Response to Comment on: Lipsky et al. Developing and Validating a Risk Score for Lower-Extremity Amputation in Patients Hospitalized for a Diabetic Foot Infection. Diabetes Care 2011;34:1695–1700

We thank Drs. Aragón-Sánchez and Lázaro-Martínez (1) for their insightful comments and questions about our article (2). As we noted, we used virtually the same database of patients as we did for a previously published article (3) in which we investigated mortality, length of stay, and cost associated with skin and soft tissue infections (of both the foot and other anatomic sites) in persons with diabetes. Because some patients hadmultiple codes for infections, we devised a hierarchy based on associated mortality risk to ensure that patients were counted only once and were classified into the highest mortality risk group. In our current article concerning a lower-extremity amputation (LEA) risk score (2), we excluded 12 of the 3,030 cases whose procedure codes indicated they had some type of upper-limb amputation (5 ofwhomalso had a foot infection). The remaining 3,018 patients included in the LEA risk score study consisted of 2,215 with “foot” infections and 803 with “nonfoot” infections. Our label of nonfoot infection may have created confusion. This group consisted of 605 infections of legs, thighs, knees, calves, or lower extremity not otherwise specified; 112 surgical-site infections; and 86 infections without specific reference to lower extremity. When we excluded these last 86 cases, the model statistics and LEA risk score strata remained virtually the same as those published. Hence, although we could not rule out potential misclassification of patients using ICD-9 codes, any effect on the risk score model fit would be minimal. Patients with infections of lower extremity above the malleoli constitute a population that may be similar to but, we believe, separate from foot infections in persons with diabetes. Of note, in those with a nonfoot infection, the LEA rate was 15.1% (121/803). We should have been more explicit in our article about the fact that we excluded patients with osteomyelitis (n 5 3,475). These patients represent a population with a much higher risk of LEA (4) who should be addressed as a distinctive cohort. Regrettably, LEA is a default treatment in some centers for these infections, so we are currently using our database to attempt to identify patients with diabetic lower extremity osteomyelitis who are at low risk and may be spared LEA. Unfortunately, most studies using retrospectively defined patient cohorts must depend on administrative data such as ICD-9 diagnosis codes, which can be vague or overlapping. In the Fincke study cited (5), only 69% of cases were classifiable into one of their predefined (but nonvalidated) groups. The authors also note that the ICD-9 codes for some of these diagnoses (e.g., gangrene, ulcer) will include noninfectious cases, and patients can have multiple codes for anatomic locations. While many retrospective studies have used only ICD-9 codes to define infections, we added the requirement of a positive culture from the lesion to identify our study cohort. As we stated, we would like to see our LEA risk score validated in a prospective study that defines patient population using predefined clinical criteria rather than ICD-9 codes.