A Predictive Model for Venous Ulceration in Older Adults: Results of a Retrospective Cohort Study

Login toDownload PDF version
Paul Y. Takahashi, MD; Anupam Chandra, MD; Stephen S. Cha, MS; and Sarah J. Crane, MD

Abstract 

Although venous leg ulcers (VLU) are a common health concern for many older adults, knowledge about common risk factors and optimal screening methods remains limited. To determine risk factors and develop a predictive model for future venous ulceration, a retrospective cohort study of clinical outpatients >60 years old in a primary care panel in Olmsted County, MN (N = 12,650) was conducted.

The primary outcome was a new diagnosis of VLU within 2 years of the study date (January 1, 2005). Risk factors included demographic and comorbid health risk factors. The average age of study participants was 72.7 years (SD 8.9) and the incidence of VLU was 1.7% — 8.6 cases per 1,000 patient-years. The most significant risk factor was a history of venous ulceration (OR 19.4; 95% CI 14.5–25.9). In the final multivariate logistic regression model, a history of venous ulceration (OR 18.66; CI 13.96–24.96, P < 0.0001), renal insufficiency (OR 2.24; 95% CI 1.60–3.15, P <.0001), cataracts (OR 1.64; CI 1.16–2.33, P = 0.01), blindness (OR 2.53; 95% CI 1.05–6.11, P = 0.04), and pressure ulceration (OR 2.36; 95% CI 1.26–4.45, P = 0.01) remained significant as risk factors for VLU. Using the odds ratios from the predictors in the final multivariable mode, a scoring model for the entire cohort was created and a scoring model for each subject was run for the final model. The area under the receiver operator curve was 0.797 with a cutoff score of 3. Prior venous ulceration remains the most important risk factor for future venous ulceration. Further study and application of the risk assessment model to substantiate its clinical value are warranted.

Please address correspondence to: Paul Y. Takahashi, MD, Mayo Clinic College of Medicine, 200 First Street SW, Rochester, MN 55905; email: takahashi.paul@mayo.edu.

     Venous ulcers are a common health concern for many older adults. Estimates of venous ulcer prevalence vary; however, according to Rucker,1 venous ulcers may be present in 1% of the US population. The point prevalence in a Swedish cohort was 2.4 cases out of 1,000 persons in 2008.2 In a retrospective cohort study of 300,000 inpatients in China, venous ulcer prevalence was more than 6%.3

     Venous ulcers can be painful and consume a great deal of patient, family, and caregiver effort.4 The cornerstone of venous ulcer therapy is edema control through external compression of the edematous legs.5 For direct wound management, clinicians focus on controlling exudate to prevent maceration around the ulcer,6 a time-consuming, painful, and expensive endeavor for clinician and patient. A recent Cochrane review7 of venous ulcer compression estimated an individual venous wound costs 814 to 1994 euros ($1,100 to $2,800 US). These facts indicate that the development of a risk scoring system or model to predict venous ulceration would be helpful for targeting preventive care toward higher-risk patients.

     Although several predictors of venous ulceration have been identified that would guide the creation of a new model or risk scoring system, many of the risk factors were observed in younger populations and may not be relevant to patients >60 years old. It is important to consider venous disease in older versus younger people because functional status may be lower in older adults. In a retrospective study8 of 62 patients (average age 57 years) undergoing venous surgery, a previous history of venous ulceration was a significant (P = 0.02) predictor of future venous ulceration. In a prospective study (n = 229), patients with venous insufficiency also developed venous ulceration much more frequently, with venous refill time a good indicator for venous ulceration.9 These studies provide a basis for future work; however, many questions remain about potential venous ulcer risks in an outpatient, older population.

     The objective of this retrospective cohort practice study of community-dwelling older adults in one outpatient community was to determine the association between demographic, venous, vascular, and comorbid health factors and the development of venous ulceration in order to create a prevention-oriented model that would help predict venous ulceration in this cohort.

Methods

     This retrospective cohort study involved adult outpatients within the division of Primary Care Internal Medicine (PCIM) at Mayo Clinic, Rochester, MN. Participants included in the analysis comprised all adults >60 years who were assigned to a PCIM primary care provider on the study date (January 1, 2005). The Mayo Clinic, Rochester is an academic medical center with both faculty physician and training healthcare providers. The Mayo Clinic Institutional Review Board (IRB) reviewed and approved the protocol.

     The authors conducted all aspects of the research on this project in accordance with the principles of the Declaration of Helsinki,10 as well as in adherence to Minnesota state statue with regard to medical record use and privacy.11

     All participants were community-dwelling or lived in an assisted-living facility within Olmsted County. Residents living within a skilled nursing facility or who did not provide informed consent were excluded (eg, patients with signed forms that request their medical records not be used for medical research).

