ORIGINAL

ARTICLE

Cellulitis Recurrence Score: A tool for predicting recurrence of lower limb cellulitis Evelyn Yuxin Tay, MRCP,a Stephanie Fook-Chong, MSc,b Choon Chiat Oh, MRCP,c Thamotharampillai Thirumoorthy, FRCP,c Shiu Ming Pang, FRCP,c and Haur Yueh Lee, MRCPc Singapore Background: Cellulitis is the most common skin and soft tissue infection and is associated with frequent recurrences. Objectives: An objective of our study was to identify factors for recurrence in patients who present with a first episode of lower-limb cellulitis. A secondary aim was to formulate a score based on observed clinical risk factors that might predict recurrence within a year. Methods: Dermatology referral forms and national computerized records were reviewed from 2003 to 2012. Demographics, coexistent dermatoses, local factors, and comorbidities were reviewed. Results: A total of 102 (45.3%) of 225 patients had recurrence. Multivariate analysis showed that lymphedema (P \.0005), chronic venous insufficiency (P \.0005), peripheral vascular disease (P = .002), and deep vein thrombosis (P = .008) predicted for recurrence. The Cellulitis Recurrence Score (CRS) was constructed based on these factors. CRS $ 2 was associated with a positive predictive value of 83.6% and negative predictive value of 67.5%. Model performance was good (Hosmer-Lemeshow statistic, P = .753). Limitations: This is a retrospective study limited to an inpatient cohort. Conclusion: Lymphedema, chronic venous insufficiency, peripheral vascular disease, and deep vein thrombosis were risk factors. CRS is reliable for predicting recurrence, and early interventions should be considered in patients with CRS $ 2. ( J Am Acad Dermatol http://dx.doi.org/10.1016/j.jaad.2014.08.043.) Key words: cellulitis; erysipelas; recurrence; lymphedema; chronic venous insufficiency; stasis dermatitis; peripheral vascular disease; deep vein thrombosis.

C

ellulitis is an acute, pyogenic inflammation of the dermis and subcutaneous tissue,1 with an incidence of 16.4 to 24.6 per 1000 person-years.2 Cellulitis accounts for 14% of visits to the emergency department and contributes up to 7% of hospital admissions in the United States.3,4 Cellulitis contributes to an average length-of-stay of 6 days and a hospital bill of at least $13,000 US dollars per admission.5 The all-cause mortality rate for patients admitted for cellulitis is 7.2%.6 Microbiologic analyses from blood cultures, aspirates, and skin biopsies isolate various groups of Streptococci in75% to 90% of cases.7

Recurrence is common, occurring at 8% to 20% per year.8,9 Cellulitis contributes to postcellulitic edema, setting up a vicious cycle, which further predisposes subjects to recurrent cellulitis.10 Measures such as the administration of prophylactic antibiotics11 and reduction of modifiable risk factors

From the National Skin Centre, Singapore,a Health Services Research and Biostatistics, Division of Research, Singapore General Hospital,b and Dermatology Unit,c Singapore General Hospital. Funding sources: None. Conflicts of interest: None declared. Accepted for publication August 26, 2014.

Reprint requests: Haur Yueh Lee, MRCP, Dermatology Unit, Singapore General Hospital, Outram Road, Singapore 169608. E-mail: [email protected]. Published online October 16, 2014. 0190-9622/$36.00 Ó 2014 by the American Academy of Dermatology, Inc. http://dx.doi.org/10.1016/j.jaad.2014.08.043

Abbreviations used: CRS: Cellulitis Recurrence Score CVI: Chronic venous insufficiency DVT: Deep vein thrombosis

