Medical complications following shoulder arthroplasty: a review of the evidence

Abstract

Introduction

Shoulder arthroplasty (SA) procedures are increasing in number worldwide. Medical complications (MC) following SA have not been widely documented in the literature. Knowledge of these complications is needed to obtain informed consent.

Methods

The aim of this review article was to summarise the literature regarding systemic complications and mortality after SA to aid the consenting process.

Findings

We found that the literature regarding MC after SA was variable in quality. The rate of MC was 12% and the rate of mortality was 0.01%. The most common MC were cardiovascular, genitourinary and respiratory. The patients most at risk were the elderly, females, fracture patients and those with HIV or hepatitis C. During inpatient stay, reverse arthroplasty and blood transfusion were associated with MC. We recommend that our data are used to obtain informed consent for SA.

Conclusions

To reduce risk, we encourage optimisation of patients prior to surgery with correction of anaemia, arranging necessary clinician reviews and appropriate postoperative nursing care.

Introduction

Joint registries record an increasing number of shoulder arthroplasties (SA).¹ This is likely due to improved techniques and better outcomes for a range of shoulder pathologies.^2,3

Each arthroplasty procedure requires informed consent. This must include a shared decision to proceed after consideration of the risks and complications of the procedure.^4,5 A complication may be defined as ‘any undesirable or unintended event that occurs to the patient as a direct result of the operation which would not have occurred had the operation gone as well as could reasonably be hoped’.⁶

The literature regarding complications will guide the discussion. Currently, much of the literature regarding SA concentrates on the 6% risk of regional and surgical complications.⁵ It is the opinion of the authors that medical complications (MC) of SA are poorly explored in the literature, and of those available the majority focus on venous thromboembolism (VTE).^7–9 The limited available data demonstrate that MC after SA are associated with increased rates of hospital stay, readmission and mortality.^10–12

This appears to contrast with the MC of lower limb arthroplasty (LLA) which appear to have been studied in greater depth.^13–16 Consequently, it is the opinion of the authors that shoulder surgeons tend to extrapolate the systemic risks of SA from LLA-specific data, which may not be representative.^17–19

We aimed to aid the consenting process by summarising the literature regarding MC after SA. We defined MC as all acute medical pathologies recorded for patients in the peri- and postoperative periods. This included, but was not limited to, urinary tract infections (UTI), blood loss and transfusions, myocardial infarctions (MI), hospital-acquired pneumonia (HAP) and VTE.

Overall incidence of MC following SA

We found the rate of MC following SA was around 12% with large data sets reporting a range between 2% and 37% (Table 1).^{10,12,20–26,28–34} These total percentages were either taken directly from the article cited or represent an average of percentages for complications from each body system.

Table 1 .

Incidence of medical complications after shoulder arthroplasty (Percentages are either taken directly from the article or represent an average of percentages for complications different body systems)

Authors	Database	Years tested	Follow-up	No. of patients	Systemic complications % (n)
Anakwenze et al10	NIS	2006–2010	NTS	125,766	4.5 (5,704)
Bala et al20	USMSAF	2005–2012	90 days	145,761	37.2 (54,275)
Basques et al21	USMSAF	2005–2012	90 days	123,347	6.2 (7,685)
Boddapati et al22	ACS NSQIP	2005–2015	30 days	10,867	6.1 (660)
Cancienne et al23	PearlDiver	2005–2012	2 years	22,968	13.4 (3,066)
Craig et al12	NHS	1998–2017	90 days	58,054	8.9 (5,184)
Farng et al24	OSHPD	1995–2005	90 days	15,288	2.6 (403)
Grier et al25	USMSAF	2005–2010	2 years	42,087	3.4 (1,429)
Hambright et al26	NIS	1988–2005	NTS	25,398	10.0 (2,542)
Herrick et al27	PHDUSA	2010–2015	30 days	75,146	23.5 (17,645)
Issa et al28	NIS	1998–2010	NTS	226,170	4.8 (10,912)
Jiang et al29	NIS	2010–2011	NTS	19,497	11.0 (2,154)
Knapp et al30	USMSAF	2012–2016	30 days	12,445	6.1 (760)
Stundner et al31	PHDUSA	2007–2011	30 days	17,157	20.6 (3,529)
Tobert et al32	NIS	2002–2011	30 days	178,590	22.4 (40,025)
Westermann et al33	NIS	2011	NTS	66,485	16.5 (10,994)
Young et al34	NIS	2002–2011	NTS	422,372	12.8% (54,274)

