Application of the Caprini Risk Assessment Model in Evaluation of Non–Venous Thromboembolism Complications in Plastic and Reconstructive Surgery Patients

Background: The Caprini Risk Assessment Model is used to categorize patient risk for venous thromboembolism (VTE) events; its predictive associations have been repeatedly corroborated. Calculating scores involves consideration of systemic factors that may predict other postoperative complications.

Objective: This study investigates whether Caprini scores can be applied to non-VTE complications.

Methods: The authors undertook a retrospective chart review of 1598 encounters for a series of complex reconstructive and body contouring operations at an academic medical institution. Input variables included Caprini score components, patient comorbidities, and prophylactic use of antithrombotic drugs. Output variables were postoperative complications. Tests for proportions were performed on percentile data. Nonpercentile data were treated with comparison of means (t test). Odds ratios for complications were calculated for stratified risk groups and compared.

Results: The overall complication rate was 28.03%. Deep vein thrombosis (DVT) incidence was 1.50%. Differences in age, body mass index (BMI), operation time, hypertension, diabetes, renal disease, and cancer were statistically significant between patients who experienced complications and those who did not. For DVT versus DVT-free patients, differences in sex, BMI, operation time, smoking status, diabetes, hypertension, and prior DVT were significant. Caprini scores identified 628 encounters as low risk (0–4) and 970 as high risk (>5). Dehiscence, infection, necrosis, seroma, hematoma, and overall complication rate significantly increased the incidence for the high-risk group.

Conclusions: Caprini scores can be used as valuable predictors for some non-VTE postoperative complications (dehiscence, infection, seroma, hematoma, and necrosis). In addition to VTE events, clinicians should pay special attention to clinical signs indicative of the complications listed above when dealing with high-risk, high–Caprini score patients.

Level of Evidence: 4

The Caprini Risk Assessment Model (RAM) is an ordinal scoring tool used to quantify and categorize a patient's risk for venous thromboembolism (VTE—an aggregate disease that includes both pulmonary embolism [PE] and deep vein thrombosis [DVT]) in the postoperative setting.¹ It was developed initially in response to older predictive indices that failed to attain wide acceptance, most likely because they required extensive laboratory testing^2,3 or separated patients into broad treatment groups based on general categories that failed to account for individual differences relevant to the incidence of disease.⁴ In time, use of the Caprini RAM has expanded to include the nonsurgical practice of medicine, as well as numerous subspecialties of surgery.^5–8

In plastic surgery, Davison et al⁹ first proposed an adaptation of the Caprini RAM. Subsequent studies have identified additional factors that may demonstrate a positive association with the incidence of VTE events¹⁰ but have largely corroborated the results of the original study.^11,12 However, despite strong support for its utility in VTE risk assessment, the Caprini RAM has not been thoroughly analyzed for correlation with any of the other potential complications of surgical procedures.

This sort of investigation is necessary because up to 98 000 Americans die yearly due to preventable medical errors,¹³ among them the failure to administer appropriate thromboembolism prophylaxis.¹⁴ Wound and medical complications still affect up to 4.6% and 4.9% of patients, and the overall incidence of postoperative complication is 10.9%.¹⁵ Because of these significant risks, classification of an operation as a plastic surgery procedure has been found to be a significant predictor for intraoperative complications at outpatient surgery centers.¹⁶

In an attempt to address the lack of investigation into model-based analyses of non-VTE complications, we conducted a retrospective chart review of patients who underwent a series of reconstructive and body contouring procedures. The study attempted to analyze the appropriateness of utilizing the Caprini RAM for evaluating a patient's risk for non-VTE complications following operation.

Methods

A retrospective chart review was undertaken subsequent to approval by the Institutional Review Board at UT Southwestern Medical Center (Dallas, Texas). The study population consisted of a subset of 1801 unique plastic surgery encounters encompassing a series of complex reconstructive and body contouring procedures as determined by Current Procedural Terminology (CPT) codes. The patient cohort was consecutive for the CPT codes analyzed. Members of the UT Southwestern Plastic Surgery Department saw these patients from January 2008 to January 2012 (Table 1). Nineteen faculty surgeons performed these operations at facilities associated with UT Southwestern Medical Center (Parkland Memorial Hospital, St Paul University Hospital, Zale Lipshy University Hospital, or the Outpatient Surgery Center, all in Dallas, Texas).

