Combining ecological and individual variables to reduce confounding by indication: case study--subarachnoid hemorrhage treatment

Practice Patterns and Outcomes of Potassium Repletion Thresholds during Critical Illness

Rationale: Potassium repletion is common in critically ill patients. However, practice patterns and outcomes related to different intensive care unit (ICU) potassium repletion strategies are unclear. Objectives: 1) Describe potassium repletion practices in critically ill adults; 2) compare the effectiveness of potassium repletion strategies; and 3) compare effectiveness and safety of specific potassium repletion thresholds on patient outcomes. Methods: This was a retrospective analysis of the PINC AI Healthcare Database (2016-2022), including all critically ill adults admitted to an ICU on Hospital Day 1 and with a serum potassium concentration measured on Hospital Day 2. We determined the frequency of potassium repletion (any formulation) at each measured serum potassium concentration in each ICU, then classified ICUs as having threshold-based (a large increase in potassium repletion rates at a specific serum potassium concentration) or probabilistic (linear relationship between serum concentration and the repletion probability) patterns of repletion. Between patients in threshold-based and probabilistic repletion ICUs, we compared outcomes (primary outcome: potassium repletion frequency). We reported unadjusted percentages per exposure group and the adjusted odds ratios (from hierarchical regression models) for each outcome. Among patients in threshold-based ICUs with the most common repletion thresholds (3.5 mEq/L and 4.0 mEq/L), we conducted regression discontinuity analyses to examine the effectiveness of potassium repletion at each potassium threshold. Results: We included 190,490 patients in 88 ICUs; 35.0% received at least one dose of potassium on the same calendar day. Rates of potassium repletion were similar between 22 threshold-based strategy ICUs (33.5%) and 22 probabilistic strategy ICUs (36.4%). There was no difference in the adjusted risk of potassium repletion between patients admitted to threshold-based strategy ICUs versus probabilistic strategy ICUs (adjusted odds ratio, 1.09; 95% confidence interval [CI], 0.76-1.57). In regression discontinuity analysis, crossing the 3.5 mEq/L threshold from high to low potassium levels resulted in a 39.1% (95% CI, 23.7-42.4) absolute increase in potassium repletion but no change in other outcomes. Similarly, crossing the 4.0 mEq/L threshold resulted in a 36.4% (95% CI, 22.4-42.2) absolute increase in potassium repletion but no change in other outcomes. Conclusions: Potassium repletion is common in critically ill patients and occurs over a narrow range of "normal" potassium levels (3.5-4.0 mEq/L); use of a threshold-based repletion strategy to guide potassium repletion in ICU patients is not associated with clinically meaningful differences in outcomes.

Variation in Sedative and Analgesic Use During the COVID-19 Pandemic and Associated Outcomes

BACKGROUND

Providing analgesia and sedation is an essential component of caring for many mechanically ventilated patients. The selection of analgesic and sedative medications during the COVID-19 pandemic, and the impact of these sedation practices on patient outcomes, remain incompletely characterized.

RESEARCH QUESTION

What were the hospital patterns of analgesic and sedative use for patients with COVID-19 who received mechanical ventilation (MV), and what differences in clinical patient outcomes were observed across prevailing sedation practices?

STUDY DESIGN AND METHODS

We conducted an observational cohort study of hospitalized adults who received MV for COVID-19 from February 2020 through April 2021 within the Society of Critical Care Medicine Discovery Viral Infection and Respiratory Illness Universal Study (VIRUS) COVID-19 Registry. To describe common sedation practices, we used hierarchical clustering to group hospitals based on the percentage of patients who received various analgesic and sedative medications. We then used multivariable regression models to evaluate the association between hospital analgesia and sedation cluster and duration of MV (with a placement of death [POD] approach to account for competing risks).

RESULTS

We identified 1,313 adults across 35 hospitals admitted with COVID-19 who received MV. Two clusters of analgesia and sedation practices were identified. Cluster 1 hospitals generally administered opioids and propofol with occasional use of additional sedatives (eg, benzodiazepines, alpha-agonists, and ketamine); cluster 2 hospitals predominantly used opioids and benzodiazepines without other sedatives. As compared with patients in cluster 2, patients admitted to cluster 1 hospitals underwent a shorter adjusted median duration of MV with POD (β-estimate, -5.9; 95% CI, -11.2 to -0.6; P = .03).

INTERPRETATION

Patients who received MV for COVID-19 in hospitals that prioritized opioids and propofol for analgesia and sedation experienced shorter adjusted median duration of MV with POD as compared with patients who received MV in hospitals that primarily used opioids and benzodiazepines.

Patient Outcomes After Long-Term Acute Care Hospital Closures

Importance

Long-term acute care hospitals (LTCHs) are common sites of postacute care for patients recovering from severe respiratory failure requiring mechanical ventilation (MV). However, federal payment reform led to the closure of many LTCHs in the US, and it is unclear how closure of LTCHs may have affected upstream care patterns at short-stay hospitals and overall patient outcomes.

Objective

To estimate the association between LTCH closures and short-stay hospital care patterns and patient outcomes.

Design, Setting, and Participants

This retrospective, national, matched cohort study used difference-in-differences analysis to compare outcomes at short-stay hospitals reliant on LTCHs that closed during 2012 to 2018 with outcomes at control hospitals. Data were obtained from the Medicare Provider Analysis and Review File, 2011 to 2019. Participants included Medicare fee-for-service beneficiaries aged 66 years and older receiving MV for at least 96 hours in an intensive care unit (ie, patients at-risk for prolonged MV) and the subgroup also receiving a tracheostomy (ie, receiving prolonged MV). Data were analyzed from October 2022 to June 2023.

Exposure

Admission to closure-affected hospitals, defined as those discharging at least 60% of patients receiving a tracheostomy to LTCHs that subsequently closed, vs control hospitals.

Main Outcomes and Measures

Upstream hospital care pattern outcomes were short-stay hospital do-not-resuscitate orders, palliative care delivery, tracheostomy placement, and discharge disposition. Patient outcomes included hospital length of stay, days alive and institution free within 90 days, spending per days alive within 90 days, and 90-day mortality.

Results

Between 2011 and 2019, 99 454 patients receiving MV for at least 96 hours at 1261 hospitals were discharged to 459 LTCHs; 84 LTCHs closed. Difference-in-differences analysis included 8404 patients (mean age, 76.2 [7.2] years; 4419 [52.6%] men) admitted to 45 closure-affected hospitals and 45 matched-control hospitals. LTCH closure was associated with decreased LTCH transfer rates (difference, -5.1 [95% CI -8.2 to -2.0] percentage points) and decreased spending-per-days-alive (difference, -$8701.58 [95% CI, -$13 323.56 to -$4079.60]). In the subgroup of patients receiving a tracheostomy, there was additionally an increase in do-not-resuscitate rates (difference, 10.3 [95% CI, 4.2 to 16.3] percentage points) and transfer to skilled nursing facilities (difference, 10.0 [95% CI, 4.2 to 15.8] percentage points). There was no significant association of closure with 90-day mortality.

Conclusions and Relevance

In this cohort study, LTCH closure was associated with changes in discharge patterns in patients receiving mechanical ventilation for at least 96 hours and advanced directive decisions in the subgroup receiving a tracheostomy, without change in mortality. Further studies are needed to understand how LTCH availability may be associated with other important outcomes, including functional outcomes and patient and family satisfaction.

Effects of hospital preference for endovascular repair on postoperative mortality after elective abdominal aortic aneurysm repair: analysis of the Dutch Surgical Aneurysm Audit

BACKGROUND

Increased use of endovascular aneurysm repair (EVAR) and reduced open surgical repair (OSR), has decreased postoperative mortality after elective repair of abdominal aortic aneurysms (AAAs). The choice between EVAR or OSR depends on aneurysm anatomy, and the experience and preference of the vascular surgeon, and therefore differs between hospitals. The aim of this study was to investigate the current mortality risk difference (RD) between EVAR and OSR, and the effect of hospital preference for EVAR on overall mortality.

METHODS

Primary elective infrarenal or juxtarenal aneurysm repairs registered in the Dutch Surgical Aneurysm Audit (2013-2017) were analysed. First, mortality in hospitals with a higher preference for EVAR (high-EVAR group) was compared with that in hospitals with a lower EVAR preference (low-EVAR group), divided by the median percentage of EVAR. Second, the mortality RD between EVAR and OSR was determined by unadjusted and adjusted linear regression and propensity-score (PS) analysis and then by instrumental-variable (IV) analysis, adjusting for unobserved confounders; percentage EVAR by hospital was used as the IV.

RESULTS

A total of 11 997 patients were included. The median hospital rate of EVAR was 76.6 per cent. The overall mortality RD between high- and low-EVAR hospitals was 0.1 (95 per cent -0.5 to 0.4) per cent. The OSR mortality rate was significantly higher among high-EVAR hospitals than low-EVAR hospitals: 7.3 versus 4.0 per cent (RD 3.3 (1.4 to 5.3) per cent). The EVAR mortality rate was also higher in high-EVAR hospitals: 0.9 versus 0.7 per cent (RD 0.2 (-0.0 to 0.6) per cent). The RD following unadjusted, adjusted, and PS analysis was 4.2 (3.7 to 4.8), 4.4 (3.8 to 5.0), and 4.7 (4.1 to 5.3) per cent in favour of EVAR over OSR. However, the RD after IV analysis was not significant: 1.3 (-0.9 to 3.6) per cent.

