Amy S. Nowacki

Understanding Equivalence and Noninferiority Testing

Increasingly, the goal of many studies is to determine if new therapies have equivalent or noninferior efficacies to the ones currently in use. These studies are called equivalence/noninferiority studies, and the statistical methods for their analysis require only simple modifications to the traditional hypotheses testing framework. Nevertheless, important and subtle issues arise with the application of such methods. This article describes the concepts and statistical methods involved in testing equivalence/noninferiority. The aim is to enable the clinician to understand and critically assess the growing number of articles utilizing such methods.

The association between personal health record use and diabetes quality measures.

BackgroundElectronic personal health records (PHRs) have the potential to empower patients in self-management of chronic diseases, which should lead to improved outcomes.ObjectiveTo measure the association between use of an advanced electronic medical record-linked PHR and diabetes quality measures in adults with diabetes mellitus (DM).DesignRetrospective audit of PHR use and multivariable regression analyses.Patients10,746 adults 18–75-years of age with DM seen at least twice at the office of their primary care physician at the Cleveland Clinic from July 2008 through June 2009.Main MeasuresPHR use was measured as number of use days. Diabetes quality measures were: hemoglobin A1c (HbA1c), LDL cholesterol, blood pressure, body mass index (BMI), HbA1c testing, ACEi/ARB use and/or microalbumin testing, pneumococcal vaccination, foot and dilated eye examination, and smoking status.Key ResultsCompared to non-users, PHR users were younger, had higher incomes and educational attainment, were more likely to identify as Caucasian, and had better unadjusted and adjusted diabetes quality measure profiles. Adjusted odds ratio of HbA1c testing was 2.06 (p < 0.01) and most recent HbA1c was 0.29% lower (p < 0.01). Among PHR users, increasing number of login days was generally not associated with more favorable diabetes quality measure profiles.ConclusionsPHR use, but not intensity of use, was associated with improved diabetes quality measure profiles. It is likely that better diabetes profiles among PHR users is due to higher level of engagement with their health among those registered for the PHR rather than PHR use itself. PHR use was infrequent. To maximize value, next-generation PHRs must be designed to engage patients in everyday diabetes self-management.

Strategies for handling missing data in electronic health record derived data.

Electronic health records (EHRs) present a wealth of data that are vital for improving patient-centered outcomes, although the data can present significant statistical challenges. In particular, EHR data contains substantial missing information that if left unaddressed could reduce the validity of conclusions drawn. Properly addressing the missing data issue in EHR data is complicated by the fact that it is sometimes difficult to differentiate between missing data and a negative value. For example, a patient without a documented history of heart failure may truly not have disease or the clinician may have simply not documented the condition. Approaches for reducing missing data in EHR systems come from multiple angles, including: increasing structured data documentation, reducing data input errors, and utilization of text parsing / natural language processing. This paper focuses on the analytical approaches for handling missing data, primarily multiple imputation. The broad range of variables available in typical EHR systems provide a wealth of information for mitigating potential biases caused by missing data. The probability of missing data may be linked to disease severity and healthcare utilization since unhealthier patients are more likely to have comorbidities and each interaction with the health care system provides an opportunity for documentation. Therefore, any imputation routine should include predictor variables that assess overall health status (e.g. Charlson Comorbidity Index) and healthcare utilization (e.g. number of encounters) even when these comorbidities and patient encounters are unrelated to the disease of interest. Linking the EHR data with other sources of information (e.g. National Death Index and census data) can also provide less biased variables for imputation. Additional methodological research with EHR data and improved epidemiological training of clinical investigators is warranted.

