Applying propensity score methods in medical research: pitfalls and prospects

High Mortality in Severe Sepsis and Septic Shock Patients with Do-Not-Resuscitate Orders in East Asia

BACKGROUND

Severe sepsis is a potentially deadly illness and always requires intensive care. Do-not-resuscitate (DNR) orders remain a debated issue in critical care and limited data exist about its impact on care of septic patients, particularly in East Asia. We sought to assess outcome of severe sepsis patients with regard to DNR status in Taiwan.

METHODS

A retrospective cohort study was conducted in intensive care units (ICUs) between 2008 and 2010. All severe sepsis patients were included for analysis. Primary outcome was association between DNR orders and ICU mortality. Volume of interventions was used as proxy indicator to indicate aggressiveness of care.

RESULTS

Sixty-seven (9.4%) of 712 patients had DNR orders on ICU admission, and these patients were older and had higher disease severity compared with patients without DNR orders. Notably, DNR patients experienced high ICU mortality (90%). Multivariate analysis revealed that the presence of DNR orders was independently associated with ICU mortality (odds ratio: 6.13; 95% confidence interval: 2.66-14.10). In propensity score-matched cohort, ICU mortality rate (91%) in the DNR group was statistically higher than that (62%) in the non-DNR group (p <0.001). Regarding ICU interventions, arterial and central venous catheterization were more commonly used in DNR patients than in non-DNR patients.

CONCLUSIONS

From the Asian perspective, septic patients placed on DNR orders on ICU admission had exceptionally high mortality. In contrast to Western reports, DNR patients received more ICU interventions, reflecting more aggressive approach to dealing with this patient population. The findings in some ways reflect differences between East and West cultures and suggest that DNR status is an important confounder in ICU studies involving severely septic patients.

A Comparison of Health Plan- and Provider-Delivered Chronic Care Management Models on Patient Clinical Outcomes

BACKGROUND

The real world implementation of chronic care management model varies greatly. One aspect of this variation is the delivery mode. Two contrasting strategies include provider-delivered care management (PDCM) and health plan-delivered care management (HPDCM).

OBJECTIVE

We aimed to compare the effectiveness of PDCM vs. HPDCM on improving clinical outcomes for patients with chronic diseases.

DESIGN

We used a quasi-experimental two-group pre-post design using the difference-in-differences method.

PATIENTS

Commercially insured patients, with any of the five chronic diseases-congestive heart failure, chronic obstructive pulmonary disease, coronary heart disease, diabetes, or asthma, who were outreached to and engaged in either PDCM or HPDCM were included in the study.

MAIN MEASURES

Outreached patients were those who received an attempted or actual contact for enrollment in care management; and engaged patients were those who had one or more care management sessions/encounters with a care manager. Effectiveness measures included blood pressure, low density lipoprotein (LDL), weight loss, and hemoglobin A1c (for diabetic patients only). Primary endpoints were evaluated in the first year of follow-up.

KEY RESULTS

A total of 4,000 patients were clustered in 165 practices (31 in PDCM and 134 in HPDCM). The PDCM approach demonstrated a statistically significant improvement in the proportion of outreached patients whose LDL was under control: the proportion of patients with LDL < 100 mg/dL increased by 3 % for the PDCM group (95 % CI: 1 % to 6 %) and 1 % for the HPDCM group (95 % CI: -2 % to 5 %). However, the 2 % difference in these improvements was not statistically significant (95 % CI: -2 % to 6 %). The HPDCM approach showed 3 % [95 % CI: 2 % to 6 %] improvement in overall diabetes care among outreached patients and significant reduction in obesity rates compared to PDCM (4 %, 95 % CI: 0.3 % to 8 %).

CONCLUSIONS

Both care management delivery modes may be viable options for improving care for patients with chronic diseases. In this commercially insured population, neither PDCM nor HPDCM resulted in substantial improvement in patients' clinical indicators in the first year. Different care management strategies within the provider-delivered programs need further investigation.

Effect of Lipophilic and Hydrophilic Statins on Breast Cancer Risk in Thai Women: A Cross-sectional Study

Background: Statins are proposed as a chemoprevention agent for breast cancer due to their anti-inflammatory effect. The effects of lipophilic and hydrophilic statins on breast cancer risk might be different due to their different pharmacologic properties. Therefore, this study aimed to assess a casual-effect of lipophilic and hydrophilic statins on breast cancer risk using a counterfactual framework approach.

Methods: A cross-sectional study of 15,718 women who were screened for breast cancer at Mammographic center, Ramathibodi Hospital, Bangkok, Thailand, was conducted during September 2011 to 2012. A counterfactual framework approach was applied to assess causal effects of treatments (i.e., lipophilic and hydrophilic statins) on outcome (i.e. breast cancer). Multi-logit and logistic regression models were used for treatment and outcome models, respectively. An inverse probability weight regression analysis (IPWRA) was then applied to estimate potential outcome mean (POM) and average treatment effect (ATE) by combining the outcome and treatment models.