CONCLUSION

Even though EVAR has a lower mortality rate than OSR, the overall effect is offset by the high mortality rate after OSR in hospitals with a strong focus on EVAR.

Unifying instrumental variable and inverse probability weighting approaches for inference of causal treatment effect and unmeasured confounding in observational studies

Confounding is a major concern when using data from observational studies to infer the causal effect of a treatment. Instrumental variables, when available, have been used to construct bound estimates on population average treatment effects when outcomes are binary and unmeasured confounding exists. With continuous outcomes, meaningful bounds are more challenging to obtain because the domain of the outcome is unrestricted. In this paper, we propose to unify the instrumental variable and inverse probability weighting methods, together with suitable assumptions in the context of an observational study, to construct meaningful bounds on causal treatment effects. The contextual assumptions are imposed in terms of the potential outcomes that are partially identified by data. The inverse probability weighting component incorporates a sensitivity parameter to encode the effect of unmeasured confounding. The instrumental variable and inverse probability weighting methods are unified using the principal stratification. By solving the resulting system of estimating equations, we are able to quantify both the causal treatment effect and the sensitivity parameter (i.e. the degree of the unmeasured confounding). We demonstrate our method by analyzing data from the HIV Epidemiology Research Study.

Development of a Prognostic Factor Prediction Model in Patients with Musculoskeletal Pain Treated with Homeopathy: An Individual Patient Data Meta-Analysis of Three Randomized Clinical Trials

BACKGROUND

Prognostic factor research methodology has not yet been applied to randomized clinical trial data of homeopathic medicines.

OBJECTIVES

To investigate the principle of individualization in homeopathy by developing a prognostic factor prediction model.

METHOD

A pooled, in-dividual patient data meta-analysis of 3 randomized trials -investigating the efficacy of a homeopathic gel (Spiroflor SRL®) containing Rhus toxicodendron as a key ingredient in osteoarthritis of the knee and acute low back pain. The prognostic value of a predefined set of 5 typical R. toxicodendron symptoms was investigated by assessing treatment-by-symptom interactions on pain as an outcome measure in a regression model.

RESULTS

The pooled dataset consisted of 284 patients in the Spiroflor SRL group and 275 patients in the control group. Adjusted for pain at baseline, a statistically significant effect modification for the symptoms "numbness or tingling of the affected part" (+2.0 mm VAS; p = 0.02), "amelioration by movement" (-5.6 mm VAS; p = 0.01), and "amelioration of pain by local heat" (+7.0 mm VAS; p = 0.02) was found.

CONCLUSIONS

Investigating aspects of treatment individualization in homeopathy using randomized trial data and standard meta-analytical techniques is possible. The symptom amelioration by local heat is of possible value as a homeopathic symptom (prognostic factor) predicting an increased likelihood of pain relief following treatment with the homeopathic product.

Patients with a Ruptured Abdominal Aortic Aneurysm Are Better Informed in Hospitals with an "EVAR-preferred" Strategy: An Instrumental Variable Analysis of the Dutch Surgical Aneurysm Audit

BACKGROUND

While several observational studies suggested a lower postoperative mortality after minimal invasive endovascular aneurysm repair (EVAR) in patients with a ruptured abdominal aortic aneurysm (RAAA) compared to conventional open surgical repair (OSR), landmark randomized controlled trials have not been able to prove the superiority of EVAR over OSR. Randomized controlled trials contain a selected, homogeneous population, influencing external validity. Observational studies are biased and adjustment of confounders can be incomplete. Instrumental variable (IV) analysis (pseudorandomization) may help to answer the question if patients with an RAAA have lower postoperative mortality when undergoing EVAR compared to OSR.

METHODS

This is an observational study including all patients with an RAAA, registered in the Dutch Surgical Aneurysm Audit between 2013 and 2017. The risk difference (RD) in postoperative mortality (30 days/in-hospital) between patients undergoing EVAR and OSR was estimated, in which adjustment for confounding was performed in 3 ways: linear model adjusted for observed confounders, propensity score model (multivariable logistic regression analysis), and IV analysis (two-stage least square regression), adjusting for observed and unobserved confounders, with the variation in percentage of EVAR per hospital as the IV instrument.

RESULTS

2419 patients with an RAAA (1489 OSR and 930 EVAR) were included. Unadjusted postoperative mortality was 34.9% after OSR and 22.6% after EVAR (RD 12.3%, 95% CI 8.5-16%). The RD adjusted for observed confounders using linear regression analysis and propensity score analysis was, respectively, 12.3% (95% CI 9.6-16.7%) and 13.2% (95%CI 9.3-17.1%) in favor of EVAR. Using IV analysis, adjusting for observed and unobserved confounders, RD was 8.9% (95% CI -1.1-18.9%) in favor of EVAR.

CONCLUSIONS

Adjusting for observed confounders, patients with an RAAA undergoing EVAR had a significant better survival than OSR in a consecutive large cohort. Adjustment for unobserved confounders resulted in a clinical relevant RD. An "EVAR preference strategy" in patients with an RAAA could result in lower postoperative mortality.

Association of hospital transfusion use and infection-related rehospitalizations among patients receiving hemodialysis: A retrospective cohort study

INTRODUCTION

Red blood cell transfusions have been associated with infection risk. We investigated whether hospital transfusions are associated with infections in maintenance hemodialysis patients requiring transfusions for chronic anemia.

METHODS

In this retrospective cohort study, hemodialysis patients who experienced an incident hospitalization during 2012-2013 were identified from the Medicare end-stage renal disease database. Hospital transfusions were first categorized into one of five groups based on adjusted likelihood of administering red blood cell transfusions during inpatient hospital stays that occurred over the previous year (2011) among the general Medicare cohort. Next, in a patient-level analysis, patients were categorized according to transfusion use at the incident hospitalization hospital. Outcomes were infection-related rehospitalization and a composite of infection-related hospitalization and all-cause mortality during the 60 days following hospital discharge. We estimated adjusted rate ratios for the association between hospital transfusion use and risk of rehospitalization or the composite endpoint using Poisson regression models.

FINDINGS

The study included 1578 hospitals and 61,455 hemodialysis patients. Patient characteristics were balanced across hospital transfusion use groups. The overall transfusion rate was 16.0%. The overall 30-day infection-related hospitalization rate (95% confidence interval) per 100 patient-months was 8.8 (8.6-9.1); rates did not differ by transfusion use group. Rate ratios for infection-related rehospitalization were 1.00 (0.91-1.10) over 30 days and 0.98 (0.91-1.05) over 60 days comparing the lowest and highest transfusion use groups.

DISCUSSION

We found no differences in risk of infection-related rehospitalization for patients receiving maintenance hemodialysis across the varying blood transfusion rates of US hospitals.

Treatment Setting and Outcomes of Cystic Fibrosis Pulmonary Exacerbations

RATIONALE

There are important gaps in knowledge of the optimal treatment of cystic fibrosis pulmonary exacerbations. Previous observational studies comparing inpatient with outpatient treatment have suffered from methodologic weaknesses, especially indication bias.

OBJECTIVES

We analyzed data from the Epidemiologic Study of Cystic Fibrosis using techniques to control for indication bias to determine whether there is an advantage to inpatient treatment of cystic fibrosis pulmonary exacerbations.

METHODS

We identified typical pulmonary exacerbations in patients ages 6 years and older during the 3-year observation period ending in 2005. In our primary analysis, we used the instrumental variables method, implemented using two-stage least squares regression, to evaluate the effect of the proportion of total time that intravenous treatment was administered on an inpatient (versus outpatient) basis on the likelihood of return of percent predicted forced expiratory volume in 1 second to greater than or equal to 90% of baseline post-treatment. We also evaluated two other indicators of treatment setting, three other measures of treatment response, and two alternative modeling techniques, and we also looked for differences between children and adults.

RESULTS

Our final analysis included 4,497 pulmonary exacerbations in 2,773 individual patients at 75 sites. We calculated the mean proportion of intravenous treatment time that was provided in the hospital setting at each site. The median across sites was 0.581 (interquartile range, 0.396-0.753). The median treatment success rate across sites was 74.2% (interquartile range, 67.9 to 79.2%). Univariate analysis and two-stage least squares models showed a positive relationship between treatment success and proportion of inpatient treatment days. Our primary model revealed an absolute increase of 9.08% (95% confidence interval, 2.55-15.61; P = 0.006) in the achievement of a return of percent predicted forced expiratory volume in 1 second to greater than or equal to 90% of baseline comparing complete inpatient treatment with no inpatient treatment. Treatment response was not related to duration of intravenous therapy. Similar results were found for all our modeling techniques and outcomes.

CONCLUSIONS

Patients with cystic fibrosis treated at sites with more reliance on inpatient treatment were more likely to achieve successful forced expiratory volume in 1 second recovery. There was no relationship between treatment duration and recovery of forced expiratory volume in 1 second.

