Predictors of extended length of stay after unicompartmental knee arthroplasty

Abstract

Purpose

To identify factors that independently predict extended length of stay after unicompartmental knee arthroplasty (UKA) surgery (defined as length of stay longer than 3 days), and to identify factors predicting early post-operative complications.

Methods

A retrospective analysis of all patients undergoing UKA from January 2016–January 2019 at our institution was performed. Clinical notes were reviewed to determine the following information: Patient age (years), gender, American Society of Anesthesiologists (ASA) grade, weight (kg), height (meters), body mass index (BMI), co-morbidities, indication for surgery, surgeon, surgical volume, surgical technique (navigated or patient-specific instrumentation), implant manufacturer, estimated blood loss (ml), application of tourniquet during the surgery, application of drain, hospital length of stay (days) and surgical complications.

Results

Multivariate regression analysis showed that ASA 3–4 vs. ASA 1–2 [OR 4.4 (CI; 1.8–10.8, p = 0.001)] and a history of cardiovascular disease [OR 2.8 (CI; 1.4–5.5), p = 0.004)] were significant independent predictors of prolonged length of stay. Hosmer-Lemeshow goodness of fit of the model showed a p-value of 0.214. Nagelkerke R-Square was 0.2. For complications, multivariate regression analysis showed that ASA 3–4 vs. ASA 1–2 [OR 5.8 (CI; 1.7–20.7)] and high BMI (BMI >30) [OR 4.3 (CI; 1.1–17.1)] were significant independent predictors of complications. Hosmer-Lemeshow goodness of fit was 0.89 and Nagelkerke R-Square was 0.2. Patients treated with robotics (Navio) techniques had shorter length of stay median 51 h (IQR; 29–96) when compared to other techniques 72 h (IQR; 52–96), p = 0.008.

Conclusion

Based on the results of our study, high ASA grade (≥3) appears to be the most important factor excluding eligibility for fast-track UKA. Any number of co-morbidities may increase ASA, but in and of themselves, apart from a history of cardiovascular disease, they should not be seen as contraindications. Appropriate patient selection, technical tools and details during the surgery could facilitate fast track surgery.

1. Introduction

Length of hospital stay (LOS) is an important determinant of overall cost following joint arthroplasty surgery.¹ Whilst improvements in the efficiency of post-operative care have led to a substantial reduction in LOS over the last two decades, growing financial pressures on healthcare systems have prompted renewed efforts to predict and reduce LOS.² Furthermore, questions have surfaced around eligibility criteria for ‘fast-track’ or day-case joint arthroplasty surgery which is predicated by the ability of hospitals to identify patients at risk of prolonged LOS and mitigate this risk appropriately.^{3, 4, 5, 6}

Unicompartmental knee arthroplasty (UKA) is a less invasive alternative to total knee arthroplasty (TKA) associated with lower post-operative morbidity and faster return to function.^{7, 8, 9} The mean length of stay following UKA in England and Wales has been reported as 4.1 days compared to 5.5 days for TKA.¹⁰ It has therefore attracted particular attention as a focus for efforts to safely achieve low LOS.¹¹ However, whilst a number of patient, surgeon and hospital-related factors have been shown to predict LOS after TKA, there is a paucity of literature regarding UKA.^12,13 The difference in demographics between patients undergoing these procedures, with UKA patients typically being younger with fewer co-morbidities, justifies the need for separate analyses.¹⁰

In our institution, LOS <3 days is defined as ‘fast track’. There is also evidence that LOS >3 days is associated with an increased risk of complications, including readmission, after arthroplasty.¹⁴ Information regarding surgical complications is of use to clinicians and policymakers seeking to identify patients at high risk of delayed discharge, guiding pre-emptive intervention and defining eligibility criteria for fast-track UKA.¹⁵

Our primary aim was to identify factors that independently predict extended length of stay after UKA, defined as LOS of more than 3 days (72 h). Our secondary aim was to identify factors predicting in-hospital surgical complications following UKA surgery.

2. Materials and methods

Following institutional approval, a retrospective analysis of all patients undergoing primary UKA surgery from January 2016 to 2019, was performed. All patients undergoing primary UKA were included in the study. Exclusion criteria included patients undergoing bilateral UKAs or revision surgery for any reason. A total number of 155 patients were identified and included for further analysis.