     Data collection. Health science research personnel extracted data from the electronic medical record (EMR) of the patients enrolled in the study. The abstractors were blinded to the study hypothesis and were not involved with the analysis or interpretation of the data. The EMR contains all medical diagnosis, surgical interventions, and demographic information for each patient. Mayo Clinic Rochester maintains EMR information within one electronic system for clinical use, billing, and medical archives through the Rochester Epidemiology Project.12 Using this administrative electronic information, the authors collected predictor variables of demographics, venous disease, arterial vascular risk factors, and other comorbid risk factors that occurred before the study date (January 1, 2005) and any subsequent event. The medical provider made the final clinical diagnosis within the EMR. Previous research using this EMR has shown that the accuracy of medical diagnosis for all conditions varies; however, for some (eg, deep venous thrombosis), the specificity of the medical record is as high as 99%.13

     Outcome variables. The primary outcome was the development of a new venous ulcer within 2 years of the study date. The diagnosis of venous ulceration was made clinically by medical providers and included in the research database when documented in the medical record and/or billed as a venous ulcer.

     Predictor variables. Demographic variables collected included age, gender, and marital status as of the day of study. Venous disease risk factors included both a history of venous ulceration and a history of venous insufficiency. Arterial vascular comorbid health conditions included coronary artery disease (CAD), stroke, peripheral vascular disease (PAD), renal insufficiency, congestive heart failure (CHF), previous history of myocardial infarction (MI), hypertension, hyperlipidemia, stroke, and combined cardiac outcomes of a history of CAD/CHF/MI. Other comorbid medical illnesses abstracted included a history of cancer, diabetes, depression, dementia, rheumatoid arthritis, degenerative arthritis, peripheral neuropathy, hypothyroidism, pressure ulcer, fall, hip fracture, chronic obstructive pulmonary disease (COPD), and cataracts. All comorbid health conditions were recognized before the day of this study.

     Data analysis. Health Research Services personnel (blinded to the study hypothesis) directly entered all information via electronic abstraction into a Microsoft Excel version 2003 spreadsheet (Microsoft, Redmond, WA) for data entry, retrieval, and analysis. The investigators analyzed the final information using SAS 9.13 software (SAS Institute Inc., Cary, NC). Initial analysis included unadjusted analysis of individual predictor variables (demographics, venous disease, arterial vascular comorbid health disease, and other comorbid health illness) and the development of venous ulceration using either Pearson chi-square tests or two-sample t-tests. The age-adjusted odds ratios and the 95% confidence interval (CI) and P values for individual variables were obtained using a logistic regression model.

     All significant variables from the unadjusted risk variables were placed in a multivariate model. The multivariate logistic model was constructed using all factors with a P value <0.10 and the authors used a stepwise elimination approach. Factors with P <0.05 in the final model were considered significant.

     Following completion of the multivariate model, the authors constructed a model with all significant predictors. The scoring system is based upon the final multivariable model, with venous ulceration having the highest predictor variable at 19, blindness having a predictor of 3, and renal insufficiency, cataracts, and pressure ulcer each having a score of 2. The final model was validated using a Bayesian approach, rerunning the final model 1,000 times to confirm the final variables were still significant. The final model included a scoring system based upon the odds ratios from the final risk factors in the multivariate model.

     Using this model, the authors created a receiver operator curve (ROC) to demonstrate the optimum cutpoint to reach the maximum of both sensitivity and specificity. The ROC curve facilitates determination of the accuracy of the model using both sensitivity and specificity. The area under the curve (AUC) and its standard error (SE) also were obtained.14

Results

     Of the >60 years of age population (13,316), 664 (5%) refused research authorization. Records of all 12,650 patients >60 years old who gave consent were used for the study on January 1, 2005. Of the 12,650 participants, 215 (1.7%) developed a new venous ulcer over the subsequent 2 years. The incidence of venous ulceration was 8.6 cases per 1,000 patient-years. Of the 12,435 patients without venous ulcers, 701 (6%) died compared to 15 of 215 (7%) who had developed venous ulceration. This difference was not significant.

