Objective To directly compare the performance and externally validate the three most studied prediction tools for osteoporotic fractures—QFracture, FRAX, and Garvan—using data from electronic health records.Design Retrospective cohort study.Setting Payer provider healthcare organisation in Israel.Participants 1 054 815 members aged 50 to 90 years for comparison between tools and cohorts of different age ranges, corresponding to those in each tools’ development study, for tool specific external validation.Main outcome measure First diagnosis of a major osteoporotic fracture (for QFracture and FRAX tools) and hip fractures (for all three tools) recorded in electronic health records from 2010 to 2014. Observed fracture rates were compared to probabilities predicted retrospectively as of 2010.Results The observed five year hip fracture rate was 2.7% and the rate for major osteoporotic fractures was 7.7%. The areas under the receiver operating curve (AUC) for hip fracture prediction were 82.7% for QFracture, 81.5% for FRAX, and 77.8% for Garvan. For major osteoporotic fractures, AUCs were 71.2% for QFracture and 71.4% for FRAX. All the tools underestimated the fracture risk, but the average observed to predicted ratios and the calibration slopes of FRAX were closest to 1. Tool specific validation analyses yielded hip fracture prediction AUCs of 88.0% for QFracture (among those aged 30-100 years), 81.5% for FRAX (50-90 years), and 71.2% for Garvan (60-95 years).Conclusions Both QFracture and FRAX had high discriminatory power for hip fracture prediction, with QFracture performing slightly better. This performance gap was more pronounced in previous studies, likely because of broader age inclusion criteria for QFracture validations. The simpler FRAX performed almost as well as QFracture for hip fracture prediction, and may have advantages if some of the input data required for QFracture are not available. However, both tools require calibration before implementation.

No abstract

AimsTo identify clinically meaningful clusters of patients with similar glycated hemoglobin (HbA1c) trajectories among patients with type 2 diabetes.MethodsA retrospective cohort study using unsupervised machine learning clustering methodologies to determine clusters of patients with similar longitudinal HbA1c trajectories. Stability of these clusters was assessed and supervised random forest analysis verified the clusters’ reproducibility. Clinical relevance of the clusters was assessed through multivariable analysis, comparing differences in risk for a composite outcome (macrovascular and microvascular outcomes, hypoglycemic events, and all-cause mortality) at HbA1c thresholds for each cluster.ResultsAmong 60,423 patients, three clusters of HbA1c trajectories were generated: stable (n = 45,679), descending (n = 6,084), and ascending (n = 8,660) trends, which were reproduced with 99.8% accuracy using a random forest model. In the clinical relevance assessment, HbA1c levels demonstrated a J-shape association with the risk for outcomes. HbA1c level thresholds for minimizing outcomes’ risk differed by cluster: 6.0–6.4% for the stable cluster, <8.0% for the descending cluster, and <9.0 for the ascending cluster.ConclusionsBy applying unsupervised machine learning to longitudinal HbA1c trajectories, we have identified clusters of patients who have distinct risk for diabetes-related complications. These clusters can be the basis for developing individualized models to personalize glycemic targets.

The PPSV23 vaccine is effective against the most severe invasive forms of pneumococcal disease, but the lack of effectiveness of PPSV23 in protecting against all-cause HTP should be considered for future vaccine policies.

AimsThis study assesses the attributable impact of adherence to oral glucose medications as a risk factor for poor glycemic control in population subgroups of a large general population, using an objective medication adherence measure.MethodsUsing electronic health records data, adherence to diabetes medications over a two-year period was calculated by prescription-based Medication Possession Ratios for adults with diabetes diagnosed before January 1, 2010. Glycemic control was determined by the HbA1c test closest to the last drug prescription during 2010–2012. Poor control was defined as HbA1c>75 mmol/mol (9.0%). Medication adherence was categorized as “good” (>80%), “moderate” (50–80%), or “poor” (<50%). Logistic regression models assessed the role medication adherence plays in the association between disease duration, age, and poor glycemic control. We calculated the change in the attributable fraction of glucose control if the non-adherent diabetic medication population would become adherent by age-groups.ResultsAmong 228,846 diabetes patients treated by oral antiglycemic medication, 46.4% had good, 28.8% had moderate, and 24.8% had poor adherence. Good adherence rates increased with increasing disease duration, while glycemic control became worse. There was a strong inverse association between adherence level and poor control (OR = 2.50; CI = 2.43–2.58), and adherence was a significant mediator between age and poor control.ConclusionsA large portion of the diabetes population is reported to have poor adherence to oral diabetes medications, which is strongly associated with poor glycemic control in all disease durations. While poor adherence does not mediate the poorer glycemic control seen in patients with longer-standing disease, it is a significant mediator of poor glycemic control among younger diabetes patients. A greater fraction of poorly controlled younger patients, compared to older patients, could be prevented if at least 80% adherence to their medications was achieved. Therefore, our results suggest that interventions to improve adherence should focus on this younger sub-group.