All information was obtained using the CERNER electronic record system with clinical notes reviewed to determine the following information: patient age (years), gender, ASA grade, weight (kg), height (meters), BMI, co-morbidities, indication for surgery, surgeon, surgical volume, surgical technique (navigated or patient-specific instrumentation), implant manufacturer, estimated blood loss (ml), application of tourniquet during the surgery, application of drain, hospital length of stay (days) and surgical complications.

All patients underwent a standardized arthroplasty pathway including pre-operative patient education presentations from nursing staff, occupational therapy & physiotherapy. All patients had an anaesthetic pre-operative assessment to optimize relevant medical co-morbidities prior to surgery. All operations were performed under general or spinal anaesthesia with standard antibiotic prophylaxis. As standard, multi-modal pain relief was offered as part of an enhanced recovery programme with admission to an elective orthopaedic ward. Patients all underwent post-operative rehabilitation under the same orthopaedic physiotherapy team. All patients were admitted on the day of surgery. Length of stay (LOS) was defined as the surgical start time to discharge from hospital ward.

2.1. Factors analysed

Statistical comparison of the impact of individual pre-operative factors on LOS was performed by categorization of pre-operative factors into high and low data point groups. Patient age was categorized as old (age >75 years) and young (age ≤75 years). ASA was categorized as low ASA (ASA 1–2) and high ASA (ASA 3–4). BMI was calculated for every patient based on the equation for BMI = weight (kg)/height²(m). Patients with BMI>30 were defined as high BMI. Patients with BMI≤30 were defined as low BMI.

Co-morbidities were categorized as cardiovascular (including hypertension, previous myocardial infarction, atrial fibrillation, arrhythmia or congestive cardiac failure), respiratory (including asthma, chronic obstructive pulmonary disease, a history of lung tuberculosis or lung sarcoidosis), neuro-psychiatric (including stroke, schizophrenia, depression, Parkinson's disease) and immune system disorders. Diabetes mellitus was defined as having either tablet or insulin treatment, or a self-defined history of diabetes. Surgeons with more than 10 surgeries/year during the time period were classified as high volume, whilst those with fewer than 10 surgeries/year were classified as low volume.¹⁶ Length of stay (LOS) was measured in days, defined as full days (24 h), from admission to discharge. Fast track was defined as length of stay ≤3 days and long stayers were defined as patients with LOS >3 days.

Surgical complications were defined as complications which could be directly related to the surgery including: wound infection, venous thromboembolism, acute coronary syndromes following surgery, urinary tract infection, hospital acquired pneumonia and implant related complications.

2.2. Statistical analysis

Scale variables with normal distribution are presented as mean (standard deviation, SD). Non-parametric variables and scale variables without normal distribution are presented as median (inter-quartile range, IQR). Normality of distribution was tested using the Shapiro-Wilk Test. Proportions are presented as n; number (%). Univariate comparative analysis of scale variables with normal distribution was performed using the Student-T test. Analysis of non-parametric variables or non-normally distributed scale variables was conducted using the Mann-Whitney U test. Categorical data were analysed using Fisher's exact test. Multivariate regression analysis was conducted to analyze factors affecting LOS and post-operative complications. Receiver Operating Characteristic (ROC) was used to illustrate factors analysed in the multivariate analysis section. A p-value < 0.05 was considered as statistically significant throughout.

3. Results

A total of 155 patients underwent UKA in our institution over a 3 year period dating from January 2016 to 2019. The mean patient age was 66.¹⁰ Male to female ratio was 1.2:1. Medial side Osteoarthritis (OA) was the stated indication for UKA in 135 out of 155 (87%) of patients and the indication in the remaining 13% was lateral side OA. Two surgeons performed a combined total of 129 (83%) of procedures. Technical choice; 98 out of 155 (63%) of cases were performed using computer navigation or robotic-assisted surgery (Navio, Smith and Nephew or Brainlab, Depuy Synthes) and the remainder (37%) performed using patient-specific instrumentation (PSI) (Embody UK, London) or jig-based technique. The general characteristics of our cohort is presented in Table 1.

Table 1.

General patient characteristics.

