Sebastian Schneeweiss

High-dimensional propensity score adjustment in studies of treatment effects using health care claims data

Background: Adjusting for large numbers of covariates ascertained from patients’ health care claims data may improve control of confounding, as these variables may collectively be proxies for unobserved factors. Here, we develop and test an algorithm that empirically identifies candidate covariates, prioritizes covariates, and integrates them into a propensity-score-based confounder adjustment model. Methods: We developed a multistep algorithm to implement high-dimensional proxy adjustment in claims data. Steps include (1) identifying data dimensions, eg, diagnoses, procedures, and medications; (2) empirically identifying candidate covariates; (3) assessing recurrence of codes; (4) prioritizing covariates; (5) selecting covariates for adjustment; (6) estimating the exposure propensity score; and (7) estimating an outcome model. This algorithm was tested in Medicare claims data, including a study on the effect of Cox-2 inhibitors on reduced gastric toxicity compared with nonselective nonsteroidal anti-inflammatory drugs (NSAIDs). Results: In a population of 49,653 new users of Cox-2 inhibitors or nonselective NSAIDs, a crude relative risk (RR) for upper GI toxicity (RR = 1.09 [95% confidence interval = 0.91–1.30]) was initially observed. Adjusting for 15 predefined covariates resulted in a possible gastroprotective effect (0.94 [0.78–1.12]). A gastroprotective effect became stronger when adjusting for an additional 500 algorithm-derived covariates (0.88 [0.73–1.06]). Results of a study on the effect of statin on reduced mortality were similar. Using the algorithm adjustment confirmed a null finding between influenza vaccination and hip fracture (1.02 [0.85–1.21]). Conclusions: In typical pharmacoepidemiologic studies, the proposed high-dimensional propensity score resulted in improved effect estimates compared with adjustment limited to predefined covariates, when benchmarked against results expected from randomized trials.

Relationship Between Selective Cyclooxygenase-2 Inhibitors and Acute Myocardial Infarction in Older Adults

Background—Although cyclooxygenase-2 inhibitors (coxibs) were developed to cause less gastrointestinal hemorrhage than nonselective nonsteroidal antiinflammatory drugs (NSAIDs), there has been concern about their cardiovascular safety. We studied the relative risk of acute myocardial infarction (AMI) among users of celecoxib, rofecoxib, and NSAIDs in Medicare beneficiaries with a comprehensive drug benefit. Methods and Results—We conducted a matched case-control study of 54 475 patients 65 years of age or older who received their medications through 2 state-sponsored pharmaceutical benefits programs in the United States. All healthcare use encounters were examined to identify hospitalizations for AMI. Each of the 10 895 cases of AMI was matched to 4 controls on the basis of age, gender, and the month of index date. We constructed matched logistic regression models including indicators for patient demographics, healthcare use, medication use, and cardiovascular risk factors to assess the relative risk of AMI in patients who used rofecoxib compared with persons taking no NSAID, taking celecoxib, or taking NSAIDs. Current use of rofecoxib was associated with an elevated relative risk of AMI compared with celecoxib (odds ratio [OR], 1.24; 95% CI, 1.05 to 1.46; P = 0.011) and with no NSAID (OR, 1.14; 95% CI, 1.00 to 1.31; P = 0.054). The adjusted relative risk of AMI was also elevated in dose-specific comparisons: rofecoxib ≤25 mg versus celecoxib ≤200 mg (OR, 1.21; 95% CI, 1.01 to 1.44; P = 0.036) and rofecoxib >25 mg versus celecoxib >200 mg (OR, 1.70; 95% CI, 1.07 to 2.71; P = 0.026). The adjusted relative risks of AMI associated with rofecoxib use of 1 to 30 days (OR, 1.40; 95% CI, 1.12 to 1.75; P = 0.005) and 31 to 90 days (OR, 1.38; 95% CI, 1.11 to 1.72; P = 0.003) were higher than >90 days (OR, 0.96; 95% CI, 0.72 to 1.25; P = 0.8) compared with celecoxib use of similar duration. Celecoxib was not associated with an increased relative risk of AMI in these comparisons. Conclusions—In this study, current rofecoxib use was associated with an elevated relative risk of AMI compared with celecoxib use and no NSAID use. Dosages of rofecoxib >25 mg were associated with a higher risk than dosages ≤25 mg. The risk was elevated in the first 90 days of use but not thereafter.