Results: Breast cancer risks were 0.0072 (95% CI: 0.0055, 0.0089), 0.0051 (95% CI: 0.0008, 0.0095), and 0.0038 (95% CI: 0.002, 0.0056) for non-statin users, hydrophilic, and lipophilic statin users, respectively. The estimated risk differences were -0.0021 (95% CI: -0.0067, 0.0026) and -0.0034 (95% CI: -0.0059, -0.0009) for hydrophilic and lipophilic statins respectively. The number needed to treat for hydrophilic and lipophilic statins were 2.1 (95% CI: -2.6, 6.7) and 3.4 (95% CI: 1.0, 5.9) per 1000 subjected, respectively.

Conclusions: Our results suggested that using lipophilic statin could significantly reduce risk of breast cancer in Thai women.

Variable Selection for Confounder Control, Flexible Modeling and Collaborative Targeted Minimum Loss-Based Estimation in Causal Inference

This paper investigates the appropriateness of the integration of flexible propensity score modeling (nonparametric or machine learning approaches) in semiparametric models for the estimation of a causal quantity, such as the mean outcome under treatment. We begin with an overview of some of the issues involved in knowledge-based and statistical variable selection in causal inference and the potential pitfalls of automated selection based on the fit of the propensity score. Using a simple example, we directly show the consequences of adjusting for pure causes of the exposure when using inverse probability of treatment weighting (IPTW). Such variables are likely to be selected when using a naive approach to model selection for the propensity score. We describe how the method of Collaborative Targeted minimum loss-based estimation (C-TMLE; van der Laan and Gruber, 2010 [27]) capitalizes on the collaborative double robustness property of semiparametric efficient estimators to select covariates for the propensity score based on the error in the conditional outcome model. Finally, we compare several approaches to automated variable selection in low- and high-dimensional settings through a simulation study. From this simulation study, we conclude that using IPTW with flexible prediction for the propensity score can result in inferior estimation, while Targeted minimum loss-based estimation and C-TMLE may benefit from flexible prediction and remain robust to the presence of variables that are highly correlated with treatment. However, in our study, standard influence function-based methods for the variance underestimated the standard errors, resulting in poor coverage under certain data-generating scenarios.

A propensity-matched cohort study of vancomycin-associated nephrotoxicity in neonates

BACKGROUND

The incidence of nephrotoxicity among vancomycin-treated neonates has been reported to range from 2% to 20%. These widely varying estimates have led to confusion and controversy regarding the safety of vancomycin among neonates.

OBJECTIVE

Evaluate the incidence of nephrotoxicity among neonates receiving vancomycin concomitantly with gentamicin.

DESIGN

Retrospective observational cohort study using propensity score matching to provide covariate balance between neonates who did or did not receive vancomycin based on factors known to be related to the development of renal dysfunction.

SETTING

Hospitals (n=22) throughout the Intermountain West, including a quaternary care children's hospital.

PATIENTS

Neonates ≤44 postmenstrual weeks (median gestational age: 31 (IQR 28-36) weeks) receiving intravenous gentamicin with or without exposure to vancomycin from January 2006 to December 2012.

MAIN OUTCOME MEASURES

Nephrotoxicity based on the modified Acute Kidney Injury Network criteria for acute kidney injury (AKI) or serum creatinine concentration ≥1.5 mg/dL persisting for ≥48 h.

RESULTS

The final cohort was comprised of 1066 neonates (533 receiving vancomycin and gentamicin vs 533 receiving gentamicin). In a propensity score-matched cohort that was well balanced across 16 covariates, AKI was not associated with vancomycin use (16 neonates receiving vancomycin vs 7 controls experienced AKI; OR 1.5; 95% CI 0.6 to 4.0). However, the presence of a patent ductus arteriosus, concomitant non-steroidal anti-inflammatory drug use, ≥1 positive blood cultures, low birth weight and higher severity of illness and risk of mortality scores were associated with an increased risk of nephrotoxicity.

CONCLUSIONS

These results corroborate several earlier reports and much anecdotal evidence describing the infrequent occurrence of nephrotoxicity in neonates receiving concomitant vancomycin and gentamicin.

Impact of Collaborative Care for Underserved Patients with PTSD in Primary Care: a Randomized Controlled Trial

BACKGROUND

The effectiveness of collaborative care of mental health problems is clear for depression and growing but mixed for anxiety disorders, including posttraumatic stress disorder (PTSD). We know little about whether collaborative care can be effective in settings that serve low-income patients such as Federally Qualified Health Centers (FQHCs).

OBJECTIVE

We compared the effectiveness of minimally enhanced usual care (MEU) versus collaborative care for PTSD with a care manager (PCM).

DESIGN

This was a multi-site patient randomized controlled trial of PTSD care improvement over 1 year.

PARTICIPANTS

We recruited and enrolled 404 patients in six FQHCs from June 2010 to October 2012. Patients were eligible if they had a primary care appointment, no obvious physical or cognitive obstacles to participation, were age 18-65 years, planned to continue care at the study location for 1 year, and met criteria for a past month diagnosis of PTSD.

MAIN MEASURES

The main outcomes were PTSD diagnosis and symptom severity (range, 0-136) based on the Clinician-Administered PTSD Scale (CAPS). Secondary outcomes were medication and counseling for mental health problems, and health-related quality of life assessed at baseline, 6 months, and 12 months.

