Dealing with missing data in observational health care outcome analyses

Handling missing values in the analysis of between-hospital differences in ordinal and dichotomous outcomes: a simulation study

Missing data are frequently encountered in registries that are used to compare performance across hospitals. The most appropriate method for handling missing data when analysing differences in outcomes between hospitals with a generalised linear mixed model is unclear. We aimed to compare methods for handling missing data when comparing hospitals on ordinal and dichotomous outcomes. We performed a simulation study using data from the Multicentre Randomised Controlled Trial of Endovascular Treatment for Acute Ischaemic Stroke in the Netherlands (MR CLEAN) Registry, a prospective cohort study in 17 hospitals performing endovascular therapy for ischaemic stroke in the Netherlands. The investigated methods for handling missing data, both case-mix adjustment variables and outcomes, were complete case analysis, single imputation, multiple imputation, single imputation with deletion of imputed outcomes and multiple imputation with deletion of imputed outcomes. Data were generated as missing completely at random (MCAR), missing at random and missing not at random (MNAR) in three scenarios: (1) 10% missing data in case-mix and outcome; (2) 40% missing data in case-mix and outcome; and (3) 40% missing data in case-mix and outcome with varying degree of missing data among hospitals. Bias and reliability of the methods were compared on the mean squared error (MSE, a summary measure combining bias and reliability) relative to the hospital effect estimates from the complete reference data set. For both the ordinal outcome (ie, the modified Rankin Scale) and a common dichotomised version thereof, all methods of handling missing data were biased, likely due to shrinkage of the random effects. The MSE of all methods was on average lowest under MCAR and with fewer missing data, and highest with more missing data and under MNAR. The 'multiple imputation, then deletion' method had the lowest MSE for both outcomes under all simulated patterns of missing data. Thus, when estimating hospital effects on ordinal and dichotomous outcomes in the presence of missing data, the least biased and most reliable method to handle these missing data is 'multiple imputation, then deletion'.

Time of coronary revascularization: methodology of a mediation analysis study

BACKGROUND

The advantage of coronary artery bypass grafting (CABG) over percutaneous coronary intervention (PCI), established in trials, may not be generalizable to populations in which the method of treatment determines the time to treatment. We sought to describe the methodology of a population-based observational study for assessing how changes in time to treatment may affect the comparative effectiveness of these 2 methods of coronary revascularization.

METHODS

We propose a framework of causal mediation analysis to compare the outcomes of choosing CABG over PCI, if patients selected for either method waited the same amount of time had they undergone a PCI. We will include patients who underwent a first-time, nonurgent isolated CABG or single-session PCI for multivessel or left main coronary artery disease from January 2001 to December 2016, in British Columbia. We will use absolute risk difference as a measure of the total effect of choosing CABG over PCI and partition it into the direct effect of the treatment choice and the effect mediated by the treatment-specific timing.

INTERPRETATION

Understanding how time to treatment mediates the relation between method of revascularization and outcomes will have implications for treatment selection, resource allocation and planning benchmarks. Findings on the benefits and risks of performing PCI or CABG within a certain time will guide multidisciplinary teams in determining the appropriate revascularization method for individual patients.

Understanding and Predicting Cognitive Improvement of Young Adults in Ischemic Stroke Rehabilitation Therapy

Accurate early predictions of a patient's likely cognitive improvement as a result of a stroke rehabilitation programme can assist clinicians in assembling more effective therapeutic programs. In addition, sufficient levels of explainability, which can justify these predictions, are a crucial requirement, as reported by clinicians. This article presents a machine learning (ML) prediction model targeting cognitive improvement after therapy for stroke surviving patients. The prediction model relies on electronic health records from 201 ischemic stroke surviving patients containing demographic information, cognitive assessments at admission from 24 different standardized neuropsychology tests (e.g., TMT, WAIS-III, Stroop, RAVLT, etc.), and therapy information collected during rehabilitation (72,002 entries collected between March 2007 and September 2019). The study population covered young-adult patients with a mean age of 49.51 years and only 4.47% above 65 years of age at the stroke event (no age filter applied). Twenty different classification algorithms (from Python's Scikit-learn library) are trained and evaluated, varying their hyper-parameters and the number of features received as input. Best-performing models reported Recall scores around 0.7 and F1 scores of 0.6, showing the model's ability to identify patients with poor cognitive improvement. The study includes a detailed feature importance report that helps interpret the model's inner decision workings and exposes the most influential factors in the cognitive improvement prediction. The study showed that certain therapy variables (e.g., the proportion of memory and orientation executed tasks) had an important influence on the final prediction of the cognitive improvement of patients at individual and population levels. This type of evidence can serve clinicians in adjusting the therapeutic settings (e.g., type and load of therapy activities) and selecting the one that maximizes cognitive improvement.

Do routine hospital data accurately record comorbidity in advanced kidney disease populations? A record linkage cohort study

BACKGROUND

Routine healthcare datasets capturing clinical and administrative information are increasingly being used to examine health outcomes. The accuracy of such data is not clearly defined. We examine the accuracy of diagnosis recording in individuals with advanced chronic kidney disease using a routine healthcare dataset in England with comparison to information collected by trained research nurses.

METHODS

We linked records from the Access to Transplant and Transplant Outcome Measures study to the Hospital Episode Statistics dataset. International Classification of Diseases (ICD-10) and Office for Population Censuses and Surveys Classification of Interventions and Procedures (OPCS-4) codes were used to identify medical conditions from hospital data. The sensitivity, specificity, positive and negative predictive values were calculated for a range of diagnoses.

RESULTS

Comorbidity information was available in 96% of individuals prior to starting kidney replacement therapy. There was variation in the accuracy of individual medical conditions identified from the routine healthcare dataset. Sensitivity and positive predictive values ranged from 97.7 and 90.4% for diabetes and 82.6 and 82.9% for ischaemic heart disease to 44.2 and 28.4% for liver disease.

CONCLUSIONS

Routine healthcare datasets accurately capture certain conditions in an advanced chronic kidney disease population. They have potential for use within clinical and epidemiological research studies but are unlikely to be sufficient as a single resource for identifying a full spectrum of comorbidities.

Long-term Outcomes Following Mechanical or Bioprosthetic Aortic Valve Replacement in Young Women

Background

Studies performed to date reporting outcomes after mechanical or bioprosthetic aortic valve replacement (AVR) have largely neglected the young female population. This study compares long-term outcomes in female patients aged < 50 years undergoing AVR with either a mechanical or bioprosthetic valve.