1

J AM ACAD DERMATOL

2 Tay et al

than 3 months’ duration, consistent with at least stage II of the grading of the International Society of Lymphology.12 Chronic kidney disease was taken as a glomerular filtration rate of less than 60 mL/min/ 1.73m2 for at least 3 months, which corresponds to stage 3 on the Kidney Disease Outcomes Quality Initiative Chronic Kidney Disease classification. Liver METHODS cirrhosis was defined by the The study was conducted presence of fibrosis detected based on subjects referred to CAPSULE SUMMARY on imaging. A body mass the inpatient dermatology index of $ 25kg/m2 was consult service at our 1500Recurrence of lower limb cellulitis is considered obese. bed tertiary hospital. This common. Statistical analysis was perstudy was reviewed and Clinical factors that predict recurrences formed using R version 3.01 approved by the ethical reare lymphedema, venous insufficiency, (R Foundation for Statistical view board of our institution deep vein thrombosis, and peripheral Computing, Vienna, Austria). (Protocol: 2012/1003/E). vascular diseases. Statistical significance was set Inclusion criteria were paat P \ .05 for all testing. For tients with lower extremity Cellulitis Recurrence Score is a scoring univariate analyses, differcellulitis at or below the level system based on presence or absence of ences between the 2 groups of the buttocks, aged 18 lymphedema, venous insufficiency, deep were assessed using the x2 years and older and with vein thrombosis, and peripheral vascular test for association for catethe diagnosis of cellulitis disease. It is useful for predicting gorical variables and 2made by a consultant dermarecurrence in lower limb cellulitis. sample Student t test or tologist (TT, SMP, HYL). Mann-Whitney U for continSubjects with necrotizing fasuous variables. The multivariate model for prediction ciitis or with a secondarily infected dermatologic of recurrent cellulitis was constructed by considering process such as folliculitis, septic bursitis with overall factors emerging as significant (at significance level lying cellulitis, septic arthritis with surrounding of P \.05) in the univariate analysis, using backward cellulitis, carbuncles, and furuncles were excluded. stepwise variable selection with entry and exit criteria Inpatient consultation and national computerized of P \.01 and P [.05, respectively. medical records were retrospectively reviewed for Scores were constructed by converting into the period of March 2003 to March 2012 by an integers the regression coefficients of independently independent doctor (ET) to ensure that the clinical predictive factors in the multivariate logistic features were in concordance with cellulitis. regression model. Presence of any independent A case of recurrence was defined as a second risk factor was awarded the corresponding integer episode of cellulitis occurring in the same limb score. A total score, named Cellulitis Recurrence within 1 year of the first episode, with documented Score (CRS), was calculated by summing up the clearance of cellulitis between the 2 episodes. This scores for each independent variable. definition was either by (1) follow-up documentaThe bootstrap technique (resampling without tion of clearance in the medical records, (2) the replacement from original dataset) with 200 absence of any further hospital inpatient and outparesamples was used to generate the c-statistics for tient consultations for the same episode, (3) absence assessing discrimination and the bias-corrected of documentation of failure of therapy after the initial calibration curve for assessing calibration. The episode, and (4) absence of prescription of further c-statistics and calibration curves were measures of courses of antibiotics within 2 months of the initial the performance of the prediction tool.13 A c-statistic episode. Data on demographics, the presence of between 0.7 and 0.9 is an indicator of fair to good coexistent dermatoses, local factors predisposing to discrimination, and a calibration curve close to the cellulitis, comorbidities, and laboratory investigaideal y = x line is an indicator of good calibration . tions were analyzed. Chronic venous insufficiency (CVI) referred to the presence of at least 2 of the following: sclerotic skin, RESULTS varicose veins, pigmented purpuric dermatoses, or During the study period, 225 patients were the detection of venous insufficiency on Doppler referred to the dermatology inpatient consultation studies. Lymphedema was defined as a chronic, service with cellulitis. Table I summarizes their progressive swelling of the affected limb for more baseline characteristics. have shown promise in reducing the rate of recurrence. We seek to identify factors for recurrence in patients presenting with their first episode of lower limb cellulitis and propose a validated scoring system predictive of the risk of recurrence.

d

d

d

J AM ACAD DERMATOL

Tay et al 3

Table I. Characteristics of 225 patients referred to the inpatient dermatology consult service for cellulitis Patient characteristics*