     Significant demographic and health history differences between patients who did and did not develop a venous ulcer were observed (see Table 1). Patients with a history of previous venous ulceration or venous insufficiency were found to have the highest risk of future venous ulceration with odds ratios of 19.4 (95% CI 14.5–25.9, P <0.001) and 19.4 (95% CI 14.5–25.9, P <0.001), respectively (see Table 2). Thus, the presence of either previous venous ulceration or venous insufficiency increased risk of future venous ulceration 19-fold. Of the vascular disease and other comorbid risk factors evaluated, older adults with a history of pressure ulcers had the highest risk, with an odds ratio of 3.68 (95% CI 2.07–6.54, P <0.001). Other noteworthy factors included age, diabetes, rheumatoid arthritis, degenerative arthritis, peripheral neuropathy, falls, and cataracts.

     Of the initially significant factors found with univariate analysis, the remaining predictor variables in the multiple variable model were prior venous ulceration, renal insufficiency, cataracts, blindness, and pressure ulceration. After adjustment, a history of a venous ulcer had an odds ratio of 18.7 (95% CI 14.0–25.0). Prior venous ulceration remained significant following validation with an odds ratio of 19.0 (SD 2.9) and was significant in all models. Renal insufficiency was also significant in 98% of models. Pressure ulcers, cataracts, and blindness all had odds ratios >1.7; the significance of blindness changed in 49% of the models.

     After the Bayesian validation, the final model scoring system using previous venous ulcer (19 points), renal insufficiency (3 points), cataracts (2 points), and pressure ulcer (2 points) was applied to the entire cohort of 12,650 patients. Each patient was scored based upon the presence of the noted risk factor. Using ROC, the optimal cutoff total score was 3 with a sensitivity of 0.63 and a specificity of 0.88 (see Figure 1). The area under the curve was 0.797 with a standard error of 0.018 (Figure 1).

Discussion

     This study presents novel findings regarding risk factors for venous ulceration in a cohort of older community-dwelling patients. The incidence of 8.6 cases per 1,000 patient-years in patients >60 years of age is a novel finding.2 Previous venous ulceration remains the strongest predictor of future venous ulceration both in the univariate and multivariate models. Presence of prior venous ulceration with a score of 19 on the instrument automatically places a patient at high risk, inferring the need for appropriate preventive measures.

     In a review of other cohort studies15 of patients of all ages without an emphasis on older adults, up to 70% of patients with an initial venous ulcer developed a subsequent venous ulcer. Venous insufficiency is the primary physiological reason for venous ulceration; univariate analysis determined an odds ratio of 19.4 (95% CI 14.5–25.9) for future ulceration in patients with venous insufficiency. These results indicate a 19-fold increased risk of venous ulceration in patients with venous insufficiency compared to patients without venous insufficiency. However, this was not an independent risk factor after multivariate analysis and was not included in the final model. Ulceration of the leg often involves venous incompetence. In a cross-sectional study of 463 lower extremity ulcers, 72% of subjects had venous insufficiency as documented by ultrasound.16

     In the current study, pressure ulcers also had a strong association with future venous ulceration, with an odds ratio of 3.68 (95% CI 2.07–6.54). Pressure ulcers have not previously been described in association with venous ulceration. The final model of significant risk factors of prior venous ulcers, pressure ulceration, blindness, cataracts, and venous insufficiency performed well when placed in a ROC with a cutoff score of three. The area under the curve was 0.797 with a SE of 0.018, an optimal sensitivity of 0.63, and a specificity of 0.88. These parameters indicate a well-functioning model. The well-defined curve in Figure 1 demonstrates a robust model as indicated by the high area under the curve. The high specificity of 0.88 indicates that if a patient has a low score, the risk of development of a venous ulcer is low. However, a high score on the predictive instrument does not necessarily mean the patient will develop a venous ulcer.

     Demographic risk factors were not related to venous ulceration in the final model. In the initial univariate model, age was a risk factor for venous ulcer development with an odds ratio of 1.05 (95% CI 1.04–1.06). However, after adjustment, age was not included in the final model. Although previous cohort studies in a broad community17 have shown that advancing age is associated with venous ulceration, within the narrow age range of the current study age did not influence venous ulcer development.

     In this study, gender also was not associated with venous ulcer development. As shown in previous retrospective cohort studies,17-19 in general practice more women than men have venous insufficiency. In one cohort study,19 the incidence of venous ulcers was higher in younger women; however, in men age 60 to 70 years, the incidence of venous ulcers may be higher. Age homogeneity may explain the lack of gender difference in the current study. Marital status also was not associated with venous ulceration.

     Of the vascular risk factors, only renal insufficiency was included in the final multivariate model. The presence of renal disease has been noted to affect healing in both venous and ischemic ulcers.20 The remaining vascular illnesses of combined heart disease (coronary disease, MI, CHF, stroke, PAD, and hyperlipidemia) were not risk factors for venous ulceration on unadjusted analysis. Hypertension was borderline significant (P = 0.09) in univariate analysis and was not significant after adjustment with other variables. One retrospective cohort study21 involving long-term care residents found CHF was associated with venous ulceration. The concurrence of arterial and venous disease in lower extremity ulceration is common in many patients; thus, there may be a potential association between vascular disease and venous insufficiency.2,22 One explanation in the current cohort may be that underlying vascular disease may not have been recognized or was not reported. These findings reinforce the importance of venous disease and edema, not underlying ischemic disease, as primary risk factors of venous ulceration.

     Cataracts and blindness were significantly related to venous ulcer development in both unadjusted analysis and in the final model. The presence of cataracts was important, with an initial odds ratio of 1.94 (95% CI 1.36–2.77). After the stepwise multivariate model, the odds ratio was 1.64 (95% CI 1.16–2.34). This association has not been previously reported. Blindness had the lowest concurrence of significance using the Bayesian validation. The possible reasons for cataracts and blindness as risk factors are unclear; however, they may relate to the challenges of skin care with decreased vision or potential leg trauma. Ultimately, cataracts as a risk factor for venous ulceration should be considered by providers.

     Diabetes, degenerative arthritis, falls, peripheral neuropathy, and inflammatory arthritis were significant on univariate evaluation but were not significant after multivariate analysis. Previous large epidemiological studies in patients with rheumatoid arthritis often show an association between rheumatoid arthritis and venous ulceration, confirming current findings on univariate analysis but differing after adjustment.23

     Cancer, depression, dementia, hypothyroidism, and chronic obstructive pulmonary disease were not associated with venous ulceration. These findings differ from previous studies evaluating the relationship between depression and venous ulceration. Depression and anxiety have previously been reported as significant factors in patients with venous ulceration.24,25 Given these differences, it would be beneficial to validate this model in different populations.

Study Limitations

     This study’s strengths enhance the validity of the multivariate model and the findings but the retrospective cohort design and data collection method present limitations. Mayo Clinic maintains a robust data system and comprehensive EMR that collects demographic, medical utilization, and comorbid health data. Administrative data systems suffer inherent weaknesses, including potential missing or miscoded information — possibly, mortality or significant comorbid illnesses could be missed. Previous cross-sectional survey studies12,13 evaluating the accuracy of the medical record at Mayo Clinic, Rochester estimate accuracy at 97% for specific conditions. Generally, the medical record may be more accurate than survey or self-report.26

     A second concern involves the incident diagnosis of venous ulceration. It is possible that some patients developed a venous ulcer and did not seek medical attention; thus, this model represents the more serious cases of venous ulceration that prompted patients to seek medical attention. The diagnosis of venous ulceration was made clinically by the provider without verification of venous insufficiency with diagnostic testing; thus, potential for misclassification and nonspecific billing codes exists. However, the authors note that in clinical practice, most venous ulcers are diagnosed without benefit of diagnostic tests.

     The ability to generalize this study beyond the study county must be tempered. The population of Olmsted County is primarily Northern European and more than 90% of the population is Caucasian.27 However, despite this potential study limitation, many of the risk factors in this older population confirm previous risk factors in different age and ethnic populations.

Conclusion

     During a 2-year follow up, the incidence of venous ulcers in this population of community-dwelling persons 60 years or older was 1.7% or 8.6 cases per 1,000 patient-years. Significant differences between persons who did and did not develop a venous ulcer were observed. A history of previous venous ulcers was the strongest predictor of future ulceration; thus, patients with prior venous ulceration should be considered at high risk for future venous ulceration and provided aggressive edema control. Renal insufficiency, blindness, cataracts, and a history of pressure ulceration also were important factors for the development of venous ulcers. In clinical practice, a history of venous disease should encourage aggressive prevention measures. The information was used to develop a predictive model to help guide treatment, particularly in patients with edema. These findings can help providers target preventive measures through edema management in high-risk older adults. Further studies to prospectively assess use of the model and subsequent aggressive intervention have the potential to enhance clinical care for patients at risk for venous ulceration.

References: 

1. Ruckley CV. Socioeconomic impact of chronic venous insufficiency and leg ulcers. Angiology. 1997;48(1):67–69.

2. Forssgren A, Fransson I, Nelzen O. Leg ulcer point prevalence can be decreased by broad-scale intervention: a follow-up cross-sectional study of a defined geographical population. Acta Derm Venereol. 2008;88(3):252–256.

3. Fu X, Sheng Z, Cherry GW, Li Q. Epidemiological study of chronic dermal ulcers in China. Wound Repair Regen. 1998;6(1):21–27.

4. de Araujo T, Valencia I, Federman DG, Kirsner RS. Managing the patient with venous ulcers. Ann Intern Med. 2003;138(4):326–334.

5. Cullum N, Nelson EA, Fletcher AW, Sheldon TA. Compression for venous leg ulcers. Cochrane Database of Systematic Reviews. 2009;(1):CD000265.

6. Takahashi PY, Kiemele LJ, Jones JP Jr. Wound care for elderly patients: advances and clinical applications for practicing physicians. Mayo Clin Proc. 2004;79(2):260–267.

7. O'Meara S, Cullum NA, Nelson EA. Compression for venous leg ulcers. Cochrane Database Syst Rev. 2009(1):CD000265.

8. Magnusson MB, Nelzen O, Volkmann R. Leg ulcer recurrence and its risk factors: a duplex ultrasound study before and after vein surgery. Eur J Vasc Endovasc Surg. 2006;32(4):453–461.

9. Kulkarni SR, Gohel MS, Wakely C, Minor J, Poskitt KR, Whyman MR. The Ulcerated Leg Severity Assessment score for prediction of venous leg ulcer healing. Br J Surg. 2007;94(2):189–193.

10. Association WM. Declaration of Helsinki: Ethical Principles for Medical Research Involving Human Subjects. Paper presented at the World Medical Association 18th General Assembly. Helsinki, Finland. 1964.

11. Bonham PA. Assessment and management of patients with venous, arterial, and diabetic/neuropathic lower extremity wounds. AACN Clinical Issues. 2003;14(4):442–456; quiz 548–550.

12. Melton LJ III. History of the Rochester Epidemiology Project. Mayo Clin Proc. 1996;71(3):266–274.

13. Leibson CL, Needleman J, Buerhaus P, et al. Identifying in-hospital venous thromboembolism (VTE): a comparison of claims-based approaches with the Rochester Epidemiology Project VTE cohort. Med Care. 2008;46(2):127–132.

14. Pepe MS. The Statistical Evaluation of Medical Tests for Classification and Prediction. New York, NY: Oxford Press; 2003.

15. Finlayson K, Edwards H, Courtney M. Factors associated with recurrence of venous leg ulcers: a survey and retrospective chart review. Int J Nurs Stud. 2009;46(8):1071–1078.

16. Nelzen O, Bergqvist D, Lindhagen A. Leg ulcer etiology — a cross-sectional population study. J Vasc Surg. 1991;14(4):557–564.

17. Baker SR, Stacey MC, Jopp-McKay AG, Hoskin SE, Thompson PJ. Epidemiology of chronic venous ulcers. Br J Surg. 1991;78(7):864–867.

18. Boccalon H, Janbon C, Saumet JL, Tafani A, Roux T, Vilain C. Characteristics of chronic venous insufficiency in 895 patients followed in general practice. Int Angiol. 1997;16(4):226–234.

19. Heit JA, Rooke TW, Silverstein MD, et al. Trends in the incidence of venous stasis syndrome and venous ulcer: a 25-year population-based study. J Vasc Surg. 2001;33(5):1022–1027.

20. Treiman GS, Oderich GS, Ashrafi A, Schneider PA. Management of ischemic heel ulceration and gangrene: an evaluation of factors associated with successful healing. J Vasc Surg. 2000;31(6):1110–1118.

21. Wipke-Tevis DD, Rantz MJ, Mehr DR, et al. Prevalence, incidence, management, and predictors of venous ulcers in the long-term-care population using the MDS. Adv Skin Wound Care. 2000;13(5):218–224.

22. Etufugh CN, Phillips TJ. Venous ulcers. Clin Dermatol. 2007;25(1):121–130.

23. Baker SR, Stacey MC, Singh G, Hoskin SE, Thompson PJ. Aetiology of chronic leg ulcers. Eur J Vasc Surg. 1992;6(3):245–251.

24. Barron GS, Jacob SE, Kirsner RS. Dermatologic complications of chronic venous disease: medical management and beyond. Ann Vasc Surg. 2007;21(5):652–662.

25. Jones J, Barr W, Robinson J, Carlisle C. Depression in patients with chronic venous ulceration. Br J Nurs. 2006;15(11):S17–S23.

26. Brown ML, Quinonez LG, Schaff HV, Sundt TM III. What are patients really telling us? Comparison of survey responses and the medical record in cardiovascular surgical patients. Ann Thorac Surg. 2008;86(6):1888–1896.

27. Bureau UC. US Census. Available at: www.census.gov.Accessed December 1, 2008.

Section: