Our study should be viewed in the context of its strengths and limitations. Some other studies reported lower incidence of VTE with SCDs compared to NONE but the results were not statistically significant. Since these patients were considerably less mobile, the results may not be reproducible in our study. The CLOTS III trial reported significant effectiveness of SCDs in DVT prevention in immobile patients with acute stroke. In another review, the strength of the evidence was insufficient to determine the effectiveness of SCDs for thromboprophylaxis in high-risk medical patients because of limited data. The overall incidence of symptomatic VTE was 4000 individuals were analyzed and there was no significant difference with the use of compressive and pneumatic devices compared to no treatment or use of anticoagulants. Propensity score matching matched the SCD and NONE groups with no statistical difference in VTE incidence. In comparison to NONE, SCD patients had significant differences in risk factors for VTE, including higher CCI, higher prevalence of cancer and obesity, and longer LOS. Our large retrospective study of 30,824 medically ill patients demonstrated a similar incidence of VTE with SCDs only compared to the NONE group. A multivariable logistic model with a subset of the covariate mix was applied using information criteria for model selection. We also performed a risk-adjusted analysis for VTE incidence with an indicator of SCD use. We used conditional logistic regression to obtain the adjusted OR and 95% CI for the association of SCDs with VTE incidence. In the matched sample, we examined the quality of the matching by comparing the standardized mean differences and variance ratios between SCD and NONE. The SAS procedure PSMATCH was used for matching. This greedy matching algorithm, which proceeded sequentially with SCD patients selected in random order of propensity scores and matched to a unique NONE patient, resulted in 10,071 unique pairs. A randomly chosen SCD patient was matched to one NONE patient in the common region of propensity scores extended by 0.25 times the pooled estimate of the standard deviation of the logits of propensity scores in the two groups. We followed published principles and guidelines to form treated and non-treated pairs based on their propensity scores. Figure 2 depicts the adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for the binary variables. The c-statistic was 0.707, indicating an acceptable level of discrimination between SCD and NONE patients. A spline function of a continuous variable is a smooth function composed of polynomial pieces connected at interior points called knots in the range of the variable. The variables included in the final model for propensity scores were sex, any type of cancer, comorbidities, and three continuous variables modeled by splines: age (6 terms), log-transformed LOS (3 terms), and CCI (4 terms). We experimented with different specifications, especially for LOS and CCI, with the same qualitative conclusion. There are features of randomness in the selection of treated patients and their matches that could lead to different models for assessing the propensity scores. For each patient, we estimated the propensity score (likelihood of receiving SCD) from a multivariable logistic regression model. Since patients were not randomly assigned to receive SCDs, propensity score analysis was performed. Differences between the SCD group and NONE group were compared using t-tests or Wilcoxon rank sum tests for continuous variables and chi-square tests for categorical variables.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |