The following is the abstract of a paper (Majumdar et al., 2006), which we discussed in the exercise Statins in pneumonia in Week 8:
Objectives To determine whether statins reduce mortality or need for admission to intensive care in patients admitted to hospital with community acquired pneumonia; and to assess whether previously reported improvements in sepsis related outcomes were a result of the healthy user effect.
Design Population based prospective cohort study.
Setting Six hospitals in Capital Health, Edmonton, Alberta, Canada.
Participants Adults admitted to hospital with pneumonia and categorised according to use of statins for at least one week before admission and during hospital stay.
Main outcome measures Composite of in-hospital mortality or admission to an intensive care unit.
Results Of 3415 patients with pneumonia admitted to hospital, 624 (18%) died or were admitted to an intensive care unit. Statin users were less likely to die or be admitted to an intensive care unit than non-users (50/325 (15%) v 574/3090 (19%), odds ratio 0.80, P=0.15). After more complete adjustment for confounding, however, the odds ratios changed from potential benefit (0.78, adjusted for age and sex) to potential harm (1.10, fully adjusted including propensity scores, 95% confidence interval 0.76 to 1.60).
Conclusions Statins are not associated with reduced mortality or need for admission to an intensive care unit in patients with pneumonia; reports of benefit in the setting of sepsis may be a result of confounding.
The authors explain the term "propensity score" as follows:
"Finally, we included a propensity score for statin use. We used multivariate logistic regression to construct a score that reflected a patient's likelihood of being prescribed a statin (variables available on request). We expected that our propensity score would incorporate variables reflecting healthy user status. We calculated rates of our composite outcome across fifths of increasing propensity and tested for trend using chi-squared tests. We then entered the score, as a continuous variable, into the models."
1. What is meant by "adjustment for confounding"?
2. What statistical method would be used for the adjustment and why?
3. How would a chi-squared test be used to check the propensity scores?
4. After using splitting into fifths to check the propensity scores, the authors say that they entered the score into their regression models as a continuous variable. What are the advantages and disadvantages of using score as a continuous variable rather than as a categorical variable with five categories? (This is a pretty difficult question, by the way!)
Reference: Majumdar SR, McAlister FA, Eurich DT, Padwal RS, Marrie TJ. (2006) Statins and outcomes in patients admitted to hospital with community acquired pneumonia: population based prospective cohort study. British Medical Journal 333, 999-1001.
To Applied Biostatistics index.
To Martin Bland's M.Sc. index.
This page maintained by Martin Bland.
Last updated: 12 December, 2006.