Gender Male Female Race Chinese Indian Malay Others Coexistent dermatoses Tinea pedis Toe web intertrigo Feet eczema Lymphedema Asteatotic eczema Onychomycosis Lichen simplex chronicus Psoriasis Lichen amyloidosis Local factors Chronic venous insufficiency Surgery involving lymphatics or blood vessels Ipsilateral deep vein thrombosis Ipsilateral orthopedic implant in situ Previous saphenous venectomy Previous abscess drainage surgery Previous high tie ligation Comorbidities Obesity Missing data BMI $ 25 BMI \ 25 Diabetes mellitus Ischemic heart disease Chronic kidney disease Active malignancy Liver cirrhosis Chronic lung disease

N (%)

147 (65.3) 78 (34.7) 155 35 33 2

(68.9) (15.6) (14.7) (0.9)

56 56 37 32 32 23 12 10 7

(24.9) (24.9) (16.4) (14.2) (14.2) (10.2) (5.3) (4.4) (3.1)

74 (32.9) 43 (19.1) 26 23 22 20 13

(11.6) (10.2) (9.8) (8.9) (5.8)

114 58 53 75 62 27 16 14 8

(50.7) (25.8) (23.6) (33.3) (27.6) (12.0) (7.1) (6.2) (3.6)

BMI, Body mass index. *Age at first presentation, 57.1 6 16.9 y.

One hundred two patients (45.3%) had recurrence of cellulitis in the same limb within 1 year of the first episode. The mean duration of follow-up was 73.1 6 37.4 months. In patients with recurrence, the mean time to recurrence was 10.9 6 8.9 months, and the mean number of episodes of cellulitis was 3.9 6 4.6. Table II shows the results of the univariate analysis. On multivariate logistic regression analysis, 4 independent risk factors remained: (1) lymphedema, (2) CVI, (3) history of chronic deep vein thrombosis (DVT) in the ipsilateral limb, and

(4) peripheral vascular disease (Table III). We tested for an association between DVT and CVI and CVI and lymphedema, but there was a low kappa measurement of agreement between these factors (0.159 and 0.253, respectively). The bootstrap-corrected c-statistic for the model containing the above 4 covariates was 0.78, indicating good discrimination. Discrimination is the ability of a model to separate those who had recurrent cellulitis from those who did not. Calibration is the correspondence between the predicted probability versus the observed probability of recurrence and is assessed by a calibration plot, which showed good performance of the model (Fig 1). Results of the HosmerLemeshow test (0.753), a goodness-of-fit test comparing observed versus expected frequencies, indicated that there was a good fit. The mean absolute error, which is the average absolute difference between the predicted and bias-corrected actual probabilities, was low at 0.039. A score of 1 each was assigned if there was presence of ipsilateral DVT or CVI. Two points were added if there was lymphedema, and 3 points were assigned if there was peripheral vascular disease as a risk factor. Based on the presence or absence of these 4 factors, patients would get a CRS of 0 (minimum) to 7 (maximum). A plot of patients’ predicted probability versus their total risk score was examined to determine the creation of low, moderate, high, and very high risk groups of cellulitis recurrence (Fig 2). Patients were stratified into the 4 risk categories as follows: low (CRS = 0), moderate (CRS = 1), high (CRS = 2), and very high (CRS = 3-7) risk (Table IV). A cut-off score of CRS $ 2 is strongly predictive with a positive predictive value of 83.6% (95% confidence interval [CI], 71.2%-92.2%) and a negative predictive value of 67.5% (95% CI, 59.8%-74.5%).

DISCUSSION This was a retrospective cohort study with inherent flaws. First, there were missing data on obesity such that we were unable to validate if this was an important risk factor.14 Second, recurrences were based on review of hospital records. Fortunately, our national medical records capture prescriptions from hospitals and primary health clinics, and the antibiotic prescriptions in primary care were reviewed. Third, because the diagnosis of cellulitis was made on clinical grounds,1,7 we only included subjects whose diagnoses were made by an inpatient team of attending dermatologists. Previous studies on cellulitis had also depended on assessments by dermatologists, who were deemed more likely to be correct in their diagnoses.11,15-17

J AM ACAD DERMATOL

4 Tay et al

Table II. Results from univariate analysis of risk factors for recurrent lower extremity cellulitis Variable

Age, mean 6 SD (ys) Dermatologic conditions Tinea pedis, n (%) Intertrigo, n (%) Onychomycosis, n (%) Asteatotic eczema, n (%) Local factors Lymphedema, n (%) Chronic venous insufficiency, n (%) Ipsilateral deep vein thrombosis, n (%) Peripheral vascular disease, n (%) Previous abscess drainage of lower extremity, n (%) Previous surgery involving lymphatics or blood supply, n (%) Comorbidities Chronic kidney disease, n (%) Ischemic heart disease, n (%) Liver cirrhosis, n (%) Diabetes mellitus, n (%) Active malignancy, n (%) Obesity, no. (%) Laboratory markers Mean total white count (3106 L) Mean albumin (g/dL) Mean C-reactive protein (mg/dL)

Recurrence (n = 102)

No recurrence (n = 123)

P value

58.8 6 16.1

55.8 6 17.5

.175*

Odds ratio (95% CI)

1.01 (1.00-1.03)

32 32 15 19

(31.4) (31.4) (14.7) (18.6)

24 24 9 13

(19.5) (19.5) (7.3) (10.6)

.041 .042 .079 .089

1.89 1.89 2.18 1.94

(1.02-3.48) (1.02-3.48) (0.91-5.22) ( 0.91-4.15)

27 50 19 13 14

(26.5) (49.0) (18.6) (12.7) (13.7)

5 24 7 1 6

(4.1) (19.4) (5.7) (0.8) (4.9)

.001 .001 .003 .006 .020

8.50 3.97 3.79 17.8 3.10

(3.13-23.0) (2.20-7.16) (1.53-9.44) (2.29-138.) (1.15-8.40)

18 (17.6)

25 (20.3)

.259

0.840 (0.429-1.65)

20 35 9 39 8 28

7 27 5 36 8 30

.001 .039 .141 .155 .697 .934

4.04 1.89 2.28 1.50 1.22 0.969

.143* .436* .645*

1.041 (0.986, 1.099) 1.021 (0.969, 1.076) 1.001 (0.998, 1.004)

(19.6) (34.3) (8.8) (38.2) (7.8) (51.9)

11.9 6 0.57 31.4 6 0.796 94.9 6 11.7

(5.7) (22.0) (4.1) (29.3) (6.5) (52.6)

13.7 6 0.778 31.0 6 0.799 110 6 14.7

(1.63-10.0) (1.03-3.35) (0.740-7.05) (0.857-2.61) (0.442-3.38) (0.46-2.04)

P values in bold are those statistically significant at P \ .05 on univariate analysis. *Laboratory markers and age were analyzed as continuous data and were not dichotomized. The odds ratio for these variables represent the increase in odds of cellulitis recurrence per one unit increase in the laboratory marker.

Table III. Results from multivariate analysis of risk factors for recurrent lower extremity cellulitis and the corresponding score assigned for each significant factor Covariate

Chronic venous insufficiency* Ipsilateral deep vein thrombosisy Lymphedema*y Peripheral vascular disease

Odds ratio (95% CI)

3.45 3.81 9.18 26.90

(1.79-6.66) (1.41-10.31) (3.22-26.16) (3.30-218.92)

P value

B coefficient

Score

\ .0005 .008 \ .0005 .002

1.2 1.3 2.2 3.3

1 1 2 3

*Kappa measurement of agreement between chronic venous insufficiency and lymphedema = 0.159. y Kappa measurement of agreement between ipsilateral deep vein thrombosis and lymphedema = 0.253.

In addition, 93.3% of our patients had cellulitic episodes with unilateral involvement, 89.2% of the cohort had increased C-reactive protein levels and 49.5% had increased total white count, which were consistent with previous studies.7 We observed that with an increase in the number of predisposing factors, there was a cumulative increase in the recurrence risk. Hence, we proposed the CRS, a scoring system comprising the 4 factors— lymphedema, peripheral vascular disease, CVI and previous DVT, which were found to be significant in multivariate analysis. The CRS showed good discriminatory ability and calibration. Increase in CRS was

associated with an incremental increase in the risk of recurrence. In lymphedema, each episode of cellulitis further damages the lymphatic system, which propagates a vicious cycle in cellulitis.18 In CVI, accumulation of fluid within the dermis and subcutaneous tissue impairs tissue metabolism and reduces local immunity.10 The development of stasis dermatitis with venous insufficiency predisposes to fissuring, scratching, and contact dermatitis, further compromising the cutaneous barrier. In our study, toe web intertrigo and tinea pedis were not significant factors for recurrence. This study

J AM ACAD DERMATOL

Tay et al 5

Table IV. CRS and rate of recurrence in low-, moderate-, high- and very high-risk groups

Low risk (n = 117)

Moderate risk (n = 52)

High risk (n = 25)

Very high risk (n = 30)

CRS 0 1 2 3-7 26 (22.2) 29 (55.8) 19 (76.0) 27 (90.0) Patients with cellulitis recurrence for each risk group, n (%) Fig 1. Smoothed nonparametric calibration plot of actual versus predicted probability of cellulitis recurrence. The dashed line indicates an ideal calibration curve, the dotted line represents the calibration curve derived from the original sample, and the solid line shows the biascorrected calibration curve developed from bootstrap resamples.

Fig 2. Plot of probability of cellulitis recurrence against CRS. A higher CRS correlated with a greater risk of recurrence. This graph was used to decide on the stratification of CRS scores from low- to very high-risk for cellulitis recurrence.

may have been limited by the small sample size. Nonetheless, this result is in agreement with previous case-control studies of recurrent cellulitis.10,15,19 Although disruption of the cutaneous barrier in toe web intertrigo and tinea pedis are established risk factors for acute cellulitis, the association with recurrent cellulitis is less clear. Many studies that evaluated recurrence were conducted by dermatologists or infectious disease specialists who may have actively looked for and treated intertrigo and tinea pedis during the initial episode of cellulitis. In addition, antibiotics that were used to treat the index episode of cellulitis may have concurrently reduced the pathogenic bacterial load in the toe webs.19

CRS, Cellulitis Recurrence Score.

The utility of a predictive score is manifold; patients identified to be at a high risk of recurrence would benefit from close follow-up and easy access to care. These patients should also be evaluated and treated for concomitant risk factors such as intertrigo, tinea pedis, peripheral vascular disease, and CVI. Some aspects of this study may have implications for future research. Although controversial, studies on the potential role of prophylactic antibiotics in reducing recurrence in cellulitis and surgical options in reducing lymphedema are ongoing.11,20,21 With a CRS $ 2, an early and holistic approach to address the risk factors that are associated with recurrent cellulitis should be adopted. REFERENCES 1. Swartz MN. Cellulitis. N Engl J Med 2004;350:904-12. 2. Figtree M, Konecny P, Jennings Z, Goh C, Krilis SA, Miyakis S. Risk stratification and outcome of cellulitis admitted to hospital. J Infect 2010;60:431-9. 3. Dong SL, Kelly KD, Oland RC, Holroyd BR, Rowe BH. ED management of cellulitis: a review of five urban centers. Am J Emerg Med 2001;19:535-40. 4. McCall N, Harlow J, Dayhoff D. Rates of hospitalization for ambulatory care sensitive conditions in the Medicare 1 Choice population. Health Care Financing Rev 2001;22:127-45. 5. Lipsky BA, Weigelt JA, Gupta V, Killian A, Peng MM. Skin, soft tissue, bone, and joint infections in hospitalized patients: epidemiology and microbiological, clinical, and economic outcomes. Infect Control Hosp Epidemiol 2007; 28:1290-8. 6. Carratala J, Roson B, Fernandez-Sabe N, Shaw E, del Rio O, Rivera A, et al. Factors associated with complications and mortality in adult patients hospitalized for infectious cellulitis. Eur J Microbiol Infect Dis 2003;22:151-7. 7. Hirschmann JV, Raugi GJ. Lower limb cellulitis and its mimics. J Am Acad Dermatol 2012;67:163.e1-12. 8. Bergkvist Pl, Sjobeck K. Relapse of erysipelas following treatment with prednisolone or placebo in addition to antibiotics: a 1-year follow-up. Scand J Infect Dis 1998;30: 206-7. 9. Mcnamara DR, Tlyjeh IM, Berbari EF, Lahr BD, Martinez J, Mirzoyev SA, et al. A predictive model of recurrent lower extremity cellulitis in a population-based cohort. Arch Intern Med 2007;167:709-15.

6 Tay et al

10. Cox NH. Oedema as a risk factor for multiple episodes of cellulitis/ erysipelas of the lower leg: a series with community follow-up. Br J Dermatol 2006;155:947-50. 11. Thomas KS, Crook AM, Nunn AJ, Foster KA, Mason JM, Chalmers JR, et al. Penicillin to prevent recurrent leg cellulitis. N Engl J Med 2013;368:1695-703. 12. International Society of Lymphology. The diagnosis and treatment of peripheral lymphedema: 2013 Consensus Document of the International Society of Lymphology. Lymphology 2013;46:1-11. 13. Farrell FE. Regression modeling strategies with applications to linear models, logistic regression and survival analysis. 1st ed. New York: Springer Verlag NY Inc; 2001. 14. Lewis SD, Peter GS, Gomez-Marin O, Bisno AL. Risk factors for recurrent lower extremity cellulitis in a U.S. veterans medical center population. Am J Med Sci 2006;332:304-7. 15. Roujeau JC, Sigurgeirsson B, Korting HC, Kerl H, Paul C. Chronic dermatomycoses of the foot as risk factors for acute bacterial cellulitis of the leg: a case-control study. Dermatology 2004;209:301-7.

J AM ACAD DERMATOL

16. Dupuy A, Benchikhi H, Ruojeau JC, Bernard P, Vaillant L, Chosidow O, et al. Risk factors for erysipelas of the leg (cellulitis): case-control study. BMJ 1999;318:1591-4. 17. Thomas K, Crook A, Foster K, Mason J, Chalmers J, Bourke J, et al. Prophylactic antibiotics for the prevention of cellulitis (erysipelas) of the leg: results of the UK Dermatology Clinical Trials Network’s PATCH II trial. Br J Dermatol 2012; 166:169-78. 18. Al-Niaimi F, Cox N. Celllulitis and lymphedema: a vicious cycle. J Lymphedema 2009;4:38-42. 19. Leclerc S, Texeria A, Mahe E, Descamps V, Crickx B, Chosidow O. Recurrent erysipelas: 47 cases. Dermatology 2007;214:52-7. 20. Oh CC, Ko HC, Lee HY, Safdar N, Maki DG, Chlebicki MP. Antibiotic prophylaxis for preventing recurrent cellulitis: a systematic review and meta-analysis. J Infect 2014;69: 26-34. 21. Becker C, Vasile JV, Levine JL, Batista BN, Studinger RM, Chen CM, et al. Microlymphatic surgery for the treatment of iatrogenic lymphedema. Clin Plast Surg 2012;39:385-98.

Cellulitis Recurrence Score: a tool for predicting recurrence of lower limb cellulitis.

Cellulitis is the most common skin and soft tissue infection and is associated with frequent recurrences...
272KB Sizes 0 Downloads 8 Views