Methods

In this propensity-matched study, we compared outcomes after mechanical AVR (n = 57) and bioprosthetic AVR (n = 57) between 2004 and 2018. The primary outcome of this study is survival. Secondary outcomes include the rate of reoperation, stroke, myocardial infarction, rehospitalization for heart failure, and incidence of serious adverse events. Outcomes were measured over 15 years, with a median follow-up of 7.8 years.

Results

In patients receiving a mechanical AVR vs a bioprosthetic AVR, overall survival at median follow-up was equivalent, at 93%. There is a lower rate of reoperation in patients receiving a mechanical AVR vs a bioprosthetic AVR (1.8% vs 8.8%). The rate of new-onset atrial fibrillation was significantly higher in the mechanical AVR group vs the bioprosthetic AVR group (18.2% vs 7.3%). No significant difference was seen in the rate of serious adverse events.

Conclusions

These results provide contemporary data demonstrating equivalent long-term survival between mechanical and bioprosthetic AVR, with higher rates of new atrial fibrillation after mechanical AVR, and higher rates of reoperation after bioprosthetic AVR. These results suggest that either valve type is safe, and that preoperative assessment and counselling, as well as the follow-up, medical treatment and indications for intervention, must be a collaborative decision-making process between the clinician and the patient.

Impact of missing data on bias and precision when estimating change in patient-reported outcomes from a clinical registry

BACKGROUND

Clinical registries, which capture information about the health and healthcare use of patients with a health condition or treatment, often contain patient-reported outcomes (PROs) that provide insights about the patient's perspectives on their health. Missing data can affect the value of PRO data for healthcare decision-making. We compared the precision and bias of several missing data methods when estimating longitudinal change in PRO scores.

METHODS

This research conducted analyses of clinical registry data and simulated data. Registry data were from a population-based regional joint replacement registry for Manitoba, Canada; the study cohort consisted of 5631 patients having total knee arthroplasty between 2009 and 2015. PROs were measured using the 12-item Short Form Survey, version 2 (SF-12v2) at pre- and post-operative occasions. The simulation cohort was a subset of 3000 patients from the study cohort with complete PRO information at both pre- and post-operative occasions. Linear mixed-effects models based on complete case analysis (CCA), maximum likelihood (ML) and multiple imputation (MI) without and with an auxiliary variable (MI-Aux) were used to estimate longitudinal change in PRO scores. In the simulated data, bias, root mean squared error (RMSE), and 95% confidence interval (CI) coverage and width were estimated under varying amounts and types of missing data.

RESULTS

Three thousand two hundred thirty (57.4%) patients in the study cohort had complete data on the SF-12v2 at both occasions. In this cohort, mixed-effects models based on CCA resulted in substantially wider 95% CIs than models based on ML and MI methods. The latter two methods produced similar estimates and 95% CI widths. In the simulation cohort, when 50% of the data were missing, the MI-Aux method, in which a single hypothetical auxiliary variable was strongly correlated (i.e., 0.8) with the outcome, reduced the 95% CI width by up to 14% and bias and RMSE by up to 50 and 45%, respectively, when compared with the MI method.

CONCLUSIONS

Missing data can substantially affect the precision of estimated change in PRO scores from clinical registry data. Inclusion of auxiliary information in MI models can increase precision and reduce bias, but identifying the optimal auxiliary variable(s) may be challenging.

Total and Cause-Specific Mortality After Percutaneous Coronary Intervention: Observations From the Alberta Provincial Project for Outcome Assessment in Coronary Heart Disease Registry

Background

Patients undergoing percutaneous coronary intervention (PCI) are increasingly older and have a higher comorbidity burden. This study evaluated trends in 30-day, 1-year, and 2-year total and cause-specific mortality using a large, contemporary cohort of patients who underwent PCI in Alberta, Canada.

Methods

We used the Alberta Provincial Project for Outcome Assessment in Coronary Heart Disease (APPROACH) registry to identify patients aged ≥ 20 years who underwent PCI between 2005 and 2013. All patients were followed until death or being censored by August 2016. Cause of death was from the Vital Statistics database and classified as cardiac or noncardiac. Multivariable logistic regression was used to calculate predicted mortality at 30 days, 1 year, and 2 years post-PCI.

Results

Of the 35,602 patients who underwent PCI, 5284 (14.8%) had died. Mean (standard deviation) follow-up was 74.9 (35.1) months. Over the study period, patients were older and more likely to undergo PCI for an acute coronary syndrome indication. Thirty-day (2005: 1.3%; 2013: 3.2%; P < 0.001), 1-year (2005: 2.7%; 2013: 5.7%; P < 0.001), and 2-year (2005: 4.5%; 2013: 7.5%; P < 0.001) predicted mortality after PCI increased over the study period. Cardiac cause of death dominated in the short-term, but the proportion of noncardiac deaths increased as time from PCI to death increased (30 days = 11.5%, 1 year = 31.5%, 2 years = 39.6%; P < 0.001).

Conclusions

In this population-based study, we found all-cause mortality at 30 days, 1 year, and 2 years after PCI increased over time. Cardiac causes of death dominate in the short-term after PCI; however, noncardiac cause becomes a major driver of mortality in the long-term.

The agreement between chronic diseases reported by patients and derived from administrative data in patients undergoing joint arthroplasty

BACKGROUND

This study examined the agreement between patient-reported chronic diseases and hospital administrative records in hip or knee arthroplasty patients in England.

METHODS

Survey data reported by 676,428 patients for the English Patient Reported Outcome Measures (PROMs) programme was linked to hospital administrative data. Sensitivity and specificity of 11 patient-reported chronic diseases were estimated with hospital administrative data as reference standard.

RESULTS

Specificity was high (> 90%) for all 11 chronic diseases. However, sensitivity varied by disease with the highest found for 'diabetes' (87.5%) and 'high blood pressure' (74.3%) and lowest for 'kidney disease' (18.8%) and 'leg pain due to poor circulation' (26.1%). Sensitivity was increased for diseases that were given as specific examples in the questionnaire (e.g. 'parkinson's disease' (65.6%) and 'multiple sclerosis' (69.5%), compared to 'diseases of the nervous system' (20.9%)).

CONCLUSIONS

Patients can give information about the presence of chronic diseases that is consistent with chronic diseases derived from hospital administrative data if the description in the patient questionnaire is precise and if the disease is familiar to most patients and has significant impact on their life. Such patient questionnaires need to be validated before they are used for research and service evaluation projects.