General Characteristics
Total N (%)	155 (100)
Age, Mean (SD)	66 (10)
Female, N(%)	70 (45)
Height cm, Mean(SD)	165 (10)
Weight kg, Mean(SD)	82 (17)
BMI, Median(IQR)	29 (26–34)
Blood volume, Median(IQR)	4700 (4200–5100)
ASA Score
1 N (%)	20 (13)
2 N (%)	102 (66)
3 N (%)	31 (20)
4 N (%)	2 (1)
Pre-op Hemoglobin, Mean(SD)	135 (14)
Post-op Hemoglobin, Mean(SD)	122 (13)
Pre-op Hematocrit, Mean(SD)	0.41 (0.04)
Post-op Hematocrit Mean(SD)	0.36 (0.04)
Blood loss, Median(IQR)	600 (400–830)
Blood loss percentage, Mean(SD)	13 (6.5)
Tranexamic Acid
Yes, N (%)	53 (34)
No, N (%)	98 (63)
Missing, N (%)	4 (3)
Tourniquet
Yes, N (%)	148 (95)
No, N (%)	7 (5)
Side
Right	73(47)
Left	82(53)
Medial	135(87)
Lateral	20(13)
Technique
Navio	67(43)
Brain Lab	31(20)
Jig	51(33)
PSI	6(4)
Manufacturer
Accuris	89(57)
Journey	19(12)
Oxford Uni	46(30)
Sigma	1(1)

Fast track patients (LOS ≤3 days) represented 85 out of 155 (55%) of cases. Compared to long stayers, fast track patients had a significantly lower mean age [64(10) vs 69(10), p = 0.003], lower BMI [28(26–32) vs 31(26–36), p = 0.038] and lower proportion of ASA grade 3–4 patients [8(9%) vs 25(36%), p < 0.001). There was no significant difference in estimated blood loss [(13%(±7%) vs 13%(±6%), p = 0.526] or tourniquet time [78 min (29 min) vs 74 min (23 min)] between the two groups (Table 2).

Table 2.

Comparison of Fast Track patients where the length of stay is ≤ 3 days to long stayers (>3days).

Total n(%)	Fast track 85(55)	Long stay 70(45)	p-value
Age, Mean (SD)	64(10)	69(10)	0.003
Female, N(%)	42(50)	28(40)	0.260
Height cm, Mean(SD)	167(10)	163(10)	0.015
Weight kg, Mean(SD)	81(16)	83(17)	0.513
BMI, Median(IQR)	28(26–32)	31(26–36)	0.038
ASA Score
1–2 N (%)	77(91)	45(64)	<0.001
3–4 N (%)	8(9)	25(36)
Blood loss percentage, Mean(SD)	13(6)	13(7)	0.526
Tranexamic Acid
Yes, N (%)	52(61)	46(66)	0.854
No, N (%)	31 (36.5)	22(31)
Missing, N (%)	2(2.5)	2(3)
Tourniquet
Yes, N (%)	80(94)	68(97)	0.458
No, N (%)	5(6)	2(3)
Tourniquet Time Mean(%)	74(23)	78(29)	0.372
Drain	4(5)	1(1)	0.065
Side N (%)
Right	45(53)	28(40)	0.145
Left	40(47)	42(60)
Medial N (%)	71(83.5)	64(91)	0.158
Lateral N (%)	14(16.5)	6(9)
Technique N (%)
Navio	41(48)	26(37)	0.187
Brain Lab	14(16.5)	17(24)
Jig	25(29)	26(37)
PSI	5(6)	1(1)
Manufacturer N (%)
Accuris	47(55)	42(60)	0.643
Journey	12(14)	7(10)
Oxford Uni	26(31)	20(29)
Sigma	0(0)	1(1)
High Volume Surgeon N (%)	69(81)	60 (86)	0.521
Age>75 N(%)	12(14)	21(30)	0.014
BMI>30 N(%)	28(35)	35(52)	0.045
Any Comorbidity N(%)	60(48)	64(52)	0.001
Cardiovascular Disease N(%)	35(42)	49(58)	<0.001
Respiratory Disease N(%)	5(29)	12(71)	0.037
Neuro-Psychiatric Disease N (%)	4(50)	4(50)	1.000
Immune System Disease N (%)	21(43)	28(57)	0.056
Diabetes N (%)	7(8)	18(26)	0.004
Hyperlipidemia N (%)	17(50)	17(50)	0.562
Hypertension N (%)	32(38)	43(61)	0.004
Any complication N (%)	2(2)	11(16)	0.003
Readmission N (%)	3 (4)	4 (6)	0.702

Overall 13 out of 155 (8%) of patient developed a surgical complication consisting of two wound infections, two cellulitis, two pulmonary embolisms, one DVT, two electrolyte imbalances, two cardiac anginas, one urinary tract infection and one patient with excessive cement that required removal.

