Real-world evidence from the first online healthcare analytics platform-Livingstone. Validation of its descriptive epidemiology module.

Incidence and prevalence are key epidemiological determinants characterizing the quantum of a disease. We compared incidence and prevalence estimates derived automatically from the first ever online, essentially real-time, healthcare analytics platform-Livingstone-against findings from comparable peer-reviewed studies in order to validate the descriptive epidemiology module. The source of routine NHS data for Livingstone was the Clinical Practice Research Datalink (CPRD). After applying a general search strategy looking for any disease or condition, 76 relevant studies were first retrieved, of which 10 met pre-specified inclusion and exclusion criteria. Findings reported in these studies were compared with estimates produced automatically by Livingstone. The published reports described elements of the epidemiology of 14 diseases or conditions. Lin's concordance correlation coefficient (CCC) was used to evaluate the concordance between findings from Livingstone and those detailed in the published studies. The concordance of incidence values in the final year reported by each study versus Livingstone was 0.96 (95% CI: 0.89-0.98), whilst for all annual incidence values the concordance was 0.93 (0.91-0.94). For prevalence, concordance for the final annual prevalence reported in each study versus Livingstone was 1.00 (0.99-1.00) and for all reported annual prevalence values, the concordance was 0.93 (0.90-0.95). The concordance between Livingstone and the latest published findings was near perfect for prevalence and substantial for incidence. For the first time, it is now possible to automatically generate reliable descriptive epidemiology from routine health records, and in near-real time. Livingstone provides the first mechanism to rapidly generate standardised, descriptive epidemiology for all clinical events from real world data.

A genetic programming approach to development of clinical prediction models: A case study in symptomatic cardiovascular disease.

BACKGROUND: Genetic programming (GP) is an evolutionary computing methodology capable of identifying complex, non-linear patterns in large data sets. Despite the potential advantages of GP over more typical, frequentist statistical approach methods, its applications to survival analyses are rare, at best. The aim of this study was to determine the utility of GP for the automatic development of clinical prediction models. METHODS: We compared GP against the commonly used Cox regression technique in terms of the development and performance of a cardiovascular risk score using data from the SMART study, a prospective cohort study of patients with symptomatic cardiovascular disease. The composite endpoint was cardiovascular death, non-fatal stroke, and myocardial infarction. A total of 3,873 patients aged 19-82 years were enrolled in the study 1996-2006. The cohort was split 70:30 into derivation and validation sets. The derivation set was used for development of both GP and Cox regression models. These models were then used to predict the discrete hazards at t = 1, 3, and 5 years. The predictive ability of both models was evaluated in terms of their risk discrimination and calibration using the validation set. RESULTS: The discrimination of both models was comparable. At time points t = 1, 3, and 5 years the C-index was 0.59, 0.69, 0.64 and 0.66, 0.70, 0.70 for the GP and Cox regression models respectively. At the same time points, the calibration of both models, which was assessed using calibration plots and the generalization of the Hosmer-Lemeshow test statistic, was also comparable, but with the Cox model being better calibrated to the validation data. CONCLUSION: Using empirical data, we demonstrated that a prediction model developed automatically by GP has predictive ability comparable to that of manually tuned Cox regression. The GP model was more complex, but it was developed in a fully automated way and comprised fewer covariates. Furthermore, it did not require the expertise normally needed for its derivation, thereby alleviating the knowledge elicitation bottleneck. Overall, GP demonstrated considerable potential as a method for the automated development of clinical prediction models for diagnostic and prognostic purposes.

Risk of cardiovascular events, arrhythmia and all-cause mortality associated with clarithromycin versus alternative antibiotics prescribed for respiratory tract infections: a retrospective cohort study.

OBJECTIVE: To determine whether treatment with clarithromycin for respiratory tract infections was associated with an increased risk of cardiovascular (CV) events, arrhythmias or all-cause mortality compared with other antibiotics. DESIGN: Retrospective cohort design comparing clarithromycin monotherapy for lower (LRTI) or upper respiratory tract infection (URTI) with other antibiotic monotherapies for the same indication. SETTING: Routine primary care data from the UK Clinical Practice Research Datalink and inpatient data from the Hospital Episode Statistics (HES). PARTICIPANTS: Patients aged ≥35 years prescribed antibiotic monotherapy for LRTI or URTI 1998-2012 and eligible for data linkage to HES. MAIN OUTCOME MEASURES: The main outcome measures were: adjusted risk of first-ever CV event, within 37 days of initiation, in commonly prescribed antibiotics compared with clarithromycin. Secondarily, adjusted 37-day risks of first-ever arrhythmia and all-cause mortality. RESULTS: Of 700 689 treatments for LRTI and eligible for the CV analysis, there were 2071 CV events (unadjusted event rate: 29.6 per 10 000 treatments). Of 691 998 eligible treatments for URTI, there were 688 CV events (9.9 per 10 000 treatments). In LRTI and URTI, there were no significant differences in CV risk between clarithromycin and all other antibiotics combined: OR=1.00 (95% CI 0.82 to 1.22) and 0.82 (0.54 to 1.25), respectively. Adjusted CV risk in LRTI versus clarithromycin ranged from OR=1.42 (cefalexin; 95% CI 1.08 to 1.86) to 0.92 (doxycycline; 0.64 to 1.32); in URTI, from 1.17 (co-amoxiclav; 0.68 to 2.01) to 0.67 (erythromycin; 0.40 to 1.11). Adjusted mortality risk versus clarithromycin in LRTI ranged from 0.42 to 1.32; in URTI, from 0.75 to 1.43. For arrhythmia, adjusted risks in LRTI ranged from 0.68 to 1.05; in URTI, from 0.70 to 1.22. CONCLUSIONS: CV events were more likely after LRTI than after URTI. When analysed by specific indication, CV risk associated with clarithromycin was no different to other antibiotics.