Expanding the impact of a longstanding Canadian cardiac registry through data linkage: challenges and opportunities

The Alberta Provincial Project for Outcome Assessment in Coronary Heart Disease (APPROACH) began as a province-wide inception cohort of all adult Alberta residents undergoing cardiac catheterization for ischemic heart disease. Strengths of the APPROACH initiative include the prospective collection of detailed clinical, procedural, and treatment information, measured at point-of-care. While this aspect of APPROACH provides data users with several advantages over use of typical administrative data, the ability to link APPROACH with data from multiple other sources has provided several unique opportunities to measure cardiovascular care and outcomes. As of June 2018, clinical information has been collected by APPROACH on over 240,000 adult Alberta residents. Linkage of this rich clinical data to administrative health data (eg. Vital statistics, hospitalizations, ambulatory events, prescription medications), secondary use clinical data (e.g. laboratory, ECG, rehabilitation, EMR, imaging) and other data sources (eg. Geospatial, crime data, meteorological) allows better study of the determinants of a patient's health trajectory. This paper describes applied examples of work that has leveraged the potential of linking several datasets with the APPROACH registry.

Statistical primer: sample size and power calculations-why, when and how?

When designing a clinical study, a fundamental aspect is the sample size. In this article, we describe the rationale for sample size calculations, when it should be calculated and describe the components necessary to calculate it. For simple studies, standard formulae can be used; however, for more advanced studies, it is generally necessary to use specialized statistical software programs and consult a biostatistician. Sample size calculations for non-randomized studies are also discussed and two clinical examples are used for illustration.

Predicting operative mortality in octogenarians for isolated coronary artery bypass grafting surgery: a retrospective study

Background

Available cardiac surgery risk scores have not been validated in octogenarians. Our objective was to compare the predictive ability of the Society of Thoracic Surgeons (STS) score, EuroSCORE I, and EuroSCORE II in elderly patients undergoing isolated coronary artery bypass grafting surgery (CABG).

Methods

All patients who underwent isolated CABG (2002 – 2008) were identified from the Alberta Provincial Project for Outcomes Assessment in Coronary Heart Disease (APPROACH) registry. All patients aged 80 and older (n = 304) were then matched 1:2 with a randomly selected control group of patients under age 80 (n = 608 of 4732). Risk scores were calculated. Discriminatory accuracy of the risk models was assessed by plotting the areas under the receiver operator characteristic (AUC) and comparing the observed to predicted operative mortality.

Results

Octogenarians had a significantly higher predicted mortality by STS Score (3 ± 2% vs. 1 ± 1%; p < 0.001), additive EuroSCORE (8 ± 3% vs. 4 ± 3%; p < 0.001), logistic EuroSCORE (15 ± 14% vs. 5 ± 6%; p < 0.001), and EuroSCORE II (4 ± 3% vs. 2 ± 2%; p < 0.001) compared to patients under age 80 years. Observed mortality was 2% and 1% for patients age 80 and older and under age 80, respectively (p = 0.323). AUC revealed areas for STS, additive and logistic EuroSCORE I and EuroSCORE II, respectively, for patients age 80 and older (0.671, 0.709, 0.694, 0.794) and under age 80 (0.829, 0.750, 0.785, 0.845).

Conclusion

All risk prediction models assessed overestimated surgical risk, particularly in octogenarians. EuroSCORE II demonstrated better discriminatory accuracy in this population. Inclusion of new variables into these risk models, such as frailty, may allow for more accurate prediction of true operative risk.

Need of informatics in designing interoperable clinical registries

Clinical registries are designed to collect information relating to a particular condition for research or quality improvement. Intuitively, informatics in the area of data management and extraction plays a central role in clinical registries. Due to various reasons such as lack of informatics awareness or expertise, there may be little informatics involvement in designing clinical registries. In this paper, we studied a clinical registry from two critical perspectives, data quality and interoperability, where informatics can play a role. We evaluated these two aspects of an existing registry, Gynecology Surgery Registry, by mapping data elements and value sets, used in the registry, to a standardized terminology, SNOMED-CT. The results showed that majority of the values are ad-hoc and only 6 of 91 procedures in the registry could be mapped to the SNOMED-CT. To tackle this issue, we assessed the feasibility of automated data abstraction process, by training machine learning classifiers, based on existing manually extracted data. These classifiers achieved a reasonable average F-measure of 0.94. We concluded that more informatics engagement is needed to improve the interoperability, reusability, and quality of the registry.

Acute coronary syndrome patients admitted to a cardiology vs non-cardiology service: variations in treatment & outcome

Background

Specialized cardiology services have contributed to reduced mortality in acute coronary syndromes (ACS).  We sought to evaluate the outcomes of ACS patients admitted to non-cardiology services in Southern Alberta.

Methods

Retrospective chart review performed on all troponin-positive patients in the Calgary Health Region identified those diagnosed with ACS by their attending team. Patients admitted to non-cardiology and cardiology services were compared, using linked data from the Alberta Provincial Project for Outcomes Assessment in Coronary Heart Disease (APPROACH) registry and the Strategic Clinical Network for Cardiovascular Health and Stroke.

Results

From January 1, 2007 to December 31, 2008, 2105 ACS patients were identified, with 1636 (77.7%) admitted to cardiology and 469 (22.3%) to non-cardiology services. Patients admitted to non-cardiology services were older, had more comorbidities, and rarely received cardiology consultation (5.1%). Cardiac catheterization was underutilized (5.1% vs 86.4% in cardiology patients (p < 0.0001)), as was evidence-based pharmacotherapy (p < 0.0001). Following adjustment for baseline comorbidities, 30-day through 4-year mortality was significantly higher on non-cardiology vs. cardiology services (49.1% vs. 11.0% respectively at 4-years, p < 0.0001).

Conclusion

In a large ACS population in the Calgary Health Region, 25% were admitted to non-cardiology services. These patients had worse outcomes, despite adjustment for baseline risk factor differences. Although many patients were appropriately admitted to non-cardiology services, the low use of investigations and secondary prevention medications may contribute to poorer patient outcome. Further research is required to identify process of care strategies to improve outcomes and lessen the burden of illness for patients and the health care system.

Quality of life following coronary artery bypass graft surgery vs. percutaneous coronary intervention in diabetics with multivessel disease: a five-year registry study

Aims

The aim of this study is to investigate the long-term relationship between revascularization technique and health status in diabetics with multivessel disease.