Full Coverage for Preventive Medications after Myocardial Infarction

BACKGROUND Adherence to medications that are prescribed after myocardial infarction is poor. Eliminating out-of-pocket costs may increase adherence and improve outcomes. METHODS We enrolled patients discharged after myocardial infarction and randomly assigned their insurance-plan sponsors to full prescription coverage (1494 plan sponsors with 2845 patients) or usual prescription coverage (1486 plan sponsors with 3010 patients) for all statins, beta-blockers, angiotensin-converting-enzyme inhibitors, or angiotensin-receptor blockers. The primary outcome was the first major vascular event or revascularization. Secondary outcomes were rates of medication adherence, total major vascular events or revascularization, the first major vascular event, and health expenditures. RESULTS Rates of adherence ranged from 35.9 to 49.0% in the usual-coverage group and were 4 to 6 percentage points higher in the full-coverage group (P<0.001 for all comparisons). There was no significant between-group difference in the primary outcome (17.6 per 100 person-years in the full-coverage group vs. 18.8 in the usual-coverage group; hazard ratio, 0.93; 95% confidence interval [CI], 0.82 to 1.04; P=0.21). The rates of total major vascular events or revascularization were significantly reduced in the full-coverage group (21.5 vs. 23.3; hazard ratio, 0.89; 95% CI, 0.90 to 0.99; P=0.03), as was the rate of the first major vascular event (11.0 vs. 12.8; hazard ratio, 0.86; 95% CI, 0.74 to 0.99; P=0.03). The elimination of copayments did not increase total spending (

Patterns of cardiovascular risk in rheumatoid arthritis

66,008 for the full-coverage group and

A combined comorbidity score predicted mortality in elderly patients better than existing scores

71,778 for the usual-coverage group; relative spending, 0.89; 95% CI, 0.50 to 1.56; P=0.68). Patient costs were reduced for drugs and other services (relative spending, 0.74; 95% CI, 0.68 to 0.80; P<0.001). CONCLUSIONS The elimination of copayments for drugs prescribed after myocardial infarction did not significantly reduce rates of the trials primary outcome. Enhanced prescription coverage improved medication adherence and rates of first major vascular events and decreased patient spending without increasing overall health costs. (Funded by Aetna and the Commonwealth Fund; MI FREEE ClinicalTrials.gov number, NCT00566774.).

Risk of death associated with the use of conventional versus atypical antipsychotic drugs among elderly patients

Background: Although it is known that rheumatoid arthritis is associated with an increased risk of cardiovascular disease (CVD), the pattern of this risk is not clear. This study investigated the relative risk of myocardial infarction, stroke and CVD mortality in adults with rheumatoid arthritis compared with adults without rheumatoid arthritis across age groups, sex and prior CVD event status. Methods: We conducted a cohort study among all residents aged ⩾18 years residing in British Columbia between 1999 and 2003. Residents who had visited the doctor at least thrice for rheumatoid arthritis (International Classification of Disease = 714) were considered to have rheumatoid arthritis. A non-rheumatoid arthritis cohort was matched to the rheumatoid arthritis cohort by age, sex and start of follow-up. The primary composite end point was a hospital admission for myocardial infarction, stroke or CVD mortality. Results: 25 385 adults who had at least three diagnoses for rheumatoid arthritis during the study period were identified. During the 5-year study period, 375 patients with rheumatoid arthritis had a hospital admission for myocardial infarction, 363 had a hospitalisation for stroke, 437 died from cardiovascular causes and 1042 had one of these outcomes. The rate ratio for a CVD event in patients with rheumatoid arthritis was 1.6 (95% confidence interval (CI) 1.5 to 1.7), and the rate difference was 5.7 (95% CI 4.9 to 6.4) per 1000 person-years. The rate ratio decreased with age, from 3.3 in patients aged 18–39 years to 1.6 in those aged ⩾75 years. However, the rate difference was 1.2 per 1000 person-years in the youngest age group and increased to 19.7 per 1000 person-years in those aged ⩾75 years. Among patients with a prior CVD event, the rate ratios and rate differences were not increased in rheumatoid arthritis. Conclusions: This study confirms that rheumatoid arthritis is a risk factor for CVD events and shows that the rate ratio for CVD events among subjects with rheumatoid arthritis is highest in young adults and those without known prior CVD events. However, in absolute terms, the difference in event rates is highest in older adults.

Incidence and predictors for chronicity of headache in patients with episodic migraine

OBJECTIVE To develop and validate a single numerical comorbidity score for predicting short- and long-term mortality, by combining conditions in the Charlson and Elixhauser measures. STUDY DESIGN AND SETTING In a cohort of 120,679 Pennsylvania Medicare enrollees with drug coverage through a pharmacy assistance program, we developed a single numerical comorbidity score for predicting 1-year mortality, by combining the conditions in the Charlson and Elixhauser measures. We externally validated the combined score in a cohort of New Jersey Medicare enrollees, by comparing its performance to that of both component scores in predicting 1-year mortality, as well as 180-, 90-, and 30-day mortality. RESULTS C-statistics from logistic regression models including the combined score were higher than corresponding c-statistics from models including either the Romano implementation of the Charlson Index or the single numerical version of the Elixhauser system; c-statistics were 0.860 (95% confidence interval [CI]: 0.854, 0.866), 0.839 (95% CI: 0.836, 0.849), and 0.836 (95% CI: 0.834, 0.847), respectively, for the 30-day mortality outcome. The combined comorbidity score also yielded positive values for two recently proposed measures of reclassification. CONCLUSION In similar populations and data settings, the combined score may offer improvements in comorbidity summarization over existing scores.