KEY RESULTS

Patients in both conditions improved similarly over the 1-year evaluation period. At 12 months, PTSD diagnoses had an absolute decrease of 56.7% for PCM patients and 60.6% for MEU patients. PTSD symptoms decreased by 26.8 and 24.2 points, respectively. MEU and PCM patients also did not differ in process of care outcomes or health-related quality of life. Patients who actually engaged in care management had mental health care visits that were 14% higher (p < 0.01) and mental health medication prescription rates that were 15.2% higher (p < 0.01) than patients with no engagement.

CONCLUSIONS

A minimally enhanced usual care intervention was similarly effective as collaborative care for patients in FQHCs.

[Disease management programs: Difficulties in the analysis of benefit]

After an introduction to the theme with an overview of the implementation of the Disease Management Programs (DMP), accompanying documentation, present utilization and costs of the programs, the present article is primarily devoted to the issue of the analysis of the benefits of DMP. Following an assessment of the legally specified evaluation requirements, in the absence of a prospective, randomized, controlled trial (RCT), the results of three studies are first summarized, which, with the application of propensity score matching, utilize the routine data of the statutory health insurance schemes to form a control group. The overview concludes with a look at the planned changes in evaluation and the intended expansion of the DMP to programs for other chronic illnesses.

Impact of use of angiotensin II receptor blocker on all-cause mortality in hemodialysis patients: prospective cohort study using a propensity-score analysis

BACKGROUND

It is controversial whether treatment with an angiotensin II receptor blocker (ARB) or a calcium channel blocker (CCB) improves prognosis of hemodialysis (HD) patients.

METHODS

This study was designed as a multicenter prospective cohort study. HD patients (n = 1071) were enrolled from 22 institutes in January 2009 and followed up for 3 years. Patients with missing data, kidney transplantation or retraction of consent during the follow-up period (n = 204) were excluded, and 867 patients contributed to analysis of mortality. Propensity score (PS) for use of ARB and that for CCB was calculated using a multiple logistic regression model.

RESULTS

ARB and CCB were prescribed in 45.6 and 54.7 % of patients at enrollment. During the 3-year follow-up period, all-cause mortality and cardiovascular mortality rates were 18.8 and 5.1 %, respectively. Kaplan-Meier curves showed that all-cause and cardiovascular mortality rates were lower in the ARB group than in the non-ARB group, though the mortality rates were similar in the CCB group and non-CCB group. In PS-stratified Cox regression analysis, ARB treatment was associated with 34 and 45 % reduction of all-cause death and cardiovascular death, respectively. In PS matching analysis, ARB treatment was associated with a significant reduction (46 % reduction) in the risk of all-cause death. A significant impact of CCB treatment on all-cause or cardiovascular mortality was not detected in PS analysis.

CONCLUSIONS

The use of an ARB, but not a CCB, is associated with reduced all-cause and cardiovascular mortalities in patients on HD.

Fluoroquinolones versus β-Lactam/β-Lactamase Inhibitors in Outpatients with Chronic Obstructive Pulmonary Disease and Pneumonia: A Nationwide Population-Based Study

Background

Studies on the association between antibiotic treatment and outcomes in outpatients with chronic obstructive pulmonary disease (COPD) and pneumonia are scarce. This study aimed to evaluate the effectiveness of fluoroquinolones and β-lactam/β-lactamase inhibitors for pneumonia in COPD outpatients.

Methods

We conducted a retrospective cohort study and identified 4,851 episodes of pneumonia among COPD outpatients treated with fluoroquinolones or β-lactam/β-lactamase inhibitors from the Taiwan National Health Insurance Research Database during 2002–2011. Using the propensity score analysis, 1,296 pairs of episodes were matched for the demographic and clinical characteristics. The primary outcome was pneumonia/empyema-related hospitalization or emergency department (ED) visits, and the secondary outcomes were treatment failure, all-cause mortality and medical costs within 30 days.

Results

Compared with episodes treated with β-lactam/β-lactamase inhibitors, episodes treated with fluoroquinolones had similar clinical outcomes. The rates of pneumonia/empyema-related hospitalization or ED visits were 3.9% and 3.5% in the fluoroquinolone and β-lactam/β-lactamase inhibitor groups, respectively (adjusted hazard ratio [aHR], 1.11; 95% confidence interval [CI], 0.74–1.66). The percentage of treatment failure and all-cause mortality were 28.2% versus 31.3% (adjusted odds ratio, 0.86; 95% CI, 0.73–1.02) and 0.5% versus 0.4% (aHR, 1.40; 95% CI, 0.45–4.41) in the fluoroquinolone and β-lactam/β-lactamase inhibitor groups, respectively. The medical expenditures, including total medical costs (528 versus 455 US dollars) and pneumonia-related costs (202 vs. 155 USD) were also balanced between the two treatment groups (both P >0.05).

Conclusions

For pneumonia in COPD outpatients, fluoroquinolones were associated with similar clinical outcomes and medical expenditures compared with β-lactam/β-lactamase inhibitors.

Comparative effectiveness of aggressive thoracic radiation therapy and concurrent chemoradiation therapy in metastatic lung cancer

PURPOSE

We aimed to determine the comparative effectiveness of radiation dose escalation and concurrent chemoradiation therapy (CCRT) in a population-based cohort of patients with stage IV non-small cell lung cancer who underwent palliative thoracic radiation therapy (RT).