Instrumental variable analysis in the context of dichotomous outcome and exposure with a numerical experiment in pharmacoepidemiology

Background

In pharmacoepidemiology, the prescription preference-based instrumental variables (IV) are often used with linear models to solve the endogeneity due to unobserved confounders even when the outcome and the endogenous treatment are dichotomous variables. Using this instrumental variable, we proceed by Monte-Carlo simulations to compare the IV-based generalized method of moment (IV-GMM) and the two-stage residual inclusion (2SRI) method in this context.

Methods

We established the formula allowing us to compute the instrument’s strength and the confounding level in the context of logistic regression models. We then varied the instrument’s strength and the confounding level to cover a large range of scenarios in the simulation study. We also explore two prescription preference-based instruments.

Results

We found that the 2SRI is less biased than the other methods and yields satisfactory confidence intervals. The proportion of previous patients of the same physician who were prescribed the treatment of interest displayed a good performance as a proxy of the physician’s preference instrument.

Conclusions

This work shows that when analysing real data with dichotomous outcome and exposure, appropriate 2SRI estimation could be used in presence of unmeasured confounding.

Echocardiogram in the Evaluation of Hemodynamically Stable Acute Pulmonary Embolism: National Practices and Clinical Outcomes

RATIONALE

Societal guideline recommendations vary with regard to the role of routine trans-thoracic echocardiography to screen for right ventricular strain in patients with hemodynamically stable acute pulmonary embolism.

OBJECTIVE

To characterize national patterns in use of early trans-thoracic echocardiography for the evaluation of patients with hemodynamically stable acute pulmonary embolism and determine associations between trans-thoracic echocardiography use and patient outcomes.

METHODS

Retrospective cohort study using Premier, Inc. database of approximately 20% of patients hospitalized in the United States with hemodynamically stable acute pulmonary embolism between 2008 and 2011. Multivariable, risk-adjusted hierarchical regression models were used to evaluate hospital variation in use of trans-thoracic echocardiography for pulmonary embolism and associations between hospital trans-thoracic echocardiography rates and patient outcomes. Patient-level trans-thoracic echocardiography exposure was used in sensitivity analyses.

RESULTS

We identified 64,037 patients (mean age, 61.7 years; 54% women; 68% white) hospitalized at 363 U.S. hospitals. Trans-thoracic echocardiography rates for hemodynamically stable acute pulmonary embolism varied widely among hospitals (median trans-thoracic echocardiography rate, 41.4%; range, 0-89%; interquartile range, 32.7-51.7%). Hospital rates of trans-thoracic echocardiography were not associated with significant differences in risk-adjusted mortality (trans-thoracic echocardiography rate quartile 4 vs. quartile 1: odds ratio, 0.88; 95% confidence interval, 0.69-1.13) or use of thrombolytics (odds ratio, 1.28; 95% confidence interval, 0.84-1.96), but rates of intensive care unit admission (odds ratio, 1.57; 95% confidence interval, 1.18-2.07), hospital length of stay (relative risk, 1.08; 95% confidence interval, 1.03-1.15), and costs (relative risk, 1.15; 95% confidence interval, 1.07-1.23) were significantly higher at hospitals with high trans-thoracic echocardiography rates. Analyses of patient-level trans-thoracic echocardiography exposure produced similar results, except with higher rates of thrombolysis (odds ratio, 5.58; 95% confidence interval, 4.40-7.09) and bleeding (odds ratio, 1.37; 95% confidence interval, 1.24-1.51) among patients receiving trans-thoracic echocardiography.

CONCLUSIONS

Trans-thoracic echocardiography use in the evaluation of patients with hemodynamically stable acute pulmonary embolism varied widely between hospitals. Hospitals with high rates of pulmonary embolism-associated trans-thoracic echocardiography use did not achieve different patient mortality outcomes but had higher resource use and costs. Our findings support the 2016 American College of Chest Physicians guidelines for management of pulmonary embolism, which recommend selective, rather than routine, use of trans-thoracic echocardiography to risk stratify patients with hemodynamically stable pulmonary embolism.

Practice Patterns and Outcomes Associated With Use of Anticoagulation Among Patients With Atrial Fibrillation During Sepsis

IMPORTANCE

Atrial fibrillation (AF) during sepsis is associated with an increased risk of ischemic stroke during hospitalization, but risks and benefits associated with anticoagulation for AF during sepsis are unclear.

OBJECTIVE

To determine clinician practice patterns and patient risk of stroke and bleeding associated with use of anticoagulation for AF during sepsis.

DESIGN, SETTING, AND PARTICIPANTS

A retrospective cohort study using enhanced administrative claims data from approximately 20% of patients hospitalized in the United States July 1, 2010, to June 30, 2013, examined patients with AF during sepsis who did not have additional indications for therapeutic anticoagulation. Propensity score and instrumental variable analyses were used to evaluate risks of in-hospital stroke and bleeding associated with anticoagulation during sepsis.

EXPOSURES

Parenteral anticoagulants administered in doses greater than those used for prophylaxis of venous thromboembolism.

MAIN OUTCOMES AND MEASURES

Ischemic stroke and clinically significant bleeding events during hospitalization.

RESULTS

Of 113 511 patients hospitalized with AF and sepsis, 38 582 were included in our primary analysis (18 976 men and 19 606 women; mean [SD] age, 74.9 [11.7] years). A total of 13 611 patients (35.3%) received parenteral anticoagulants, while 24 971 (64.7%) did not. Hospital utilization rates of parenteral anticoagulants for AF during sepsis varied (median, 33%; 25th-75th percentile, 25%-43%). CHA2DS2VASc scores (congestive heart failure, hypertension, age ≥75 years [doubled], type 1 or type 2 diabetes, stroke or transient ischemic attack or thromboembolism [doubled], vascular disease [prior myocardial infarction, peripheral artery disease, or aortic plaque], age 65-75 years, sex category [female]) poorly discriminated the risk of ischemic stroke during sepsis (C statistic, 0.526). Among 27 010 propensity score-matched patients, rates of in-hospital ischemic stroke events did not differ significantly between patients who did (174 of 13 505 [1.3%]) and did not (185 of 13 505 [1.4%]) receive parenteral anticoagulation (relative risk [RR], 0.94; 95% CI, 0.77-1.15). Clinically significant bleeding occurred more often among patients who received parenteral anticoagulation (1163 of 13 505 [8.6%]) than patients who did not receive parenteral anticoagulation (979 of 13 505 [7.2%]; RR, 1.21; 95% CI, 1.10-1.32). Risk of ischemic stroke associated with parenteral anticoagulation did not differ significantly between patients with preexisting (RR, 1.12; 95% CI, 0.86-1.44) or newly diagnosed AF (RR, 0.85; 95% CI 0.57-1.27; P = .31 for interaction). Results were robust to multiple sensitivity analyses, including hospital utilization rates of parenteral anticoagulation for AF as an instrument for anticoagulation exposure (RR for stroke, 1.08; 95% CI, 0.62-1.90; RR for bleeding, 1.23; 95% CI, 0.88-1.72).

CONCLUSIONS AND RELEVANCE

Among patients with AF during sepsis, parenteral anticoagulation was not associated with reduced risk of ischemic stroke and was associated with higher bleeding rates.

Health Care Utilization and Treatment Persistence Associated with Oral Paliperidone and Lurasidone in Schizophrenia Treatment

BACKGROUND

Oral paliperidone and lurasidone are new second-generation antipsychotics (SGAs). Empirical evidence on the comparative costs and persistence of these 2 agents are absent in the literature.

OBJECTIVE

To assess health care use and persistence associated with the 2 new agents oral paliperidone and lurasidone and other SGAs.

METHODS

Schizophrenia patients who initiated SGA therapy were identified in the January 2007-June 2013 claims databases of a large managed care organization. Multivariate regressions using aripiprazole as the comparator were conducted. Ordinary least squares regressions were used to estimate the total medical and pharmacy costs associated with each drug. Poisson regressions were conducted to evaluate the frequency of hospitalizations and emergency department (ED) visits associated with each drug. A censored regression model was used to evaluate the comparative persistence. Sensitivity analyses using generalized linear models, two-part models, hurdle models, and instrumental variable regressions were also performed.  

RESULTS

Compared with aripiprazole, paliperidone was not associated with significantly different total costs, yet lurasidone was associated with lower total costs (-$7,052; 95% CI = -$9,221, -$4,882). Lurasidone was also associated with significantly lower medical services costs (-$5,025; 95% CI = -$7,096, -$2,955), drug costs (-$2,026; 95% CI = -$2,695, -$1,357), hospital costs (-$3,026; 95% CI = -$4,731, -$1,321), outpatient costs (-$1,999; 95% CI = -$2,536, -$1,463), and ED costs (-$2,284; 95% CI = -$3,069, -$1,499), whereas paliperidone did not have significant effects on any types of costs. Paliperidone users had fewer ED visits (-0.25; 95% CI = -0.42, -0.08), while lurasidone users had fewer hospitalizations (-5.98; 95% CI = -6.61, -5.35) and fewer ED visits (-2.51; 95% CI = -2.92, -2.10). Both paliperidone and lurasidone were associated with lower levels of treatment persistence.

CONCLUSIONS

Paliperidone does not associate with lower total costs compared with commonly used SGAs, whereas lurasidone is associated with lower total health costs. Thus, high access fees of lurasidone are not necessarily a major concern in prescription.

Instrumental variable methods for causal inference

A goal of many health studies is to determine the causal effect of a treatment or intervention on health outcomes. Often, it is not ethically or practically possible to conduct a perfectly randomized experiment, and instead, an observational study must be used. A major challenge to the validity of observational studies is the possibility of unmeasured confounding (i.e., unmeasured ways in which the treatment and control groups differ before treatment administration, which also affect the outcome). Instrumental variables analysis is a method for controlling for unmeasured confounding. This type of analysis requires the measurement of a valid instrumental variable, which is a variable that (i) is independent of the unmeasured confounding; (ii) affects the treatment; and (iii) affects the outcome only indirectly through its effect on the treatment. This tutorial discusses the types of causal effects that can be estimated by instrumental variables analysis; the assumptions needed for instrumental variables analysis to provide valid estimates of causal effects and sensitivity analysis for those assumptions; methods of estimation of causal effects using instrumental variables; and sources of instrumental variables in health studies.

Association between pre-transplant dialysis modality and patient and graft survival after kidney transplantation

BACKGROUND

Previous studies have found inconsistent associations between pre-transplant dialysis modality and subsequent post-transplant survival. We aimed to examine this relationship using the instrumental variable method and to compare the results with standard Cox regression.

METHODS

We included 29 088 patients (age >20 years) from 16 European national or regional renal registries who received a first kidney transplant between 1 January 1999 and 31 December 2008 and were on dialysis before transplantation for a period between 90 days and 10 years. Standard multivariable Cox regression examined the association of individually assigned pre-transplant dialysis modality with post-transplant patient and graft survival. To decrease confounding-by-indication through unmeasured factors, we applied the instrumental variable method that used the case-mix adjusted centre percentage of peritoneal dialysis (PD) as predictor variable.

RESULTS

Standard analyses adjusted for age, sex, primary renal disease, donor type, duration of dialysis, year of transplantation and country suggested that PD before transplantation was associated with better patient [hazard ratio, HR (95% CI) = 0.83 (0.76-0.91)] and graft survival (HR (95% CI) 0.90 (0.84-0.96)) when compared with haemodialysis (HD). In contrast, the instrumental variable analysis showed that a 10% increase in the case-mix adjusted centre percentage of patients on PD was neither associated with post-transplant patient survival [HR (95% CI = 1.00 (0.97-1.04)] nor with graft survival [HR (95% CI) = 1.01 (0.98-1.04)].

CONCLUSIONS

The instrumental variable method failed to confirm the associations found in standard Cox regression between pre-transplant dialysis modality and patient and graft survival after transplantation. The lack of association in instrumental variable analysis may be due to better control of residual confounding.

Preference-based instrumental variable methods for the estimation of treatment effects: assessing validity and interpreting results

Observational studies of drugs and medical procedures based on administrative data are increasingly used to inform regulatory and clinical decisions. However, the validity of such studies is often questioned because available data may not contain measurements of many important prognostic variables that guide treatment decisions. Recently, approaches to this problem have been proposed that use instrumental variables (IV) defined at the level of an individual health care provider or aggregation of providers. Implicitly, these approaches attempt to estimate causal effects by using differences in medical practice patterns as a quasi-experiment. Although preference-based IV methods may usefully complement standard statistical approaches, they make assumptions that are unfamiliar to most biomedical researchers and therefore the validity of such analyses can be hard to evaluate. Here, we propose a simple framework based on a single unobserved dichotomous variable that can be used to explore how violations of IV assumptions and treatment effect heterogeneity may bias the standard IV estimator with respect to the average treatment effect in the population. This framework suggests various ways to anticipate the likely direction of bias using both empirical data and commonly available subject matter knowledge, such as whether medications or medical procedures tend to be overused, underused, or often misused. This approach is described in the context of a study comparing the gastrointestinal bleeding risk attributable to different non-steroidal anti-inflammatory drugs.

Blood pressure levels and mortality risk among hemodialysis patients in the Dialysis Outcomes and Practice Patterns Study

KDOQI practice guidelines recommend predialysis blood pressure <140/90 mm Hg; however, most prior studies had found elevated mortality with low, not high, systolic blood pressure. This is possibly due to unmeasured confounders affecting systolic blood pressure and mortality. To lessen this bias, we analyzed 24,525 patients by Cox regression models adjusted for patient and facility characteristics. Compared with predialysis systolic blood pressure of 130-159 mm Hg, mortality was 13% higher in facilities with 20% more patients at systolic blood pressure of 110-129 mm Hg and 16% higher in facilities with 20% more patients at systolic blood pressure of ≥160 mm Hg. For patient-level systolic blood pressure, mortality was elevated at low (<130 mm Hg), not high (≥180 mm Hg), systolic blood pressure. For predialysis diastolic blood pressure, mortality was lowest at 60-99 mm Hg, a wide range implying less chance to improve outcomes. Higher mortality at systolic blood pressure of <130 mm Hg is consistent with prior studies and may be due to excessive blood pressure lowering during dialysis. The lowest risk facility systolic blood pressure of 130-159 mm Hg indicates this range may be optimal, but may have been influenced by unmeasured facility practices. While additional study is needed, our findings contrast with KDOQI blood pressure targets, and provide guidance on optimal blood pressure range in the absence of definitive clinical trial data.

Emergency department case volume and patient outcomes in acute exacerbations of chronic obstructive pulmonary disease

OBJECTIVES

The objective was to determine whether emergency department (ED) case volume of acute exacerbation of chronic obstructive pulmonary disease (AECOPD) is associated with patient outcomes in AECOPD.

METHODS

The authors analyzed the 2007 Nationwide Emergency Department Sample (NEDS), the largest publicly available all-payer ED database in the United States. ED visits for AECOPD were identified with a principal diagnosis of International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code 491.21. EDs were categorized into quartiles by ED case volume of AECOPD. The primary outcome measures were early inpatient mortality (within the first 3 days of admission) and hospital length of stay (LOS).

RESULTS

The 2007 NEDS sample contained 126,045 ED visits for AECOPD from 946 U.S. EDs; 58% were hospitalized. Of these, the overall inpatient mortality rate was 2.0%, the early inpatient mortality 0.6%, and the median hospital LOS 4 days. Early inpatient mortality was lower in the highest-volume EDs (0.47%), compared with the lowest-volume EDs (1.13%). In a multivariable analysis adjusting for 37 patient and hospital characteristics, early inpatient mortality remained lower in patients admitted through the highest-volume EDs (adjusted odds ratios [ORs] = 0.51; 95% confidence interval [CI] = 0.32 to 0.82), compared with the lowest-volume EDs; however, the hospital LOS in the highest-volume EDs was slightly longer (adjusted difference in LOS = 0.53 day; 95% CI = 0.29 to 0.77). The volume threshold for reduced early mortality was approximately 200 cases per year.

CONCLUSIONS

ED patients who are hospitalized for AECOPD have an approximately 50% reduction in early inpatient mortality if they were admitted from an ED that handles a large volume of AECOPD cases.

The effect of bivalirudin on costs and outcomes of treatment of ST-segment elevation myocardial infarction

BACKGROUND

Bivalirudin is commonly used during percutaneous coronary intervention (PCI) rather than unfractionated heparin. The higher cost of bivalirudin may be offset if it reduces costly bleeding complications and/or length of stay. We sought to assess the effect of using bivalirudin on the costs of care among patients with ST-segment elevation myocardial infarction (STEMI) undergoing PCI.

METHODS

We analyzed data from 64,872 patients treated in 1 of 278 hospitals. The effect of overall hospital use of bivalirudin on clinical and economic outcomes was assessed using multivariable regression, based on average hospital use of treatments.

RESULTS

The use of bivalirudin among patients with STEMI treated with PCI varied widely across hospitals, with a median of 6.9% (interquartile range 2.3%-18.6%). After controlling for patient and hospital characteristics, use of bivalirudin rather than heparin and a glycoprotein IIb/IIIa inhibitor reduced bleeding (odds ratio 0.47, P < .001), length of stay (-0.47 days, P < .03), and hospital costs (-14%, P < .04).

CONCLUSIONS

Use of bivalirudin among patients with STEMI treated with PCI appears to reduce bleeding and overall costs.

Association of dialysis facility-level hemoglobin measurement and erythropoiesis-stimulating agent dose adjustment frequencies with dialysis facility-level hemoglobin variation: a retrospective analysis

BACKGROUND

A key goal of anemia management in dialysis patients is to maintain patients' hemoglobin (Hb) levels consistently within a target range. Our aim in this study was to assess the association of facility-level practice patterns representing Hb measurement and erythropoiesis-stimulating agent (ESA) dose adjustment frequencies with facility-level Hb variation.

METHODS

This was a retrospective observational database analysis of patients in dialysis facilities affiliated with large dialysis organizations as of July 01, 2006, covering a follow-up period from July 01, 2006 to June 30, 2009. A total of 2,763 facilities representing 436,442 unique patients were included. The predictors evaluated were facility-level Hb measurement and ESA dose adjustment frequencies, and the outcome measured was facility-level Hb variation.