International guidelines recommend treatment with statins for patients with preexisting ischemic heart disease to prevent additional cardiovascular events but differ regarding target levels of low-density lipoprotein cholesterol (LDL-C). Trial data on this question are inconclusive and observational data are lacking. OBJECTIVE To assess the relationship between levels of LDL-C achieved with statin treatment and cardiovascular events in adherent patients with preexisting ischemic heart disease. DESIGN, SETTING, AND PARTICIPANTS Population-based observational cohort study from 2009 to 2013 using data from a health care organization in Israel covering more than 4.3 million members. Included patients had ischemic heart disease, were aged 30 to 84 years, were treated with statins, and were at least 80% adherent to treatment or, in a sensitivity analysis, at least 50% adherent. Patients with active cancer or metabolic abnormalities were excluded. EXPOSURES Index LDL-C was defined as the first achieved serum LDL-C measure after at least 1 year of statin treatment, grouped as low (Յ70.0 mg/dL), moderate (70.1-100.0 mg/dL), or high (100.1-130.0 mg/dL). MAIN OUTCOMES AND MEASURES Major adverse cardiac events included acute myocardial infarction, unstable angina, stroke, angioplasty, bypass surgery, or all-cause mortality. The hazard ratio of adverse outcomes was estimated using 2 Cox proportional hazards models with low vs moderate and moderate vs high LDL-C, adjusted for confounders and further tested using propensity score matching analysis. RESULTS The cohort with at least 80% adherence included 31 619 patients, for whom the mean (SD) age was 67.3 (9.8) years. Of this population, 27% were female and 29% had low, 53% moderate, and 18% high LDL-C when taking statin treatment. Overall, there were 9035 patients who had an adverse outcome during a mean 1.6 years of follow-up (6.7 per 1000 persons per year). The adjusted incidence of adverse outcomes was not different between low and moderate LDL-C (hazard ratio [HR], 1.02; 95% CI, 0.97-1.07; P = .54), but it was lower with moderate vs high LDL-C (HR, 0.89; 95% CI, 0.84-0.94; P < .001). Among 54 884 patients with at least 50% statin adherence, the adjusted HR was 1.06 (95% CI, 1.02-1.10; P = .001) in the low vs moderate groups and 0.87 (95% CI, 0.84-0.91; P = .001) in the moderate vs high groups. CONCLUSIONS AND RELEVANCE Patients with LDL-C levels of 70 to 100 mg/dL taking statins had lower risk of adverse cardiac outcomes compared with those with LDL-C levels between 100 and 130 mg/dL, but no additional benefit was gained by achieving LDL-C of 70 mg/dL or less. These population-based data do not support treatment guidelines recommending very low target LDL-C levels for all patients with preexisting heart disease.

BackgroundWith increasing diabetes prevalence worldwide, an impending diabetes “pandemic” has been reported. However, definitions of incident cases and the population at risk remain varied and ambiguous. This study analyzed trends in mortality and screening that contribute to diabetes prevalence and incidence, distinguishing between new incident cases and newly detected cases.MethodsIn an integrated provider-and-payer-system covering 53% of Israel’s population, a composite diabetes case-finding algorithm was built using diagnoses, lab tests, and antidiabetic medication purchases from the organization’s electronic medical record database. Data were extracted on adult members aged 26+ each year from January 1, 2004 through December 31, 2012. Rates of diabetes prevalence, incidence, screening, and mortality were reported, with incidence rates evaluated among the total, “previously-screened,” and “previously-unscreened” at-risk populations.ResultsThere were 343,554 diabetes cases in 2012 (14.4%) out of 2,379,712 members aged 26+. A consistent but decelerating upward trend in diabetes prevalence was observed from 2004–2012. Annual mortality rates among diabetics decreased from 13.8/1000 to 10.7/1000 (p = 0.0002). Total population incidence rates declined from 13.3/1000 in 2006 to 10.8/1000 in 2012 (p < 0.0001), with similar incidence trends (13.2/1000 to 10.2/1000; p = 0.0007) among previously-screened at-risk members, and a rise in testing rates from 53.0% to 66.7% (p = 0.0004). The previously-unscreened group decreased 28.6%, and the incidence rates within this group remained stable.ConclusionsThe increase in diabetes prevalence is decelerating despite declining mortality and increasing testing rates. A decline in previously-screened incident cases and a shrinking pool of previously-unscreened members suggests that diabetes trends in Israel are moving toward equilibrium, rather than a growing epidemic.Electronic supplementary materialThe online version of this article (doi:10.1186/s12963-014-0032-y) contains supplementary material, which is available to authorized users.

Background Disease-specific guidelines are not aligned with multimorbidity care complexity. Meeting all guideline-recommended care for multimorbid patients has been estimated but not demonstrated across multiple guidelines. Objective Measure guideline-concordant care for patients with multimorbidity; assess in what types of care and by whom (clinician or patient) deviation from guidelines occurs and evaluate whether patient characteristics are associated with concordance. Methods A retrospective cohort study of care received over 1 year, conducted across 11 primary care clinics within the context of multimorbidity-focused care management program. Patients were aged 45+ years with more than two common chronic conditions and were sampled based on either being new (≤6 months) or veteran to the program (≥1 year). Measures Three guideline concordance measures were calculated for each patient out of 44 potential guideline-recommended care processes for nine chronic conditions: overall score; referral score (proportion of guideline-recommended care referred) and patient-only score (proportion of referred care completed by patients). Guideline concordance was stratified by care type. Results 4386 care processes evaluated among 204 patients, mean age = 72.3 years (standard deviation = 9.7). Overall, 79.2% of care was guideline concordant, 87.6% was referred according to guidelines and patients followed 91.4% of referred care. Guideline-concordant care varied across care types. Age, morbidity burden and whether patients were new or veteran to the program were associated with guideline concordance. Conclusions Patients with multimorbidity do not receive ~20% of guideline recommendations, mostly due to clinicians not referring care. Determining the types of care for which the greatest deviation from guidelines exists can inform the tailoring of care for multimorbidity patients.