Aprotinin during Coronary-Artery Bypass Grafting and Risk of Death

Background: Public health advisories have warned that the use of atypical antipsychotic medications increases the risk of death among elderly patients. We assessed the short-term mortality in a population-based cohort of elderly people in British Columbia who were prescribed conventional and atypical antipsychotic medications. Methods: We used linked health care utilization data of all BC residents to identify a cohort of people aged 65 years and older who began taking antipsychotic medications between January 1996 and December 2004 and were free of cancer. We compared the 180-day all-cause mortality between residents taking conventional antipsychotic medications and those taking atypical antipsychotic medications. Results: Of 37 241 elderly people in the study cohort, 12 882 were prescribed a conventional antipsychotic medication and 24 359 an atypical formulation. Within the first 180 days of use, 1822 patients (14.1%) in the conventional drug group died, compared with 2337 (9.6%) in the atypical drug group (mortality ratio 1.47, 95% confidence interval [CI] 1.39–1.56). Multivariable adjustment resulted in a 180-day mortality ratio of 1.32 (1.23–1.42). In comparison with risperidone, haloperidol was associated with the greatest increase in mortality (mortality ratio 2.14, 95% CI 1.86–2.45) and loxapine the lowest (mortality ratio 1.29, 95% CI 1.19–1.40). The greatest increase in mortality occurred among people taking higher (above median) doses of conventional antipsychotic medications (mortality ratio 1.67, 95% CI 1.50–1.86) and during the first 40 days after the start of drug therapy (mortality ratio 1.60, 95% CI 1.42–1.80). Results were confirmed in propensity score analyses and instrumental variable estimation, minimizing residual confounding. Interpretation: Among elderly patients, the risk of death associated with conventional antipsychotic medications is comparable to and possibly greater than the risk of death associated with atypical antipsychotic medications. Until further evidence is available, physicians should consider all antipsychotic medications to be equally risky in elderly patients.

The Comparative Safety of Analgesics in Older Adults With Arthritis

The authors followed 532 consecutive patients with episodic migraine (<15 days/month) for 1 year. Sixty-four patients (14%) developed chronic headache (≥15 days/month). The odds ratios for developing CH were 20.1 (95% CI 5.7 to 71.5) comparing patients with a “critical” (10 to 14 days/month) vs “low” (0 to 4 days/month) and 6.2 (95% CI 1.7 to 26.6) in patients with an “intermediate” (6 to 9 days/month) vs “low” headache frequency and 19.4 (95% CI 8.7 to 43.2) comparing patients with and without medication overuse.

Association Between Disease-Modifying Antirheumatic Drugs and Diabetes Risk in Patients With Rheumatoid Arthritis and Psoriasis

BACKGROUND Aprotinin (Trasylol) is used to mitigate bleeding during coronary-artery bypass grafting (CABG). Accumulating evidence suggests that this practice increases mortality. METHODS Using electronic administrative records of the Premier Perspective Comparative Database, we studied hospitalized patients with operating-room charges for the use of aprotinin (33,517 patients) or aminocaproic acid (44,682 patients) on the day CABG was performed. We tabulated the numbers of patients with a hospital-discharge status of death and performed three types of analyses: a multivariable logistic-regression analysis (primary analysis); propensity-score matching in the highly selected subcohort of patients who received full amounts of the study drug, who underwent CABG by surgeons who performed 50 or more CABG surgeries during the study period, and for whom information on 10 additional covariates was available because the surgery occurred on hospital day 3 or later; and an instrumental-variable analysis of data from patients whose surgeons showed a strong preference for one of the two study drugs. RESULTS In all, 1512 of the 33,517 aprotinin recipients (4.5%) and 1101 of the 44,682 aminocaproic acid recipients (2.5%) died. After adjustment for 41 characteristics of patients and hospitals, the estimated risk of death was 64% higher in the aprotinin group than in the aminocaproic acid group (relative risk, 1.64; 95% confidence interval [CI], 1.50 to 1.78). In the first 7 days after surgery, the adjusted relative risk of in-hospital death in the aprotinin group was 1.78 (95% CI, 1.56 to 2.02). The relative risk in a propensity-score-matched analysis was 1.32 (95% CI, 1.08 to 1.63). In the instrumental-variable analysis, the use of aprotinin was found to be associated with an excess risk of death of 1.59 per 100 patients (95% CI, 0.14 to 3.04). Postoperative revascularization and dialysis were more frequent among recipients of aprotinin than among recipients of aminocaproic acid. CONCLUSIONS Patients who received aprotinin alone on the day of CABG surgery had a higher mortality than patients who received aminocaproic acid alone. Characteristics of neither the patients nor the surgeons explain the difference, which persisted through several approaches to control confounding.