Variables

The independent variables included all information points necessary to calculate a Caprini score under the guidelines of the 2005 Caprini RAM.¹⁷ Patient comorbidity information, including diabetes, smoking status, coronary artery disease (CAD), cancer, hypertension (HTN), human immunodeficiency virus (HIV)/AIDS, and chronic obstructive pulmonary disease (COPD), was also noted during data collection. Also, if thromboembolism prophylaxis was given, type (heparin, warfarin, or low-molecular-weight heparin), frequency, and number of days administered were noted.

In addition to newly diagnosed VTE events within the 30-day follow-up period, outcome variables included other potential postoperative complications. For data analysis, wound complications were defined by 1 or more of the following: infection, dehiscence, erythema, necrosis, seroma, hematoma, or delayed wound healing, as recorded in the postoperative chart or follow-up note by the attending surgeon. Because not all operations in the study involved free or pedicled flaps, the specific complication of flap failure was not analyzed separately. However, incidences of the complications of interest (infection, dehiscence, etc) were included whether they were localized to the donor site, recipient site, or the flap used for the operation. Similarly, a category labeled “other complication” was deemed to be an overgeneralized collection of issues aggregating things not classifiable elsewhere, such as urinary tract infections or ear infections; this category was subsequently excluded from further analysis.

Statistical Analysis

Caprini scores were recorded as both ordinal values (0-17) and as stratified groups. Two-tailed tests for proportions were performed on percentile data to look for differences in rates of complication in female sex, proportion of patients with a body mass index (BMI) >25 and 40, history of major surgery 1 month prior, smoking, diabetes, COPD or other pulmonary disease, HTN, CAD, history of myocardial infarction or other cardiovascular disease, renal disorders, cancer, HIV/AIDS, and a history of DVT. Nonpercentile data (age, BMI, and operative time) were treated with a comparison of means using the Student t test.

For correlation of complications with Caprini score, patient encounters were separated into “low-risk” and “high-risk” categories based on Caprini score, with a value of 0 to 4 constituting low risk and a value equal to or greater than 5 constituting high risk in line with previously constructed classifications and prophylactic guidelines.¹⁸ These low- and high-risk groups were compared for complication rates using simple odds ratios. Odds ratios were calculated for the event of any overall complication and the individual complications of DVT, dehiscence, infection, erythema, seroma, hematoma, necrosis, and delayed wound healing. Sensitivity and specificity were also calculated utilizing the risk stratification as a 2 × 2 contingency table for each of the noted complications.

All patient data were recorded in Microsoft Excel (Microsoft, Redmond, Washington). Tests for proportions, comparison of means, sensitivity, specificity, and associated significance levels were calculated using standard statistical formulas. Odds ratios and associated significance levels were calculated using the MedCalc statistical software program (MedCalc Software, Ostend, Belgium). P values were considered statistically significant at the α = .05 level.

Results

From the original 1801 patient encounters initially identified for the study, 203 patients were excluded because inconsistencies and deficiencies in medical records prevented calculation of a reliable Caprini score. These patients were subsequently excluded from further analysis, giving a final cohort of 1598 encounters. Patients in this cohort ranged in age from 14 to 86 years, with an average age of 49.79 years. There were 1289 women and 309 men. Most patients underwent breast reconstruction (n = 402) or flap-based (n = 491) procedures, with excision lipectomy (n = 271), suction-assisted lipectomy (n = 226), delayed breast prosthesis (n = 177), and facial wrinkle removal (n = 31) comprising the remainder. Overall complication incidences for these procedures were 158, 170, 70, 28, 17, and 5 cases, respectively, resulting in an overall complication rate of 28.03% (Table 2). The overall rate of DVT was 1.50%.

When comparing individual risk factors between patients who experienced any complication versus complication-free patients, age, BMI, operation time, hypertension, renal disease, and cancer were identified to be statistically significant. Interestingly, for women, oral contraceptive pill (OCP) use was associated with a lower complication rate. For DVT versus DVT-free patients, sex, BMI, operation time, smoking status, diabetes, hypertension, and prior history of DVT were identified as significant (Table 3). For the low- and high-risk stratifications, 628 encounters were identified as low risk, and 970 were identified as high risk. Dehiscence, infection, necrosis, seroma, hematoma, and overall complication rate had significantly increased the incidence for the high-risk group (Table 4).

The risk-stratified comparison did not find DVT incidence significant between high- and low-risk groups (odds ratio [OR], 1.58; P = .310). To control for the use of chemoprophylaxis potentially confounding the rate of DVT, patient encounters were noted on the use of heparin (low molecular weight or unfractionated) or warfarin perioperatively; those without any chemoprophylaxis were isolated and analyzed separately. No form of prophylaxis was noted in 1024 patient encounters; from this population, 473 encounters were considered low risk and 551 encounters were considered high risk (Table 5). Calculations from this population showed an increased odds ratio and statistical significance (Table 6).

Discussion

Traditional Caprini Scores: Venous Thromboembolism Analysis

Venous thromboembolism, DVT, and PE are well-defined surgical complications. Mortality associated with these thrombotic events ranges from 7% to 65% in trauma patients,¹⁹ 25% to 33% in general surgery procedures,²⁰ and 41% to 85% following total hip arthroplasty, total knee arthroplasty, and other orthopedic procedures.⁴ Compounding the problem is the high incidence of symptomatically silent thrombosis events, which can range from 16% in new diagnoses²¹ to 35% in patients with previous DVT events.²² The high risk of the disease and efficacy of preventative measures have led to multiple government agencies calling for implementation of prophylactic measures,²³ spearheaded by the Office of the Surgeon General.²⁴ The American College of Chest Physicians has also released guidelines for thromboprophylaxis of at-risk patients. The ninth iteration of these guidelines provides detailed treatment advice for patients, stratifying them into different risk categories by age, type of surgery, and additional risk factors.²⁵ Although these guidelines are widely applicable, the American College of Chest Physicians does not include plastic surgeons, and no specific guidelines exist for the specialty. Furthermore, the ambiguity of classifications such as “minor,” “nonmajor,” or “major” leads to difficulties in classifying procedures that can vary widely in operation length and extent of body tissue dissected while remaining wholly in the more superficial planes of dissection.

Risk assessment models such as the Caprini scoring system are used to elucidate individual patient and procedure characteristics to provide a more comprehensive and precise analysis of risk in stratifying patients into different categories. Expanding on the original work of Clayton et al,² who pinpointed the effects of age, BMI, and varicose veins, studies by Caprini et al,¹ Davison et al,⁹ and later authors¹⁰ have identified additional factors that contribute to the risk of developing VTE complications peri- or postoperatively. Our retrospective analysis supports the conclusions drawn by previous authors and demonstrates an increased rate of VTE events concordant with rising Caprini scores. The data, however, do not demonstrate statistical significance, which likely reflects a number of contributing variables.

To some extent, this lack of significance may be a structural defect in the Caprini model, an intrinsic weakness that is in some respects unavoidable because the model is built for a broad range of surgical disciplines and not specifically tailored to plastic and reconstructive surgery. Analysis by Hatef et al¹⁰ noted that even the revised Caprini model modified by Davison was not completely predictive of VTE incidence and elucidated additional factors to increase the accuracy of the model, among them circumferential abdominoplasty and hormone therapy use. Similarly, studies published subsequent to the Venous Thromboembolism Prevention Study (VTEPS) noted length of stay as an additional independent predictor of VTE incidence not included in Caprini score calculations.¹⁸

Another factor may be a low incidence of DVT occurrence overall (1.5%) as well as within the stratified risk groups as presented in Table 5 (1%–4%). Also partly responsible for the lack of significance in the odds ratios may be an increasing use of thromboprophylactic treatment with larger Caprini scores. Fourteen percent of patients in the lowest risk category received this treatment, whereas 62% in the highest risk category received prophylaxis. This may depress incidence rates in the higher risk groups, as thromboprophylaxis has been associated with decreased VTE events.^26–28 In our 1024 patients who were not prophylaxed, however, there was a significant increase in VTE events in high- versus low-risk patients, as would be expected by the Caprini RAM.

It has been noted in other studies that several alternative and adjunct therapies work with various degrees of efficacy in lowering a patient's risk for thromboembolism,²⁹ including compression devices, total intravenous anesthesia, avoidance of prone positioning, aspirin, and more. These studies also express skepticism about the need to provide anticoagulant prophylaxis in plastic surgery. It is very possible that aspirin, compression devices, and other treatments result in a statistically significant decrease in thromboembolism incidence. However, no support is given to demonstrate that these are more efficacious than anticoagulation alone or a combined treatment, nor do they provide evidence of similar levels of efficacy with significantly lowered bleeding risks compared with the same treatment options listed above (anticoagulation or combination with thromboprophylaxis).

Additionally, the data demonstrated an unexpected correlation between OCP use and complication rate, with increasing use being correlated with a lack of complication (P = .045). This study did not investigate links between OCP use and incidence of complications, but a link between OCP use and rates of DVT has been well characterized. The Food and Drug Administration (FDA) requires all OCP manufacturers to print a “black box” warning that use of such products increases the risk of venous coagulation. Furthermore, use of OCP is included in the calculation of Caprini scores. A general relationship between contraceptive use and rates of DVT has been noted,³⁰ as well as more specific ones in outpatient plastic surgery operations,³¹ but links to other complications of surgery have not been analyzed. The data here noted that OCP use is associated with a lack of postoperative complication. We do not believe that OCP use is a direct cause of a decreased complication rate. Rather, we speculate that other factors such as selectivity for a premenopausal age and increased health maintenance/primary care interaction confounded the results and contributed to a lower complication rate for OCP users.

No separate analysis was performed based on sex, although the sample size appears large enough for analysis (assuming sufficient incidence rates). The goal of this study was not to distinguish between cohorts, but we did note no significance in total complications and positive significance for DVT based on sex. This would be an interesting topic to pursue in the future but was outside the scope of this study.

Expanding Caprini to Non-VTE Events

Although the Caprini RAM has been extensively investigated with respect to preoperative evaluation of VTE risks, to the best of our knowledge, there has been no similar exploration into predictive associations of this score with the other potential complications of surgical procedures (infection, seroma, hematoma, and delayed wound healing, to highlight a few). This is surprising because the full list of variables that comprise an individual Caprini score involves a host of inherited and acquired clinical characteristics in addition to biochemical profiles (see supplementary file at http://aes.sagepub.com/supplemental). Many of the clinical factors that are taken into consideration are systemic dysfunctions that involve multiple organ systems. The data presented here show that increasing Caprini scores are associated with increasing odds ratios for dehiscence (OR, 1.73; P = .028), infection (OR, 2.04; P = .0003), seroma (OR, 1.52; P = .045), hematoma (OR, 2.12; P = .050), and necrosis (OR, 2.83; P = .0004), with a corresponding overall OR increase of 1.69 (P < .0001). For clarification, an odds ratio is a measure of association. It represents the odds that an outcome will occur in a group that shares some characteristic or exposure, compared with the odds of that event occurring in a group that lacks the characteristic or exposure of interest. These odds ratios parallel similar, but not identical, increases in the baseline occurrence rates for the noted complications: dehiscence (3.66%–6.19%), infection (5.89%–11.34%), seroma (5.57%–8.25%), hematoma (1.43%–2.99%), necrosis (2.39%–6.49%), and general complications (21.82%–32.06%). On the basis of this, we support the use of Caprini scores for preoperative calculation of the risk for non-VTE complications, a recommendation that is borne out by independent studies across disciplines.

