Cohort study to characterise surgical site infections after open surgery in the UK’s National Health Service

Abstract

Objective

To characterise surgical site infections (SSIs) after open surgery in the UK’s National Health Service.

Design

Retrospective cohort analysis of electronic records of patients from Clinical Practice Research Datalink, linked with Hospital Episode Statistics’ secondary care datasets.

Setting

Clinical practice in the community and secondary care.

Participants

Cohort of 50 000 adult patients who underwent open surgery between 2017 and 2022.

Outcome measures

Incidence of SSI, clinical outcomes, patterns of care and costs of wound management.

Results

11% (5281/50 000) of patients developed an SSI a mean of 18.4±14.7 days after their surgical procedure, of which 15% (806/5281) were inpatients and 85% (4475/5281) were in the community after hospital discharge. The incidence of SSI varied according to anatomical site of surgery. The incidence also varied according to a patient’s risk and whether they underwent an emergency procedure. SSI onset reduced the 6 months healing rate by a mean of 3 percentage points and increased time to wound healing by a mean of 15 days per wound. SSIs were predominantly managed in the community by practice and district nurses and 16% (850/5281) of all patients were readmitted into hospital. The total health service cost of surgical wound management following SSI onset was a mean of £3537 per wound ranging from £2542 for a low-risk patient who underwent an elective procedure to £4855 for a high-risk patient who underwent an emergency procedure.

Conclusions

This study provides important insights into several aspects of SSI management in clinical practice in the UK that have been difficult to ascertain from surveillance data. Surgeons are unlikely to be fully aware of the true incidence of SSI and how they are managed once patients are discharged from hospital. Current SSI surveillance services appear to be under-reporting the actual incidence.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• This is the most comprehensive study in a decade to characterise surgical site infections (SSIs), clinical outcomes, patterns of care and costs, spanning both primary and secondary care across the UK.

• This study was undertaken using real-world evidence derived from the anonymised records of a sample of 50 000 patients in the Clinical Practice Research Datalink database linked with secondary care datasets generated by Hospital Episode Statistics.

• The estimates were derived following a systematic analysis of patients’ characteristics, clinical outcomes and community-based and secondary care resource use pertaining to wound care and SSI management contained in the patients’ electronic records.

• The data set was analysed retrospectively, and no other data sources were available to verify the completeness and accuracy of some of the data.

• The analysis was based on clinicians’ entries into their patients’ records and inevitably subject to a certain amount of imprecision, lack of detail and recording bias.

Introduction

Surgical site infections (SSIs) are among the most common healthcare-associated infections. They occur in an incision created by an invasive surgical procedure.¹ These infections are a leading cause of morbidity and have been reported to result in increased hospital costs due, in part, to revision surgery.^{2 3} SSIs can be classified based on the tissues involved: superficial SSI which is limited to skin and subcutaneous tissue; deep SSI which affects the fascial and muscular layers; and cavity space infection which occurs within an abdominal or joint cavity.

There is considerable uncertainty surrounding the incidence of SSI and their impact on wound healing, levels of healthcare resource use and health service costs of patient management. The 2017 National Health Service (NHS) Getting It Right First Time Report in General Surgery highlighted that only 8% (4 out of 50) hospitals knew their incidence of SSI.⁴ Among 134 547 surgical procedures submitted to the Public Health England Surgical Site Infection Surveillance Service in 2019/2020, a total of 1197 SSIs were detected during either inpatient stay or on readmission,⁵ indicating an overall incidence of 1%. However, the incidence of SSI varied across surgical categories from <1.0% for hip and knee replacement to 9.1% for bile duct, liver or pancreatic surgery.⁵ In 2021/2022, a total of 968 SSIs were detected during either inpatient stay or on readmission among the 115 796 surgical procedures that were submitted to the Surgical Site Infection Surveillance Service,⁶ indicating a static overall incidence of 1%. However, the incidence varied between 0.4% and 15.4% across surgical categories.⁶

The 30-day postoperative morbidity rates among those undergoing emergency surgery has been reported to be in the range of 33%–71%, depending on a patient’s individual risk and type of surgery.^7–9 SSIs have been reported to contribute to a significant proportion of the postoperative morbidity experienced by patients in the emergency setting, with estimated rates in the range of 25%–40%.^{10 11}

Evidence-based guidelines for the prevention of SSI and many studies have successfully reduced rates of wound sepsis by a combination of effective intervention and feedback of data surrounding infection rates to surgical teams.^12–16 Ten-year trends in the annual inpatient and readmission SSI risk showed that the majority of surgical categories saw an overall declining trend in risk.¹ Notwithstanding this, the majority of the published evidence has focused on the development of SSIs in the hospital setting and not considered the onset of an SSI after a postoperative patient has been discharged from hospital.

Clearly, there is considerable uncertainty surrounding the incidence of SSI and their impact on wound healing, levels of healthcare resource use and costs of surgical wound management. While SSI prevention is a high-priority goal for healthcare organisations, the current uncertainty only adds to the pressures faced by these organisations in allocating resources to meet the needs of their catchment populations. Accordingly, this study aimed to reduce some of this uncertainty by providing more robust estimates of the incidence of SSI, stratified by patient risk and surgical category.

Methods

Study design

This was a retrospective cohort analysis of the anonymised case records of patients with an SSI randomly extracted from the Clinical Practice Research Datalink (CPRD) database linked with admitted patient care data and hospital outpatient data and accident and emergency data generated by Hospital Episode Statistics (HES). The perspective of the analysis was the UK’s health services.

Clinical Practice Research Datalink

CPRD is a real-world evidence research service supporting retrospective and prospective public health and clinical studies. It is provided by the UK’s Medicines and Healthcare Products Regulatory Agency (MHRA) with support from the National Institute for Health and Care Research, as part of the Department of Health and Social Care.¹⁷ CPRD contains the electronic records of 60 million anonymised patients including 18 million currently registered patients, provided by a network of general practices from across the UK.¹⁸ CPRD comprises two databases: CPRD GOLD and CPRD Aurum. CPRD GOLD contains data contributed by general practices using Vision software,¹⁹ whereas CPRD Aurum contains data contributed by general practices using the EMIS Web electronic patient record system software.²⁰ The electronic records include patients’ demographic characteristics, details of clinician visits including all medical conditions and symptoms as well as prescribing history. Patients’ primary care records within CPRD can be individually linked to secondary care and other health and area-based datasets such as HES hospital admissions, HES hospital outpatient visits and HES accident and emergency visits.²¹ This linkage enables CPRD to provide a fuller picture of a patient’s healthcare.

Study population

The study population comprised the anonymised case records of 50 000 patients from the CPRD databases with linked data sets from HES who underwent open surgery. Patients were included in the data set if:

• They had one of 493 Read codes in CPRD GOLD or 1 of 1041 Snomed codes in CPRD Aurum for having undergone an open surgical procedure after 1 January 2017.

• Were 18 years of age or older at the time of their surgical procedure.

• Had a linked record to HES Admitted Patient Care data, HES Outpatient data and HES Accident and Emergency data.

Patients’ complete electronic records were supplied to the authors by the MHRA, which enabled analysis of data both within and outside the study period.

Sample size

A feasibility count undertaken by the MHRA indicated that between 1% and 3% of patients who had undergone an open surgical procedure developed an SSI. Depending on the ratio of records from CPRD GOLD and CPRD Aurum, it was estimated that in a sample of 50 000 patients who underwent open surgery, ~840 patients would have developed an SSI following their procedure. Power calculations indicated that if 50% of patients who developed an SSI were high-risk (n=420) and 50% were low-risk (n=420), a sample of 50 000 patients would be sufficiently large to detect a 10 percentage point difference in healing rate or SSI amelioration rate between the groups with 90% power and a type I (alpha) error of 0.05. If 70% of patients who developed an SSI were high risk (n=588) and 30% were low risk (n=252), a sample of 50 000 patients would be sufficiently large to detect a 10 percentage point difference in healing rate or SSI amelioration rate between the groups with 80% power and a type I (alpha) error of 0.05.

Study variables

On receipt of 50 000 CPRD and HES linked data sets, the following information was systematically extracted from the patients’ records from the time of their surgical procedure.

• Demographic characteristics (eg, age, gender, smoking status, body mass index).

• Comorbidities.

• Anatomical site of surgery with dates and whether it was an elective or emergency procedure.

• Clinical outcomes pertaining to wound healing, SSI and mortality with dates.

• Healthcare resource use with dates from the records of patients who developed an SSI (eg, length of hospital stay, hospital readmission, hospital outpatient visits, postdischarge community-based clinician visits, postdischarge primary care-based clinician visits, postdischarge attendance at accident and emergency units, prescribed medications and prescribed wound care products).

Patients were considered to be at high-risk of developing an SSI if they were smokers,^{22 23} or had a body mass index >35 kg/m^{24 25} or a Charlson Comorbidity Index (CCI) score^26–28 equal to or greater than 3.² All other patients were considered to be low-risk.

Data and statistical analyses

The study estimated:

• Surgical wound healing rates and time to wound healing.

• Incidence of SSI and time to infection onset following surgery.

• Location of SSI onset (ie, during initial admission or in the community postdischarge).

• Consequences of an SSI on wound healing.

• Mortality.

Patients were stratified according to their anatomical site of surgery as well as their risk and whether they underwent an elective or emergency procedure.

Kaplan-Meier analyses were undertaken to compare the distribution of surgical wound healing among different subgroups and the distribution of SSI onset. Differences between multiple groups were tested for statistical significance using a Kruskal-Wallis test or χ² test. Binary logistic regression was undertaken to identify relationships between baseline variables and clinical outcomes and logistic regression was used to investigate relationships between baseline variables and the cost of surgical wound management. The p values <0.05 were considered statistically significant and have been reported. All p values ≥0.05 were not considered statistically significant and these numerical values have not been reported. All statistical analyses were performed using IBM SPSS V.23 Statistics (IBM).

The use of individual healthcare resources was quantified for all the patients who developed an SSI. These quantities were then used to estimate the mean utilisation of each resource attributable to surgical wound management following SSI onset, stratified by anatomical site of surgery, a patient’s risk and whether the procedure was elective or emergency.

Cost of surgical wound management following SSI onset

The health service cost of surgical wound management for each patient was estimated by assigning unit costs at 2020/2021 prices^29–31 to the quantity of healthcare resources used by individual patients. The total cost of utilisation of each healthcare resource was then combined in order to estimate the mean total direct cost of wound management following SSI onset, from the perspective of the UK’s health services. These costs were stratified by a patient’s anatomical site of surgery, their risk and whether they underwent an elective or emergency procedure. Accordingly, the study does not consider the cost of surgical wound management among patients who did not develop an SSI.

Patient and public involvement

Patients and members of the public were not directly involved in this study. The study population was limited to the anonymised records of patients in CPRD GOLD and CPRD Aurum.

Results

Patients’ baseline characteristics

The study population comprised 50 000 patients who had undergone open surgery, of whom 20 000 were obtained from CPRD GOLD and 30 000 from CPRD Aurum. There was some variance in the mean age of patients between those undergoing different types of surgery, with those who underwent an obstetric/gynaecological procedure being younger (mean 45.9±15.3 years per patient) and those who underwent a vascular procedure being older (mean 71.4±11.3 years per patient) (table 1 and online supplemental table 1). Additionally, the mean age of patients classified as low-risk was lower than that of high-risk patients (table 1). The ratio of males to females varied according to surgical category (table 1 and online supplemental table 1). The percentage of high-risk patients and percentage of emergency procedures also varied according to surgical category. There were more low-risk patients and more high-risk patients among those who underwent an obstetric/gynaecological procedure (55%) and vascular procedure (85%), respectively (table 1 and online supplemental table 1). There were more elective procedures and more emergency procedures among patients who underwent breast surgery (98%) and orthopaedic surgery (26%), respectively (table 1 and online supplemental table 1).There were no other differences in patients baseline characteristics when stratified by anatomical site of surgery.

Table 1.

Baseline characteristics

	All	Low-risk, elective	Low-risk, emergency	High-risk, elective	High-risk, emergency
No of patients	50 000	13 000 (26%)	4500 (9%)	26 000 (52%)	6500 (13%)
Mean age per patient (years)	59.8±17.7	50.4±13.6	40.0±14.3	67.0±13.7	62.8±21.9
Percentage male	42	39	41	44	43
Mean body mass index per patient (kg/m2)	27.9±5.4	26.9±4.0	25.9±4.0	28.9±5.8	27.0±6.1
Percentage smokers	14	0	0	19	34
Percentage ex-smokers	29	29	25	30	23
Percentage non-smokers	57	71	75	51	43
Percentage high-risk patients	65	0	0	100	100
Percentage emergency procedures	22	0	100	0	100
Mean age-adjusted Charlson Comorbidity Index score	3.9±3.0	1.7±1.2	0.9±1.1	5.3±2.8	4.4±3.2
Percentage who had abdominal surgery	12	11	4	14	8
Percentage who had breast surgery	1	1	<1	1	<1
Percentage who had cardiothoracic surgery	9	8	6	9	9
Percentage who had obstetric and gynaecological surgery	23	37	36	17	10
Percentage who had orthopaedic surgery	51	41	53	53	70
Percentage who had vascular surgery	3	1	1	4	4

Patients’ comorbidities are summarised in online supplemental table 2. It is noticeable that patients’ mean CCI score was lower among those who underwent an obstetric/gynaecological procedure (mean 2.3±2.9 per patient) and higher among those who underwent an abdominal (mean 5.0±3.3 per patient) or vascular procedure (mean 5.2±2.5 per patient) (online supplemental table 1).

Surgical wound healing, SSI and mortality

Wound healing and SSI were clinical observations documented in the patient’s record by their managing clinician, but not necessarily confirmed by a specialist, and it is unknown if the clinicians who managed these patients used any consistent definition. Furthermore, if a wound was not recorded as being healed it was considered to be unhealed. This assumption was supported by continued healthcare professional visits for wound care and the continued prescribing of wound care products and antibiotics. On this basis, over 90% of all the surgical wounds had healed by 6 months following the surgical procedure. The time to wound healing was a mean of 50.9±52.7 days per wound (table 2). The shortest time to healing was observed among patients who had undergone an obstetric/gynaecological procedure (mean 39.6±38.1 days per wound) and longest among those who underwent a cardiothoracic procedure (mean 76.0±72.4 days per wound) (online supplemental table 3).

Table 2.

Clinical outcomes

	All	Low-risk, elective	Low-risk, emergency	High-risk, elective	High-risk, emergency	P value
Percentage of all wounds that healed by 6 months	98	98	98	97	97	ns
Mean time to heal for all patients (days)	50.9±52.7	43.7±44.5	49.4±47.5	53.2±55.1	57.1±59.1	0.001
Percentage who developed an SSI	11	10	11	11	10	ns
Mean time to SSI onset (days)	18.4±14.7	17.7±13.8	17.3±13.6	18.2±14.4	21.4±17.7	ns
Percentage of wounds that healed by 6 months among those who did not develop an SSI	98	99	98	97	97	ns
Percentage of wounds that healed by 6 months among those who developed an SSI	95	97	96	95	92	ns
Mean time to heal among those who did not develop an SSI (days)	49.4±51.1	43.0±43.6	49.4±46.7	51.2±53.4	54.9±56.8	0.001
Mean time to heal among those who developed an SSI (days)	64.3±63.3	48.5±51.8	50.6±65.4	69.8±73.5	77.1±63.3	0.001
Estimated increase in mean time to heal potentially due to an SSI (days)	14.9	5.5	1.2	18.6	22.2
30 day mortality rate	1	<1	<1	1	3	0.01

ns, not significant; SSI, surgical site infection.