RESULTS

First to 99th percentile ranges for facility-level Hb measurement and ESA dose adjustment frequencies were approximately once per month to once per week and approximately once per 3 months to once per 3 weeks, respectively. Facility-level Hb measurement and ESA dose adjustment frequencies were inversely associated with Hb variation. Modeling results suggested that a more frequent Hb measurement (once per week rather than once per month) was associated with approximately 7% to 9% and 6% to 8% gains in the proportion of patients with Hb levels within a ±1 and ±2 g/dL range around the mean, respectively. Similarly, more frequent ESA dose adjustment (once per 2 weeks rather than once per 3 months) was associated with approximately 6% to 9% and 5% to 7% gains in the proportion of patients in these respective Hb ranges.

CONCLUSIONS

Frequent Hb measurements and timely ESA dose adjustments in dialysis patients are associated with lower facility-level Hb variation and an increase in proportion of patients within ±1 and ±2 g/dL ranges around the facility-level Hb mean.

Effect of household and village characteristics on financial catastrophe and impoverishment due to health care spending in Western and Central Rural China: A multilevel analysis

OBJECTIVE

The study aimed to examine the effect of household and community characteristics on financial catastrophe and impoverishment due to health payment in Western and Central Rural China.

METHODS

A household survey was conducted in 2008 in Hebei and Shaanxi provinces and the Inner Mongolia Autonomous Region using a multi-stage sampling technique. Independent variables included village characteristics, household income, chronic illness status, health care use and health spending. A composite contextual variable, named village deprivation, was derived from socio-economic status and availability of health care facilities in each village using factor analysis. Dependent variables were whether household health payment was more than 40% of household's capacity to pay (catastrophic health payment) and whether household per capita income was put under Chinese national poverty line (1067 Yuan income per year) after health spending (impoverishment). Mixed effects logistic regression was used to assess the effect of the independent variables on the two outcomes.

RESULTS

Households with low per capita income, having elderly, hospitalized or chronically ill members, and whose head was unemployed were more likely to incur financial catastrophe and impoverishment due to health expenditure. Both catastrophic and impoverishing health payments increased with increased village deprivation. However, the presence of a village health clinic had no effect on the two outcomes, nor did household enrollment in the New Rural Cooperative Medical Scheme (national health insurance).

CONCLUSIONS

Village deprivation independently increases the risk for financial hardship due to health payment after adjusting for known household-level factors. This suggests that policy makers need to view the individual, household and village as separate units for policy targeting.

Instrumental variable methods in comparative safety and effectiveness research

Instrumental variable (IV) methods have been proposed as a potential approach to the common problem of uncontrolled confounding in comparative studies of medical interventions, but IV methods are unfamiliar to many researchers. The goal of this article is to provide a non-technical, practical introduction to IV methods for comparative safety and effectiveness research. We outline the principles and basic assumptions necessary for valid IV estimation, discuss how to interpret the results of an IV study, provide a review of instruments that have been used in comparative effectiveness research, and suggest some minimal reporting standards for an IV analysis. Finally, we offer our perspective of the role of IV estimation vis-à-vis more traditional approaches based on statistical modeling of the exposure or outcome. We anticipate that IV methods will be often underpowered for drug safety studies of very rare outcomes, but may be potentially useful in studies of intended effects where uncontrolled confounding may be substantial.

Greater Epoetin alfa (EPO) doses and short-term mortality risk among hemodialysis patients with hemoglobin levels less than 11 g/dL

PURPOSE

We examined the association between high doses of Epoetin alfa (EPO), which are used to raise and maintain hemoglobin (Hb) levels within target ranges for hemodialysis patients, and short-term mortality risk using multivariable regression and an instrumental variable (IV) analysis.

METHODS

We identified 32 734 patients receiving hemodialysis in 786 facilities from a large US dialysis provider between July 2000 and March 2002 who received care for >4 consecutive months, and had an Hb < 11 g/dL in the third month. We assessed dose titrations following the Hb < 11 g/dL and characterized facilities based on the percentage of patients with dose titrations >25% (instrument). We assessed deaths during the subsequent 90 days and evaluated the EPO dose-mortality association using conventional linear and IV regression.

RESULTS

The study population had a mean (SD) age of 60.4 (15.0) years; 48% were white, 42% were black and 51% were male. In unadjusted analyses, high EPO doses were associated with 90-day mortality risk (Risk Difference, RD = 3.0 per 100 persons, 95%CI:2.3-3.6); mortality risk was attenuated after adjustment for confounding (RD = 1.5 per 100 persons, 95%CI:0.8-2.2) and not associated with high EPO dose in the pooled IV analysis, though confidence intervals (CI) were wide (RD = -0.4 per 100 persons, 95%CI:-3.2-2.4).

CONCLUSIONS

The difference in risk estimates between the adjusted linear regression and the IV regression suggests that the short-term mortality related to EPO dosing may be largely attributable to confounding-by-indication for higher doses. The IV method, which was employed to address the possibility of residual confounding, yielded near null though imprecise effect estimates.

Instrumental variables II: instrumental variable application-in 25 variations, the physician prescribing preference generally was strong and reduced covariate imbalance

OBJECTIVE

An instrumental variable (IV) is an unconfounded proxy for a study exposure that can be used to estimate a causal effect in the presence of unmeasured confounding. To provide reliably consistent estimates of effect, IVs should be both valid and reasonably strong. Physician prescribing preference (PPP) is an IV that uses variation in doctors' prescribing to predict drug treatment. As reduction in covariate imbalance may suggest increased IV validity, we sought to examine the covariate balance and instrument strength in 25 formulations of the PPP IV in two cohort studies.

STUDY DESIGN AND SETTING

We applied the PPP IV to assess antipsychotic medication (APM) use and subsequent death among two cohorts of elderly patients. We varied the measurement of PPP, plus performed cohort restriction and stratification. We modeled risk differences with two-stage least square regression. First-stage partial r(2) values characterized the strength of the instrument. The Mahalanobis distance summarized balance across multiple covariates.

RESULTS

Partial r(2) ranged from 0.028 to 0.099. PPP generally alleviated imbalances in nonpsychiatry-related patient characteristics, and the overall imbalance was reduced by an average of 36% (+/-40%) over the two cohorts.

CONCLUSION

In our study setting, most of the 25 formulations of the PPP IV were strong IVs and resulted in a strong reduction of imbalance in many variations. The association between strength and imbalance was mixed.

Job strain and ischemic heart disease: a prospective study using a new approach for exposure assessment

BACKGROUND

Prolonged psychosocial load at the workplace may increase the risk of ischemic heart disease (IHD), but the issue is still unsettled. We analyzed the association between psychosocial workload and risk of IHD using a new approach allocating measures of psychosocial load to individuals based on the average exposure level in minor work units.

METHODS

Cohort study of 18,258 Danish public service workers in 1106 work units; 79% were women; 108 subjects with history of cardiovascular disease were excluded from the follow-up. The outcome was hospitalization due to IHD (angina pectoris or myocardial infarction) during the period 2002 to end of 2007.

RESULTS

During 87,428 person-years at risk (mean follow-up = 4.82 years), 101 subjects were admitted to a hospital due to IHD. Neither job strain (synergy of job demands and job control) nor general job dissatisfaction were related to IHD risk. However, compared with others, subjects who were allocated to the low job control category, had an increased risk of IHD, Hazard Ratio (95% CI) = 2.0 (1.1 to 3.6).

CONCLUSIONS

The findings presented do not lend support to the hypothesis that high job strain and job dissatisfaction are important determinants for IHD among Danish predominantly female public service workers.

Psychosocial work environment and antidepressant medication: a prospective cohort study

Background

Adverse psychosocial work environments may lead to impaired mental health, but it is still a matter of conjecture if demonstrated associations are causal or biased. We aimed at verifying whether poor psychosocial working climate is related to increase of redeemed subscription of antidepressant medication.

Methods

Information on all antidepressant drugs (AD) purchased at pharmacies from 1995 through 2006 was obtained for a cohort of 21,129 Danish public service workers that participated in work climate surveys carried out during the period 2002–2005. Individual self-reports of psychosocial factors at work including satisfaction with the work climate and dimensions of the job strain model were obtained by self-administered questionnaires (response rate 77,2%). Each employee was assigned the average score value for all employees at his/her managerial work unit [1094 units with an average of 18 employees (range 3–120)]. The risk of first-time AD prescription during follow-up was examined according to level of satisfaction and psychosocial strain by Cox regression with adjustment for gender, age, marital status, occupational status and calendar year of the survey.

Results

The proportion of employees that received at least one prescription of ADs from 1995 through 2006 was 11.9% and prescriptions rose steadily from 1.50% in 1996 to the highest level 6.47% in 2006. ADs were prescribed more frequent among women, middle aged, employees with low occupational status and those living alone. None of the measured psychosocial work environment factors were consistently related to prescription of antidepressant drugs during the follow-up period.

Conclusion

The study does not indicate that a poor psychosocial work environment among public service employees is related to prescription of antidepressant pharmaceuticals. These findings need cautious interpretation because of lacking individual exposure assessments.