METHODS AND MATERIALS

The cohort consisted of 27,063 patients in the National Cancer Database with stage IV non-small cell lung cancer treated with thoracic RT between 20 and 55 Gy in 2004 to 2011. High- versus intermediate- vs low-dose (HD vs ID vs LD, respectively) RT was defined as biologically effective dose above 50 Gy, between 35 and 50 Gy, and below 35 Gy, respectively. Among patients who received any chemotherapy, separate analyses were performed to examine the impact of CCRT on overall survival (OS).

RESULTS

The median follow-up was 3.9 months for the entire cohort and 18 months for surviving patients. The 5 most common treatment schemes were 30/10 (Gy/fraction, 23% of entire cohort), 35/14 (8%), 37.5/15 (7%), 40/20 (3%), and 50/20 (3%). On multivariable analysis, the survival hazard ratios (HRs) for HD and ID compared with LD RT were 0.37 and 0.51, respectively (P < .0001). Propensity score matching found a superior survival benefit for ID and HD (HR, 0.41 and 0.57 for HD and ID RT, respectively, vs LD, P < .0001). Among those who received any chemotherapy (59% of total), the median OS for patients treated with CCRT (19% of total) was 5.3 versus 5.6 months (P = .667). On multivariable analysis, the HR for CCRT was 1.01 (P = .46).

CONCLUSIONS

The delivery of higher-dose RT but not concurrent chemotherapy was associated with a significant improvement of OS. This population-based study supports higher-dose palliative regimens and motivates prospective study of escalation beyond a biologically effective dose of 35 Gy.

Pharmacogenomic and pharmacogenetic-guided therapy as a tool in precision medicine: current state and factors impacting acceptance by stakeholders

Pharmacogenetic/pharmacogenomic (PGx) testing is currently available for a wide range of health problems including cardiovascular disease, cancer, diabetes, autoimmune disorders, mental health disorders and infectious diseases. PGx contributes important information to the field of precision medicine by clarifying appropriate treatments for specific disease subtypes. Tangible benefits to patients including improved outcomes and reduced total health care costs have been observed. However, PGx-guided therapy faces many barriers to full integration into clinical practice and acceptance by stakeholders, whether practitioner, patient or payer. Each stakeholder has a unique perspective on the role of PGx testing, although all are similarly challenged with demonstrating or appraising its cost-to-benefit value. Coverage by insurers is a critical step in achieving widespread adoption of PGx testing. The acceleration of adoption of precision medicine in general and for PGx testing in particular will be determined by how quickly robust evidence can be accumulated that shows a return on investment for payers in terms of real dollars, for clinicians in terms of patient clinical responses, and for patients in terms of economic, health and quality of life outcomes. Trends in PGx testing utilization and uptake by payers in real-world practice are discussed; the role of pharmacoeconomics in assessing cost-effectiveness is highlighted using a case study in psychiatric care, and several issues that will affect adoption of PGx testing in the United States (US) over the next few years are reviewed.

Propensity scores: a practical method for assessing treatment effects in pain and symptom management research

When conducting research on pain and symptom management interventions for seriously ill individuals, randomized controlled trials are not always feasible or ethical to conduct. Secondary analyses of observational data sets that include information on treatments experienced and outcomes for individuals who did and did not receive a given treatment can be conducted, but confounding because of selection bias can obscure the treatment effect in which one is interested. Propensity scores provide a way to adjust for observable characteristics that differ between treatment and comparison groups. This article provides conceptual guidance in addition to an empirical example to illustrate two areas of propensity score analysis that often lead to confusion in practice: covariate selection and interpretation of resultant treatment effects.

Methods for constructing and assessing propensity scores

OBJECTIVES

To model the steps involved in preparing for and carrying out propensity score analyses by providing step-by-step guidance and Stata code applied to an empirical dataset.

STUDY DESIGN

Guidance, Stata code, and empirical examples are given to illustrate (1) the process of choosing variables to include in the propensity score; (2) balance of propensity score across treatment and comparison groups; (3) balance of covariates across treatment and comparison groups within blocks of the propensity score; (4) choice of matching and weighting strategies; (5) balance of covariates after matching or weighting the sample; and (6) interpretation of treatment effect estimates.

EMPIRICAL APPLICATION

We use data from the Palliative Care for Cancer Patients (PC4C) study, a multisite observational study of the effect of inpatient palliative care on patient health outcomes and health services use, to illustrate the development and use of a propensity score.

CONCLUSIONS

Propensity scores are one useful tool for accounting for observed differences between treated and comparison groups. Careful testing of propensity scores is required before using them to estimate treatment effects.

Antibiotic Prescriptions for Upper Respiratory Infection in the Emergency Department: A Population-Based Study

Study Objective. Antibiotics prescriptions for upper respiratory infections (URI) are not uncommon, but the benefits for these groups had seldom been evaluated. We aimed to utilize a sampled National Health Insurance (NHI) claims data containing one million beneficiaries to explore if the use of antibiotics could reduce the possibility of unscheduled returns. Methods. We identified patients presented to ambulatory clinics with the discharged diagnoses of URI. The prescriptions of antibiotics were identified. We further matched each patient in the antibiotic group to the patient in the control group by selected covariates using a standard propensity score greedy-matching algorithm. The risks of unscheduled revisits were compared between the two groups. Results. A total of 6915140 visits were identified between 2005 and 2010. The proportions of antibiotics prescriptions are similar among these years, ranging from 9.99% to 13.38 %. In the propensity score assignment, 9190 patients (4595 in each group) were further selected. The odds ratio of unscheduled revisits among antibiotics group and control group was 0.92 (95% CI, 0.70–1.22) with P value equal to 0.569. Conclusions. Overall, antibiotics prescriptions did not seem to decrease the unscheduled revisits in patients presented to the ED with URI. Emergency physicians should reduce the unnecessary prescriptions and save antibiotics to patients with real benefits.

Effectiveness of tigecycline-based versus colistin- based therapy for treatment of pneumonia caused by multidrug-resistant Acinetobacter baumannii in a critical setting: a matched cohort analysis

Background

Colistin and tigecycline have both been shown good in vitro activity among multi-drug resistant Acinetobacter baumannii (MDRAB). A comparative study of colistin versus tigecycline for MDRAB pneumonia is lacking.

Methods

The study enrolled adults with MDRAB pneumonia admitted to intensive care units at a referral medical center during 2009–2010. Since there were no standardized minimum inhibitory concentration (MIC) interpretation criteria of tigecycline against A. baumannii, MIC of tigecycline was not routinely tested at our hospital. During the study periods, MIC of colistin was not routinely tested also. We consider both colistin and tigecycline as definite treatments of MDRAB pneumonia. Patients who received tigecycline were selected as potential controls for those who had received colistin. We performed a propensity score analysis, by considering the criteria of age, gender, underlying diseases, and disease severity, in order to match and equalize potential prognostic factors and severity in the two groups.

Results

A total of 294 adults with MDRAB pneumonia were enrolled, including 119 who received colistin and 175 who received tigecycline. We matched 84 adults who received colistin with an equal number of controls who received tigecycline. The two well matched cohorts share similar characteristics: the propensity scores are colistin: 0.37 vs. tigecycline: 0.37, (P = .97); baseline creatinine (1.70 vs. 1.81, P = .50), and the APACHE II score (21.6 vs. 22.0, P = .99). The tigecycline group has an excess mortality of 16.7% (60.7% vs. 44%, 95% confidence interval 0.9% – 32.4%, P = .04). The excess mortality of tigecycline is significant only among those with MIC >2 μg/mL (10/12 vs. 37/84, P = .01), but not for those with MIC ≦ 2 μg/mL (4/10 vs. 37/84, P = .81).

Conclusions

Our data disfavors the use of tigecycline-based treatment in treating MDRAB pneumonia when tigecycline and colistin susceptibilities are unknown, since choosing tigecycline-based treatment might result in higher mortality. The excess mortality of tigecycline-based group may be related to higher MIC of tigecycline (> 2 μg/mL). Choosing tigecycline empirically for treating MDRAB pneumonia in the critical setting should be cautious.

What is the most adapted indication of prophylactic pancreatic duct stent within the high-risk group of post-endoscopic retrograde cholangiopancreatography pancreatitis? Using the propensity score analysis

BACKGROUND

Conducting randomized controlled trial (RCT) for each of the risk factors associated with prophylactic pancreatic duct stent (PPDS) for post-endoscopic retrograde cholangiopancreatography (ERCP) pancreatitis (PEP) is difficult owing to the volume of cases and ethical considerations. In this study, we tried to reveal the degree of preventive effects of PPDS for each individual risk factor within the high-risk group of PEP using the propensity score analysis.

METHODS

The clinical data of 1131 ERCP practices performed at Kobe University Hospital from April 2006 to February 2009 were collected prospectively. We investigated their clinical characteristics including the risk factors of PEP, the use of PPDS and complications of ERCP. We conducted the stratification analysis using the propensity score matching analysis.

RESULTS

In 210 propensity score-matched ERCPs, PPDS proved to be effective in preventing PEP in patients with a history of pancreatitis (odds ratio 0.11, 95% CI 0.01-0.76, P = 0.01) and cases of difficult cannulation (requiring more than 30 min) (odds ratio 0.13, 95% CI 0.01-1.14, P = 0.08).

CONCLUSIONS

Patients with a history of pancreatitis and cases of difficult cannulation are strongly recommended for PPDS placement. The propensity score analysis can be adapted to the ERCP-related analysis with many procedure-related factors with using retrospective data, and may be adapted to investigate the matters that are unsuitable for RCT by volume and ethical issue.

Early Endoscopy Can Shorten the Duration of Hospitalization in Suspected Variceal Hemorrhage

Background. Endoscopic treatment in patients with acute variceal hemorrhage should be performed within 12 hours, but the recommendation is based on the experts’ opinion. Objective. Our study investigated if time to endoscopy was a significant factor that could alter the outcomes of patients with suspected variceal hemorrhage. The primary outcome was the length of hospital stay and the mortality. The secondary outcomes were complications during observation periods. Methods. Patients were included if variceal hemorrhage was suspected in the emergency department. Patients were further divided into early (receiving endoscopy within 12 hours) and delayed groups (receiving endoscopy after 12 hours), and each patient was matched using a standard propensity score greedy-matching algorithm. The primary and secondary outcomes were compared accordingly. Results. 1442 patients met our inclusion criteria and therefore were enrolled for further analysis. In the Cox regression model, log time to endoscopy was not a significant factor. In the propensity score assignment, 566 patients (283 in each group) were further selected into the subcohort (P value = 0.8001). Kaplan-Meier curves showed a discharge benefit favoring the early endoscopy over the control group. The mortality rates and complications were not statistically different between the two groups (P value = 0.0045). Conclusion. Early endoscopy before 12 hours in patients with suspected variceal hemorrhage could result in shorter length of hospitalization without increasing the mortalities and complications. The results would help emergency physicians in decisions making when these patients are encountered.