An estimated 11% of all patients developed an SSI a mean of 18.4±14.7 days after the surgical procedure (table 2). Of these, 15% of patients developed an SSI while an inpatient a mean of 3.9±7.1 days after their surgical procedure and 85% of patients developed an SSI after discharge from hospital a mean of 20.9±14.4 days after their surgical procedure. The incidence of SSI onset varied between different surgical procedures, ranging from 7% of patients who underwent an orthopaedic procedure to 17% of those who underwent a vascular procedure (online supplemental table 3). Moreover, the percentage of SSIs developed in the community varied according to anatomical site of surgery, ranging from 67% among those who underwent vascular surgery to 90% among patients who underwent an obstetric/gynaecological procedure.

Kaplan-Meier analyses showed a significant difference in distribution of SSI onset when stratified by anatomical site of surgery (figure 1; log rank (Mantel-Cox) p<0.001). The incidence of SSI onset also varied between inpatients and postoperative patients who had been discharged from hospital, when stratified by anatomical site of surgery (figure 2). However, a patient’s risk and whether the surgical procedure was elective or emergency did not affect the SSI incidence (table 2, online supplemental table 3 and figure 2).

Figure 1.

Kaplan-Meier analysis of SSI onset stratified by anatomical site of surgery (log rank (Mantel-Cox) p<0.001). SSI, surgical site infection.

Figure 2.

Incidence of SSI stratified by location of onset. SSI, surgical site infection.

The overall incidence of SSI remained relatively stable between 2017 and 2022, although this varied according to anatomical site of surgery. There was a statistically significant decreasing trend in the incidence of SSI among those who underwent cardiothoracic and breast surgery and a statistically significant increasing trend in the incidence of SSI among those who underwent obstetric/gynaecological surgery (online supplemental figure 1). Additionally, between 2017 and 2022, the incidence of SSI increased among those who underwent an emergency procedure but it was stable among those who underwent an elective procedure (online supplemental figure 2).

The percentage of surgical wounds that healed by 6 months was reduced by a mean of 3 percentage points among those who developed an SSI, but this varied according to anatomical site of surgery. Additionally, the onset of an SSI increased the time to wound healing by a mean of 15 days per wound, ranging from no increase among those who underwent breast surgery to an increase of a mean of 29 days among those who underwent vascular surgery (table 2 and online supplemental table 3). Time to wound healing also increased to a much greater extent among high-risk patients (mean of >19 days per wound) than among low-risk patients (mean of >1 day per wound) (table 2). Kaplan-Meier analysis indicated that the healing distributions between all the groups were significantly different (figure 3; log rank (Mantel-Cox) p<0.001). The graph shows that the healing curves for the groups in which patients developed an SSI have been shifted to the right compared with the other groups.

Figure 3.

Kaplan-Meier analysis of wound healing stratified by surgery type and presence of an SSI (log rank (Mantel-Cox) p<0.001). SSI, surgical site infection.

The overall rate of wound dehiscence was 3% and the 30-day mortality rate was a mean of 1%, but this varied according to anatomical site of surgery (table 2 and online supplemental table 3).

Logistic regression identified the risk factors in this study’s cohort which impacted on wound healing (online supplemental table 4). These included the presence of an SSI as well as several comorbidities (online supplemental table 4). Logistic regression also identified the risk factors in this study’s cohort which potentially increased the probability of a patient developing an SSI (online supplemental table 5). These included anatomical site of surgery, several comorbidities, body mass index, high-risk and emergency surgery. Interestingly, patients with an immunological comorbidity had a reduced Odds of developing an SSI (online supplemental table 5). (Only 1% of patients in the cohort had an immunological comorbidity of which the majority were a range of different autoimmune disorders and immune deficiencies. It could be argued that there was an enhanced focus on SSI prevention in this small cohort which may have resulted in better outcomes than would be expected, but that would only be a hypothesis). These regression analyses indicated that the relative risk of developing an SSI is impacted on by surgical category, comorbidities and baseline characteristics which combine to generate a tailored risk for each patient individually. Table 3 indicates which patients are most likely to develop an SSI, according to surgical category and whether they are of high-risk or low-risk and whether they undergo an elective or emergency procedure.

Table 3.

Percentage of patients who developed an SSI stratified by surgical category and risk

	Low-risk, elective	Low-risk, emergency	High-risk, elective	High-risk, emergency	Most likely to develop an SSI
Abdominal	9%	17%	12%	16%	Emergency procedure
Breast	10%	33%	10%	33%	Emergency procedure
Cardiothoracic	11%	14%	13%	18%	Emergency procedure
Obstetrics and gynaecology	16%	19%	14%	22%	Emergency procedure
Orthopaedic	5%	4%	9%	6%	High-risk patient
Vascular	11%	13%	18%	19%	High-risk patient

SSI, surgical site infection.

Length of hospital admission

The length of hospital admission among those who did not develop an SSI was a mean of 6.2±7.8 days per patient. This increased by between 0 and 2 days (depending on anatomical site of surgery) among those who developed an SSI (table 4 and online supplemental table 6), and the time between hospital discharge and SSI onset was a mean of 11.5±16.5 days. However, the length of hospital admission among the 15% of patients who developed an SSI while an inpatient was a mean of 14.2±16.9 days, compared with a mean of 5.6±4.5 days among those who developed an SSI in the community postdischarge. The longer duration of hospital stay could have potentially increased the probability of a patient developing an SSI. Conversely, a patient diagnosed with an SSI may have stayed longer, particularly if it was a deep space infection, especially if they required radiological or surgical source control.

Table 4.

Hospital admission metrics

	All	Low-risk, elective	Low-risk, emergency	High-risk, elective	High-risk, emergency	P value
Mean length of hospital admission for all patients (days)	6.3±7.9	4.6±5.1	5.5±6.9	6.3±7.7	10.3±11.6	0.001
Mean length of hospital admission for patients who did not develop an SSI (days)	6.2±7.8	4.6±5.1	5.4±6.8	6.3±7.7	10.2±11.5	0.001
Mean length of hospital admission for patients who developed an SSI (days)	7.0±8.4	4.9±5.5	6.5±7.5	7.0±8.1	11.2±12.2	0.001
Estimated additional length of hospital admission potentially due to an SSI (days)	1.0	0	1.0	1.0	1.0	ns
Mean time between hospital discharge and SSI onset (days)	11.5±16.5	12.8±14.9	10.9±16.1	11.3±16.3	10.2±20.3	ns
Percentage who developed an SSI as an inpatient	15	13	17	14	23	0.01
Percentage who developed an SSI in the community after discharge	85	87	83	86	77	0.01

Healthcare resources associated with surgical wound management following SSI onset

All patients who developed an SSI were predominantly managed in the community by practice nurses and district nurses (table 5 and online supplemental table 7). A total of 16% of all the patients who developed an SSI were readmitted into hospital, ranging from 11% of those who underwent breast surgery to 24% of those who underwent abdominal surgery. This rate was unaffected by location of SSI onset (ie, as an inpatient or in the community postdischarge) (table 5 and online supplemental table 7). A total of 23% of all patients attended an accident and emergency unit for their SSI and 72% of patients visited their general practitioner (GP). However, the actual percentages and number of visits varied according to surgical category (online supplemental table 7) and risk (table 5).

Table 5.

Healthcare resource use associated with surgical wound management following an SSI, stratified by risk

	All		Low-risk, elective		Low-risk, emergency		High-risk, elective		High-risk, emergency		SSI onset after discharge		SSI onset as inpatient
	%	N	%	N	%	N	%	N	%	N	%	N	%	N
Accident and emergency visits	23	1.5	16	1.4	20	1.4	23	1.5	42	1.6	23	1.5	24	1.6
District nurse visits	95	17.7	95	15.0	96	15.3	94	18.6	94	21.1	95	17.5	95	18.9
GP visits	72	3.1	78	2.6	76	2.5	71	3.4	64	3.5	73	3.1	70	3.6
Hospital outpatient visits	64	2.9	51	2.5	44	3.1	72	2.9	67	3.4	64	2.8	65	3.2
Practice nurse visits	82	5.5	86	4.5	84	4.6	81	5.9	74	6.5	82	5.4	78	5.9
Readmissions into hospital among all patients	16	1.6	10	1.6	13	1.4	17	1.7	26	1.5	16	1.6	18	1.7
Readmissions into hospital among patients whose SSI developed while an inpatient	18	1.7	12	2.0	14	1.5	18	1.8	28	1.5
Readmissions into hospital among patients whose SSI developed following hospital discharge	16	1.6	10	1.5	13	1.4	17	1.7	25	1.6
Prescriptions for analgesics	42	3.3	32	2.7	22	2.6	49	3.5	46	3.4	42	3.3	41	3.6
Prescriptions for antibiotics	76	1.6	75	1.2	68	1.3	71	1.3	59	1.3	71	1.3	67	1.4

%, percentage of patients who utilised a resource.

GP, general practitioner; N, Mean amount of resource use per patient who used the resource; SSI, surgical site infection.

Health service cost of surgical wound management following SSI onset

The total health service cost of wound management following SSI onset was estimated to be a mean of £3537 per wound (table 6). However, this varied nearly twofold between low-risk patients who underwent an elective procedure and high-risk patients who underwent an emergency procedure, due in part to the increased cost of hospital readmission and district nurse visits (table 6). The principal cost driver was hospital readmissions, which accounted for up to 47% of the total cost of wound management. This was followed by district nurse visits, which accounted for up to a further 31% of the total cost (table 6 and online supplemental table 8).

Table 6.