Absence of an association between glucose levels and surgical site infections in patients undergoing craniotomies for brain tumors

OBJECT In select patient populations, hyperglycemia has been shown to increase the risk of surgical site infection (SSI), whereas stringent glucose control has improved outcomes. To date, no study has focused on whether SSIs in patients with brain tumors undergoing resection are associated with hyperglycemia. METHODS The authors performed a retrospective chart review of patients who underwent a craniotomy after receiving a diagnosis of brain tumor. From 2001 to 2008, 2485 patients underwent a craniotomy for tumor resection at the Brain Tumor & Neuro-Oncology Center at the Cleveland Clinic. Fifty-seven of these patients (2.3%) developed SSIs postoperatively. A matched case-control study design was used, with 57 patients who developed SSIs after craniotomy (cases) matched with 57 patients who did not develop SSIs (controls). The results were analyzed using both univariate and multivariate conditional logistic regression. RESULTS Glucose level was not a significant factor in postoperative SSI (p = 0.83) after adjusting for duration of surgery and adherence to antibiotic prophylaxis. However, duration of surgery was significantly associated with postoperative SSI (p = 0.047). CONCLUSIONS For patients who undergo craniotomy for definitive resection of a brain tumor, duration of surgery described more variation in the model to predict SSI than blood glucose levels.

Predicting C5 palsy via the use of preoperative anatomic measurements

BACKGROUND CONTEXT C5 nerve root palsy (C5P) is a relatively rare complication after anterior and posterior cervical decompression surgery that leads to a variety of debilitating symptoms. The precise etiology remains obscure, and a clear understanding of preoperative risk factors for C5P development does not exist. PURPOSE To determine whether postoperative C5P can be predicted from preoperative anteroposterior diameter (APD), foraminal diameter (FD), and/or cord-lamina angle (CLA). STUDY DESIGN Retrospective review. PATIENT SAMPLE Consecutive patients who underwent either anterior or posterior decompression surgery at C4-C5 for cervical spondylotic myelopathy. OUTCOME MEASURES Development of C5P. METHODS Blinded reviewers retrospectively assessed magnetic resonance images for each included patients C4-C5 interspace, including the midline APD, the left and right FDs, and the left and right CLA. Multivariable logistic regression was used to model the probability of palsy on the basis of one or more predictors. A jackknife validation was performed to internally validate the model and assess its generalizability. RESULTS A total of 98 patients fit the inclusion criteria; 12% had developed symptoms of C5 palsy postoperatively. Using the three variables in a predictor-model, we found that the odds ratio of having palsy for APD, FD, and CLA was 0.3, 0.02, and 1.4, respectively. For every 1-mm increase in APD and FD, the odds of developing palsy decrease 69% (p<.0001) and decrease 98% (p<.0003), respectively. In contrast, for every 1-degree increase in CLA, the odds of developing palsy increase by 43% (p<.0001). The receiver-operating characteristic curve for this three-variable model predicting development of palsy has an area under the curve (concordance index) of 0.97. After implementing a jackknife validation, the area under the curve was 95%. CONCLUSIONS This study is the first to use the combination of APD, FD, and CLA to predict development of postoperative C5 palsy after decompression surgery for patients with spondylotic myelopathy. This prediction formula may allow for better patient selection and to prepare patients that have an increased probability of developing this complication.

Cervical arthroplasty: a critical review of the literature

BACKGROUND CONTEXT Cervical disc arthroplasty (CDA) is a motion-preserving procedure that is an alternative to fusion. Proponents of arthroplasty assert that it will maintain cervical motion and prevent or reduce adjacent segment degeneration. Accordingly, CDA, compared with fusion, would have the potential to improve clinical outcomes. Published studies have varying conclusions on whether CDA reduces complications and/or improves outcomes. As many of these previous studies have been funded by CDA manufacturers, we wanted to ascertain whether there was a greater likelihood for these studies to report positive results. PURPOSE To critically assess the available literature on cervical arthroplasty with a focus on the time of publication and conflict of interest (COI). STUDY DESIGN/SETTING Review of the literature. METHODS All clinical articles about CDA published in English through August 1, 2013 were identified on Medline. Any article that presented CDA clinical results was included. Study design, sample size, type of disc, length of follow-up, use of statistical analysis, quality-of-life (QOL) outcome scores, COI, and complications were recorded. A meta-analysis was conducted stratifying studies by COI and publication date to identify differences in complication rates reported. RESULTS Seventy-four studies were included that investigated 8 types of disc prosthesis and 22 met the criteria for a randomized controlled trial (RCT). All Level Ib RCTs reported superior quality-of-life outcomes for CDA versus anterior cervical discectomy and fusion (ACDF) at 24 months. Fifty of the 74 articles (68%) had a disclosure section, including all Level Ib RCTs, which had significant COIs related to the respective studies. Those studies without a COI reported mean weighted average adjacent segment disease rates of 6.3% with CDA and 6.2% with ACDF. In contrast, the reverse was reported by studies with a COI, for which the averages were 2.5% with CDA and 6.3% with ACDF. Those studies with a COI (n=31) had an overall weighted average heterotopic ossification rate of 22%, whereas those studies with no COI (n=43) had a rate of 46%. CONCLUSIONS Associated COIs did not influence QOL outcomes. Conflicts of interest were more likely to be present in studies published after 2008, and those with a COI reported greater adjacent segment disease rates for ACDF than CDA. In addition, heterotopic ossification rates were much lower in studies with COI versus those without COI. Thus, COIs did not affect QOL outcomes but were associated with lower complication rates.

