The aim was to evaluate the effect of control selection on risk factor analysis for extended-spectrum beta-lactamase-producing Klebsiella pneumoniae (ESBL-KP) infections. Four contemporaneous case-control studies were conducted prospectively with 372 patients: Study 1 (ESBL-KP-infected vs non-infected); Study 2 (ESBL-KP-infected vs non-ESBL-KP-infected); Study 3 (all KP-infected vs non-infected); Study 4 (non-ESBL-KP-infected vs non-infected). Time at risk (TAR, i.e. duration of hospital stay) was the most significant risk factor [Study 1: odds ratio (OR): 5.74 (95% CI: 2.26-14.59; P<0.001); Study 2: 3.52 (1.47-8.43; P=0.005); Study 3: 2.68 (1.57-4.58; P<0.001)]; central venous catheterisation (CVC) was a risk factor in Study 1: 5.31 (1.67-16.82; P=0.005) and Study 3: 2.10 (1.04-4.27; P=0.04). Prior use of cephalosporins (PUC) was a risk factor only in studies with non-infected patients as controls [Study 1: 5.64 (1.90-16.72; P=0.002) and Study 3: 4.60 (2.09-10.13; P<0.001)]. The ORs were uniformly lower with 'non-ESBL-KP-infected' (TAR: 3.52; CVC: 2.07; PUC: 1.97) compared with 'non-infected' patients (TAR: 5.74; CVC: 5.31; PUC: 5.64) as control groups. Selection of control patients has a crucial role in the evaluation of risk factors for ESBL-KP infections. A consistent underestimation of the magnitude of the risk factors is observed when the control group is defined by the non-ESBL-KP-infected patients.