We thank Pannucci et al for their interest and for discussing the limitation of missing inputs for the Caprini risk assessment model (RAM) in the National Surgical Quality Improvement Program, which we also addressed in our article.

The Caprini score, as published in the American College of Chest Physicians (ACCP) guidelines, the original Caprini 2005 RAM cited by Dr Pannucci, the original article of Stroud et al, and our article , all use the risk groups 0-1, 2, 3-4, and ≥5. Our central conclusion was that the Caprini score has limited utility among gynecologic oncology patients as ∼98% fall into the highest risk category (≥5). Pannucci et al instead divided their population using the groupings 0-2, 3-4, 5-6, 7-8, and >8. We agree that when the highest risk group (score ≥5) is further subdivided, the Caprini score could be a useful tool (see Figure 1 from our article where we performed a similar analysis). However, risk scores and guidelines are to be used with their accompanying published risk strata. Furthermore, while their figure suggests a trend, when confidence intervals are calculated or statistical comparisons performed there is significant overlap between all groups. Additionally, their data only include laparotomy patients and do not address minimally invasive surgery, a key population in which the Caprini score overestimates venous thromboembolism risk and a prevalent form of surgery for gynecologic cancers.

Missing data are important; however, this limitation is less significant than suggested. First, if a Caprini input was missing, the patient received 0 points. If we had these data, patient scores would only be higher, categorizing more as highest risk and reinforcing our conclusions. Second, some missing risk factors are uncommon. An immobilizing plaster cast is not highly prevalent in this population. Data regarding very rare inputs would not change our conclusions. Third, validation studies performed with retrospective medical record review are not free from the limitation of missing data. Lack of documented information regarding a risk factor is often mistaken for the absence of that risk factor. Furthermore, risk factors, such as varicose veins, may not be uniformly documented in the medical record. We agree missing data for model inputs are a limitation, but for the reasons outlined, we disagree this limitation calls our conclusions into question.

We agree that the Caprini RAM could be a useful tool among gynecologic oncology patients if the highest risk group is substratified. However, we reiterate our conclusion that the 2005 Caprini RAM applied using its original published groupings, as recommended by the ACCP, has limited utility in gynecologic oncology patients.