Risk factors, mortality, and timing of ischemic and hemorrhagic stroke with left ventricular assist devices

BACKGROUND Stroke is a major cause of mortality after left ventricular assist device (LVAD) placement. METHODS Prospectively collected data of patients with HeartMate II (n = 332) and HeartWare (n = 70) LVADs from October 21, 2004, to May 19, 2015, were reviewed. Predictors of early (during index hospitalization) and late (post-discharge) ischemic and hemorrhagic stroke and association of stroke subtypes with mortality were assessed. RESULTS Of 402 patients, 83 strokes occurred in 69 patients (17%; 0.14 events per patient-year [EPPY]): early ischemic stroke in 18/402 (4%; 0.03 EPPY), early hemorrhagic stroke in 11/402 (3%; 0.02 EPPY), late ischemic stroke in 25/402 (6%; 0.04 EPPY) and late hemorrhagic stroke in 29/402 (7%; 0.05 EPPY). Risk of stroke and death among patients with stroke was bimodal with highest risks immediately post-implant and increasing again 9-12 months later. Risk of death declined over time in patients without stroke. Modifiable stroke risk factors varied according to timing and stroke type, including tobacco use, bacteremia, pump thrombosis, pump infection, and hypertension (all p < 0.05). In multivariable analysis, early hemorrhagic stroke (adjusted odds ratio [aOR] 4.3, 95% confidence interval [CI] 1.0-17.8, p = 0.04), late ischemic stroke (aOR 3.2, 95% CI 1.1-9.0, p = 0.03), and late hemorrhagic stroke (aOR 3.7, 95% CI 1.5-9.2, p = 0.005) predicted death, whereas early ischemic stroke did not. CONCLUSIONS Stroke is a leading cause and predictor of death in patients with LVADs. Risk of stroke and death among patients with stroke is bimodal, with highest risk at time of implant and increasing risk again after 9-12 months. Management of modifiable risk factors may reduce stroke and mortality rates.

Individualized Metabolic Surgery Score: Procedure Selection Based on Diabetes Severity

Objective: To construct and validate a scoring system for evidence-based selection of bariatric and metabolic surgery procedures according to severity of type 2 diabetes (T2DM). Background: Roux-en-Y gastric bypass (RYGB) and sleeve gastrectomy (SG) account for >95% of bariatric procedures in United States in patients with T2DM. To date, there is no validated model to guide procedure selection based on long-term glucose control in patients with T2DM. Methods: A total of 659 patients with T2DM who underwent RYGB and SG at an academic center in the United States and had a minimum 5-year follow-up (2005–2011) were analyzed to generate the model. The validation dataset consisted of 241 patients from an academic center in Spain where similar criteria were applied. Results: At median postoperative follow-up of 7 years (range 5–12), diabetes remission (HbA1C <6.5% off medications) was observed in 49% after RYGB and 28% after SG (P < 0.001). Four independent predictors of long-term remission including preoperative duration of T2DM (P < 0.0001), preoperative number of diabetes medications (P < 0.0001), insulin use (P = 0.002), and glycemic control (HbA1C < 7%) (P = 0.002) were used to develop the Individualized Metabolic Surgery (IMS) score using a nomogram. Patients were then categorized into 3 stages of diabetes severity. In mild T2DM (IMS score ⩽25), both procedures significantly improved T2DM. In severe T2DM (IMS score >95), when clinical features suggest limited functional &bgr;-cell reserve, both procedures had similarly low efficacy for diabetes remission. There was an intermediate group, however, in which RYGB was significantly more effective than SG, likely related to its more pronounced neurohormonal effects. Findings were externally validated and procedure recommendations for each severity stage were provided. Conclusions: This is the largest reported cohort (n = 900) with long-term postoperative glycemic follow-up, which, for the first time, categorizes T2DM into 3 validated severity stages for evidence-based procedure selection.