The social determinants of infant mortality and birth outcomes in Western developed nations: a cross-country systematic review

Infant mortality (IM) and birth outcomes, key population health indicators, have lifelong implications for individuals, and are unequally distributed globally. Even among western industrialized nations, striking cross-country and within-country patterns are evident. We sought to better understand these variations across and within the United States of America (USA) and Western Europe (WE), by conceptualizing a social determinants of IM/birth outcomes framework, and systematically reviewing the empirical literature on hypothesized social determinants (e.g., social policies, neighbourhood deprivation, individual socioeconomic status (SES)) and intermediary determinants (e.g., health behaviours). To date, the evidence suggests that income inequality and social policies (e.g., maternal leave policies) may help to explain cross-country variations in IM/birth outcomes. Within countries, the evidence also supports neighbourhood SES (USA, WE) and income inequality (USA) as social determinants. By contrast, within-country social cohesion/social capital has been underexplored. At the individual level, mixed associations have been found between individual SES, race/ethnicity, and selected intermediary factors (e.g., psychosocial factors) with IM/birth outcomes. Meanwhile, this review identifies several methodological gaps, including the underuse of prospective designs and the presence of residual confounding in a number of studies. Ultimately, addressing such gaps including through novel approaches to strengthen causal inference and implementing both health and non-health policies may reduce inequities in IM/birth outcomes across the western developed world.

Statin therapy in the reduction of cardiovascular events in patients undergoing intermediate-risk noncardiac, nonvascular surgery

BACKGROUND

Hydroxymethylglutaryl-coenzyme A reductase inhibitors (statins) reduce perioperative cardiac events in high-risk patients undergoing cardiovascular surgery. However, there is paucity of data on the role of statins in patients undergoing intermediate-risk noncardiac, nonvascular surgery (NCNVS).

HYPOTHESIS

Statins are cardioprotective in intermediate-risk NCNVS.

METHODS

We identified a retrospective cohort of patients undergoing intermediate risk NCNVS. Our composite end point (CEP) included 30-day all-cause mortality, atrial fibrillation (AF), and nonfatal myocardial infarction (MI). A stepwise logistic regression with adjustment using propensity scores was performed to determine if statin therapy was independently associated with the risk reduction of adverse postoperative cardiovascular outcomes.

RESULTS

We identified 752 patients. Seventy-five of them (9.97%) developed composite end points; 10 (1.33%) had in-hospital nonfatal MI, 44 (5.85%) developed AF, and 35 (4.65%) died within 30 days. The 30-day all-cause mortality was 31/478 (6.48%) among statin nonusers vs 4/274 (1.45%) for statin users (P < 0.002). As compared with nonusers, patients on statin therapy had a 5-fold reduced risk of 30-day all-cause mortality. Statin therapy was associated with decreased CEP after adjusting for baseline characteristics, with a propensity score to predict use of statins (odds ratio [OR]: 0.54, 95% confidence interval [CI]: 0.30-0.97, P = 0.039). After further adjustment for propensity score, diabetes mellitus, percutaneous coronary intervention, and prior coronary artery bypass grafting, statin therapy proved beneficial (OR: 0.51, 95% CI: 0.28-0.92, P = 0.026).

CONCLUSIONS

Statin use in the perioperative period was associated with a reduction in cardiovascular adverse events and 30-day all-cause mortality in patients undergoing intermediate-risk NCNVS.

Nonresponse to interferon-α based treatment for chronic hepatitis C infection is associated with increased hazard of cirrhosis

Background

The long-term consequences of unsuccessful interferon-α based hepatitis C treatment on liver disease progression and survival have not been fully explored.

Methods and Findings

We performed retrospective analyses to assess long-term clinical outcomes among treated and untreated patients with hepatitis C virus in two independent cohorts from a United States Veterans Affairs Medical Center and a University Teaching Hospital. Eligible patients underwent liver biopsy during consideration for interferon-α based treatment between 1992 and 2007. They were assessed for the probability of developing cirrhosis and of dying during follow-up using Cox proportional hazards models, stratified by pretreatment liver fibrosis stage and adjusted for known risk factors for cirrhosis and characteristics affecting treatment selection. The major predictor was a time-dependent covariate for treatment outcome among four patient groups: 1) patients with sustained virological response to treatment; 2) treatment relapsers; 3) treatment nonresponders; and 4) never treated patients. Treatment nonresponders in both cohorts had a statistically significantly increased hazard of cirrhosis compared to never treated patients, as stratified by pretreatment liver fibrosis stage and adjusted for clinical and psychosocial risk factors that disproportionately affect patients who were ineligible for treatment (Veterans Affairs HR = 2.35, CI 1.18–4.69, mean follow-up 10 years, and University Hospital HR = 5.90, CI 1.50–23.24, mean follow-up 7.7 years). Despite their increased risk for liver disease progression, the overall survival of nonresponders in both cohorts was not significantly different from that of never treated patients.

