RKHS-based covariate balancing for survival causal effect estimation.

Survival causal effect estimation based on right-censored data is of key interest in both survival analysis and causal inference. Propensity score weighting is one of the most popular methods in the literature. However, since it involves the inverse of propensity score estimates, its practical performance may be very unstable, especially when the covariate overlap is limited between treatment and control groups. To address this problem, a covariate balancing method is developed in this paper to estimate the counterfactual survival function. The proposed method is nonparametric and balances covariates in a reproducing kernel Hilbert space (RKHS) via weights that are counterparts of inverse propensity scores. The uniform rate of convergence for the proposed estimator is shown to be the same as that for the classical Kaplan-Meier estimator. The appealing practical performance of the proposed method is demonstrated by a simulation study as well as two real data applications to study the causal effect of smoking on survival time of stroke patients and that of endotoxin on survival time for female patients with lung cancer respectively.

A community-driven hypertension treatment group in rural Honduras.

BACKGROUND: We formed a self-funded hypertension treatment group in a resource-poor community in rural Honduras. After training community health workers and creating protocols for standardized treatment, we used group membership fees to maintain the group, purchase generic medications in bulk on the local market, and hire a physician to manage treatment. We then assessed whether participation in the group improved treatment, medication adherence, and hypertension control. DESIGN: This is a program evaluation using quasi-experimental design and no control group. Using data from the 86 members of the hypertension treatment group, we analyzed baseline and follow-up surveys of members, along with 30 months of clinical records of treatment, medication adherence, and blood pressure readings. RESULTS: Our initial hypertension needs assessment revealed that at baseline, community hypertensives relied on the local Ministry of Health clinic as their source of anti-hypertensive medications and reported that irregular supply interfered with medication adherence. At baseline, hypertension group members were mainly female, overweight or obese, physically active, non-smoking, and non-drinking. After 30 months of managing the treatment group, we found a significant increase in medication adherence, from 54.8 to 76.2% (p<0.01), and hypertension control (<140/90 mmHg), from 31.4 to 54.7% (p<0.01). We also found a mean monthly decrease of 0.39 mmHg in systolic blood pressure (p<0.01). At the end of the 30-month observation period, the local Ministry of Health system had increased provision of low-cost anti-hypertensive medications and adopted the hypertension treatment group's treatment protocols. CONCLUSIONS: Formation of a self-funded, community-based hypertension treatment group in a rural, resource-poor community is feasible, and group participation may improve treatment, medication adherence, and hypertension control and can serve as a political driver for improving hypertension treatment services provided by the public system.

Early predictors of lumbar spine surgery after occupational back injury: results from a prospective study of workers in Washington State.

STUDY DESIGN: Prospective population-based cohort study. OBJECTIVE: To identify early predictors of lumbar spine surgery within 3 years after occupational back injury. SUMMARY OF BACKGROUND DATA: Back injuries are the most prevalent occupational injury in the United States. Few prospective studies have examined early predictors of spine surgery after work-related back injury. METHODS: Using Disability Risk Identification Study Cohort (D-RISC) data, we examined the early predictors of lumbar spine surgery within 3 years among Washington State workers, with new workers compensation temporary total disability claims for back injuries. Baseline measures included worker-reported measures obtained approximately 3 weeks after claim submission. We used medical bill data to determine whether participants underwent surgery, covered by the claim, within 3 years. Baseline predictors (P < 0.10) of surgery in bivariate analyses were included in a multivariate logistic regression model predicting lumbar spine surgery. The area under the receiver operating characteristic curve of the model was used to determine the model's ability to identify correctly workers who underwent surgery. RESULTS: In the D-RISC sample of 1885 workers, 174 (9.2%) had a lumbar spine surgery within 3 years. Baseline variables associated with surgery (P < 0.05) in the multivariate model included higher Roland-Morris Disability Questionnaire scores, greater injury severity, and surgeon as first provider seen for the injury. Reduced odds of surgery were observed for those younger than 35 years, females, Hispanics, and those whose first provider was a chiropractor. Approximately 42.7% of workers who first saw a surgeon had surgery, in contrast to only 1.5% of those who saw a chiropractor. The area under the receiver operating characteristic curve of the multivariate model was 0.93 (95% confidence interval, 0.92-0.95), indicating excellent ability to discriminate between workers who would versus would not have surgery. CONCLUSION: Baseline variables in multiple domains predicted lumbar spine surgery. There was a very strong association between surgery and first provider seen for the injury even after adjustment for other important variables.

Estimating incident population distribution from prevalent data.

A prevalent sample consists of individuals who have experienced disease incidence but not failure event at the sampling time. We discuss methods for estimating the distribution function of a random vector defined at baseline for an incident disease population when data are collected by prevalent sampling. Prevalent sampling design is often more focused and economical than incident study design for studying the survival distribution of a diseased population, but prevalent samples are biased by design. Subjects with longer survival time are more likely to be included in a prevalent cohort, and other baseline variables of interests that are correlated with survival time are also subject to sampling bias induced by the prevalent sampling scheme. Without recognition of the bias, applying empirical distribution function to estimate the population distribution of baseline variables can lead to serious bias. In this article, nonparametric and semiparametric methods are developed for distribution estimation of baseline variables using prevalent data.