ACS NSQIP = American College of Surgeons National Surgical Quality Improvement Program database USA; NHS = hospital episode statistics for NHS England – UK; NIS = Nationwide Inpatient Sample database USA; NTS = no time specified; OSHPD = California Office of Statewide Health Planning and Development; PearlDiver insurance database USA (http://www.pearldiverinc.com); PHDUSA = Premier Healthcare Database USA; USMSAF =  United States Medicare Standard Analytical Files database.

This wide range was attributable to:

• •variation in coding and reporting criteria for complications between studies;
• •variation in follow-up period after SA;
• •variable inclusion of high-risk patients such as those undergoing SA for fractures and trauma in different studies and data sets.²⁰

By contrast, the majority of existing MC data for LLA were derived from elective surgical data, which demonstrates overall MC rates of 1–8% for elective hip and knee arthroplasty.^35–37 Systemic complication rates following fractured neck of femur were higher with 30-day systemic complication rates up to 45%.³⁸

Risk factors for systemic complications after SA

Both patient and procedural factors were associated with increased MC.

Age

Age ≥80 years was an independent risk factor for both MC and mortality following SA.^10,28 Older patients were at risk of pulmonary embolism (PE). Patients with PE were significantly older than those without (72 versus 69 years) and the odds ratio (OR) of having a PE increased significantly with every decade of life.³⁴

Gender

Females were at higher risk of MC following SA.^12,30 This was notable for postoperative UTI (OR 0.58, p=0.032) and blood transfusion (OR 0.49, p=<0.001).³⁰

Race

In the literature, there was no significant difference in the rate of MC following SA when comparing Caucasian, African American and American Hispanic patients.^28,34,39

By contrast, some LLA data demonstrated higher rates of PE in African American patients although there were no differences found for other MC.⁴⁰ The overall rate of PE following SA (see below) may be too low to mirror the difference found in LLA.

Fracture surgery

Patients undergoing hemiarthroplasty (HA) and reverse shoulder arthroplasty (RSA) for fracture were at higher risk of 90-day complications, including VTE, when compared with those undergoing the same procedures for arthritis.²⁴

Similarly, SA patients with fractures had significantly higher rates of 30-day MC compared with those undergoing treatment for osteoarthritis Fracture patients were more at risk of HAP, MI, cardiac arrest, deep vein thrombosis (DVT), PE, transfusion, sepsis, septic shock, acute renal failure, progressive renal insufficiency, UTI and stroke and need for postoperative blood transfusions (p=<0.001).⁴¹

Systemic illness

Patients with HIV had a significantly higher risk of a wide range of MC following SA. This included acute renal failure, postoperative anaemia, arrhythmia, transfusion requirement, VTE, heart failure, MI, HAP, respiratory failure, sepsis/systemic inflammatory response syndrome and cerebrovascular accident (stroke).²⁰

Hepatitis C was also associated with a higher risk of MC.²³ For both HIV and hepatitis C, the increased complication rate was attributed to higher rates of other comorbidities rather than the effects of viral infection.²⁰

Rheumatoid arthritis was not associated with a higher risk of MC.²⁶

Type of anaesthesia

Herrick et al compared the MC of SA after three types of anaesthesia. MC rates were 13.8% for general anaesthetic (GA) alone, 12% for GA and nerve block, and 12% for nerve block alone. The ‘GA alone’ group was particularly at risk of infectious and pulmonary MC, but also of blood transfusions, transfer to intensive care and increased length of stay.²⁷

Other studies found no significant difference in the rate of cardiac, pulmonary or combined MC when comparing GA alone with GA and nerve block.³¹

Implant type

Several studies demonstrated that revision shoulder arthroplasty had higher rates of MC. This was attributed to patients being older and with more comorbidities.^10,25

RSA (compared with anatomic arthroplasty) was associated with increased rates of HAP, VTE, MI, UTI, blood transfusion and in-hospital morbidity,^29,33 although this was not demonstrated across all studies.

Comparison of total shoulder arthroplasty (TSA) with HA demonstrated a significantly higher risk of PE after TSA.³⁴ Preoperative pathology and the use of cement may be confounding factors for these results because the use of cement in RSA has also been demonstrated to have a slightly higher risk of VTE.⁴²

Blood transfusion

Rates of peri- and postoperative blood transfusion were found to be approximately 8% with reports varying from 1% to 18% (Table 2). Patients receiving blood transfusion following TSA had significantly higher rates of MI, HAP and sepsis/shock in the first 7 days following SA. They also had significantly higher rates of MC within 30 and 90 days following SA, including acute renal failure, arrhythmia, bleeding complications, DVT, heart failure, MI, PE, HAP, respiratory failure, sepsis/systemic inflammatory response syndrome, stroke and UTI.²⁵

Table 2 .

Complications by body system

Authors	Years tested	No. of patients	CVS % (n)	GUS % (n)	RS % (n)	GIS % (n)	CNS % (n)	Transfusion % (n)	Other % (n)
Anakwenze et al10	2006–2010	125,766	0.8 (993)	1.5 (1,853)	1.3 (1,611)	0.4 (472)	0.2 (234)	*	0.1 (118)
Bala et al20	2005–2012	145,761	15.9 (23,152)	4.9 (7,181)	1.6, (2,331)	*	0.4 (549)	5.6 (8,189)	7.7 (11,276)
Basques et al21	2005–2012	123,347	*	3.6 (4,442)	0.9 (1,160)	*	*	0.8 (1,020)	*
Boddapati et al22	2005–2015	10,867	*	0.8 (86)	0.7 (75)	*	0.1 (12)	3.9 (427)	0.2 (24)
Craig et al12	1998-2017	58,054	0.5 (282)	4.2 (2,442)	3.4 (1,980)	*	0.4 (218)	*	*
Farng et al24	1995–2005	15,288	*	*	*	*	0.8 (124)	*	1.2 (188)
Grier et al25	2005–2010	42,087	0.6 (267)	*	2.1 (882)	*	*	18.5 (7,794)	0.7 (280)
Hambright et al26	1988–2005	25,398	0.04 (11)	*	0.9 (219)	*	*	8.2 (2,076)	0.6 (144)
Herrick et al27	2010–2015	75,146	8.2 (6,125)	2.2 (1,649)	2.7 (2,014)	0.2 (178)	0.1 (71)	7.6 (5,721)	2.5 (1,887)
Issa et al28	1998–2010	226,170	0.9, (1,954)	2.9 (6,660)	0.6 (1,388)	*	0.04 (90)	*	0.05 (105)
Jiang et al29	2010–2011	19,497	0.3 (61)	2.5 (486)	0.5 (104)	0.3 (53)	0.3 (57)	6.7 (1,307)	*
Knapp et al30	2012–2016	12,445	0.3 (35)	0.8 (98)	0.5 (57)	*	0.1 (11)	3.3 (408)	0.5 (65)
Stundner et al31	2007–2011	17,157	7.1 (1,218)	1.2 (208)	3.0 (519)	0.3 (45)	0.1 (15)	6.1 (1,047)	2.6 (445)
Tobert et al32	2002–2011	178,590	0.3 (528)	1.2 (2,094)	1.1 (1,939)	0.3 (492)	*	8.9 (15,881)	10.2 (18,186)
Westermann et al33	2011	66,485	*	*	0.8 (549)	0.3 (168)	0.2 (131)	*	15.1 (10,067)
Young et al34	2002–2011	422,372	0.3 (1,267)	1.2 (5,068)	1.3 (5491)	0.3 (1267)	0.1 (422)	8.9 (37,591)	0.3 (1,267)
		Range	0.04–15.9%	0.8–4.9%	0.3–3.4%	0.2–0.4%	0.04–0.8%	0.8–18.5%	0.05–19.1%
		Mean	2.9%	2.3%	1.4%	0.3%	0.2%	8.1%	6.4%

CNS = central nervous system; CVS = cardiovascular system; GIS = gastrointestinal system; GUS = genitourinary system; RS = respiratory system

*Zero results excluded to account for discrepancy in reporting.

It was noted that SA patients receiving blood transfusions often had more preoperative medical comorbidities as well as longer operative times. It remains unclear whether these increased levels of MC are a direct result of the transfusions and blood products themselves. Surgeons should be mindful of preoperative patient optimisation and counselling of highly comorbid patients for SA with regard to transfusion requirement and its potential effects.²⁵

Longer hospital stay

Patients fit enough to undergo day-case SA had significantly lower rates of peri- and postoperative MC (for a 90-day period) than inpatients. This pertained to: acute kidney injury (AKI; p≤0.001), VTE (p≤0.001) and UTI (p≤0.001).²¹ This was similar to recent studies regarding comparisons of outpatient and inpatient knee arthroplasty.⁴³

The preoperative conditions associated with longer stay were preoperative narcotic use, heart disease, anaemia and a Charlson–Deyo Comorbidity Index (CDCI) score ≥2.⁴⁴

Complications by body system

It was difficult to extrapolate this information from published data due to significant inconsistencies in the reporting of MC between papers. Few studies addressed complications following SA and fewer still addressed the MC. The available data are summarised in Table 2.

Two studies specifically assessed MC following SA in a large cohort of patients. Young et al investigated patient factors associated with the development of PE and other MC that occurred by body system in >40,000 patients. Only ‘major complications’ were included: MI (0.3%), HAP (1.3%), DVT (0.2%), stroke (0.1%), ‘acute renal failure’ (1.2%), ‘gastrointestinal complication’ (0.3%), mechanical ventilation (0.3%) and blood transfusion (8.9%).³⁴

Issa et al reviewed 226,170 patients to determine MC during the postoperative inpatient stay. Of all inpatient postoperative complications, the most common complications were genitourinary (UTI and renal failure at 2.9%), with cardiovascular system (CVS) complications the second highest (0.9%).²⁸

Cardiovascular system

CVS complications were the most common MC following SA with a rate of around 3% across all studies (range 0–16%). This variation was in part because some studies reported only ischaemic cardiac events such as MI, whereas others included nonischaemic events such as arrhythmia.^{10,12,20,25–32,34}

We found the rate of MI overall was <1%, whereas the rate of all nonischaemic CVS complications was 2–16%.^20,31,44,45

These figures are comparable with those following LLA.^15,35,36

Genitourinary system

The most commonly reported genitourinary system (GUS) complications were acute renal failure (AKI) and UTI with a range between 0% and 5% (Table 2).^{10,12,20,21,27–29,31,32,34,46}

Knapp et al reported specific rates of individual GUS complications.³⁰ UTI was the most common (0.6%), followed by renal insufficiency (0.1%) and acute renal failure (0.05%). In Knapp et al’s data, when compared with other body systems, GUS complications were the most common with a significantly higher incidence in female patients.³⁰

The rate of GUS complications was slightly lower than in LLA where authors have reported an incidence of UTI between 1% and 2% and of AKI between 0% and 1%.^35–37

Respiratory system

Respiratory system (RS) complications following SA were widely reported. The overall range was 0–3%.^{10,12,20–22,25–34} HAP was common with a rate of 1–2% in studies with detailed data.^{12,20,21,25–30,32,34}

The rate of RS complications appeared to be higher than after LLA. After LLA, the RS rate was <1%.^35–37 This might be a consequence of more frequent use of GA for SA compared with regional anaesthesia for hip and knee arthroplasty. Furthermore, regional anaesthesia used in conjunction with SA has the potential for respiratory compromise because of undesired phrenic nerve paresis.⁴⁷

Gastrointestinal system

The reported rate of gastrointestinal system (GIS) complications was 0.2–0.4%. GIS complications were poorly defined with many papers reporting non-specific ‘gastrointestinal complications’.^27,31,32,34 The most common specific GIS complication was postoperative ileus.^10,29,33

These rates are comparable with those following both hip and knee arthroplasty.^35–37 However, the data were extremely limited and further study is needed to effectively compare the rates of GIS complications in SA versus LLA.