Methods and results

Using the Alberta Provincial Project for Outcomes Assessment in Coronary Heart Disease (APPROACH) registry, we captured 1319 diabetics with multivessel disease requiring revascularization for an acute coronary syndrome (January 2009–December 2012) and reported health status using the Seattle Angina Questionnaire (SAQ) at baseline, 1, 3 and 5-years [599 underwent coronary artery bypass grafting (CABG); 720 underwent percutaneous coronary intervention (PCI)]. Adjusted analyses were performed using a propensity score-matching technique. After adjustment (including baseline SAQ domain scores), 1-year mean (95% CI) SAQ scores (range 0–100 with higher scores reflecting improved health status) were significantly greater in selected domains for CABG compared to PCI (exertional capacity: 81.7 [79.5–84.0] vs. 78.8 [76.5–81.0], P = 0.07; angina stability: 83.1 [80.4–85.9] vs. 75.0 [72.3–77.8], P &lt; 0.001]; angina frequency 93.2 [91.6–95.0] vs. 90.0 [87.8–91.3], P = 0.003; treatment satisfaction: 93.6 [92.2–94.9] vs. 90.8 [89.2–92.0], P = 0.003; quality of life [QOL]: 83.8 [81.7–85.8] vs. 77.2 [75.2–79.2] P &lt; 0.001). At 3-years, these benefits were attenuated (exertional capacity: 79.3 [76.9–81.7] vs. 78.7 [76.3–81.1], P = 0.734; angina stability 79.3 [76.3–82.3] vs. 75.5 [72.5–78.5], P = 0.080; angina frequency: 93.2 [91.3–95.1] vs. 90.9 [89.0–92.8], P = 0.095; treatment satisfaction: 92.5 [91.0–94.0] vs. 91.5 [90.0–93.0] P = 0.382; QOL: 83.2 [81.1–85.2] vs. 80.3 [78.2–82.4], P = 0.057). At 5-years, majority of domains were similar (exertional capacity: 77.8 [75.0–80.6] vs. 76.3 [73.2–79.3], P = 0.482; angina stability: 78.0 [74.8–81.2] vs. 74.8 [71.4–78.2], P = 0.175; angina frequency: 94.2 [92.3–96.0] vs. 90.9 [89.0–92.9], P = 0.018; treatment satisfaction: 93.7 [92.2–95.1] vs. 92.2 [90.6–93.7], P = 0.167; QOL: 84.1 [82.0–86.3] vs. 81.1 [78.8–83.4], P = 0.058). Majority in both groups remained angina-free at 5-years (75.0% vs. 70.3%, P = 0.15).

Conclusion

Improvements in health status with CABG compared with PCI were not sustained long-term. This temporal sequence should be considered when contemplating a revascularization strategy in diabetics with multivessel disease.

Body Mass Index Is Associated With Differential Rates of Coronary Revascularization After Cardiac Catheterization

BACKGROUND

The association of obesity with coronary revascularization procedures is not clear. We examined rates of coronary artery bypass grafting (CABG) and percutaneous coronary intervention (PCI) associated with obesity while accounting for the severity of coronary disease and diabetes status.

METHODS

Patients who underwent cardiac catheterization were stratified according to coronary anatomy risks and diabetes status. Within each stratum, using normal body mass index (BMI) (18.5-24.9 kg/m²) as a reference, the associations of overweight (25.0-29.9 kg/m²), obese class I (30.0-34.9 kg/m²), obese class II (35.0-39.9 kg/m²), and obese class III (≥ 40.0 kg/m²) with the likelihood of receiving CABG and PCI were assessed while adjusting for clinical covariates.

RESULTS

Of 56,722 patients analyzed, overall use of revascularization was higher in the overweight, obese class I, and obese class II groups (overweight: adjusted hazard ratio [aHR], 1.10; 95% confidence interval [CI], 1.06-1.13; obese class I: aHR, 1.08; 95% CI, 1.05-1.12; obese class II: aHR,1.05; 95% CI, 1.01-1.10), whereas it was lower in the obese class III group (aHR, 0.91; 95% CI, 0.85-0.97) compared with normal BMI. In the subgroup with high-risk coronary anatomy and diabetes, all obese classes had higher rates of PCI (obese class I: aHR,1.24; 95% CI, 1.08-1.42; obese class II: aHR,1.27; 95% CI, 1.07-1.49, obese class III: aHR,1.37; 95% CI, 1.12-1.67) than the normal BMI group.

CONCLUSIONS

Our results showed that BMI is associated with differential rates of coronary revascularization. In patients with high-risk coronary anatomy and diabetes, clinical appropriateness of higher rates of PCI associated with obesity warrants further investigation.

A pseudo-random patient sampling method evaluated

OBJECTIVES

To compare two human immunodeficiency virus (HIV) cohorts to determine whether a pseudo-random sample can represent the entire study population.

STUDY DESIGN AND SETTING

HIV-positive patients receiving care at eight sites in seven Asian countries. The TREAT Asia HIV Observational database (TAHOD) pseudo-randomly selected a patient sample, while TREAT Asia HIV Observational database-Low Intensity Transfer (TAHOD-LITE) included all patients. We compared patient demographics, CD4 count, and HIV viral load testing for each cohort. Risk factors associated with CD4 count response, HIV viral load suppression (<400 copies/mL), and survival were determined for each cohort.

RESULTS

There were 2,318 TAHOD patients and 14,714 TAHOD-LITE patients. Patient demographics, CD4 count, and HIV viral load testing rates were broadly similar between the cohorts. CD4 count response and all-cause mortality were consistent among the cohorts with similar risk factors. HIV viral load response appeared to be superior in TAHOD and many risk factors differed, possibly due to viral load being tested on a subset of patients.

CONCLUSION

Our study gives the first empirical evidence that analysis of risk factors for completely ascertained end points from our pseudo-randomly selected patient sample may be generalized to our larger, complete population of HIV-positive patients. However, results can significantly vary when analyzing smaller or pseudo-random samples, particularly if some patient data are not completely missing at random, such as viral load results.

Clinical and angiographic outcomes associated with surgical revascularization of angiographically borderline 50-69% coronary artery stenoses

OBJECTIVES

Coronary artery bypass grafting (CABG) improves outcomes in patients with multivessel coronary artery disease. Bypass of angiographically significant lesions ≥70% is recommended, yet little is known about the incidence/outcomes with bypasses of 50-69% angiographically borderline lesions (ABLs) without fractional flow reserve testing. The objective of this study was to investigate the incidence and outcomes of bypass of 50-69% ABLs.

METHODS

Between 2007 and 2013, 3195 patients underwent isolated first multivessel CABG. Patients with an isolated ABL of a major epicardial vessel were included. Outcomes of interest included time to all-cause mortality, and 30-day and 1-year mortality.

RESULTS

Among 350 patients with an ABL, 268 (76.6%) had the vessel containing the ABL bypassed, while 82 (23.4%) did not. The mean follow-up was 4.2 years. Patients with a bypassed ABL were older (66.1 vs 62.5 mean years, P = 0.006) but otherwise similar in sex, comorbidities, diabetes, ejection fraction and number of coronary stenoses. Cardiopulmonary bypass time was longer in patients with bypassed ABLs (104.2 vs 90.4 min, mean, P < 0.001). Unadjusted overall mortality until the end of follow-up was higher among patients with bypassed ABLs (11.6 vs 3.7%, P = 0.034). After multivariable adjustment, the association between ABL bypass and mortality was attenuated (hazard ratio 2.84, 95% confidence interval: 0.87-9.23, P = 0.080). No differences were observed in unadjusted 30-day (1.1 vs 0.0%, P = 0.336) or 1-year mortality (4.1 vs 0.0%, P = 0.062). Repeat revascularization rate of patients with bypassed ABLs was numerically higher (4.1 vs 0.0%, P = 0.107).

