Factors associated with postsurgical wound infections among breast cancer patients: A retrospective case‐control record review

Abstract

Women with non‐metastatic breast cancer will be offered surgery as their first option. Unfortunately, studies have shown that the most common postoperative complication is surgical wound infection (SWI). We investigated the prevalence of SWI in breast cancer patients and identified the factors predictive of its development. The study was conducted at the breast cancer centre in Singapore. A retrospective case‐control review of medical records was used. During the 2013 to 2016 study period, there were 657 postsurgical breast cancer patients with only 105 records eligible for the study. The sample consisted of one to four case:control (21:84), matched according to their age at the time of their surgery. Patients presenting with SWI were grouped into cases, while those without SWI were grouped into the controls. Chi‐square test and Mann‐Whitney U test were used to identify risk factors associated with SWI. Regression analysis of predictive variables from the univariate analyses was included. These variables were type of breast surgery, implants, comorbidities, previous surgery, previous chemotherapy, surgical drains, seroma, blood transfusion, surgeon department, and length of stay. The prevalence of SWI was 9%. Demographic, clinical, and comorbidities were not associated with SWI. However, multivariate analysis found that “surgeon department,” “discharged with surgical drains,” and “postoperative seroma” were predictive of SWI. Monitoring SWI is indispensable to minimise burdens on individuals and institutions. Health care professionals should identify high‐risk patients based on the identified predictive variables. A cross‐institutional record review of SWI in postoperative breast cancer patients should be conducted.

1. INTRODUCTION

Globally, breast cancer is reported as the most frequently diagnosed malignancy in both developed and developing countries. ¹ In Singapore, breast cancer screening has been widely promoted with self‐examinations and subsidised mammograms for women above the age of 50 years. Typically, wide excision or mastectomy procedures will be offered to women diagnosed with non‐metastatic breast cancer, ² with the choice to proceed with breast reconstructive procedures with implants or tissue flaps. ³ Depending on the extent of the disease, chemotherapy and radiation therapy may also be offered as neoadjuvant postsurgery. With surgery being the linchpin of breast cancer treatment, it is important to study its adverse outcomes, the most prevalent being surgical wound infection (SWI), accounting for up to 20% of hospital acquired infections. ⁴

SWI refers to infections that occur at primary or secondary incisional sites, affecting either superficial or deep tissues which can occur up to a year after the operative procedure. ⁵ The International Nosocomial Infection Control Consortium ⁶ reported a rate of 1.7% for SWI in patients after breast surgery from years 2005 to 2010 in a prospective cohort multinational multi‐centre surveillance study across 66 cities in 30 countries including Singapore. SWI was also found to be associated with a delay in adjuvant treatment such as radiation and chemotherapy. ⁷ While breast cancer surgery is categorised as a clean operation, the rate of SWI was 30% compared with 5% for other clean operations. ⁸ , ⁹ , ¹⁰

In a systematic review and meta‐analysis, 14 out of 20 possible risk factors were found significant for SWIs including increased age, hypertension, higher body mass index (BMI), diabetes mellitus, previous operation, preoperative chemoradiation, operation type, hematoma, seroma, significant intraoperative bleeding, postoperative drain, longer drainage time, and drainage tube in place. ⁹ Seroma is a serous fluid collection under the skin flaps that usually occurs after breast and axillary surgery filling the dead space and is known to contribute to SWI. ¹¹ Smoking habit, immediate reconstruction, axillary lymph node dissection or in the axillary, preoperative chemotherapy, corticosteroid usage, and prophylactic antibiotic were also found to be associated with SWI. ¹²

SWI development is detrimental for both the patient's and their caregiver's physical and mental well‐being, affecting their quality of life and their finances. ¹³ , ¹⁴ , ¹⁵ Economic effects of SWI to health care institutions contribute to substantial increase in clinical and economic burdens associated with an increased length of stay (LOS), readmissions, and subsequent surgeries required for the treatment of SWI. ¹⁶ , ¹⁷ , ¹⁸ Considering the various factors associated with SWI development, its repercussions on individuals and institutions, and the paucity of local studies on this topic in Singapore, it was timely to conduct this study to contribute to existing evidence. With surgery being the mainstay of treatment for breast cancer, which affects 29% of Singaporean women and 2.09 million women globally, it is paramount for SWIs to be minimised. ¹⁹ Examining the predictive risk factors for SWI will provide health care professionals optimal preventive measures before and after surgery for breast cancer patients. Moreover, it is imperative to discern the prevalence rate of postoperative SWIs among breast cancer patients and its predictive variables to compare and improve surgical outcomes for future patients who will undergo similar surgeries. The aims of our study were as follows: (a) to examine the prevalence of post‐breast‐cancer SWI, and (b) to investigate the factors associated with and predictive of SWI in breast cancer patients.

2. METHODS

2.1. Design

A quantitative case‐control design involving the retrospective review of medical records of breast cancer patients who underwent breast surgery was adopted. SWI cases were compared with patients without documented SWI who underwent breast surgery during the same period. Patients presenting with SWI were grouped into cases, while those without SWI were grouped into the controls.

2.2. Sample and setting

The study was conducted at a tertiary hospital in Singapore. In 2016, only 5.5% of breast cancer patients who sought surgical treatment were subsequently diagnosed with wound infection at the study site. In view of the low rates of patients with SWI, there were very few cases to select from in comparison with the larger population of patients without SWI. As such, this study did not adopt an equal case to control ratio. This is justifiable, because an equal ratio of case to control with a power calculation between 0.1 and 0.9 indicates that an increase in the number of controls would lead to a statistically relevant increase in power such as when the controls are more than four. ²⁰ , ²¹ Therefore, a 1:4 case to control ratio was adopted. In this study, the collected records were matched according to age using aforementioned ratio to minimise the influence it might have on outcomes and allow for an accurate association of SWI with plausible predictive factors. Medical records were searched for patients who fit the inclusion criteria for cases and controls. The inclusion criteria for both case and control groups included: age over 18 years, all female postoperative breast cancer patients, admitted and discharged between January 2013 and December 2016. For the case group, they should have a diagnosis of SWI within 1 year after surgery and for the control group, no diagnosis of SWI. Exclusion criteria for both groups included: all male postsurgical breast cancer patients and those already diagnosed with SWI at other body sites prior to breast surgery. Out of a total of 657 medical records reviewed, 379 were excluded for not meeting the inclusion criteria. Another 173 (4 cases and 169 controls) records were removed for not matching the 1:4 case‐control ratio required for the study. The final sample was 105 (21 cases and 84 controls) records included in the study (Figure 1).

FIGURE 1.

Inclusion of records for analysis

2.3. Outcome measures

The sociodemographic data of breast cancer patients included: age, ethnicity, marital status, height, weight, BMI, smoking status, and menopause status. Clinical data extracted were: comorbidities (diabetes, hypertension, and hyperlipidemia), type of surgery, length of surgery, length of hospital stay, previous surgery, previous chemotherapy, previous steroid therapy, blood loss, presence of surgical drains, postoperative hematoma, postoperative seroma, postoperative antibiotics, and duration of antibiotics therapy and surgeon department.

2.4. Procedure

A total of 657 medical records of postoperative breast cancer patients were retrieved from the hospital's secured medical records. The medical records were screened for eligibility, extracted onto the hardcopy data extraction form, and allocated into cases and controls matching ensuring the 1:4 ratio. Subsequent to data extraction, the records were verified by another researcher before they were entered into the study hospital secured Research Electronic Data Capture (REDCap) for analysis.

2.5. Data analysis

Data from REDCap was imported to IBM SPSS Statistics 24.0 for analysis. Descriptive statistics were used to present sociodemographic and clinical data. Continuous data were analysed using central tendency, median, and interquartile range, whereas categorical variables were analysed using frequency and percentage. Prior to data analysis, normality testing of continuous variables was carried out using Shapiro‐Wilk Test. Continuous variables that were not normally distributed were analysed using Mann‐Whitney U Test. This analytic procedure explored significant differences between cases and controls with regard to each continuous variable, adopting a P‐value of .05 for significance testing. Chi square testing was used to determine the significant differences in categorical variables between the cases and controls. ²² Variables such as serum levels of haemoglobin, haematocrit, white blood cell, glucose, and sodium were translated into dichotomous variables, with their normal values coded “1” and abnormal values coded “2.”