The impact of preoperative depression on quality of life outcomes after posterior cervical fusion.

BACKGROUND CONTEXT Posterior cervical fusion (PCF) has been shown to be an effective treatment for cervical spondylosis, but is associated with a 9% complication rate and high costs. To limit such complications and costs, it is imperative that proper selection of surgical candidates occur for those most likely to do well with the surgery. Affective disorders, such as depression, are associated with worsened outcomes after lumbar surgery; however, this effect has not been evaluated in patients undergoing cervical spine surgery. PURPOSE To assess the predictive value of preoperative depression and the health state on 1-year quality of life (QOL) outcomes after PCF. STUDY DESIGN A retrospective cohort analysis. PATIENT SAMPLE Eighty-eight patients who underwent PCF for cervical spondylosis were reviewed. OUTCOME MEASURES Preoperative and 1-year postoperative health outcomes were assessed based on the Pain Disability Questionnaire (PDQ), the Patient Health Questionnaire-9 (PHQ-9), and the EuroQol five-dimensions (EQ-5D) questionnaire. METHODS Univariable and multivariable regression analyses were performed to assess for preoperative predictors of 1-year change in health status. RESULTS Compared with preoperative health states, the PCF cohort showed statistically significant improved PDQ (87.8 vs. 73.6), PHQ-9 (7.7 vs. 6.6), and EQ-5D (0.50 vs. 0.60) scores at 1 year postoperatively. Only 10/88 (11%) patients achieved or surpassed the minimum clinically important difference for the PHQ-9 (5). Multiple linear and logistic regression analyses showed that increasing PHQ-9 and EQ-5D preoperative scores were associated with reduced 1-year postoperative improvement in health status (EQ-5D index). CONCLUSIONS Of patients who undergo PCF, those with a greater degree of preoperative depression have lower improvements in postoperative QOL compared with those with less depression. Additionally, patients with better preoperative health states also attain lower 1-year QOL improvements.

Association of laboratory test result availability and rehospitalizations in an outpatient parenteral antimicrobial therapy programme

OBJECTIVES Laboratory tests are usually requested for monitoring during outpatient parenteral antimicrobial therapy (OPAT), but these recommendations are not always followed. The purpose of this study was to determine whether rehospitalization during the OPAT course is associated with the availability of these test results to the treating physician. METHODS Electronic health records (EHRs) from all patients in the Cleveland Clinic OPAT registry with start dates from 1 January to 28 February 2011 were reviewed in a retrospective cohort study. Comprehensive data on patient and OPAT characteristics were obtained for the first OPAT course per patient. Availability of laboratory test results was defined as documentation of results of at least one recommended test in the health systems EHR. Proportions of patients rehospitalized were compared for OPAT courses with test results available and non-available. Adjustments were made for patient age, hospital length of stay, anticipated OPAT duration, OPAT site and Charlson comorbidity index score. RESULTS Four hundred patients received OPAT during the study period; 60% at home, 36% in skilled nursing facilities or long-term acute care facilities and 4% in other settings. Recommended monitoring laboratory test results were available to infectious disease physicians in 291 (73%) OPAT episodes. There were 82 patient readmissions (21%) while on OPAT. In a multivariable logistic regression model, non-availability of recommended test results was independently associated with readmissions while on OPAT (adjusted OR 2.53; 95% CI 1.36-4.73). CONCLUSIONS Non-availability of recommended test results to treating physicians for patients on OPAT is associated with increased readmissions during OPAT.