CONCLUSIONS

In an unselected cohort of patients with ABLs, bypass of borderline 50-69% lesions is frequently performed and not associated with improved long-term survival. Our findings suggest that the routine surgical revascularization of 50-69% ABLs may not be warranted.

Characterization of Missing Data in Clinical Registry Studies

Patterns of missing data are seldom well-characterized in observational research. This study examined the magnitude of, and factors associated with, missing data across multiple observational studies. Missingness was evaluated for demographic, clinical, and patient-reported outcome (PRO) data from a procedure registry (TOPS), a rare disease (cystic fibrosis) registry (Port-CF), and a comparative effectiveness registry (glaucoma, RiGOR). Generalized linear mixed effects models were fit to assess whether patient characteristics or follow-up methods predicted missingness. Data from 156,707 surgical procedures, 32,118 cystic fibrosis patients, and 2373 glaucoma patients were analyzed. Data were rarely missing for demographics, treatments, and outcomes. Missingness for clinical variables varied by registry and measure and depended on whether a variable was required. Within RiGOR, PRO forms were missing more often when collected by e-mail compared with office-based paper data collection. In Port-CF, missingness varied based on insurance status and sex. Strategic consideration of operational approaches affecting missing data should be performed prior to data collection and assessed periodically during study conduct.

Cardiac rehabilitation completion is associated with reduced mortality in patients with diabetes and coronary artery disease

AIMS

Cardiac rehabilitation (CR) reduces the risks of mortality and hospitalisation in patients with coronary artery disease and without diabetes. It is unknown whether patients with diabetes obtain the same benefits from CR.

METHODS

We retrospectively examined patients referred to a 12 week CR programme between 1996 and 2010. Associations between CR completion vs non-completion and death, hospitalisation rate and cardiac hospitalisation rate were assessed by survival analysis.

RESULTS

Over the study period, 13,158 participants were referred to CR (mean ± SD, age 59.9 ± 11.1 years, 28.9% female, 2,956 [22.5%] with diabetes). Patients with diabetes were less likely to complete CR than those without diabetes (41% vs 56%, p < .0001). Over a median follow-up of 6.6 years, there were 379 deaths in patients with diabetes vs 941 deaths among those without diabetes (12.8% vs 8.9%). Of the non-completers, patients with diabetes had a higher mortality rate compared with those without diabetes (17.7% vs 11.3%). In patients who completed CR, mortality was lower: 11.1% in patients with diabetes vs 7.0% in those without diabetes. In patients with diabetes, CR completion was associated with reduced mortality (HR 0.46 [95% CI 0.37, 0.56]), reduced hospitalisation (HR 0.86 [95% CI 0.76, 0.96]) and reduced cardiac hospitalisation (HR 0.67 [95% CI 0.54, 0.84]). The protective associations were similar to those of patients without diabetes. In multivariable adjusted analyses, all of these associations remained significant.

CONCLUSIONS

Patients with diabetes were less likely to complete CR than those without diabetes. However, patients with diabetes who completed CR derived similar apparent reductions in mortality and hospitalisation to patients without diabetes.

Predicting cardiovascular intensive care unit readmission after cardiac surgery: derivation and validation of the Alberta Provincial Project for Outcomes Assessment in Coronary Heart Disease (APPROACH) cardiovascular intensive care unit clinical prediction model from a registry cohort of 10,799 surgical cases

Introduction

In medical and surgical intensive care units, clinical risk prediction models for readmission have been developed; however, studies reporting the risks for cardiovascular intensive care unit (CVICU) readmission have been methodologically limited by small numbers of outcomes, unreported measures of calibration or discrimination, or a lack of information spanning the entire perioperative period. The purpose of this study was to derive and validate a clinical prediction model for CVICU readmission in cardiac surgical patients.

Methods

A total of 10,799 patients more than or equal to 18 years in the Alberta Provincial Project for Outcomes Assessment in Coronary Heart Disease (APPROACH) registry who underwent cardiac surgery (coronary artery bypass or valvular surgery) between 2004 and 2012 and were discharged alive from the first CVICU admission were included. The full cohort was used to derive the clinical prediction model and the model was internally validated with bootstrapping. Discrimination and calibration were assessed using the AUC c index and the Hosmer-Lemeshow tests, respectively.

Results

A total of 479 (4.4%) patients required CVICU readmission. The mean CVICU length of stay (19.9 versus 3.3 days, P <0.001) and in-hospital mortality (14.4% versus 2.2%, P <0.001) were higher among patients readmitted to the CVICU. In the derivation cohort, a total of three preoperative (age ≥70, ejection fraction, chronic lung disease), two intraoperative (single valve repair or replacement plus non-CABG surgery, multivalve repair or replacement), and seven postoperative variables (cardiac arrest, pneumonia, pleural effusion, deep sternal wound infection, leg graft harvest site infection, gastrointestinal bleed, neurologic complications) were independently associated with CVICU readmission. The clinical prediction model had robust discrimination and calibration in the derivation cohort (AUC c index = 0.799; Hosmer-Lemeshow P = 0.192). The validation point estimates and confidence intervals were similar to derivation model.

Conclusions

In a large population-based dataset incorporating a comprehensive set of perioperative variables, we have derived a clinical prediction model with excellent discrimination and calibration. This model identifies opportunities for targeted therapeutic interventions aimed at reducing CVICU readmissions in high-risk patients.

Standards should be applied in the prevention and handling of missing data for patient-centered outcomes research: a systematic review and expert consensus

OBJECTIVES

To recommend methodological standards in the prevention and handling of missing data for primary patient-centered outcomes research (PCOR).

STUDY DESIGN AND SETTING

We searched National Library of Medicine Bookshelf and Catalog as well as regulatory agencies' and organizations' Web sites in January 2012 for guidance documents that had formal recommendations regarding missing data. We extracted the characteristics of included guidance documents and recommendations. Using a two-round modified Delphi survey, a multidisciplinary panel proposed mandatory standards on the prevention and handling of missing data for PCOR.

RESULTS

We identified 1,790 records and assessed 30 as having relevant recommendations. We proposed 10 standards as mandatory, covering three domains. First, the single best approach is to prospectively prevent missing data occurrence. Second, use of valid statistical methods that properly reflect multiple sources of uncertainty is critical when analyzing missing data. Third, transparent and thorough reporting of missing data allows readers to judge the validity of the findings.