Conclusion

These unexpected findings suggest that patients who receive interferon-α based therapies but fail to clear the hepatitis C virus may have an increased hazard of cirrhosis compared to untreated patients.

Kin KeeperSM: design and baseline characteristics of a community-based randomized controlled trial promoting cancer screening in Black, Latina, and Arab women

BACKGROUND

Although breast and cervical cancer deaths have declined due to early screening, detection, and more effective treatment, racial and ethnic disparities persist. This paper describes the study design and baseline characteristics of a randomized controlled trial (RCT) evaluating the effectiveness of the Kin Keeper(SM) Cancer Prevention Intervention, a family-focused educational intervention for underserved women applied in a community-based setting to promote health literacy and screening adherence to address cancer disparities.

METHODS

Female public health community health workers (CHWs) were trained to administer the intervention. They recruited female clients from their public health program caseload and asked each to assemble two to four adult female family members for the breast and cervical cancer home-based education sessions the CHWs would deliver in English, Spanish or Arabic. We randomized the clients into the kin keeper group (treatment) or the participant client group (control).

RESULTS

Complete data were obtained on 514 Black, Latina, and Arab women. Close to half were unemployed and had yearly family income below $20,000. Thirty-four percent had no medical insurance, and 21% had diabetes. Almost 40% had no mammography in the last year. Treatment and control groups were similar on most sociodemographics but showed differences in breast and cervical screening history.

CONCLUSIONS

This innovative study demonstrates the implementation of an RCT using community-based participatory research, while delivering cancer prevention education across woman's life span with women not connected to the health care system.

Tobacco smoke exposure and allergic sensitization in children: a propensity score analysis

BACKGROUND AND OBJECTIVE

There is conflicting evidence of the effect of environmental tobacco smoke (ETS) on the development of allergic diseases in children. Studies have shown that this relationship differs depending on maternal history of the disease. We employed the rigour of propensity score methods to assess this relationship using data from a birth cohort.

METHODS

Using n = 662 children from the Wayne County Health, Environment, Allergy and Asthma Longitudinal Study, we assessed the relationship between early-life ETS and subsequent allergic sensitization via a positive skin prick test (SPT+) or at least one specific immunoglobulin E (IgE) ≥ 0.35 (sIgE+) in children aged 2-3 years. Propensity score estimation followed by full and nearest neighbour matching was compared with standard multivariable regression models.

RESULTS

Among children without a maternal history of allergic disease, ETS was positively associated with allergic sensitization in children with an adjusted odds ratio (aOR) for SPT+ of 2.32 (95% confidence interval (CI): 1.28-4.22) and the aOR for sIgE+ was 2.53 (95% CI: 1.43-4.48). Contrarily, for children with a positive maternal history, the aOR for SPT+ and sIgE+ was 0.56 (95% CI: 0.24-1.32) and 0.43 (95% CI: 0.20-0.91), respectively.

CONCLUSIONS

Using propensity score methods to rigorously control for confounding factors, ETS exposure was found to reduce the risk of allergic sensitization in children with a positive maternal history. There is a strong association between early-life ETS and the development of allergic sensitization for children aged 2-3 years without maternal history.

Intravenous contrast material-induced nephropathy: causal or coincident phenomenon?

PURPOSE

To determine the causal association and effect of intravenous iodinated contrast material exposure on the incidence of acute kidney injury (AKI), also known as contrast material-induced nephropathy (CIN).

MATERIALS AND METHODS

This retrospective study was approved by an institutional review board and was HIPAA compliant. Informed consent was waived. All contrast material-enhanced (contrast group) and unenhanced (noncontrast group) abdominal, pelvic, and thoracic CT scans from 2000 to 2010 were identified at a single facility. Scan recipients were sorted into low- (<1.5 mg/dL), medium- (1.5-2.0 mg/dL), and high-risk (>2.0 mg/dL) subgroups of presumed risk for CIN by using baseline serum creatinine (SCr) level. The incidence of AKI (SCr ≥ 0.5 mg/dL above baseline) was compared between contrast and noncontrast groups after propensity score adjustment by stratification, 1:1 matching, inverse weighting, and weighting by the odds methods to reduce intergroup selection bias. Counterfactual analysis was used to evaluate the causal relation between contrast material exposure and AKI by evaluating patients who underwent contrast-enhanced and unenhanced CT scans during the study period with the McNemar test.

RESULTS

A total of 157,140 scans among 53,439 unique patients associated with 1,510,001 SCr values were identified. AKI risk was not significantly different between contrast and noncontrast groups in any risk subgroup after propensity score adjustment by using reported risk factors of CIN (low risk: odds ratio [OR], 0.93; 95% confidence interval [CI]: 0.76, 1.13; P = .47; medium risk: odds ratio, 0.97; 95% CI: 0.81, 1.16; P = .76; high risk: OR, 0.91; 95% CI: 0.66, 1.24; P = .58). Counterfactual analysis revealed no significant difference in AKI incidence between enhanced and unenhanced CT scans in the same patient (McNemar test: χ(2) = 0.63, P = .43) (OR = 0.92; 95% CI: 0.75, 1.13; P = .46).