Evolving statistical methods to facilitate evaluation of the causal association between erythropoiesis-stimulating agent dose and mortality in nonexperimental research: strengths and limitations

Findings from randomized controlled trials examining the efficacy of therapy with erythropoiesis-stimulating agents (ESAs) to normalize hemoglobin levels in patients with chronic kidney disease or kidney failure have raised questions regarding the safety of this class of drugs. However, no trial to date has specifically assessed the safety of ESA-dosing algorithms used to achieve the lower hemoglobin targets typically using in clinical practice. Although a wealth of nonexperimental data is available for dialysis patients, analyses based on these data are more susceptible to confounding bias than randomized controlled trials. Conducting valid pharmacoepidemiologic studies of drug effects in hemodialysis patients is complicated by the extent of their comorbidities, frequent hospitalizations, various concomitant medications, and an exceedingly high mortality rate. The need for greater ESA doses for the treatment of anemia in sicker patients potentially and plausibly generates confounding by indication, the control of which is complicated by the presence of time-dependent confounding. Here, we describe sources of bias in nonexperimental studies of ESA therapy in hemodialysis patients and critically appraise analytical methods that may help minimize bias in such studies.

Rosiglitazone is associated with mortality in chronic hemodialysis patients

Recent studies have associated rosiglitazone, a thiazolidinedione drug, with adverse cardiovascular outcomes in the general population with diabetes. Using data from the Dialysis Outcomes and Practice Patterns Study in the United States, we examined cardiovascular hospitalization and mortality associated with prescription of rosiglitazone, compared with other oral hypoglycemic agents, among 2393 long-term hemodialysis patients who were followed for a median of 1.1 yr. We assessed mortality risk using Cox models in patient-level and dialysis facility-level analyses that used the facility proportion of patients on rosiglitazone as the predictor (instrumental variable approach) and adjusted the models for demographics, comorbid conditions, laboratory values, and achieved dialysis dosage. Compared with patients prescribed other oral hypoglycemic agents, patients prescribed rosiglitazone had significantly higher all-cause (hazard ratio [HR] 1.38; 95% confidence interval [CI] 1.05 to 1.82) and cardiovascular (HR 1.59; 95% CI 1.14 to 2.22) mortality, and their adjusted HR for hospitalization with myocardial infarction was 3.5-fold higher (P = 0.02). We did not observe similar associations in a secondary analysis evaluating pioglitazone. By the instrumental variable approach, facilities with more than the median adjusted percentage (6.2%) of patients who had diabetes and were prescribed rosiglitazone had significantly higher all-cause mortality (HR 1.36; 95% CI 1.15 to 1.62) and cardiovascular mortality (HR 1.42; 95% CI 1.07 to 1.88) than facilities with less than the median expected percentage prescribed rosiglitazone. Our practice-based findings suggest significant associations of rosiglitazone use with higher cardiovascular and all-cause mortality among hemodialysis patients with diabetes.

Treatment intensification and risk factor control: toward more clinically relevant quality measures

BACKGROUND

Intensification of pharmacotherapy in persons with poorly controlled chronic conditions has been proposed as a clinically meaningful process measure of quality.

OBJECTIVE

To validate measures of treatment intensification by evaluating their associations with subsequent control in hypertension, hyperlipidemia, and diabetes mellitus across 35 medical facility populations in Kaiser Permanente, Northern California.

DESIGN

Hierarchical analyses of associations of improvements in facility-level treatment intensification rates from 2001 to 2003 with patient-level risk factor levels at the end of 2003.

PATIENTS

Members (515,072 and 626,130; age >20 years) with hypertension, hyperlipidemia, and/or diabetes mellitus in 2001 and 2003, respectively.

MEASUREMENTS

Treatment intensification for each risk factor defined as an increase in number of drug classes prescribed, of dosage for at least 1 drug, or switching to a drug from another class within 3 months of observed poor risk factor control.

RESULTS

Facility-level improvements in treatment intensification rates between 2001 and 2003 were strongly associated with greater likelihood of being in control at the end of 2003 (P < or = 0.05 for each risk factor) after adjustment for patient- and facility-level covariates. Compared with facility rankings based solely on control, addition of percentages of poorly controlled patients who received treatment intensification changed 2003 rankings substantially: 14%, 51%, and 29% of the facilities changed ranks by 5 or more positions for hypertension, hyperlipidemia, and diabetes, respectively.

CONCLUSIONS

Treatment intensification is tightly linked to improved control. Thus, it deserves consideration as a process measure for motivating quality improvement and possibly for measuring clinical performance.

Instrumental variables I: instrumental variables exploit natural variation in nonexperimental data to estimate causal relationships

The gold standard of study design for treatment evaluation is widely acknowledged to be the randomized controlled trial (RCT). Trials allow for the estimation of causal effect by randomly assigning participants either to an intervention or comparison group; through the assumption of "exchangeability" between groups, comparing the outcomes will yield an estimate of causal effect. In the many cases where RCTs are impractical or unethical, instrumental variable (IV) analysis offers a nonexperimental alternative based on many of the same principles. IV analysis relies on finding a naturally varying phenomenon, related to treatment but not to outcome except through the effect of treatment itself, and then using this phenomenon as a proxy for the confounded treatment variable. This article demonstrates how IV analysis arises from an analogous but potentially impossible RCT design, and outlines the assumptions necessary for valid estimation. It gives examples of instruments used in clinical epidemiology and concludes with an outline on estimation of effects.

Facility hemodialysis vascular access use and mortality in countries participating in DOPPS: an instrumental variable analysis

BACKGROUND

Previously, the Dialysis Outcomes and Practice Patterns Study (DOPPS) has shown large international variations in vascular access practice. Greater mortality risks have been seen for hemodialysis (HD) patients dialyzing with a catheter or graft versus a native arteriovenous fistula (AVF). To further understand the relationship between vascular access practice and outcomes, we have applied practice-based analyses (using an instrumental variable approach) to decrease the treatment-by-indication bias of prior patient-level analyses.

STUDY DESIGN

A prospective observational study of HD practices.

SETTING & PARTICIPANTS

Data collected from 1996 to 2004 from 28,196 HD patients from more than 300 dialysis units participating in the DOPPS in 12 countries.

PREDICTOR OR FACTOR

Patient-level or case-mix-adjusted facility-level vascular access use. OUTCOMES/MEASUREMENTS: Mortality and hospitalization risks.

RESULTS

After adjusting for demographics, comorbid conditions, and laboratory values, greater mortality risk was seen for patients using a catheter (relative risk, 1.32; 95% confidence interval, 1.22 to 1.42; P < 0.001) or graft (relative risk, 1.15; 95% confidence interval, 1.06 to 1.25; P < 0.001) versus an AVF. Every 20% greater case-mix-adjusted catheter use within a facility was associated with 20% greater mortality risk (versus facility AVF use, P < 0.001); and every 20% greater facility graft use was associated with 9% greater mortality risk (P < 0.001). Greater facility catheter and graft use were both associated with greater all-cause and infection-related hospitalization. Catheter and graft use were greater in the United States than in Japan and many European countries. More than half the 36% to 43% greater case-mix-adjusted mortality risk for HD patients in the United States versus the 5 European countries from the DOPPS I and II was attributable to differences in vascular access practice, even after adjusting for other HD practices. Vascular access practice differences accounted for nearly 30% of the greater US mortality compared with Japan.

LIMITATIONS

Possible existence of unmeasured facility- and patient-level confounders that could impact the relationship of vascular access use with outcomes.

CONCLUSIONS

Facility-based analyses diminish treatment-by-indication bias and suggest that less catheter and graft use improves patient survival.

The survival advantage for haemodialysis patients taking vitamin D is questioned: findings from the Dialysis Outcomes and Practice Patterns Study

BACKGROUND

Retrospective studies of haemodialysis patients from large dialysis organizations in the United States have indicated that intravenous vitamin D may be associated with a survival benefit. However, patients prescribed vitamin D are generally healthier than those who are not, suggesting that treatment by indication may have biased previous findings. Additionally, no survival benefit associated with vitamin D has been shown in a recent meta-analysis in CKD patients. Because treatment-by-indication bias due to both measured and unmeasured confounders cannot be completely accounted for in standard regression or marginal structural models (MSMs), this study evaluates the association between vitamin D and mortality among participants in the Dialysis Outcomes and Practice Patterns Study (DOPPS) using standard regression and MSMs with an expanded set of covariates, as well as by instrumental variable models to minimize potential bias due to unmeasured confounders.

METHODS

Data from 38 066 DOPPS participants from 12 countries between 1996 and 2007 were analysed. Mortality risk was assessed using standard baseline and time-varying Cox regression models, adjusted for demographics and detailed comorbidities, and MSMs. In models similar to instrumental variable analysis, the facility percentage of patients prescribed vitamin D, adjusted for the patient case mix, was used to predict patient-level mortality.

RESULTS

Vitamin D prescription was significantly higher in the USA compared to other countries. On average, patients prescribed vitamin D had fewer comorbidities compared to those who were not. Vitamin D therapy was associated with lower mortality in adjusted time-varying standard regression models [relative ratio (RR) = 0.92 (95% confidence interval: 0.87-0.96)] and baseline MSMs [RR = 0.84 (0.78-0.98)] and time-varying MSMs [RR = 0.78 (0.73-0.84)]. No significant differences in mortality were observed in adjusted baseline standard regression models for patients with or without vitamin D prescription [RR = 0.98 (0.93-1.02)] or for patients in facility practices where vitamin D prescription was more frequent [RR for facilities in 75th versus 25th percentile of vitamin D prescription = 0.99 (0.94-1.04)].