CONCLUSION

We urge researchers to adopt rigorous methodology and promote good science by applying best practices to the prevention and handling of missing data. Developing guidance on the prevention and handling of missing data for observational studies and studies that use existing records is a priority for future research.

Dynamic prediction modeling approaches for cardiac surgery

BACKGROUND

The calibration of several cardiac clinical prediction models has deteriorated over time. We compare different model fitting approaches for in-hospital mortality after cardiac surgery that adjust for cross-sectional case mix in a heterogeneous patient population.

METHODS AND RESULTS

Data from >300 000 consecutive cardiac surgery procedures performed at all National Health Service and some private hospitals in England and Wales between April 2001 and March 2011 were extracted from the National Institute for Cardiovascular Outcomes Research clinical registry. The study outcome was in-hospital mortality. Model approaches included not updating, periodic refitting, rolling window, and dynamic logistic regression. Covariate adjustment was made in each model using variables included in the logistic European System for Cardiac Operative Risk Evaluation model. The association between in-hospital mortality and some variables changed with time. Notably, the intercept coefficient has been steadily decreasing during the study period, consistent with decreasing observed mortality. Some risk factors, such as operative urgency and postinfarct ventricular septal defect, have been relatively stable over time, whereas other risk factors, such as left ventricular function and surgery on the thoracic aorta, have been associated with lower risk relative to the static model.

CONCLUSIONS

Dynamic models or periodic model refitting is necessary to counteract calibration drift. A dynamic modeling framework that uses contemporary and available historic data can provide a continuously smooth update mechanism that also allows for inferences to be made on individual risk factors. Better models that withstand the effects of time give advantages for governance, quality improvement, and patient-level decision making.

Alcohol and drug use disorders among patients with myocardial infarction: associations with disparities in care and mortality

BACKGROUND

Because alcohol and drug use disorders (SUDs) can influence quality of care, we compared patients with and without SUDs on frequency of catheterization, revascularization, and in-hospital mortality after acute myocardial infarction (AMI).

METHODS

This study employed hospital discharge data identifying all adult AMI admissions (ICD-9-CM code 410) between April 1996 and December 2001. Patients were classified as having an SUD if they had alcohol and/or drug (not nicotine) abuse or dependence using a validated ICD-9-CM coding definition. Catheterization and revascularization data were obtained by linkage with a clinically-detailed cardiac registry. Analyses (controlling for comorbidities and disease severity) compared patients with and without SUDs for post-MI catheterization, revascularization, and in-hospital mortality.

RESULTS

Of 7,876 AMI unique patient admissions, 2.6% had an SUD. In adjusted analyses mortality was significantly higher among those with an SUD (odds ratio (OR) 2.02; 95%CI: 1.10-3.69), while there was a trend toward lower catheterization rates among those with an SUD (OR 0.75; 95%CI: 0.55-1.01). Among the subset of AMI admissions who underwent catheterization, the adjusted hazard ratio for one-year revascularization was 0.85 (95%CI: 0.65-1.11) with an SUD compared to without.

CONCLUSIONS

Alcohol and drug use disorders are associated with significantly higher in-hospital mortality following AMI in adults of all ages, and may also be associated with decreased access to catheterization and revascularization. This higher mortality in the face of poorer access to procedures suggests that these individuals may be under-treated following AMI. Targeted efforts are required to explore the interplay of patient and provider factors that underlie this finding.

Renal outcomes associated with invasive versus conservative management of acute coronary syndrome: propensity matched cohort study

OBJECTIVES

To examine the association of early invasive management of acute coronary syndrome with adverse renal outcomes and survival, and to determine whether the risks or benefits of early invasive management differ in people with pre-existing chronic kidney disease.

DESIGN

Propensity score matched cohort study.

SETTING

Acute care hospitals in Alberta, Canada, 2004-09.

PARTICIPANTS

10,516 adults with non-ST elevation acute coronary syndrome.

INTERVENTIONS

Participants were stratified by baseline estimated glomerular filtration rate and matched 1:1 on their propensity score for early invasive management (coronary catheterisation within two days of hospital admission).

MAIN OUTCOME MEASURES

Risks of acute kidney injury, kidney injury requiring dialysis, progression to end stage renal disease, and all cause mortality were compared between those who received early invasive treatment versus conservative treatment.

RESULTS

Of 10,516 included participants, 4276 (40.7%) received early invasive management. After using propensity score methods to assemble a matched cohort of conservative management participants with characteristics similar to those who received early invasive management (n=6768), early invasive management was associated with an increased risk of acute kidney injury (10.3% v 8.7%, risk ratio 1.18, 95% confidence interval 1.03 to 1.36; P=0.019), but no difference in the risk of acute kidney injury requiring dialysis (0.4% v 0.3%, 1.20, 0.52 to 2.78; P=0.670). Over a median follow-up of 2.5 years, the risk of progression to end stage renal disease did not differ between the groups (0.3 v 0.4 events per 100 person years, hazard ratio 0.91, 95% confidence interval 0.55 to 1.49; P=0.712); however, early invasive management was associated with reduced long term mortality (2.4 v 3.4 events per 100 person years, 0.69, 0.58 to 0.82; P<0.001). These associations were consistent among people with pre-existing reduced estimated glomerular filtration rate and with alternate definitions for early invasive management.

CONCLUSIONS

Compared with conservative management, early invasive management of acute coronary syndrome is associated with a small increase in risk of acute kidney injury but not dialysis or long term progression to end stage renal disease.

Outcomes of after-hours versus regular working hours primary percutaneous coronary intervention for acute myocardial infarction

Background

Primary percutaneous coronary intervention (PCI) is a proven therapy for acute ST-segment elevation myocardial infarction. However, outcomes associated with primary PCI may differ depending on time of day.

Methods and results

Using the Alberta Provincial Project for Outcomes Assessment in Coronary Heart Disease, a clinical data-collection initiative capturing all cardiac catheterisation patients in Alberta, Canada, the authors described and compared crude and risk-adjusted survival for ST-segment elevation myocardial infarction patients undergoing primary PCI after-hours versus regular working hours. From 1 January 1999 to 31 March 2006, 1664 primary PCI procedures were performed (54.4% after-hours). Mortalities at 30 days were 3.6% for regular hours procedures and 5.0% for after-hours procedures (p=0.16). 1-year mortalities were 6.2% and 7.3% in the regular hours and after-hours groups, respectively (p=0.35). After adjusting for baseline risk factor differences, HRs for after-hours mortality were 1.26 (95% CI 0.78 to 2.02) for survival to 30 days and 1.08 (0.73 to 1.59) for survival to 1 year. A meta-analysis of our after-hours HR point estimate with other published risk estimates for after hours primary PCI outcomes yielded an RR of 1.23 (1.00 to 1.51) for shorter-term outcomes.

