Appendix Table C6b-Update.Study quality for CLABSI which control for secular trend or confounding from the update search

Study / Infection / followup 1 year or longer / adequate statistical analysis / adherence rates analyzed / infection rates analyzed / independent of other QI efforts / Overall Quality / Univar YN / Univar Model / MultivarYN / Multivar Model / Multivar Control / Comment
Barrera, Colombia - 2011 / CAUTI, CLABSI, VAP / Yes / Yes / Yes / Yes / Uncertain / Lower / FALSE / TRUE / Poisson regression / intragroup correlation specified each hospital unit as a cluster, to control for different pt populations in each unit, temporary workers and nurse-to-patient ratio / simple before-after study that controlled for potential confounders
Burrell, Australia - 2011 / CLABSI / Yes / No / No / No / Uncertain / Lower / TRUE / chi-square / TRUE / multiple logistic regression / Patient Bundle, Physician Bundle, Line-days / Study uses a ‘lead in period’, 1 year from start of intervention, in place of baseline data. This data is compared to last 6 months of follow-up. Analysis did not control for pt. factors across the 37 ICUs.
Harris, United States - 2011 / CLABSI, VAP / Yes / Yes / No / Yes / Uncertain / Lower / FALSE / TRUE / infection: time series, regression; cost: GLM / age, gender, race, comorbidity, insurance coverage, APR-DRG score, provider specialty
Kim, United States - 2011 / CLABSI / Yes / Yes / No / Yes / Uncertain / Lower / FALSE / TRUE / Poisson regression
Lilly, United States - 2011 / CLABSI, VAP / No / Yes / Yes / Yes / Uncertain / Medium / TRUE / Fisher exact or chi-sq / TRUE / logistic regression / ICU, ICU type, admission time, acuity score, operative status / This is an unblinded nonrandomized stepped wedge design. They took the same intervention, and implemented it in 7 different ICUs at different times. There was no progression of changes to the intervention. Only 3 of the 7 ICUs had followup one or more yrs
Miller, United States - 2011 / CLABSI / Yes / No / Yes* / Yes / No / Lower / FALSE / TRUE / Generalized linear models / geographical region, average length of stay and bed capacity / Adherence to the insertion and maintenance bundles was included in the model for infections in the original analysis. Baseline patient characteristics were not assessed. Part way through the postintervention period a new chlorhexidine protocol was initiated
Render, United States - 2011 / CLABSI / Yes / Yes / No / Yes / Uncertain / Lower / TRUE / Pearsons’ correlation coefficient, standardized infection ratio, ANOVA / TRUE / Poisson GEE regression / Adherence rates were given by intervention year, however no analysis was presented save a single statistic about the inverse correlation of CLABSI rate and overall bundle compliance.
Speroff, United States - 2011 / CLABSI, VAP / Yes / Yes / No / Yes / N/A / Higher / FALSE / TRUE / Hierarchical negative binomial regression / hospital, baseline covariates / Adherence rates were calculated by follow-up survey, no baseline data. Statistical analysis did only chi-square or t-test to compare groups, but study had cluster randomized design. Did binomal negative regression analysis.
Taylor, United States - 2011 / CLABSI / Yes / Yes / No / Yes / Uncertain / Lower / FALSE / TRUE / Cox regression / gestational age, central line days, census, respiratory support days / No adherence rates reported. Infection risk was reported rather than infection rates.

Was the followup period 1 year or longer?
Was the statistical analysis adequate?
Were the baseline and postintervention adherence rates reported and analyzed statistically?
Were the baseline and postintervention infection rates reported and analyzed statistically?
Was the intervention independent of other QI efforts implemented at the same time?

C-1