More specifically, our findings are in concordance with prior research conducted in other disciplines with respect to individual complications. In particular, wound dehiscence has been shown to correlate with age and BMI.^32–34 Similarly, several studies have demonstrated a relationship between BMI, smoking, and rates of fat necrosis. The Caprini RAM captures both values, either directly (for BMI) and perhaps indirectly as well (smoking, through existence of serious lung disease, including COPD).^35,36 Likewise, independent studies conducted by Arabshahi et al³⁷ and Hedrick et al³⁸ identified multiple risk factors contributive toward surgical site infection, among them age, malignancy, BMI. The confounding effects of obesity and operation length deserve special consideration. As Arabshahi et al point out, obese patients require special surgical accommodations to account for unique physical and technical issues that can extend the length of their operation. Additional research must be conducted to elucidate whether the additional risk presented by obese patients is independent of this tendency toward longer operating times. Even more troubling, our study found a significant correlation for Caprini scores and infections despite the fact that most patients received a prophylactic antibiotics course prior to the operation. In contrast to the above, the literature on hematoma formation is less extensive and more conflicted, with support both for and against the significance of different contributory factors.^39,40

Two potential complications did not demonstrate a significantly elevated incidence in response to increasing Caprini scores: erythema and delayed wound healing. Work in the literature examining potential risk factors for these complications independently has not been successful in demonstrating a clear link between the variables used to calculate a Caprini score and the complication in question. Demonstrated associations tend to focus on extracorporeal factors such as surgeon technique, sterile field integrity, and lifestyle more than patient comorbidities.^41–43

As a risk assessment model that tabulates patient characteristics that, in addition to holding predictive associations with the risk of VTE events, also has demonstrated links to other negative complications of surgery, it is not surprising that the Caprini model exhibits a statistically significant odds ratio between stratified patient groups. Unfortunately, few preoperative steps can be taken to directly address the complications for which the Caprini model is predictive. Most patients in the study were already treated with prophylactic antibiotics, and other issues like necrosis must be dealt with as they occur. However, these results can help physicians better understand which complications to keep watch for most rigorously during postoperative monitoring. In addition, the data can guide conservations with patients during initial preoperative consultations.

Our study validates the assignment of risk in providing a gradation for the purposes of preoperative decision making with regard to both VTE prophylaxis and preoperative measures for additional complications. By subdividing a diverse patient population into more homogeneous and manageable sections, it is possible to devise standardized treatment regimens for individual risk categories. However, the study presented here was a single-institution retrospective review. The patient population exhibited a normal distribution. This resulted in small cohorts at either extreme of the stratified risk groups. Additional research across institutions and disciplines would provide an opportunity to expand the study population and validate the results of this study. Randomized controlled trials focusing on non-VTE complications would also provide higher-quality data to help design protocols and make treatment recommendations.

Conclusions

The data presented here demonstrate that the Caprini RAM can be applied preoperatively to categorize patient risk of developing multiple non-VTE complications. Patients placed in the higher risk categories are at an increased risk of having wound dehiscence, infection, seroma, hematoma, and necrosis. Considering these potential complications as a single group, a high-risk patient has a 69% increased odds of having at least 1 problem compared with low-risk individuals. This information can help physicians educate patients about the risks of their operations and helps prioritize the complications that need an especially vigilant watch.

Acknowledgments

The authors thank Roberto Cortez, Rachel Hein, Ryan Constantine, Kendall Anigian, James Jewell, Natalie Sciano, Bhavani Gannavarapu, Janeiro Okafor, and Allan Wang for their help compiling the original database. Additionally, we thank Debby Noble and the research support team at UT Southwestern for their invaluable efforts.

Disclosures

Dr Davis received grants from Convatec (Skillman, New Jersey), Innovative Therapies (Pompano Beach, Florida), Unilever (London, England), Andrew Technologies (Tustin, California), and Kensey Nash (Exton, Pennsylvania); no grant money was used for this study. Dr Kenkel is an unpaid investigator for Allergan (Irvine, California) and Ultrashape (Syneron, Yokneam, Israel), and is on the Advisory Board of Kythera (Calabasas, California) and Ulthera (Mesa, Arizona). None of the other authors has anything to disclose.

Funding

The authors received no financial support for the research, authorship, and publication of this article.

References