Central nervous system

The reported complication rate was low after SA at 0.04–0.8% (Table 2). Stroke was the most commonly given CNS complication with most studies reporting it as their sole CNS complication.^{12,20,22,24,27,28,30,31,34}

Jiang et al were the only authors to describe CNS complications more specifically, reporting an overall rate of 0.3%, which included ‘acute mental status changes’ as well as stroke.²⁹

HIV patients were at greater risk of CNS complications. Bala et al demonstrated a significantly higher risk of CNS complications in those with HIV compared with those without HIV (14.78% versus 0.39%).²⁰ This may be partly due to selection bias because HIV patients had significantly higher rates of neurological conditions prior to surgery.²⁰

The rate of CNS complications following SA were similar to LLA.^35–37 This was reassuring given the known effect of beach chair positioning during SA on cerebral perfusion.⁴⁸

Venous thromboembolism

Of all MC, VTE was most frequently reported in the literature. The reported rate of VTE following SA was <1% (range 0.1–6.0%) (Table 3). The reported rate of DVT and PE was 0.1–0.7% and 0.1–0.5%, respectively.^{8–12,20–24,28,29,32–34,46,50,51} Tashjian et al is an outlier, reporting a rate of 6%. This study examined revision shoulder arthroplasty and RSA in comparison with anatomical shoulder arthroplasty, suggesting an increased chance of VTE. The sample size for these groups was small in comparison with other papers and was likely not representative.⁴⁶ Two formal systematic reviews on VTE following SA reported a rate of 0.5%.^7,52

Table 3 .

Incidence of venous thromboembolism after shoulder arthroplasty

	Authors	Database	Years tested	No. of patients	VTE % (n)	DVT % (n)	PE % (n)
NSSA	Issa et al28	NIS	1998–2010	226,170	0.3% (715)	0.1% (234)	0.2% (481)
	Bala et al20	USA Medicare database	2005–2012	145,761	1.1% (1597)	0.7% (962)	0.4% (635)
	Sing et al8	NIS	2008–2011	315,423	0.5% (1420)	0.3% (976)	0.3% (873)
	Basques et al21	USMSAF	2005–2012	123,347	0.9% (1063)	*	*
	Day et al9	USA Medicare database	2004–2009	130,158	1.1% (1408)	0.5% (694)	0.5% (714)
	Young et al34	NIS	2002–2011	422,372	0.5% (1901)	0.2% (845)	0.3% (1,056)
	Anakwenze et al10	NIS	2006–2010	125,766	0.3% (423)	0.1% (107)	0.3% (233)
	Tobert et al32	NIS	2002–2011	178,590	0.5 (905)	0.2 (432)	0.3 (473)
	Boddapati et al22	ACS NSQIP	2005–2015	10,371	0.3 (34)	*	*
ASA	Craig et al12	NHS	1998–2017	58,054	0.5 (262)	*	0.5 (262)
	Jiang et al29	NIS	2010–2011	14,031	0.3 (46)	0.2 (25)	0.1 (21)
	Schairer et al11	State Inpatient Database (USA)	2005–2010	23,434	0.3 (70)	*	*
	Westermann et al33	NIS	2011	44,793	0.1 (38)	*	*
	Sing et al8	NIS	2008–2011	90,181	0.4 (346)	*	*
	Farng et al24	OSHPD	1995–2005	15,288	0.6 (91)	*	*
	Cancienne et al49	PearlDiver	2010–2014	5,165	3.0 (154)	1.9 (97)	1.3 (67)
	Jameson et al50	NHS	2005–2008	12,358	0.2 (30)	0.1 (11)	0.2 (23)
	Johnson et al51	Retrospective analysis^	1999–2011	225	1.3 (3)	0.9 (2)	0.4 (1)
	Tashjian et al46	ACS NSQIP	2011–2016	337	1.8 (6)	*	*
RSA	Jiang et al29	NIS	2010–2011	5,466	0.7 (40)	0.5 (27)	0.2 (13)
	Schairer et al11	State Inpatient Database (USA)	2005–2010	2,784	0.4 (11)	*	*
	Westermann et al33	NIS	2011	21,692	0.2 (41)	*	*
	Sing et al8	NIS	2008–2011	5,173	0.4 (19)	*	*
	Johnson et al51	Retrospective analysis^	1999–2011	176	1.7 (3)	1.1 (2)	0.6 (1)
	Tashjian et al46	ACS NSQIP	2011–2016	112	2.7 (3)	*	*
ReSA	Sing et al8	NIS	2008–2011	7,590	0.2 (17)	*	*
	Johnson et al51	Retrospective analysis^	1999–2011	51	3.9 (2)	2.0 (1)	2.0 (1)
	Tashjian et al46	ACS NSQIP	2011–2016	84	6.0 (5)	*	*
	Boddapati et al22	ACS NSQIP	2005–2015	496	0.4 (2)	*	*