CONCLUSIONS

Vitamin D was associated with a survival benefit in models prone to bias due to unmeasured confounding. In agreement with a meta-analysis of randomized controlled studies, no difference in mortality was observed in instrumental variable models that tend to be more independent of unmeasured confounding. These findings indicate that a randomized controlled trial of vitamin D and clinical outcomes in haemodialysis patients are needed and can be ethically conducted.

Mortality risk for dialysis patients with different levels of serum calcium, phosphorus, and PTH: the Dialysis Outcomes and Practice Patterns Study (DOPPS)

BACKGROUND

Abnormalities in serum calcium, phosphorus, and parathyroid hormone (PTH) concentrations are common in patients with chronic kidney disease and have been associated with increased morbidity and mortality. No clinical trials have been conducted to clearly identify categories of calcium, phosphorus, and PTH levels associated with the lowest mortality risk. Current clinical practice guidelines are based largely on expert opinions, and clinically relevant differences exist among guidelines across countries. We sought to describe international trends in calcium, phosphorus, and PTH levels during 10 years and identify mortality risk categories in the Dialysis Outcomes and Practice Patterns Study (DOPPS), an international study of hemodialysis practices and associated outcomes.

STUDY DESIGN

Prospective cohort study.

PARTICIPANTS

25,588 patients with end-stage renal disease on hemodialysis therapy for longer than 180 days at 925 facilities in DOPPS I (1996-2001), DOPPS II (2002-2004), or DOPPS III (2005-2007).

PREDICTORS

Serum calcium, albumin-corrected calcium (Ca(Alb)), phosphorus, and PTH levels.

OUTCOMES

Adjusted hazard ratios for all-cause and cardiovascular mortality calculated using Cox models.

RESULTS

Distributions of mineral metabolism markers differed across DOPPS countries and phases, with lower calcium and phosphorus levels observed in the most recent phase of DOPPS. Survival models identified categories with the lowest mortality risk for calcium (8.6 to 10.0 mg/dL), Ca(Alb) (7.6 to 9.5 mg/dL), phosphorus (3.6 to 5.0 mg/dL), and PTH (101 to 300 pg/mL). The greatest risk of mortality was found for calcium or Ca(Alb) levels greater than 10.0 mg/dL, phosphorus levels greater than 7.0 mg/dL, and PTH levels greater than 600 pg/mL and in patients with combinations of high-risk categories of calcium, phosphorus, and PTH.

LIMITATIONS

Because of the observational nature of DOPPS, this study can only indicate an association between mineral metabolism categories and mortality.

CONCLUSIONS

Our results provide important information about mineral metabolism trends in hemodialysis patients in 12 countries during a decade. The risk categories identified in the DOPPS cohort may be relevant to efforts at international harmonization of existing clinical guidelines for mineral metabolism.

Influence of physician specialty on outcomes after acute ischemic stroke

BACKGROUND

Specialist care has been shown to improve outcomes for several complex medical conditions. For patients with ischemic stroke, prior studies have suggested that admission to the care of neurologists is associated with better outcomes, but these studies may have incompletely controlled for confounding prognostic differences.

OBJECTIVE

The objective of this study was to evaluate whether admission to the care of a neurologist is associated with improvement in outcomes of stroke patients after controlling for initial prognostic differences.

DESIGN

This was a retrospective cohort study.

SETTING

Participating in the study were 113 U.S. academic hospitals.

PATIENTS

Demographic and clinical data for all ischemic stroke patients admitted through emergency departments from 1997 to 1999 were collected from an administrative database.

MEASUREMENTS

In traditional analyses, we evaluated attending physician specialty as a predictor of in-hospital mortality. In grouped-treatment (GT) analyses, a method based on the instrumental variable approach that bypasses selection bias, the hospital rate of stroke admission to neurologists was used as the predictor. We used generalized estimating equations for all analyses, adjusting for demographics, urgency, comorbid illness severity, and treatment volume.

RESULTS

Of 26,925 ischemic stroke patients, 60% were admitted to the care of neurologists. In univariate analysis, risk of in-hospital mortality in cases admitted to neurologists (4.6%) was lower than that for those admitted to generalists (9.5%; P < .001). Adjustment for individual-level characteristics did not alter the association (0.60 OR, 95% CI, 0.50-0.72; P < .001). However, no advantage to neurologist admission was demonstrated in GT analysis, with mortality rates similar at hospitals admitting different proportions of ischemic stroke cases to neurologists (1.02 OR, 95% CI, 0.79-1.30; P = .90).

CONCLUSIONS

Differences in ischemic stroke outcomes between neurologists and generalists may be a result of differences in initial prognosis because outcomes are no better at hospitals that admit patients to the care of neurologists more frequently.

The impact of different rehabilitation strategies after major events in the elderly: the case of stroke and hip fracture in the Tuscany region

Background

On a regional level, our aims were to describe rehabilitation patterns for elderly patients with stroke and hip fracture and to investigate mortality risk during the 6-month post acute period.

Methods

Data sources included administrative data relative to patients aged 65+ resident in Tuscany admitted in hospital for stroke or hip fracture between 2001 and 2003, traced up to 3 years before and 6 months following index admission. The study design involves computerized linkage of administrative data, and an exploratory analysis of the association between rehabilitation patterns and 6-month mortality, adjusting for clinical, demographic, and acute-related care characteristics using multivariate Cox regression.

Results

Rehabilitation patterns vary greatly across Tuscany with considerable cost implications. Six month mortality risk for stroke patients is significantly lower among residents of Local Health Authorities where patients are more frequently rehabilitated, specifically in extra-hospital settings.

Conclusion

Our study, targeting two crucial conditions for elderly patients, found a high variability of rehabilitation patterns across a region, albeit coherent between the two pathologies, associated with remarkable differences in average expenditure. Differences in hazard rates for 6-month mortality after stroke at population level were also found. These results need to be confirmed and further investigated through a more robust information framework.

Combining individual and ecological data to determine compositional and contextual socio-economic risk factors for suicide

The social and economic characteristics of geographic areas are associated with their suicide rates. The extent to which these ecological associations are due to the characteristics of the people living in the areas (compositional effects) or the influence of the areas themselves on risk (contextual effects) is uncertain. Denmark's Medical Register on Vital Statistics and its Integrated Database for Longitudinal Labour Market Research were used to identify suicides and 20 matched controls per case in 25-60-year-old men and women between 1982 and 1997. Individual and area (municipality) measures of income, marital and employment status were obtained. There were 9011 suicides and 180,220 controls. Individual-level associations with these risk factors were little changed when controlling for contextual effects. In contrast, ecological associations of increased suicide risk with declining area levels of employment and income and increasing proportions of people living alone were much attenuated after controlling for compositional effects. We found no consistent evidence that associations with individual-level risk factors differed depending on the areas' characteristics (cross-level interactions). This analysis suggests the ecological associations to be attributed to characteristics of the residents rather than area influences on risk.

Evaluating short-term drug effects using a physician-specific prescribing preference as an instrumental variable

BACKGROUND

Postmarketing observational studies of the safety and effectiveness of prescription medications are critically important but fraught with methodological problems. The data sources available for such research often lack information on indications and other important confounders for the drug exposure under study. Instrumental variable methods have been proposed as a potential approach to control confounding by indication in nonexperimental studies of treatment effects; however, good instruments are hard to find.

METHODS

We propose an instrument for use in pharmacoepidemiology that is based on a time-varying estimate of the prescribing physician's preference for one drug relative to a competing therapy. The use of this instrument is illustrated in a study comparing the effect of exposure to COX-2 inhibitors with nonselective, nonsteroidal antiinflammatory medications on gastrointestinal complications.

RESULTS

Using conventional multivariable regression adjusting for 17 potential confounders, we found no protective effect due to COX-2 use within 120 days from the initial exposure (risk difference = -0.06 per 100 patients; 95% confidence interval = -0.26 to 0.14). However, the proposed instrumental variable method attributed a protective effect to COX-2 exposure (-1.31 per 100 patients; -2.42 to -0.20) compatible with randomized trial results (-0.65 per 100 patients; -1.08 to -0.22).

CONCLUSIONS

The instrumental variable method that we have proposed appears to have substantially reduced the bias due to unobserved confounding. However, more work needs to be done to understand the sensitivity of this approach to possible violations of the instrumental variable assumptions.