Comparison of the robotic (NAVIO) technique vs. other techniques showed shorter length of stay, median 51 h (IQR; 29–96) when compared to other techniques, median 72 h (IQR; 52–96), p = 0.008. Table 3 is illustrating comparison of the groups based on technical details.

Table 3.

Illustrates comparison of Robotic (NAVIO) technique vs. other techniques.

Total numbers (%)	Robotic (NAVIO)	Other Surgical Approaches	p-value
	67 (43)	88 (57)
Age; Mean (SD)	68 (10)	65 (10)	0.073
Female	34 (51)	36 (41)	0.256
LOS hours; Median (IQR)	51(29–96)	72(50–96)	0.008
Blood loss Percentage	13(8–17)	12(9–18)	0.658
BMI	28(25–33)	30(27–34)	0.277
ASA
1–2	54 (81)	68 (77)	0.383
3–4	13 (19)	20 (23)
Tourniquet Time	79(65–92)	71(60–90)	0.061
Complications	6 (9)	7 (8)	0.523

Multivariate regression analysis using stepwise fashion with Wald correction showed that ASA3-4 vs. ASA 1–2 [OR 4.4 (CI; 1.8–10.8, p = 0.001)] and a history of cardiovascular disease [OR 2.8 (CI; 1.4–5.5), p = 0.004)] were significant independent predictors of prolonged length of stay. Hosmer-Lemeshow goodness of fit of the model showed a p-value of 0.214. Nagelkerke R-Square was 0.2.

For complications, multivariate regression analysis showed that ASA 3–4 vs. ASA 1–2 [OR 5.8 (CI; 1.7–20.7)] and High BMI (BMI >30) [OR 4.3 (CI; 1.1–17.1)] were significant independent predictors of complications. Hosmer-Lemeshow goodness of fit was 0.89 and Nagelkerke R-Square was 0.2. ROC-curves with all variables with uneven distribution between the groups are presented in Fig. 1, Fig. 2.

Fig. 1.

ROC curves representing factors affecting length of stay following UKR.

Fig. 2.

ROC curves representing factors predicting complication following UKR.

4. Discussion

The demand for UKA procedures, in the setting of unicompartmental knee arthritis, continues to increase within the United Kingdom.^17,18 Whilst predictive factors for increased LOS have been studied extensively in THR and TKA procedures there is little evidence in this area for UKA patients.^7,19 Assessment of pre-operative and patient demographic factors impacting on LOS is essential to aid planning of long stay patients and the inevitably high costs this incurs to healthcare institutions.

In this study, we found that independent predictors of increased LOS after UKA were high ASA grade (ASA 3–4 vs ASA 1–2) and a history of cardiovascular disease. Other literature in this area, specific to UKA and LOS, is limited to two similar studies conducted previously.^7,20 Haughom et al.²⁰ in an analysis of the U.S. National Surgical Quality Improvement Program (NSQIP) database in 2014, identified patient-related risk factors for LOS >4 days after UKA as ASA grade ≥3, increased BMI and a history of COPD. An earlier study by Brown et al.⁷ found that increasing age, BMI and increased Charlson Morbidity Index²¹ all positively predicted increased LOS following UKA.

High ASA grade ≥3 was found to be a positive predictor of increased LOS in our own study and Haughom et al.’s study, this finding has also been replicated in both TKA and THR populations.^{1, 22, 23, 24} In relation to this, increasing co-morbidities have thus been associated with longer hospital stays in numerous studies of primary joint arthroplasty.^25,26 Interestingly, it is notable that the only discrete co-morbidity to independently predict LOS in our study was cardiovascular disease. Brown et al.⁷ also found increasing age as a separate predictor of increased LOS a finding which is consistent with TKA studies^2,22,25,27, however, our study and Haughom et al. did not come to this conclusion. One explanation for this may be due to the difference in demographics between patients undergoing these procedures. UKA patients are typically better surgical candidates with fewer co-morbidities and this may impact more on LOS than age alone.

BMI was a predictor of increased length of stay in both Brown et al. and Haughom et al.’s studies surprisingly BMI was not found to be a significant risk factor in this study. Patients with increased body weight may find post-operative mobilisation more difficult and are more susceptible to post-operative complications²⁸ thus increasing LOS. There have been equivocal findings of the impact of gender on LOS in TKA patient populations, with some studies demonstrating that women tend to have a longer LOS than men^22,29,30, whilst other studies report no significant differences.^31,32 Gender does not appears to be a significant predictor on LOS in UKA patients with this and aforementioned studies finding no significant differences.