Natural History of Patients Taking Rifaximin-α for Recurrent Hepatic Encephalopathy and Risk of Future Overt Episodes and Mortality: A Post-hoc Analysis of Clinical Trials Data.

PURPOSE: Hepatic encephalopathy (HE) is a complication of cirrhosis signaling decompensation and is associated with mortality. There has been little characterization of HE once an incident episode has occurred and of what effect the transition to overt HE might have on outcomes. We characterized the relationships between the number of previous HE episodes and risk of subsequent episodes and mortality to better understand the natural history of HE. METHODS: Data on 321 patients from a 24-month, open-label, nonrandomized trial evaluating the long-term safety profile and tolerability of twice-daily rifaximin-α 550 mg were analyzed. Patients were followed for a mean of 1.5 years (total follow-up of 467 years). FINDINGS: There were a total of 334 HE episodes and 75 deaths, corresponding to unadjusted event rates of 715 HE episodes and 161 deaths per 1000 years. There was a direct association between rate of subsequent HE episodes and number of prior HE episodes; the risk of subsequent HE episodes was elevated for each additional HE episode (hazard ratio = 1.23; 95% CI, 1.19-1.29). There was a nonlinear, nonmonotonic relationship between risk of death and number of prior HE episodes; risk initially increased, then decreased, and finally plateaued as the number of prior HE episodes increased. IMPLICATIONS: Patients with a larger number of previous, overt HE episodes had a greater risk for subsequent episodes. However, mortality risk decreased after the third episode of HE. A plausible hypothesis to explain this finding is that risk of mortality may be reduced in patients receiving long-term rifaximin-α therapy.

Estimation of health-related utility (EQ-5D index) in subjects with seasonal allergic rhinoconjunctivitis to evaluate health gain associated with sublingual grass allergen immunotherapy.

BACKGROUND: Grass allergen immunotherapy (AIT) reduces symptom severity in seasonal allergic rhinoconjunctivitis (ARC) but its impact on general health-related utility has not been characterised for the purposes of economic evaluation. The aim of this study was to model the preferred measure of utility, EQ-5D index, from symptom severity and estimate incremental quality adjusted life years (QALYs) associated with SQ-standardised grass immunotherapy tablet (GRAZAX®, 75,000 SQ-T/2,800 BAU, ALK, Denmark). METHODS: Data were analysed from five consecutive pollen seasons in a randomised placebo controlled trial of GRAZAX®. Binomial and Gaussian mixed effects modelling related weekly EQ-5D index score to daily symptom and medication scores (DSS & DMS respectively). In turn, daily EQ-5D index was estimated from ARC symptoms and medication use. RESULTS: DSS and DMS were the principal predictors of 'perfect' health (EQ-5D = 1.000; binomial) and 'imperfect' health (EQ-5D < 1.000; Gaussian). Each unit increase in DSS and DMS reduced the odds of 'perfect' health (EQ-5D = 1.000) by 27% and 16% respectively, and reduced 'imperfect' health by 0.17 and 0.13, respectively. Gender remained the only other significant main fixed effect (Male odds ratio [OR] = 1.82). Incremental estimated EQ-5D index utility for GRAZAX® was observed from day -30 to day +70 of the pooled pollen season; mean daily utility for GRAZAX® = 0.938 units (95%CI 0.932-0.943) vs. 0.914 (0.907-0.921) for placebo, an incremental difference of 0.0238 (p < 0.001). This translates into an incremental 0.0324 Quality Adjusted Life Years over the five year study period. CONCLUSIONS: ARC symptoms and medication use are the main predictors of EQ-5D index. The incremental QALYs observed for GRAZAX® may not fully describe the health benefits of this treatment, suggesting that economic modelling may be conservative.

External validation of the UKPDS risk engine in incident type 2 diabetes: a need for new type 2 diabetes-specific risk equations.

OBJECTIVE: To evaluate the performance of the UK Prospective Diabetes Study Risk Engine (UKPDS-RE) for predicting the 10-year risk of cardiovascular disease end points in an independent cohort of U.K. patients newly diagnosed with type 2 diabetes. RESEARCH DESIGN AND METHODS: This was a retrospective cohort study using routine health care data collected between April 1998 and October 2011 from ∼350 U.K. primary care practices contributing to the Clinical Practice Research Datalink (CPRD). Participants comprised 79,966 patients aged between 35 and 85 years (388,269 person-years) with 4,984 cardiovascular events. Four outcomes were evaluated: first diagnosis of coronary heart disease (CHD), stroke, fatal CHD, and fatal stroke. RESULTS: Accounting for censoring, the observed versus predicted 10-year event rates were as follows: CHD 6.1 vs. 16.5%, fatal CHD 1.9 vs. 10.1%, stroke 7.0 vs. 10.1%, and fatal stroke 1.7 vs. 1.6%, respectively. The UKPDS-RE showed moderate discrimination for all four outcomes, with the concordance index values ranging from 0.65 to 0.78. CONCLUSIONS: The UKPDS stroke equations showed calibration ranging from poor to moderate; however, the CHD equations showed poor calibration and considerably overestimated CHD risk. There is a need for revised risk equations in type 2 diabetes.