Paradoxical results in the study of risk factors of chronic obstructive pulmonary disease (COPD) re-admission

We have previously reported an apparently paradoxical association between medical care related factors and an increased risk of chronic obstructive pulmonary disease (COPD) re-hospitalisation, in a cohort of 346 COPD subjects from Barcelona, Spain. Confounding by severity or by indication is a plausible explanation. We tested the confounding effect of severity-related variables on these paradoxical associations. Forced expiratory volume in one second (FEV1), arterial oxygen pressure (PO2) and previous COPD admissions were associated with: (1) the presence of medical care related factors, and (2) re-admission during follow-up. Risks of readmission associated with most of the medical care related factors were reduced after adjustment for the severity variables. The risk associated with long-term oxygen therapy use changed from a crude OR of 2.36 (95% CI: 1.79-3.11) to an adjusted OR of 1.38 (0.95-2.00), while that associated with anticholinergics use varied from 3.52 (2.37-5.21) to 2.10 (1.32-3.36)). We concluded that the excess risk of COPD re-admission associated with medical care related factors might be partially due to confounding by indication. Residual confounding may still account for part of the remaining excess risk. True adverse effects of some pharmacological treatments cannot be excluded.

Impact of hospital-related factors on outcome after treatment of cerebral aneurysms

BACKGROUND AND PURPOSE

The goal of this study was to examine the impact of hospital characteristics on outcome after the treatment of ruptured and unruptured cerebral aneurysms.

METHODS

We identified all discharges in New York State from 1995 through 2000 with a principal diagnosis of subarachnoid hemorrhage (SAH) or unruptured cerebral aneurysm (UCA) in patients who were treated by aneurysm clipping, wrapping, or endovascular coiling. An adverse outcome was defined as in-hospital death or discharge to a rehabilitation hospital or long-term facility. We examined the effect of hospital factors, including the rate of endovascular therapy and overall procedural volume, on outcome, length of stay, and total charges.

RESULTS

There were 2200 (36.9%) and 3763 (63.1%) admissions for attempted treatment of UCA and SAH, respectively. The 10 highest-volume hospitals performed half of all the procedures. Overall, hospital volume was associated with fewer adverse outcomes and lower in-hospital mortality for both UCA (adverse outcome: odds ratio [OR], 0.89; P<0.0001; mortality: OR, 0.94; P=0.002 for each 10 additional procedures performed per year) and SAH (adverse outcome: OR, 0.94; P=0.03; mortality: OR, 0.95; P=0.005). Use of endovascular therapy (each additional 10% of cases performed endovascularly) was associated with fewer adverse outcomes after treatment of unruptured aneurysm (0.83, P=0.026). Hospital volume had more of an effect on outcome after aneurysm clipping than after endovascular therapy.

CONCLUSIONS

Hospital procedural volume and the propensity of a hospital to use endovascular therapy are both independently associated with better outcome. Improvement in outcome could be achieved by a program of regionalization and selective referral for the treatment of cerebral aneurysms.

In-hospital Mortality and Morbidity after Surgical Treatment of Unruptured Intracranial Aneurysms in the United States, 1996–2000: The Effect of Hospital and Surgeon Volume

OBJECTIVE

We sought to determine the risk of adverse outcome after contemporary surgical treatment of patients with unruptured intracranial aneurysms in the United States. Patient, surgeon, and hospital characteristics were tested as potential outcome predictors, with particular attention to the surgeon's and hospital's volume of care.

METHODS

We performed a retrospective cohort study with the Nationwide Inpatient Sample, 1996 to 2000. Multivariate logistic and ordinal regression analyses were performed with endpoints of mortality, discharge other than to home, length of stay, and total hospital charges.

RESULTS

We identified 3498 patients who were treated at 463 hospitals, and we identified 585 surgeons in the database. Of all patients, 2.1% died, 3.3% were discharged to skilled-nursing facilities, and 12.8% were discharged to other facilities. The analysis adjusted for age, sex, race, primary payer, four variables measuring acuity of treatment and medical comorbidity, and five variables indicating symptoms and signs. The statistics for median annual number of unruptured aneurysms treated were eight per hospital and three per surgeon. High-volume hospitals had fewer adverse outcomes than hospitals that handled comparatively fewer unruptured aneurysms: discharge other than to home occurred after 15.6% of operations at high-volume hospitals (20 or more cases/yr) compared with 23.8% at low-volume hospitals (fewer than 4 cases/yr) (P = 0.002). High surgeon volume had a similar effect (15.3 versus 20.6%, P = 0.004). Mortality was lower at high-volume hospitals (1.6 versus 2.2%) than at hospitals that handled comparatively fewer unruptured aneurysms, but not significantly so. Patients treated by high-volume surgeons had fewer postoperative neurological complications (P = 0.04). Length of stay was not related to hospital volume. Charges were slightly higher at high-volume hospitals, partly because arteriography was performed more frequently than at hospitals that handled comparatively fewer unruptured aneurysms.

CONCLUSION

For patients with unruptured aneurysms who were treated in the United States between 1996 and 2000, surgery performed at high-volume institutions or by high-volume surgeons was associated with significantly lower morbidity and modestly lower mortality.

Approaches to combat with confounding by indication in observational studies of intended drug effects

There has been a resurgence of controversy about the usefulness of observational data to study the efficacy of drugs. Nearly every week a researcher makes some criticism of clinical trials or justifies observational research into intended effects, with other researchers offering a contradictory viewpoint. Literature reviews are not useful in this regard because the contradictory studies will not usually be carried out. Some methods are discussed which may have potential utility in the study of intended effects. There may be a marginal role for statistical techniques such as propensity scores and confounder scores. More promising techniques may include ecological analyses, restriction of subjects and blinded prospective review. Because it is currently unknown when the observational study of drug efficacy is possible, we should arguably always carry out a study of the determinants of prescribing first, and possibly consider using the various techniques that are outlined in this article.

Characteristics of academic medical centers and ischemic stroke outcomes

BACKGROUND AND PURPOSE

Data supporting the efficacy of stroke center characteristics are limited.

METHODS

A questionnaire detailing stroke treatment practices was sent to 42 academic medical centers in the University Health Systems Consortium. In-hospital mortality of all emergency department admissions for ischemic stroke at these institutions was evaluated in a database of discharge abstracts during 1997-1999. Institutional characteristics were evaluated as predictors of in-hospital mortality after adjustment for age, sex, race, hospital treatment volume of ischemic stroke, and admission status (emergent, urgent, elective). Length of stay (LOS), total hospital charges, and frequency of tissue plasminogen activator (tPA) administration were evaluated as secondary outcomes. We used a multivariable method called generalized estimating equations, which broadens confidence intervals to adjust for clustering of variables at institutions.

RESULTS

Thirty-two institutions completed the questionnaire, and 29 of these were included in the database of discharge abstracts. In-hospital deaths occurred in 758 (7.0%) of the 10 880 ischemic stroke patients admitted through the emergency department. In-hospital deaths were less frequent at hospitals with a vascular neurologist (odds ratio [OR] 0.51; 95% CI, 0.36 to 0.74; P<0.0001) and at those with guidelines stating that only neurologists could administer tPA (OR, 0.65; 95% CI, 0.49 to 0.88; P=0.004). There was a trend toward fewer deaths at hospitals with a dedicated stroke team available by pager (OR, 0.76; 95% CI, 0.56 to 1.04; P=0.09). The presence of a dedicated neurological intensive care unit, stroke unit, and written clinical pathway for stroke were not significantly associated with in-hospital death. LOS was shorter at hospitals with a vascular neurologist (P=0.01).

CONCLUSIONS

Academic medical centers with a vascular neurologist and those with written guidelines limiting tPA administration to neurologists had lower rates of in-hospital mortality for ischemic stroke patients.

Treatment of unruptured cerebral aneurysms in California

BACKGROUND AND PURPOSE

The impact of endovascular therapy on treatment outcomes of unruptured cerebral aneurysms has not been studied in a defined geographic area.

METHODS

All primary diagnoses of unruptured aneurysms were retrieved from a statewide database of hospital discharges in California from January 1990 through December 1998. Admissions for initial treatment and all follow-up care were combined to reflect the entire course of therapy. An adverse outcome was defined as an in-hospital death or discharge to nursing home or rehabilitation hospital at any point during the treatment course. Multivariable analyses were performed with generalized estimating equations with adjustment for age, sex, ethnicity, source of admission, year of treatment, hospital volume, and clustering of observations at institutions.

RESULTS

A total of 2069 patients were treated for unruptured aneurysms. Adverse outcomes were more frequent in the 1699 patients treated with surgery (25%) than in those treated with endovascular therapy (10%; P:<0.001). The difference persisted after multivariable adjustment (surgery versus endovascular therapy: odds ratio for adverse outcomes, 3.1; 95% CI, 2.5 to 4.0; P:<0.001). Adverse outcomes declined from 1991 to 1998 in patients treated with endovascular therapy (P:<0.005) but not for surgery. In-hospital deaths occurred in 3.5% of surgical cases and 0.5% of endovascular cases (P:=0.003), and the difference remained significant after adjustment (odds ratio, 6.3; 95% CI, 3.5 to 11.4; P:<0.001). Total length of stay and hospital charges were greater in surgical cases (both P:<0.001). Results were similar in a confirmatory analysis focusing on treatment differences between institutions. Institutional treatment volume was also associated with outcome but did not account for the differences between surgery and endovascular therapy.

CONCLUSIONS

In California, endovascular therapy of unruptured aneurysms is associated with less risk of adverse outcomes and in-hospital death, lower hospital charges, and shorter hospital stays compared with surgery. Differences between therapies became more distinct through the years. Uncontrolled differences in prognosis of patients receiving endovascular therapy and surgery cannot be ruled out in this study of discharge abstracts.