BackgroundMachine-learning classifiers mostly offer good predictive performance and are increasingly used to support shared decision-making in clinical practice. Focusing on performance and practicability, this study evaluates prediction of patient-reported outcomes (PROs) by eight supervised classifiers including a linear model, following hip and knee replacement surgery.MethodsNHS PRO data (130,945 observations) from April 2015 to April 2017 were used to train and test eight classifiers to predict binary postoperative improvement based on minimal important differences. Area under the receiver operating characteristic, J-statistic and several other metrics were calculated. The dependent outcomes were generic and disease-specific improvement based on the EQ-5D-3L visual analogue scale (VAS) as well as the Oxford Hip and Knee Score (Q score).ResultsThe area under the receiver operating characteristic of the best training models was around 0.87 (VAS) and 0.78 (Q score) for hip replacement, while it was around 0.86 (VAS) and 0.70 (Q score) for knee replacement surgery. Extreme gradient boosting, random forests, multistep elastic net and linear model provided the highest overall J-statistics. Based on variable importance, the most important predictors for post-operative outcomes were preoperative VAS, Q score and single Q score dimensions. Sensitivity analysis for hip replacement VAS evaluated the influence of minimal important difference, patient selection criteria as well as additional data years. Together with a small benchmark of the NHS prediction model, robustness of our results was confirmed.ConclusionsSupervised machine-learning implementations, like extreme gradient boosting, can provide better performance than linear models and should be considered, when high predictive performance is needed. Preoperative VAS, Q score and specific dimensions like limping are the most important predictors for postoperative hip and knee PROMs.Electronic supplementary materialThe online version of this article (10.1186/s12911-018-0731-6) contains supplementary material, which is available to authorized users.

The EQ-5D-5L is a widely used generic instrument to measure health-related quality of life. This study evaluates health perception in a representative sample of the general German population from 2015. To compare results over time, a component analysis technique was used that separates changes in the description and valuation of health states. The whole sample and also subgroups, stratified by sociodemographic parameters as well as disease affliction, were analyzed. In total, 2040 questionnaires (48.4% male, mean age 47.3 year) were included. The dimension with the lowest number of reported problems was self-care (93.0% without problems), and the dimension with the highest proportion of impairment was pain/discomfort (71.2% without problems). Some 64.3% of the study population were identified as problem-free. The visual analog scale (VAS) mean for all participants was 85.1. Low education was connected with significantly lower VAS scores, but the effect was small. Depression, heart disease, and diabetes had a strong significant negative effect on reported VAS means. Results were slightly better than those in a similar 2012 survey; the most important driver was the increase in the share of the study population that reported to be problem-free. In international comparisons, health perception of the general German population is relatively high and, compared with previous German studies, fairly stable over recent years. Elderly and sick people continue to report significant reductions in perceived health states.

Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS) is an unexplained chronic, debilitating illness characterized by fatigue, sleep disturbances, cognitive dysfunction, orthostatic intolerance and gastrointestinal problems. Using ultra performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS), we analyzed the plasma proteomes of 39 ME/CFS patients and 41 healthy controls. Logistic regression models, with both linear and quadratic terms of the protein levels as independent variables, revealed a significant association between ME/CFS and the immunoglobulin heavy variable (IGHV) region 3-23/ 30. Stratifying the ME/CFS group based on self-reported irritable bowel syndrome (sr-IBS) status revealed a significant quadratic effect of immunoglobulin lambda constant region 7 on its association with ME/CFS with sr-IBS whilst IGHV3-23/30 and immunoglobulin kappa variable region 3-11 were significantly associated with ME/CFS without sr-IBS. In addition, we were able to predict ME/CFS status with a high degree of accuracy (AUC = 0.774-0.838) using a panel of proteins selected by 3 different machine learning algorithms: Lasso, Random Forests, and XGBoost. These algorithms also identified proteomic profiles that predicted the status of ME/CFS patients with sr-IBS (AUC = 0.806-0.846) and ME/CFS without sr-IBS (AUC = 0.754-0.780). Our findings are consistent with a significant association of ME/CFS with immune dysregulation and highlight the potential use of the plasma proteome as a source of biomarkers for disease.

Health-related quality of life (HRQoL) is a key measure for evaluating health status in populations. Using the recent EQ-5D-5L for measurement, this study analyzed quality of life results and their stability over consecutive population surveys. Three cross-section surveys for representative samples of the general German population from 2012, 2013, and 2014 were evaluated using the EQ-5D-5L descriptive system and valuation by the Visual Analog Scale (VAS). Aggregated sample size reached 6074. The dimension with the highest prevalence of problems was pain/discomfort (31.7%). Compared with 2012 (59.3%), the percentage of participants in the best health state increased slightly in 2013 (63.4%) and 2014 (62%). Over the 3-year period, diabetes and heart disease had the strongest negative influence on mean VAS result. The number of reported chronic diseases cumulatively reduced mean VAS. Extreme problems in one or more dimensions were stated by only 0.1%–0.2% of patients. Of the potential 247 health states with a problem score ≥20, only six were observed in the aggregated sample. HRQoL results were fairly stable over the 3 years, but the share of the population with no problems was not. Results from the aggregated sample may serve as updated reference values for the general German population.

BackgroundChronic obstructive pulmonary disease (COPD) is a leading cause of mortality and of loss of disability-adjusted life years worldwide. It often is accompanied by the presence of comorbidity.ObjectivesTo systematically review the influence of COPD comorbidity on generic health-related quality of life (HRQoL).MethodsA systematic review approach was used to search the databases Pubmed, Embase and Cochrane Library for studies evaluating the influence of comorbidity on HRQoL in COPD. Identified studies were analyzed according to study characteristics, generic HRQoL measurement instrument, COPD severity and comorbid HRQoL impact. Studies using only non-generic instruments were excluded.Results25 studies met the selection criteria. Seven studies utilized the EQ-5D, six studies each used the SF-36 or SF-12. The remaining studies used one of six other instruments each. Utilities were calculated by four EQ-5D studies and one 15D study. Patient populations covered both early and advanced stages of COPD and ranged from populations with mostly stage 1 and 2 to studies with patients classified mainly stage 3 and 4. Evidence was mainly created for cardiovascular disease, depression and anxiety as well as diabetes but also for quantitative comorbid associations. Strong evidence is pointing towards the significant negative association of depression and anxiety on reduced HRQoL in COPD patients. While all studies found the occurrence of specific comorbidities to decrease HRQoL in COPD patients, the orders of magnitude diverged. Due to different patient populations, different measurement tools and different concomitant diseases the study heterogeneity was high.ConclusionsFacilitating multimorbid intervention guidance, instead of applying a parsimony based single disease paradigm, should constitute an important goal for improving HRQoL of COPD patients in research and in clinical practice.