Conclusions

After-hours primary PCI was not associated with a statistically significant increase in mortality. However, a meta-analysis of this study with other published after-hours outcome studies yields an RR that leaves some questions about unexplored factors that may influence after-hours primary PCI care.

Methods for using data abstracted from medical charts to impute longitudinal missing data in a clinical trial

OBJECTIVE

To describe a method for imputing missing follow-up blood pressure data in a clinical hypertension trial using blood pressures abstracted from medical charts.

METHODS

We tested a two-step method. In the first, a longitudinal mixed-effects model was estimated on blood pressures abstracted from medical charts. In the second, the patient-specific fitted values from this model at follow-up were used to impute blood pressures missing at follow-up in the trial. Simulations that imposed alternative missing data mechanisms on observed trial data were used to compare this approach to imputation approaches that do not incorporate data from charts.

RESULTS

For data that are missing at random, incorporating the fitted values from chart-based longitudinal models leads to estimates of the trial-based blood pressures that are unbiased and have lower mean squared deviation than do blood pressures imputed without the chart-based data. For data that are missing not at random, incorporating fitted values ameliorates but does not eliminate the inherent missing data bias.

CONCLUSIONS

Incorporating chart data into an imputation algorithm via the use of longitudinal mixed-effects model is an efficient way to impute longitudinal data that are missing from a randomized trial.

Exploring the sources of 'missingness' in brain tissue monitoring datasets: an observational cohort study

OBJECTIVE

To investigate the frequency/category of missing electronic data in export files from patients admitted to the intensive care unit (ICU) with severe traumatic brain injury (TBI).

RESEARCH DESIGN

Observational cohort study.

METHODOLOGY

Patient data-streams for brain temperature (T(br)) and intracranial pressure (ICP) were analysed. Missing data was classified as: missing completely at random (MCAR), missing at random (MAR), missing not at random (MNAR).

RESULTS

Sixty-two patients were studied; 60% of missing T(br) and ICP data events were attributed to electronic data acquisition 'faults'. Missing data rate ranged from 9-43% (median 25%). Cross-reference of missing data to clinical observation sheets and medical case notes shows that disconnection of sensors from monitors during critical events are common. Conclusions and implications for further research: Of concern for clinical management of patients with TBI is the detection of sudden changes in ICP or brain temperature; the parameters which forewarn of impending intracranial catastrophes. Missing data occurred at critical times of the patients stay in the ICU. This work should alert research clinicians of the need to scrutinize monitored physiological data to establish the percentage of missing values in order to obviate bias in the interpretation of results.

Patterns and effects of missing comorbidity data for patients starting renal replacement therapy in England, Wales and Northern Ireland

BACKGROUND

Renal Registries play a key role in assessing quality of care and outcomes of renal replacement therapy and comparisons of outcomes between groups should adjust for differences in comorbidities. This study aimed to describe patterns of missing comorbidity data and differences in survival between patients with comorbidity data returned and those with missing comorbidity data.

METHODS

Trends in comorbidity data returns by year (1998-2006) and within centres were examined using descriptive statistics. Survival of patients was described using Kaplan-Meier graphs (log-rank tests) and hazard ratios were calculated using Cox proportional hazard models. Last follow-up was at 31 December 2007. A range of sensitivity analyses were carried out, including multiple imputation.

RESULTS

Among 34,059 patients, there were 62% who had no comorbidity data. The completeness of comorbidity data increased markedly from 17% in 1998 to 47% in 2003, but had fallen back to 37% by the year 2006. Those with a missing comorbidity generally do considerably worse than those without the comorbidity and in most cases more closely follow the survival curve of those with the comorbidity. Multiple imputation analysis suggested that those with missing information on comorbidity have higher prevalence of comorbidity than seen in those with available data. Treating missing comorbidity entries as indication of absent comorbidity (i.e. a tick only if yes policy) would lead to an attenuation of the effect of comorbidity on survival.

CONCLUSIONS

Missing data lead to difficulties in performing between centre comparisons. A 'tick if present policy' in comorbidity data collection should be discouraged. Much more work is needed to fully understand why levels of missing comorbidity data are so high and to identify strategies to improve recording.

Quality of data collection in a large HIV observational clinic database in sub-Saharan Africa: implications for clinical research and audit of care

Background

Observational HIV clinic databases are now widely used to answer key questions related to HIV care and treatment, but there has been no systematic evaluation of their quality of data. Our objective was to evaluate the completeness and accuracy of recording of key data HIV items in a large routine observational HIV clinic database.

Methods

We looked at the number and rate of opportunistic infections (OIs) per 100 person years at risk in the 24 months following antiretroviral therapy (ART) initiation in 559 patients who initiated ART in 2004-2005 and enrolled into a research cohort. We compared this with data in a routine clinic database for the same 559 patients, and a further 1233 patients who initiated ART in the same period. The Research Cohort database was considered as the reference "gold standard" for the assessment of data accuracy. A crude percentage of underreporting of OIs in the clinic database was calculated based on the difference between the OI rates reported in both databases.

We reviewed 100 clinic patient medical records to assess the accuracy of recording of key data items of OIs, ART toxicities and ART regimen changes.

Results

The overall incidence rate per 100 person years at risk for the initial OI in the 559 patients in the research cohort and clinic databases was 24.1 (95% CI: 20.5-28.2) and 13.2 (95% CI: 10.8-16.2) respectively, and 10.4 (95% CI: 9.1-11.9) for the 1233 clinic patients. This represents a 1.8- and 2.3-fold higher rate of events in the research cohort database compared with the same 599 patients and 1233 patients in the routine clinic database, or a 45.1% and 56.8% rate of underreporting, respectively. The combined error rate of missing and incorrect items from the medical records' review was 67% for OIs, 52% for ART-related toxicities, and 83% and 58% for ART discontinuation and modification, respectively.

Conclusions

There is a high rate of underreporting of OIs in a routine HIV clinic database. This has important implications for the use and interpretation of routine observational databases for research and audit, and highlights the need for regular data validation of these databases.

Time of day and outcomes of nonurgent percutaneous coronary intervention performed during working hours

BACKGROUND

During daytime working hours, outcomes may be worse when percutaneous coronary intervention (PCI) is performed later in the day because of operator fatigue and differences in process of care.