Low pre-operative haemoglobin is also a commonly reported predictor of increased LOS in THR and TKA studies.^22,27,33 This however, was not found to influence LOS in our UKA population. UKA procedures are minimally invasive and thus have lower levels of blood loss allowing avoidance of post-operative anaemia and time-consuming blood transfusions. Less blood loss and fewer blood transfusions mean the impact of pre-operative haemoglobin on LOS in UKA is not as significant as in TKA patient populations. Additionally, we found that patients operated with the robotic (Navio) technique had shorter hospital stay compared to all other techniques together. However, in a multivariate regression analysis our results were indicating the view that patient-related factors were more important in determining suitability for fast-track UKA than surgeon or hospital-related factors.^34,35

Whilst the demand for reduced LOS is predominantly financial with the aim to free up healthcare resources, institutions must ensure this does not decrease the quality of care received and that patients are not at increased risk of further complications or readmissions (3 from Dauty 2008). It is therefore imperative to analyze independent predictors of complications alongside those of decreased LOS. In our study high ASA grade and high BMI were significant independent predictors of complications as has been shown elsewhere.^20,22 However, co-morbidities including diabetes and immunodeficiency did not predict complications, in contrast to their effect on the short-term risk of complications after TKA.³⁶ This may be due to the low overall number of complications and inadequate powering of the study to assess the impact of such co-morbidities, or possibly due to the fact that such co-morbidities are comparatively less common in patients undergoing UKA compared to TKA. Nonetheless, coupled with the analysis of LOS it does suggest that many discrete co-morbidities taken in isolation should not be considered contraindications to fast-track UKA.

With regards to readmission rates our study found no significant difference between fast track and long stay patients. Several studies have analysed risk factors for readmission following primary joint arthroplasty, with the most common risk factors for readmission being significant co-morbidities, discharge destination and increased inpatient length of stay.^{37, 38, 39} A further study by Ricciardi et al.⁴⁰, in 2017, looked at readmission rates in a large cohort of 21,864 patients undergoing THR/TKA over a 4 year period. They found that found that patients with a length of stay shorter than anticipated (<2 days for THA and <3 days for TKA) was predictive of 30-day readmission. A balance, is therefore needed, between ensuring safe home discharge and reducing inpatient LOS with constant evaluation of complication and readmission rates when implementing rapid discharge pathways.

In recent years, reduction in the length of stay in patients undergoing UKA has gained considerable interest(4). This has led to the introduction of ‘fast track’, ‘enhanced recovery’ and outpatient surgery pathways aimed at reducing cost and improving the safety of UKA. A natural progression was to develop a day-case surgery pathway for UKAs. Several authors^4,41,42 to date have reported on the feasibility and safety of UKA as a day-case procedure. Complication and readmission rates were found to be comparable to established fast-track pathways.^4,41,42 Patient selection for day-case candidates such as low ASA grade and absence of cardiovascular diseases may improve the success of these pathways.

4.1. Strengths and limitations

Whilst there is an abundance of literature available on risk factors for prolonged LOS in TKA surgery, there is a relatively paucity of literature available on risk factors in UKA surgery. This study reports these independent risk factors with a large cohort of patients highlighting an important message to help identify those patients at higher risk for prolonged hospital admissions. The study does however, have limitations. These include the retrospective study design, potential recall bias relying on electronic notes to provide most of the data and (compared to registry studies) the relatively small number of patients included. Additionally there were certain factors which were not included in the study such as pre-operative mobility, cognitive function, post-operative pain, nausea and vomiting which may influence LOS but were not reliably reported on our electronic notes system.

5. Conclusion

Based on the results of our study, high ASA grade (≥3) appears to be the most important factor excluding eligibility for fast-track UKA. Any number of co-morbidities may increase ASA, but in and of themselves, apart from a history of cardiovascular disease, they should not be seen as contraindications. Appropriate patient selection and technical tools and details during the surgery could facilitate fast track surgery.

Contribution

All the authors have participated during the preparation and writing of the manuscript. All the authors have read and accepted the final format of the manuscript.

Conflicts of interest

The authors declare no conflict of interests.

Appendix A. Supplementary data

The following is the Supplementary data to this article: