidence of cancer in PAD patients and to quantify the corresponding effect size. Materials and methods: Between January 1st, 2008 and December 31st, 2017, we analysed health insurance claims data from Germany's second-largest insurance fund, BARMER. Symptomatic PAD patients suffering from intermittent claudication (IC) or chronic limb-threatening ischaemia (CLTI) were stratified by gender at index treatment. PAD patients were then followed until an incident cancer diagnosis was recorded. To adjust for age and gender, standardized incidence ratios (SIR) were computed using the 2012 German standard population as reference. Results: 96,528 PAD patients (47% female, 44% IC, mean age 72 years) were included in the current study. When compared to the overall population, female and male PAD patients have a significantly increased risk of incident cancer of the lung (SIR 3.5 vs. 2.6), bladder (SIR 3.2 vs. 4.0), pancreas (SIR 1.4 vs. 1.6), and colon (SIR 1.3 vs. 1.3). During ten years of follow-up, some 7% of males and 4% of females developed lung cancer. For bladder, colon and pancreas cancer, the cumulative hazards were 1% vs. 3.2%, 2.2% vs. 2.8%, and 0.7% vs. 0.9%, respectively. Conclusions: Patients suffering from symptomatic PAD face a markedly higher risk for incident cancer in the long-term follow-up. The cancer risk increased continuously for certain types and PAD was strongly associated with cancer of the lung, bladder, pancreas, and colon. Taking these results into account, PAD patients could benefit from secondary and tertiary screening. These results also emphasize the impact of common risk factors such as tobacco smoke as target for health prevention.

Editor's Choice - Optimal Pharmacological Treatment of Symptomatic Peripheral Arterial Occlusive Disease and Evidence of Female Patient Disadvantage: An Analysis of Health Insurance Claims Data.

OBJECTIVE: Optimal pharmacological treatment (OPT) for peripheral arterial occlusive disease (PAOD) includes prescription of lipid lowering drugs, antithrombotics, and antihypertensives to symptomatic patients affected by intermittent claudication or chronic limb threatening ischaemia. This study sought to determine sex disparities and time trends in prescription of OPT in this population (clinicaltrials.gov NCT03909022). METHODS: Using data from the second largest insurance fund in Germany, BARMER, data on patients with an index admission for symptomatic PAOD between 1 January 2010 and 30 June 2018 with follow up until the end of 2018 were analysed. Sex disparities in post-discharge prescription status six months after index admission were tested and adjusted for patient and healthcare variables using bivariable tests and logistic regression analysis. Time trends in the prescription prevalence of OPT were analysed and tested. RESULTS: There were 83 867 patients (mean age 71.9 years and 45.8% women) eligible for inclusion in the study. When compared with men, women had lower rates of prior outpatient care for PAOD (39.8% vs. 47.0%), were admitted more often with ischaemic rest pain (13.9% vs. 10.4%) and were older (74 vs. 70 y). After discharge, women had a lower rate of prescriptions for lipid lowering drugs (52.4% vs. 59.9%), while they received antihypertensive drugs more often (86.7% vs. 84.1%). We found evidence for a lower prescription prevalence of OPT in females (37.0% vs. 42.7%). Differences in patient and healthcare variables (e.g. demographics, comorbidities, prior treatment) between women and men explained 56% of this gap. The sex prescription gap did not narrow over time despite an overall upward trend in prescription prevalence for both women and men. CONCLUSION: Although presenting older and with more severe symptoms at the index admission for PAOD, women have a lower prescription prevalence of OPT compared with men, particularly with respect to lipid lowering drugs.

Editor's Choice - Comorbidity Patterns Among Patients with Peripheral Arterial Occlusive Disease in Germany: A Trend Analysis of Health Insurance Claims Data.

OBJECTIVE: Patients suffering from peripheral arterial occlusive disease (PAOD) are a central target population for multidisciplinary vascular medicine. This study aimed to highlight trends in treatment patterns and comorbidities using up to date longitudinal patient related data from Germany. METHODS: This study is a retrospective health insurance claims data analysis of patients insured by the second largest health insurance provider in Germany, BARMER. All PAOD patient hospitalisations between 2008 and 2016 were included. The comorbidities were categorised with Elixhauser groups using WHO ICD-10 codes and summarised as the linear van Walraven score (vWS). A trend analysis of the comorbidities was performed after standardisation by age and sex. RESULTS: A total of 156 217 patients underwent 202 961 hospitalisations (49.4% for chronic limb threatening ischaemia in 2016) with PAOD during the study period. Although the estimated annual incidence of PAOD among the BARMER cohort decreased slightly (- 4.4%), an increase was observed in the prevalence of PAOD (+ 23.1%), number of hospitalisations (+ 25.1%), peripheral vascular interventions (PVI, + 61.1%), and disease related reimbursement costs (+ 31%) from 2008 to 2016. Meanwhile, the number of major amputations decreased (- 15.1%). The proportion of patients aged 71-80 years increased about +10% among PAOD patients and the mean vWS also increased by two points during the study period. Considerable increases were found in the rates of hypertension, renal failure, and hypothyroidism, whereas the rates of diabetes and congestive heart failure decreased over time. CONCLUSION: Increasing numbers of PVI performed on these ageing and sicker patients lead to rising costs but correlate with decreasing major amputation rates.

Is the Rise in the Prevalence of Renal Replacement Therapy at Older Ages the Price for Living Longer?

BACKGROUND: Renal replacement therapy (RRT) is one of the most expensive in renal medicine. Cross-sectional studies suggest that life expectancy increases in the general population are associated with a higher burden of RRT. This study tests this hypothesis in a prospective setting among people aged 75+ living in Western Europe. METHODS: We gathered sex-specific data for 11 Western European countries in 2005-2014. RRT prevalence on country level was extracted from the ERA-EDTA registry, while data on population size and life expectancy for the 75+ age group came from the Eurostat database. GDP per capita was extracted from the OECD database. To measure the association between RRT prevalence and life expectancy, we performed Poisson regression models separately for each country and for all countries combined. To adjust for confounding, GDP per capita as well as time and country-fixed effects were included. RESULTS: Our analysis revealed that living longer coincides with rising RRT prevalence at ages 75+ in Western Europe between 2005 and 2014. On average, a 1-year increase in life expectancy was associated with a roughly 20% increase in RRT prevalence [(95% CI) 21-23% in men and 19-22% in women]. However, after adjustments for confounding were made, the association became insignificant among women and became weaker among men, falling to a level of 11% [(95% CI) 6-17%]. CONCLUSION: Living longer was not necessarily associated with a higher burden of RRT in Western European countries.

Forecasting differences in life expectancy by education.

Forecasts of life expectancy (LE) have fuelled debates about the sustainability and dependability of pension and healthcare systems. Of relevance to these debates are inequalities in LE by education. In this paper, we present a method of forecasting LE for different educational groups within a population. As a basic framework we use the Li-Lee model that was developed to forecast mortality coherently for different groups. We adapted this model to distinguish between overall, sex-specific, and education-specific trends in mortality, and extrapolated these time trends in a flexible manner. We illustrate our method for the population aged 65 and over in the Netherlands, using several data sources and spanning different periods. The results suggest that LE is likely to increase for all educational groups, but that differences in LE between educational groups will widen. Sensitivity analyses illustrate the advantages of our proposed method.

Health inequalities in the Netherlands: trends in quality-adjusted life expectancy (QALE) by educational level.

BACKGROUND: Quality-adjusted life expectancy (QALE) has been proposed as a summary measure of population health because it encompasses multiple health domains as well as length of life. However, trends in QALE by education or other socio-economic measure have not yet been reported. This study investigates changes in QALE stratified by educational level for the Dutch population in the period 2001-2011. METHODS: Using data from multiple sources, we estimated mortality rates and health-related quality of life (HRQoL) as functions of age, gender, calendar year and educational level. Subsequently, predictions from these regressions were combined for calculating QALE at ages 25 and 65. QALE changes were decomposed into effects of mortality and HRQoL. RESULTS: In 2001-2011, QALE increased for men and women at all educational levels, the largest increases being for highly educated resulting in a widening gap by education. In 2001, at age 25, the absolute QALE difference between the low and the highly educated was 7.4 healthy years (36.7 vs. 44.1) for men and 6.3 healthy years (39.5 vs. 45.8) for women. By 2011, the QALE difference increased to 8.1 healthy years (38.8 vs. 46.9) for men and to 7.1 healthy years (41.3 vs. 48.4) for women. Similar results were observed at age 65. Although the gap was largely attributable to widening inequalities in mortality, widening inequalities in HRQoL were also substantial. CONCLUSIONS: In the Netherlands, population health as measured by QALE has improved, but QALE inequalities have widened more than inequalities in life expectancy alone.

Quantifying the contribution of changes in healthcare expenditures and smoking to the reversal of the trend in life expectancy in the Netherlands.

BACKGROUND: Since 2001 the Netherlands has shown a sharp upturn in life expectancy (LE) after a longer period of slower improvement. This study assessed whether changes in healthcare expenditure (HCE) explain this reversal in trends in LE. As an alternative explanation, the impact of changes in smoking behavior was also evaluated. METHODS: To quantify the contribution of changes in HCE to changes in LE, we estimated a health-production function using a dynamic panel regression approach with data on 19 OECD countries (1980-2009), accounting for temporal and spatial correlation. Smoking-attributable mortality was estimated using the indirect Peto-Lopez method. RESULTS: As compared to 1990-1999, during 2000-2009 LE in the Netherlands increased by 1.8 years in females and by 1.5 years in males. Whereas changes in the impact of smoking between the two periods made almost no contribution to the acceleration of the increase in LE, changes in the trend of HCE added 0.9 years to the LE increase between 2000 and 2009. The exceptional reversal in the trend of LE and HCE was not found among the other OECD countries. CONCLUSION: This study suggests that changes in Dutch HCE, and not in smoking, made an important contribution to the reversal of the trend in LE; these findings support the view that investments in healthcare are increasingly important for further progress in life expectancy.