METHODS

Using the APPROACH database, we analyzed 2,492 consecutive nonurgent PCI procedures performed during working hours. Patients undergoing PCI for acute coronary syndromes were excluded. Patients were separated into 2 groups based on whether PCI was started in the morning (7:00 am-12:00 pm, n = 1,446) or after noon (12:01 pm-6:00 pm, n = 1,037). Outcomes included procedural complications; target vessel revascularization (TVR); and death at 7 days, 30 days, and 1 year.

RESULTS

Patients undergoing PCI in the afternoon were more likely to have heart failure, reduced ejection fraction, and Canadian Cardiovascular Society class IV or atypical angina symptoms; more likely to be inpatients; less likely to have stable angina; and less likely to receive glycoprotein IIb/IIIa inhibitors. Patients undergoing PCI in the afternoon had significantly higher unadjusted rates of the composite of death and TVR at 7 days (0.9% vs 0.3%, P = .04) and 30 days (2.0% vs 1.0%, P = .04) and death at 1 year (2.2% vs 1.1%, P = .03) compared with PCI performed in the morning. After multivariate adjustment, the differences in the composite of death and TVR at 30 days and at 1 year were not statistically significant.

CONCLUSION

Patients undergoing nonurgent PCI during working hours after noon had higher rates of TVR in the first 30 days and death at 1 year. Further study is required to determine whether patient characteristics, operator fatigue, differences in process of care, or a combination of these factors accounts for the difference in outcomes.

Association of a cancer diagnosis with vulnerability and frailty in older Medicare beneficiaries

BACKGROUND

Few studies have evaluated the independent effect of a cancer diagnosis on vulnerability and frailty, which have been associated with adverse health outcomes in older adults.

METHODS

We used data in the 2003 Medicare Current Beneficiary Survey from a nationally representative sample of 12,480 community-dwelling elders. Multivariable logistic regression models were used to evaluate whether cancer was independently associated with vulnerability and frailty. Measures of vulnerability and frailty included disability, geriatric syndromes, self-rated health, and scores on two assessment tools for elderly cancer patients-the Vulnerable Elders Survey-13 (VES-13) and the Balducci frailty criteria. All statistical tests were two-sided.

RESULTS

Diagnosis of a non-skin cancer was reported by 18.8% of the respondents. Compared with respondents without a cancer history, respondents with a personal history of cancer had a statistically significantly higher prevalence of limitations in activities of daily living (31.9% vs 26.9%), limitations in instrumental activities of daily living (49.5% vs 42.3%), geriatric syndromes (60.8% vs 53.9%), low self-rated health (27.4% vs 20.9%), score of 3 or higher on the VES-13 (45.8% vs 39.5%), and satisfying criteria for frailty as defined by Balducci (79.6% vs 73.4%) (P < .001 for all characteristics). After adjustment for confounders, a cancer diagnosis was found to be associated with low self-rated health (adjusted odds ratio [OR] = 1.46, 95% confidence interval [CI] = 1.30 to 1.64; relative risk [RR] = 1.33), limitations in activities of daily living (adjusted OR = 1.19, 95% CI = 1.06 to 1.33; RR = 1.13), limitations in instrumental activities of daily living (adjusted OR = 1.25, 95% CI = 1.13 to 1.38; RR = 1.13), a geriatric syndrome (adjusted OR = 1.27, 95% CI = 1.15 to 1.41; RR = 1.11), VES-13 score of 3 or higher (adjusted OR = 1.26, 95% CI = 1.13 to 1.41; RR = 1.14), and frailty (adjusted OR = 1.46, 95% CI = 1.29 to 1.65; RR = 1.09) as defined by Balducci criteria.

CONCLUSION

Diagnosis of a non-skin cancer was associated with increased levels of having disability, having geriatric syndromes, and meeting criteria for vulnerability and frailty.

The relationship between body mass index, treatment, and mortality in patients with established coronary artery disease: a report from APPROACH

Aims Our objective was to examine the association between body mass index (BMI) and survival according to the type of treatment in individuals with established coronary artery disease (CAD). Methods and results Patients with CAD were identified in the Alberta Provincial Project for Outcome Assessment in Coronary Heart Disease (APPROACH) registry between January 2001 and March 2006. Analyses were conducted separately by treatment strategy [medical management only, percutaneous coronary intervention (PCI), or coronary artery bypass grafting (CABG)]. Patients were grouped according to six BMI categories. Multivariable-adjusted hazard ratios (HRs) for mortality were calculated using the Cox regression with the referent group for all analyses being normal BMI (18.5-24.9 kg/m(2)). The cohort included 31 021 patients with a median follow-up time of 46 months. In the medically managed only group, BMIs of 25.0-29.9 and 30.0-34.9 kg/m(2) were associated with significantly lower mortality compared with normal BMI patients (adjusted HR 0.72; 95% CI 0.63-0.83 and adjusted HR 0.82; 95% CI 0.69.0-0.98, respectively). In the CABG group, BMI of 30.0-34.9 kg/m(2) had the lowest risk of mortality (adjusted HR 0.75; 95% CI 0.61-0.94), whereas in the PCI group, BMI of 35.0-39.9 kg/m(2) had the lowest risk of mortality (adjusted HR 0.65; 95% CI 0.47-0.90). Patients who were overweight or have mild or moderate obesity were also more likely to undergo revascularization procedures compared with those with normal BMI, despite having lower risk coronary anatomy. Conclusion A paradoxical association between BMI and survival exists in patients with established CAD irrespective of treatment strategy. Patients with obesity may be presenting earlier and receiving more aggressive treatment compared with those with normal BMI.

Data enhancement for co-morbidity measurement among patients referred for sleep diagnostic testing: an observational study

Background

Observational outcome studies of patients with obstructive sleep apnea (OSA) require adjustment for co-morbidity to produce valid results. The aim of this study was to evaluate whether the combination of administrative data and self-reported data provided a more complete estimate of co-morbidity among patients referred for sleep diagnostic testing.

Methods

A retrospective observational study of 2149 patients referred for sleep diagnostic testing in Calgary, Canada. Self-reported co-morbidity was obtained with a questionnaire; administrative data and validated algorithms (when available) were also used to define the presence of these co-morbid conditions within a two-year period prior to sleep testing.

Results

Patient self-report of co-morbid conditions had varying levels of agreement with those derived from administrative data, ranging from substantial agreement for diabetes (κ = 0.79) to poor agreement for cardiac arrhythmia (κ = 0.14). The enhanced measure of co-morbidity using either self-report or administrative data had face validity, and provided clinically meaningful trends in the prevalence of co-morbidity among this population.

Conclusion

An enhanced measure of co-morbidity using self-report and administrative data can provide a more complete measure of the co-morbidity among patients with OSA when agreement between the two sources is poor. This methodology will aid in the adjustment of these coexisting conditions in observational studies in this area.