ACS NSQIP = American College of Surgeons National Surgical Quality Improvement Program database USA; ASA = anatomical shoulder arthroplasty (hemi and total); DVT = deep vein thrombosis; NIS = Nationwide Inpatient Sample database USA; NSSA = non-specific shoulder arthroplasty; OSHPD = California Office of Statewide Health Planning and Development; PE = pulmonary embolism; PearlDiver insurance database USA (http://www.pearldiverinc.com); ReSA = revision shoulder arthroplasty; retrospective analysis = Division of Shoulder Surgery Department of Orthopaedic Surgery The Johns Hopkins University USA; RSA = reverse shoulder arthroplasty; USMSAF = The United States Medicare Standard Analytical Files database; VTE = venous thromboembolism.

*Zero results excluded to account for the discrepancy in reporting

Several authors sought to compare the VTE risk for different types of SA but no convincing difference was found, largely because of low sample sizes and low incidence of VTE within the studies, making it difficult to draw conclusions.

Young et al specifically investigated risk factors for PE following SA and reported that the top risk factor was SA performed for fracture. Other risk factors were: preoperative evidence of anaemia, heart failure or chronic obstructive pulmonary disease, obesity and increasing age.³⁴

Increased length of stay was also associated with VTE, although this may be a product of the pre-existing medical comorbidities already stated.^22,51

Day et al found a significantly increased risk of PE, both as an inpatient and within 3 months of discharge in SA performed for fracture.⁹ They also determined that the rates of VTE were higher for all LLA compared with SA apart from PE following total hip arthroplasty where they found no significant difference.⁹

Mortality following SA

The reported mortality rate following SA was 0.01% with a range of 0–1.3%.^{10,12,20,22,24,26–34,50,53} Mortality was higher for papers reporting 90-day rather than 30-day mortality (Table 4),^12,24,50 and also for fracture patients.^24,50 For instance, Farng et al demonstrated a mortality rate of 2.9% in trauma patients compared with 0.5% for elective patients at 90-day follow-up in a cohort of 15,000 patients.²⁴

Table 4 .

Postoperative mortality rate following shoulder arthroplasty

Paper	No. of patients	Length of follow-up	Deaths % (n)
Anakwenze et al10	125,766	NTS/IP	0.1 (88)
Bala et al20	145,761	30-day	0.04 (55)
Boddapati et al22	10,867	30-day	0.2 (18)
Craig et al12	25,011	90-day	0.5 (118)
Farng et al24	15,288	90-day	1.3 (199)
Hambright et al26	25,398	NTS/IP	0.4 (89)
Herrick et al27	75,146	30-day	0.1 (87)
Issa et al28	226,170	NTS/IP	0.1 (238)
Jiang et al29	19,497	NTS/IP	0.1 (20)
Knapp et al30	12,445	30-day	0.2 (21)
Stundner et al31	17,157	30-day	0.03 (6)
Tobert et al32	178,590	NTS/IP	0.1 (245)
Westermann et al33	66,485	NTS/IP	0.1 (95)
Young et al34	422,372	NTS/IP	0.1 (422)
Jameson et al50	12,358	90-day	1.0 (118)
McCormick et al53	101,323	NTS/IP	0.1 (113)

NTS/IP = no time specified/inpatient.

Craig et al reported mortality to be 0.2% at 30 days and 0.5% at 90 days in a study excluding trauma patients. These results question whether 30-day, 90-day or longer duration studies were required to capture mortality and other MC.^12,24

As expected, other risk factors for mortality after SA included preoperative comorbidities and the occurrence of postoperative respiratory distress or MI.¹⁰

McCormick et al examined the effect of age and gender on mortality and found that patients over the age of 84years were at the highest risk with an inpatient mortality rate of 0.28%. They did not demonstrate a difference in mortality between male and female patients.⁵³

Preoperative identification of patients at high risk of systemic complications

Several studies have suggested screening tools to identify patients at risk of MC after SA.