CONCLUSION

Following adjustment for presumed risk factors, the incidence of CIN was not significantly different from contrast material-independent AKI. These two phenomena were clinically indistinguishable with established SCr-defined criteria, suggesting that intravenous iodinated contrast media may not be the causative agent in diminished renal function after contrast material administration.

SUPPLEMENTAL MATERIAL

http://radiology.rsna.org/lookup/suppl/doi:10.1148/radiol.12121823/-/DC1.

Using risk models to improve patient selection for high-risk vascular surgery

Vascular surgeons frequently perform procedures aimed at limiting death, stroke, or amputation on patients who present with diseases such as aortic aneurysms, carotid atherosclerosis, and peripheral arterial occlusive disease. However, now more than ever surgeons must balance the potential benefits associated with these interventions with the risks of physiologic insult for these elderly patients, who often have significant comorbidity burdens and the potential for costly complications. In this paper, we highlight how regional and national datasets can help surgeons identify which patients are most likely to benefit from vascular operations and which patients are most likely to suffer complications in the postoperative period. By using these guidelines to improve patient selection, our risk models can help patients, physicians, and policymakers improve the clinical effectiveness of surgical and endovascular treatments for vascular disease.

The impact of parental drinking on children’s use of health care

While a significant body of literature documents the health problems of children caused by and/or associated with parental alcohol misuse, little research has been conducted on the relationship between parental problem drinking and children's use of health care. We should expect to see an increase in children's health care if alcohol-misusing parents were responsive to their children's higher physical and mental health needs. Contrarily, it would decrease (conditional on health status) if alcohol-misusing parents were irresponsive to those needs. Analyzing a nationally representative sample of parents and children, we find a positive and significant association between parental high intensity drinking and pediatric visits for their children.We also find evidence linking parental drinking to more emergency room use. These findings suggest that the impact of parental drinking on child wellbeing should be considered when assessing the full costs of alcohol misuse.

Statin use and the presence of microalbuminuria. Results from the ERICABEL trial: a non-interventional epidemiological cohort study

BACKGROUND

Microalbuminuria (MAU) is considered as a predictor or marker of cardiovascular and renal events. Statins are widely prescribed to reduce cardiovascular risk and to slow down progression of kidney disease. But statins may also generate tubular MAU. The current observational study evaluated the impact of statin use on the interpretation of MAU as a predictor or marker of cardiovascular or renal disease.

METHODOLOGY/PRINCIPAL FINDINGS

We used cross-sectional data of ERICABEL, a cohort with 1,076 hypertensive patients. MAU was defined as albuminuria ≥20 mg/l. A propensity score was created to correct for "bias by indication" to receive a statin. As expected, subjects using statins vs. no statins had more cardiovascular risk factors, pointing to bias by indication. Statin users were more likely to have MAU (OR: 2.01, 95%CI: 1.34-3.01). The association between statin use and MAU remained significant after adjusting for the propensity to receive a statin based on cardiovascular risk factors (OR: 1.82, 95%CI: 1.14-2.91). Next to statin use, only diabetes (OR: 1.92, 95%CI: 1.00-3.66) and smoking (OR: 1.49, 95%CI: 0.99-2.26) were associated with MAU.

CONCLUSIONS

Use of statins is independently associated with MAU, even after adjusting for bias by indication to receive a statin. In the hypothesis that this MAU is of tubular origin, statin use can result in incorrect labeling of subjects as having a predictor or marker of cardiovascular or renal risk. In addition, statin use affected the association of established cardiovascular risk factors with MAU, blurring the interpretation of multivariable analyses.

Macrolide antibiotics and survival in patients with acute lung injury

BACKGROUND

Animal models suggest that immunomodulatory properties of macrolide antibiotics have therapeutic value for patients with acute lung injury (ALI). We investigated the association between receipt of macrolide antibiotics and clinical outcomes in patients with ALI.

METHODS

Secondary analysis of multicenter, randomized controlled trial data from the Acute Respiratory Distress Syndrome Network Lisofylline and Respiratory Management of Acute Lung Injury Trial, which collected detailed data regarding antibiotic use among participants with ALI.

RESULTS

Forty-seven of 235 participants (20%) received a macrolide antibiotic within 24 h of trial enrollment. Among patients who received a macrolide, erythromycin was the most common (57%), followed by azithromycin (40%). The median duration of macrolide use after study enrollment was 4 days (interquartile range, 2-8 days). Eleven of the 47 (23%) patients who received macrolides died, compared with 67 of the 188 (36%) who did not receive a macrolide (P = .11). Participants administered macrolides were more likely to have pneumonia as an ALI risk factor, were less likely to have nonpulmonary sepsis or to be randomized to low tidal volume ventilation, and had a shorter length of stay prior to trial enrollment. After adjusting for potentially confounding covariates, use of macrolide was associated with lower 180-day mortality (hazard ratio [HR], 0.46; 95% CI, 0.23-0.92; P = .028) and shorter time to successful discontinuation of mechanical ventilation (HR, 1.93; 95% CI, 1.18-3.17; P = .009). In contrast, fluoroquinolone (n = 90) and cephalosporin antibiotics (n = 93) were not associated with improved outcomes.

CONCLUSIONS

Receipt of macrolide antibiotics was associated with improved outcomes in patients with ALI.