Abstract
Background/Aims Few studies have assessed risk for C. difficile infection (CDI) among patients seeking routine care. We developed and validated a prognostic risk score for CDI in the one year following an outpatient healthcare encounter.
Methods Our development cohort included Kaiser Permanente Northwest (KPNW) patients with an index outpatient visit between 07/01/2005 and 09/30/2008. Applying Cox regression, we synthesized a priori predictors into a CDI risk score, in which a higher number of points indicates higher risk. We validated and recalibrated the risk score using a KP Colorado (KPCO) cohort and the same patient characteristics. We calculated and plotted the observed one-year CDI risk for each decile of predicted risk for both cohorts.
Results The development cohort included 359,696 KPNW patients; 591 experienced CDI, a one-year incidence of 2.1 CDIs per 1000 patients. Twenty-two predictors of CDI were included in the risk score. Of these, age 60 years or greater, hospitalization of seven or more days within the previous 60 days, liver disease, and use of cephalosporins or clindamycin within the previous 60 days contributed to an approximate doubling of risk for CDI. The risk score separated high-risk patients from low-risk patients; the highest-risk patients were nearly 50 times more likely to have CDI (observed risk of 9.5 vs. 0.2 CDIs per 1000 patients). For the validation cohort, we followed 296,550 KPCO patients; 620 experienced CDI, a one-year incidence of 2.3 CDIs per 1000 patients. The risk score validated successfully in Colorado, as it was able to discriminate the highest and lowest risk patients; predictions agreed closely with observed risk in each decile.
Discussion Our risk score successfully discriminated between patients at highest and lowest risk for CDI and provided predictions which agreed closely with observed CDI risk, making it a useful risk management tool for healthcare providers. Of importance, there was good calibration among the group of patients at the highest-risk for CDI, who are likely to benefit the most from clinician recognition of this risk.