Boddapati et al and Lovy et al included ASA grade in their analysis of complication risk in SA. They demonstrated that ASA grade 3 or 4 patients had a significantly higher risk of MC following SA.^22,54 Johnson et al found that although ASA grade >2 resulted in higher rates of surgical complications, there was no significant increase in the rate of MC with higher ASA grade, although length of stay there was significantly longer in these patients.⁵¹

An alternative to ASA grade is the CDCI. This is an age-independent score for predicting 10-year survival in patients with multiple comorbidities.^55,56 Several authors have demonstrated increasing incidence of MC following SA with increasing CDCI score. The use of the CDCI as a predictive tool in SA has not been compared with other screening tools such as ASA grade.^{5,12,24,28,50,57}

Discussion

Our review found that the risk of MC after SA was 12%.^{10,12,20–34} This was double the 6% risk of regional surgical complications.⁵ The overall mortality rate was 0.1% (1/1,000).

Levels of respiratory system MC appear higher than for elective LLA potentially due to an increased proportion of GA versus regional anaesthetic used in these patients in SA. Other body system MC appear equivalent to or lower than in LLA.^35–37

We found the following risk factors for MC: increasing age, pre-existing comorbidities, increased operative time, preoperative anaemia, fracture SA, HIV and RSA.^{19,21,22,24,25,32,33,58–60} There was inconsistent evidence of increased risk associated with female gender and increased body mass index.^{12,22,28,30,34,61,62}

The shoulder literature was not as extensive compared with LLA, but it contains some large-scale studies with valuable insight into the risks of systemic complication.^10,12,34 VTE is the best understood and most extensively studied systemic complication.^7,52 Nevertheless, there remains a lack of robust large-scale data regarding MC, particularly when compared with LLA.^13–16

Limitations

The available data were highly variable, likely due to multiple factors including skewed data from specific comorbidities (eg, HIV), variable follow-up periods, inconsistencies in coding and variability of reporting criteria for databases and within the literature. Some complications may have been considered too minor to be noted.⁶³

Many studies derived data from insurance claims in the USA, there is therefore a risk of under-reporting of milder complications and inaccuracies due to coding practice.^22,62

Finally, the majority of the literature included retrospectively collected data, meaning that fewer than 50% of complications were likely to have been recorded.⁵⁸

Recommendations

• 1.All SA patients must be fully informed of the risk of MC before their operation to ensure legally robust consent is obtained. See Appendix 1 for a short summary of MC risks for use in patient communication.
• 2.High-risk patients should be identified, and all efforts made to optimise them prior to surgery, including identifying/treating anaemia, arranging anaesthetic/cardiology/respiratory review, preparing postoperative high-dependency unit beds where appropriate and planning the appropriate SA/anaesthetic option.
• 3.Further study is needed. In particular, comparing and contrasting the risk of MC of fracture SA versus other indications and examining possible scoring systems to determine suitability for outpatient SA.
• 4.If this review proves useful, there may be a need for similar studies for other orthopaedics procedures and/or arthroplasties to ensure that surgeons performing these are fully informed of the risks of MC while consenting their patients.
• 5.The use of regional anaesthesia in combination with GA appears to have the best outcomes with studies showing reduced MC.^27,31,59,60 However, patients must be informed of the potential risks.⁴⁷
• 6.VTE is more of a risk in SA than was previously thought and surgeons must identify these high-risk patients in advance of surgery and ensure appropriate VTE prophylaxis is in place postoperatively as per National Institute for Health and Care Excellence guidelines.⁶⁴ Consideration should be given to using mechanical VTE prophylaxis in combination with aspirin due to the low risk of bleeding and then using other pharmaceutical agents such as low molecular weight heparin for patients identified as high risk.^9,50,65
• 7.Patients requiring perioperative blood transfusion should be appropriately counselled on the risks of MC associated with them and efforts made to risk stratify and optimise these patients prior to SA.²⁵
• 8.To account for developments in surgical practice, a review such as this should be performed in another 5–10 years. This will allow for the most up-to-date MC rates to be easily available for SA surgeons.