Mean cost of healthcare resource use associated with surgical wound management per patient following an SSI, stratified by risk

	All		Low-risk, elective		Low-risk, emergency		High-risk, elective		High-risk, emergency		SSI onset after discharge		SSI onset as inpatient
Accident and emergency attendances	£61.59	(2%)	£38.86	(2%)	£48.29	(2%)	£61.15	(2%)	£117.31	(2%)	£60.62	(2%)	£66.96	(2%)
Diagnostic tests and procedures	£149.36	(4%)	£98.20	(4%)	£117.72	(4%)	£154.11	(4%)	£250.47	(5%)	£138.00	(4%)	£215.00	(5%)
District nurse visits	£931.04	(26%)	£790.50	(31%)	£814.85	(29%)	£973.43	(26%)	£1099.34	(23%)	£920.04	(27%)	£992.12	(25%)
GP visits	£177.10	(5%)	£160.99	(6%)	£145.62	(5%)	£189.80	(5%)	£174.32	(4%)	£173.86	(5%)	£195.04	(5%)
Hospital outpatient visits	£411.11	(12%)	£286.45	(11%)	£296.51	(11%)	£457.54	(12%)	£529.30	(11%)	£399.18	(12%)	£477.32	(12%)
Hospital readmissions	£1462.00	(41%)	£891.00	(35%)	£1060.46	(38%)	£1595.00	(42%)	£2263.77	(47%)	£1430.66	(41%)	£1635.00	(41%)
Practice nurse visits	£98.48	(3%)	£86.30	(3%)	£84.68	(3%)	£104.42	(3%)	£105.65	(2%)	£97.85	(3%)	£102.00	(3%)
Prescribed drugs	£14.19	(0%)	£10.53	(0%)	£8.77	(0%)	£16.39	(0%)	£15.39	(0%)	£14.00	(0%)	£15.00	(0%)
Wound care products	£232.26	(7%)	£179.23	(7%)	£211.69	(8%)	£243.77	(6%)	£299.30	(6%)	£228.06	(7%)	£255.54	(6%)
Total	£3537.12	(100%)	£2542.06	(100%)	£2788.59	(100%)	£3795.63	(100%)	£4854.85	(100%)	£3462.28	(100%)	£3953.99	(100%)

GP, general practitioner; SSI, surgical site infection.

The cost of wound management following SSI onset also varied according to surgical category, ranging from £2335 to £4546 per wound among those who underwent breast surgery and cardiothoracic surgery, respectively (online supplemental table 8).

The cost of wound management among patients whose infection started while they were an inpatient (£3954 per wound) was 14% more than that of managing a surgical wound if the SSI started in the community after a postoperative patient had been discharged from hospital (£3463 per wound) (table 6). The difference was principally due to marginal increases in the use of all resources (table 5). If the cost of hospital stay between SSI onset and discharge into the community is included, the cost of wound management among patients whose infection started while they were an inpatient would be increased to £6901 per wound. This would have the effect of increasing the total health service cost of surgical wound management following SSI onset by 13% (from a mean of £3537 per wound to £3988 per wound) for the whole cohort.

Linear regression indicated which covariates impacted on the cost of surgical wound management (online supplemental table 9). The analysis estimated that the cost of wound management was greater among patients with liver disease or haematological disease and that a patient’s age had minimal impact on costs. The analysis also indicated that length of hospital stay did not significantly impact on the cost of wound management following an SSI.

Discussion

This study aimed to improve understanding of the characteristics of SSIs in the UK by retrospectively analysing the anonymised electronic records of a random sample of 50 000 adult patients who had undergone open surgery, obtained from the CPRD database and HES linked secondary care data sets. To the authors’ knowledge, this is the most comprehensive study in 10 years to estimate the incidence of SSI, clinical outcomes, patterns of care and costs of surgical wound management following an SSI, spanning primary and secondary care across the UK. Patients were stratified according to their anatomical site of surgery, their risk and whether they underwent an elective or emergency procedure.

Inevitably, there were differences in patients’ baseline characteristics such as age, gender and comorbidities reflecting the different types of surgical procedures that were selected for this study. The analysis found that >90% of all surgical wounds healed within 6 months of the surgical procedure and that the time to healing was a mean of 51 days per wound. The overall incidence of SSI in this study’s cohort was 11%. However, the incidence varied to a greater extent between different surgical categories and to a lesser extent by a patient’s risk and whether the procedure was elective or emergency. Furthermore, 85% of all SSIs in this study’s cohort were developed in the community after postoperative patients had been discharged from hospital, reflecting that the time to SSI onset was a mean of 18 days per wound compared with a length of hospital stay of a mean of 7 days per initial admission. The presence of an SSI reduced the 6-month wound healing rate by 3 percentage points and increased the time to wound healing by a mean of 15 days per wound, although this varied according to anatomical site of surgery. Moreover, the time to healing increased to a much greater extent among high-risk patients (mean of >19 days per wound) than among low-risk patients (mean of >1 day per wound).

The incidence of SSI among inpatients was estimated at 1.6%. The surveillance survey of SSIs in NHS hospitals in England⁶ estimated a marginally smaller incidence between 2021 and 2022, with the exception of vascular surgery which was reported as having an incidence of 2.3%⁶ compared with our estimate of 5.6%, and abdominal surgery which was reported as having an incidence of 7.8%⁶ compared with our estimate of 2.5%. In an earlier survey in 2017 across 95 hospital Trusts, Wong et al reported a similar incidence of SSI across different specialities among inpatients to those reported in our study, with one exception³²; the authors found the incidence of SSI among women who underwent breast surgery to be a mean of 9.2%³² compared with our estimate of 1.3%. While there is inevitably some variation between studies due to different methodological designs, the major limitation of these two surveys^{6 32} is that they did not report the incidence of SSI among postoperative patients in the community after hospital discharge.

In recent years, several evidence-based guidelines for the prevention of SSIs have been published.^12–16 These guidelines focused on processes to reduce the relative risk of infection both during surgery and while the patient remains in hospital. However, the extent to which they impact on the relative risk of developing an SSI among postoperative patients after they have been discharged from hospital is unknown. Our study indicated that the incidence of SSI is around sixfold higher among postoperative patients in the community than among inpatients.

An earlier study on the health economic impact of SSIs in England suggested that the development of SSI among inpatients could have resulted in a financial loss had it not been mitigated by the hospital continuing to receive income for episodes complicated by SSI, due to an increased length of hospital stay of a median of 10 days.³ The authors also found that length of stay of the initial hospital admission was an important determinant of the overall cost of SSI management over a 2-year period between 2010 and 2012.³ This is no longer the case since patients in this study’s cohort who developed an SSI had an increased length of hospital admission of around 1 day, from a mean of 6.2 days to a mean of 7.0 days. This observation reflects changes in patient management and impacts on the determinant of cost. In our study, patients who developed an SSI were predominantly managed in the community by practice nurses and district nurses and only 16% of all these patients were readmitted into hospital.

SSIs are among the most common healthcare-associated infections and their prevention remains a global challenge. Our estimated incidence of SSI among inpatients (1.6%) is comparable to that reported in France (1.6%).³³ A higher rate of 4.9% has been reported in Germany.³⁴ The mean incidence of in-hospital SSI across Europe between 2017 and 2020 was reported to be 1.6%,^{35 36} although the incidence ranged between 0.6% and 9.5%, depending on the type of surgical procedure.³⁶ In comparison, the mean incidence of SSI in our study remained relatively static between 2017 and 2022, although this varied according to surgical category. A decreasing trend in the incidence of SSI was observed among those who underwent cardiothoracic and breast surgery and an increasing trend among those who underwent obstetric/gynaecological surgery from 2020. Additionally, there was an increasing trend from 2020 among those who underwent an emergency procedure, although it was stable among those who underwent an elective procedure. The COVID-19 pandemic in the UK started in 2020³⁶ and this may partly explain the increasing trend in SSI incidence among emergency procedures. A recent publication reported a similar observation among those who underwent emergency abdominal surgery in Uganda.³⁷ The evidence for the effect of the COVID-19 pandemic in the elective setting points towards a reduction in SSI incidence (possibly as a result of increased use of personal protective equipment, etc).^{38 39}

The total health service cost of wound management after the onset of an SSI was estimated to be a mean of £3537 per wound, and hospital readmissions and district nurse visits were the principal determinants of cost and not the increased length of the initial hospital admission. Other determinants of cost were patients’ comorbidity profile and whether the procedure was elective or emergency. The cost of wound management following SSI onset varied nearly twofold between low-risk patients who underwent an elective procedure and high-risk patients who underwent an emergency procedure, due in part to the increased cost of hospital readmission and district nurse visits. The cost of wound management among patients whose SSI started while they were an inpatient was 14% more than that of managing a wound if the SSI started in the community postdischarge. Inclusion of the cost of hospital stay between SSI onset and discharge into the community has the effect of increasing the cost of surgical wound management by 13% for the cohort as a whole. However, causality for a longer hospital stay being a result of SSI onset or due to some other medical or social reason cannot be determined in this retrospective analysis.

Other studies have also estimated the cost of surgical wound management. We previously reported that the cost of managing surgical wounds following initial presentation in the community was a mean £7345 per wound at 2015/2016 prices.⁴⁰ However, this cost varied between a mean £2000 per wound for those without any evidence of infection and a mean £5000–£11 200 per wound with a putative infection.⁴⁰ Another study found that SSIs resulted in a mean additional cost of €5450 (£4770) per wound in Germany in 2018.⁴¹ A more recent study estimated the cost of wound management associated with vascular surgery following SSI onset in one English hospital to be a mean of £3776 among inpatients at 2018/2019 prices,⁴² which is less than our own estimate of a mean of £5000 among inpatients following vascular surgery. The inevitable variance in costs reflects differences in study design and populations.

The time to wound healing is also an important factor in driving costs. The longer a wound takes to heal, the more resources are required to manage the patient. This observation is consistent with the finding from our Burden of Wounds study.⁴³ Accordingly, the cost of wound management following SSI onset can be affected by a combination of resources required to manage the infection including readmission into hospital, clinician visits for dressing changes and complexity of some treatment regimens. Less than 1% of all patients in our data set were recorded as having seen a tissue viability nurse. Furthermore, there was minimal evidence of a coordinated shared treatment plan in the patients’ records. Only 40% of all patients with an SSI received prophylactic closed incision topical negative pressure therapy. Given the nature of these wounds, this represents a surprising underutilisation of this technology in this cohort of patients. This may have resulted from either a lack of product availability, item cost considerations, skill mix and/or a failure to follow escalation pathways involving senior staff. All healthcare systems recognise the importance of managing postoperative patients in the community. Clearly, improving wound management practices may reduce the incidence of SSI in the community and could shorten the time to wound healing, leading to better outcomes for patients and cost-effective solutions for the NHS.

A review of the published literature found that hospital readmission rates and cost of wound management following SSI onset varied widely between countries. These differences reflect different surgical practices, different surveillance methods, different patient pathways and different unit resource costs, all of which makes it very difficult to make relevant comparisons across countries. Notwithstanding, surveillance is a key component in the prevention of SSIs and an important tool for monitoring the effectiveness of prevention and control measures.⁴⁴

The advantages of using the CPRD and linked HES secondary care data sets is that the patient pathways and associated resource use are based on real-world evidence reflecting the complete patient journey in clinical practice. However, the analyses were based on clinicians’ entries into their patients’ records and inevitably subject to a certain amount of imprecision and lack of detail, reporting bias and recall bias. Moreover, the computerised information in the database is collected by GPs and nursing teams for clinical care purposes and not for health economics research. Prescriptions issued by the general practice team were documented in the database, but it did not specify whether the prescriptions were dispensed or detail patient adherence with the product. Also, the CPRD may have under-recorded the use of some healthcare resources in the community since not all community records may have been linked to the general practice records.

Analysis of resource use and costs associated with managing patients who did not develop an SSI have not been estimated since they were beyond the remit of this study. Consequently, there remains some uncertainty around incremental levels of resource use and costs attributable to wound management following SSI onset. Moreover, the possibility of resource use associated with managing a patient’s underlying disease or a comorbidity being conflated with that of wound management cannot be excluded. Despite these limitations, it is the authors’ opinion that the real-world evidence contained in CPRD has provided a useful perspective on the management of SSIs across all healthcare sectors in the UK.

Three risks factors (smoking, body mass index and CCI score) have consistently been shown in the published literature to be important risk factors for SSI development.^22–28 Any one of these three were used to signify a high-risk patient in this study. It was not possible to use a broad risk scoring system such as the SSI risk score,⁴⁵ since the patients’ records did not document the required variables (eg, detailed operative variables). The analysis only considered NHS resource use and associated costs for the ‘average patient’ and was not stratified according to gender, comorbidities, disease-related factors and level of clinicians’ or surgeons’ skills. Furthermore, it was not possible to stratify the analyses according to superficial incisional SSIs, deep incisional SSIs and organ/space SSIs since this level of granularity was not contained in most of the patients’ records. For the same reason, it was not possible to classify the surgical wounds as clean, clean-contaminated, contaminated or dirty.⁴⁶ Costs incurred by non-NHS organisations (such as the provision of social care), patients’ costs and indirect societal costs as a result of patients being absent from work were also excluded from the analysis. The remit of the analysis was limited to adult patients. A separate study should be undertaken to characterise SSI following open surgery in paediatric patients.

CPRD GOLD contains data contributed by general practices using Vision software and CPRD Aurum contains data from practices using EMIS software. However, there are some differences in structure and clinical coding between these two systems and there are multiple codes for the same symptoms and procedures. Hence, the accuracy of this analysis is impacted by the way data has been coded by CPRD and NHS Digital.

Conclusion

Real-world evidence in this study provides important insights into a number of aspects of SSI management in clinical practice in the UK that have been difficult to ascertain from previous studies and surveillance data. Surgeons are unlikely to be fully aware of the true incidence of SSI, the frequency of occurrence later in the patient pathway, and how they are managed in the community once patients are discharged from hospital. The evidence suggests that current SSI surveillance is under-reporting the actual incidence, particularly among postoperative patients in the community postdischarge.

Data availability statement

All data relevant to the study are included in the article or uploaded as online supplemental information. The data sets cannot be shared as this restriction was a condition of the ethics approval. Questions concerning the data underlying the results can be sent to the corresponding author.

Ethics statements

Patient consent for publication

Not applicable.

Ethics approval

The Clinical Practice Research Datalink (CPRD) has ethical approval to collect, share and use anonymised patient data for observational research at the organisation level by the East Midlands-Derby Research Ethics Committee under approval 21/EM/0265, and all research applications considered and approved by CPRD are covered by this approval. This study’s protocol (number 21_001641) was approved by CPRD’s Research Data Governance (RDG) Process. This involved independent scientific and patient advice provided by Expert Review Committees and the Central Advisory Committee. This study is based in part on data from the CPRD obtained under licence from the UK Medicines and Healthcare products Regulatory Agency. The data are provided by patients and collected by the NHS as part of their care and support. The interpretation and conclusions contained in this study are those of the author/s alone. The study was exempted from obtaining informed consent because the database contains anonymised information on >60 million patients who would have provided informed consent to allow their anonymised data to be uploaded to the database for subsequent research studies.