INTRODUCTION Predictors of readmission for acute exacerbations of chronic obstructive pulmonary disease (AECOPD) are not well defined. Identifying modifiable predictors may help reduce the burden of readmissions. This study was done to evaluate the role of serum magnesium in frequent readmissions. METHODS One hundred patients admitted with a diagnosis of AECOPD to a tertiary care center from April 2004 to March 2006 were retrospectively followed from the time of index admission until the next admission or death. Number of admissions was calculated for the year after index admission, and frequent readmission was defined as > or =3 per year. Patients with other respiratory diseases, renal failure, and congestive heart failure were excluded. Serum magnesium was assayed at the time of admission by the colorimetric method. Logistic regression analysis was used to find independent risk factors for readmission. RESULTS The mean age of patients was 71.9 (+/-10.9 standard deviation (SD)) years. Fifty-seven were females. Ninety percent were current or ex-smokers. The median duration of time to next admission was 108 days (range 2-842). A total of 87 patients were readmitted at least once during the first year of follow-up, while 5% died; 23% had frequent readmissions; 85% had received pneumococcal vaccine within 5 years; and 29% received influenza vaccine in the current season. Frequency of readmissions was not influenced by the administration of inhaled or oral steroids, and diuretics at the time of discharge. Vaccination did not protect against frequent readmissions. The sole predictor of frequent readmissions was serum magnesium level (1.77+/-0.19 vs. 1.86+/-0.24mEq/L; adjusted odds ratio 0.003, 95% Confidence intervals <0.001-0.55; p=0.03). CONCLUSIONS Lower serum magnesium level independently predicts readmission for AECOPD. This is an easily modifiable risk factor.