To explore the factors associated with SWIs, Pearson's Chi‐square test and Mann‐Whitney U Test were used to identify variables associated with SWI development at significance level P < .05, then a multiple logistic regression using the “Backward: Wald” method was used to elucidate factors predictive of SWI. Variables eligible for inclusion in the regression comprised known risk factors and those with P‐values <.20 from the univariate analysis. ²³ , ²⁴

2.6. Ethical consideration

This study was approved by the SingHealth's Centralised Institutional Review Board (Reference number: 2017/2940). As the data will only be collected from past medical records of patients who sought treatment in the health care institution, there was no risk involved in conducting the study. Because of the retrospective nature of this study, anonymity was impossible to attain therefore, appropriate confidentiality procedures were implemented to ensure sensitive patient information are expunged. Each record extracted was coded with “CO” for controls and “CS” for cases, to disguise the participants' identities. With the absence of direct contact with the patients and the lack of intervention to be implemented, no consent was sought from the participants. The waiver of consent was approved as the study will not affect the rights and welfare of the study participants.

3. RESULTS

3.1. Demographic characteristics of cases and controls

Patients' age ranged from 31 to 73 years with a median age of 55 (IQR = 51‐65). Among the cases and controls, there were 14 and 70 Chinese, respectively. Cases and controls have a median BMI of 25.4 kg/m² (20.6‐29.95) vs 24.1 kg/m² (IQR = 22.1‐27.9) with a P‐value of .53. The sample was predominantly non‐smokers (cases = 100%, controls = 77.4%) and married (cases = 85.7%, controls = 82.1%). Of the 21 cases, 14 were menopausal (66.7%), while the controls were menopausal (67.9%). Ethnicity (P = .22), smoking (P = .26), menopause (P = .58), and marital status (P = .09) were not associated with SWI (Table 1).

TABLE 1.

Demographic data of cases and controls (n = 105)

		N (%)/median (IQR)
		Case (n = 21)	Control (n = 84)	χ 2	P‐value
Age (years)		55 (51‐65.5)	55 (50.25‐64.75)	NA	NA
Ethnic group	Chinese Malay Others a	14 (66.7) 3 (14.3) 4 (19.1)	70 (83.3) 7 (8.3) 7 (8.3)	3.049	.22
Marital status	Married Single Divorced/widowed	18 (85.7) 2 (9.5) 1 (4.8)	69 (82.1) 11 (13.1) 4 (4.8)	0.199	.09
Smoking	Yes No Not recorded	0 21 (100) 0	4 (4.8) 65 (77.4) 15 (17.9)	1.274	.26
Menopause	Yes No Not recorded	14 (66.7) 6 (28.6) 1 (4.8)	57 (67.9) 18 (21.4) 9 (10.7)	0.301	.58
BMI (kg/m2)		25.4 (20.6‐29.95)	24.1 (22.1‐27.9)	NA	.53 b

^aOthers = Arab, Burmese, and Indians.

^b P = value arrived from Mann‐Whitney U test.

3.2. Clinical and surgical characteristics of cases and controls

Diabetes mellitus (χ ² = 0.24, P = .63), hypertension (χ ² = 0.96, P = .33), and hyperlipidaemia (χ ² = 2.27, P = .13) were not significant between groups. Chemotherapy was the most common adjuvant treatment across both groups. Previous radiation was not performed among cases compared with controls (3.6%). Previous radiation was not performed among cases compared with controls (3.6%). Steroid therapy was not used in both cases and controls. Cases had a slightly longer median duration of surgery of 1.54 hours compared with 1.39 hours in the control group. Both groups had a comparable median hospital LOS of 2 days. All the clinical and surgical characteristics were not associated with SWI. However, the “Type of surgery” and “Surgical team” were found to be associated with SWI in the univariate analysis, with P‐values of .04 and .05, respectively (Table 2).

TABLE 2.

Clinical and peri‐operative data in cases and controls (N = 105)

		N (%)/case (n = 21)	Median (IQR) control (n = 84)	χ 2	P‐value (<.05)
Diabetes
Yes		5 (24)	19 (19)	0.24	.63
No		16 (76)	68 (81)
Hypertension
Yes		8 (38)	42 (50)	0.96	.33
No		13 (62)	42 (50)
Hyperlipidemia
Yes		5 (24)	35 (42)	2.27	.13
No		16 (76)	49 (58)
Length of hospitalisation (days)		2 (1‐6)	2 (1‐3)	NA	.165
Length of surgery (hh:mm)		1:54 (1:30‐3:10)	1:39 (1:20‐2:20)	NA	.205
Type of surgery	SM WE Others a	15 (71.4) 2 (9.5) 4 (19)	41 (48.8) 32 (38.1) 11 (13.1)	6.264	.044
Blood loss	Yes No	0 21 (100.0)	4 (4.8) 80 (95.2)	1.04	.308
Previous surgery	Yes No Not recorded	8 (37.6) 10 (48.0) 3 (14.4)	41 (48.4) 40 (48.0) 3 (3.6)	3.788	.15
Previous chemotherapy	Yes No	3 (14.3) 18 (85.7)	5 (6) 79 (94)	1.658	.198
Previous irradiation	Yes No	0 21 (100)	3 (3.6) 81 (96.4)	0.772	.380
Steroid therapy	Yes No	0 21 (100)	0 21 (100)	NA	NA
Surgical drains	Yes No	21 (100.0) 0	58 (69.0) 26 (31.0)	8.639	.03
Postoperative seroma	Yes No	12 (57.1) 9 (42.9)	24 (28.6) 60 (71.4)	6.195	.04
Postoperative haematoma	Yes No	0 21 (100.0)	2 (2.4) 82 (97.6)	0.51	.48*
Postoperative antibiotics given	Yes No	5 (24.0) 16 (76.0)	12 (14.3) 72 (85.7)	1.123	.30
Duration of discharged antibiotics	≤1 week >1 week None	1 (4.8) 3 (14.4) 17 (80.8)	12 (14.3) 1 (1.2) 71 (84.5)	0.327	.567
Surgeon department	GS GS/PS	16 (76.2) 5 (23.8)	74 (88.1) 10 (11.9)	1.944	.163*

Abbreviations: GS, general surgeon; PS, plastic surgeon.

^aCompletion mastectomy, nipple sparing mastectomy, skin sparing mastectomy, and wire localisation surgery.

^*Significant at P < 0.05 derived from Mann‐Whitney U Test.

During the intra‐operative period, only four (4.8%) controls were reported to have blood loss during the operation, where three (3.6%) lost less than 500 mL and one (1.2%) losing more than 500 mL of blood. Cases and controls were given intra‐operative prophylactic antibiotics. All intra‐operative factors were not associated with SWI except for surgical drains (100% of cases and 58% of controls, χ ² = 8.64, P = .003).

All case and control groups were discharged to their own homes. During the postoperative period, none of the cases had postoperative haematoma in comparison with two in the control group. Postoperative seroma, a common postoperative complication, was higher among cases (n = 12, 57.1%) than controls (n = 24, 28.1%) and was found to be associated with SWI (P = .045, χ ² = 6.19). Five (24%) cases and 12 (14.4%) controls had postoperative antibiotics, who were discharged with antibiotics requiring them for more than 1 week. All aforementioned variables were not associated with SWI except in 18 (85.6%) cases and 49 (58.3%) controls discharged with drains. Univariate analysis found that patients who were discharged with drains were associated with SWI development (χ ² = 5.45, P = .02).

3.3. Independent variables as predictors for SWI development

Variables included in the regression models were based on the results of univariate analysis as shown in Table 3. The final regression model presented in Table 4 showed that “surgical team” (P = .009), “home with surgical drains” (P = .033) and “postoperative seroma” (P = .047) were significant predictors of SWI. The Cox & Snell R square showed that the model explains 14% of the variance with SWI. Patients presenting with postoperative seroma were found to be three times more likely to develop a SWI (OR = 3.304, 95% CI: 1.018‐10.723) compared with those without seroma. Operative procedures performed by both general surgery (GS) and plastic surgery (PS) department were 11 times more likely to result in a SWI (OR = 11.337, 95% CI: 1.824‐70.452) than procedures by GS only. Finally, patients discharged with surgical drains were seven times more likely to develop SWI (OR = 7.387, 95% CI: 1.173‐46.521) compared with those without surgical drains.

TABLE 3.

Significant variables included after univariate analysis

Variable	P‐value
Postoperative seroma	.045
Home with surgical drains	.020
Operation type	.044
Previous chemotherapy	.198
Previous surgery	.150
Length of stay	.165
Surgeon department	.163

TABLE 4.

Predictors of surgical wound infection (multivariate logistic regression)

		Odds ratio (B)	95% CIa	P‐value	Cox & Snell R 2
Postoperative seroma	Yes	3.304	1.018‐10.723	.047*	0.140
	No^	1.000
Home with surgical drains	Yes	7.387	1.173‐46.521	.033*
	No^	1.000
Surgeon department	GSb^	1.000	1.824‐70.452	.009*
	GS/PSc	11.337

^ Reference group

* Significant at

p < 0.05

a Confidence interval

b General surgery

c Plastic surgery

4. DISCUSSION

This study aimed at identifying factors that significantly associate with, and predict the development of SWI among breast cancer patients. From the results of our study, many variables were not found to be associated with SWI development such as demographic and clinical characteristics of both cases and controls. We found that three main factors to predict SWI were wound drains, seroma, and surgical team. These factors were found to concur with other studies. ²⁵ , ²⁶ , ²⁷

Following breast surgery, surgical drains are inserted in both donor and transplant sites for a considerable period of time as fluid often builds up in the area where a breast is removed causing discomfort and delayed healing if not drained. Drain also reduces the chance of developing seroma. ²⁸ , ²⁹ Unfortunately, the use of drains puts patients at significant risk of developing SWI. ²⁶ This is in line with the results of a study that showed SWI was caused by the same bacteria found in the wound drainage fluid and tubing. ³⁰ Another study also identified that the predominant organism colonising the drain was from the skin flora. ¹⁰ , ¹¹ More specific to breast surgery, an analysis found that SWI occurred more frequently among subjects with greater bacterial colonisation in either drain fluid or drain tubing, as compared with those with less or no bacterial colonisation. ²⁵ , ²⁶

Wound drains act as a conduit for bacterial entry via extraluminal and intraluminal routes and antiseptic wound dressing must be adhered to. ³¹ , ³² This may not have been practiced when patients in our study are discharged home with drains, which was an independent variable found to be significantly associated with SWI development. Drain care is taught to our patients and their caregivers several times prior to their discharge; hence, it could only be assumed that their non‐compliance was because of confusion and a lack of understanding. With a median age of 55 years across case and control groups, it is common for the patients to not have English as their first language. Although patient education can also be provided in mother tongue, a language barrier would still surface and lead to miscommunication and a lack of understanding. In addition, patients as laymen, in the process of learning to care for themselves at home, may not fully comprehend the technicalities of performing certain steps.

In our study, the surgical teams were from GS and PS and we found “surgical team” to be a predictive factor for SWI development. A plausible explanation may be because of the combination of lead surgeons and the rotation of team members in the local context. With the varying team members, building rapport within the team would be a challenge and therefore contributing to the lack of familiarity. As such, teams of surgeons who performed fewer operations together were associated with a higher rate of SWI. ²⁶ , ³³

Our study found that postoperative seroma was significantly associated with the development of SWI. Our findings concurred with a trial where they found wound infection to be 1.8 times more common in patients with visible seroma. ³⁴ , ³⁵ Although the pathophysiology behind seroma formation remains idiopathic, studies have found it to be linked to the presence of dead space, disruption of the lymphatic system, and persistent exudate formation functions as a culture medium for bacterial growth and hence predictive for SWI development. ¹¹

5. LIMITATIONS

The first limitation was the missing and unavailable data, an inherent disadvantage of a retrospective case record review. This is because of the recent reinforcement of Personal Data Protection Act in Singapore. As this methodology relied on the completeness and accuracy of recorded information, the lack of it could have introduced biases that decreased the internal validity of the investigation. Another limitation is the small sample size in this study. We were only able to recruit 21 cases and 84 controls; hence, the results from this study yield a wide range for 95% CI in the multivariate logistic regression. This suggests the need for a larger sample size to minimise the production of false‐positive results, or an over‐estimation of the magnitude of an association. Matching, which accounts for confounders, can also result in the unwanted exclusion of cases when no matched control can be found. This reduced the already minimal sample size and omitting potential crucial information in the process.

6. CONCLUSION

Despite infection control measures in place, breast cancer surgical patients still developed SWI. This study provided evidence to the three factors (drain in‐situ upon discharge, seroma, and surgeon team) predictive of SWI in women with breast cancer after having had surgery. It is crucial that health care providers follow through with known measures to prevent seroma formation such as providing patient education on shoulder immobilisation immediately after surgery to decrease seroma formation. ³⁶ The use of appropriate teaching methods such as pictorial pamphlets for drain care can be used to teach patients and caregivers at home. Finally, it is paramount for surgical optimisation to be carried out with a fixed lead surgeon pairing and promotion of communication and rapport among the surgical team members. Most importantly, continuous monitoring of SWI among these women should be the mainstay of prevention and management of all health care professionals caring for patients undergoing surgery for all oncological diagnoses.

CONFLICT OF INTEREST

The authors declare no conflicts of interest.

Tan RZF, Yong B, Aloweni FAB, Lopez V. Factors associated with postsurgical wound infections among breast cancer patients: A retrospective case‐control record review. Int Wound J. 2020;17:1444–1452. 10.1111/iwj.13421