Predictive risk factors for 30-day readmissions following primary total joint replacement and modification of patient management Samantha Tayne BA, Christen A. Merrill BS, Eric L. Smith MD, William C. Mackey MD PII: DOI: Reference:
S0883-5403(14)00391-X doi: 10.1016/j.arth.2014.05.023 YARTH 54019
To appear in:
Journal of Arthroplasty
Received date: Revised date: Accepted date:
25 February 2014 8 May 2014 22 May 2014
Please cite this article as: Tayne Samantha, Merrill Christen A., Smith Eric L., Mackey William C., Predictive risk factors for 30-day readmissions following primary total joint replacement and modification of patient management, Journal of Arthroplasty (2014), doi: 10.1016/j.arth.2014.05.023
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Title: Predictive risk factors for 30-day readmissions following primary total joint replacement
RI P
T
and modification of patient management
CE
PT
ED
MA
1. Tufts University School of Medicine Boston MA 2. Department of Orthopaedic Surgery Tufts Medical Center Boston MA 3. Department of Surgery Tufts Medical Center Boston MA
NU
SC
Samantha Tayne BA1 Christen A. Merrill BS1 Eric L. Smith MD2 William C. Mackey MD3
Please address all correspondence to:
AC
Samantha Tayne Tufts Medical Center Department of Surgery 800 Washington St. South Building, 4th Floor Boston, MA 02111 Phone: 224-628-0468 Fax: 617-636-8003 Email:
[email protected] ACCEPTED MANUSCRIPT
Abstract: Centers for Medicare and Medicaid have begun to publically publish statistics on
RI P
T
readmissions following primary total hip (THA) and total knee arthroplasty (TKA). Our current study retrospectively assesses 30-day readmissions rates following total hip and
SC
knee arthroplasties, performed by a single fellowship trained surgeon at a tertiary care medical center between 2007 and 2012. Results of a univariate analysis and logistic
NU
regression model indicated female gender, high ASA class, and increased operative time to
MA
be significantly associated with higher rates of readmission (OR 4.646, OR 1.257, and OR 5.323, respectively). Readmissions most often occurred within the first week of patient
ED
discharge. Surgical complications and gastrointestinal discomfort were the most common causes for readmission. Using readmission risk we can stratify patients into tiered critical
Introduction:
CE
PT
care pathways to reduce readmissions.
AC
The Centers for Medicare and Medicaid (CMS) began publically reporting 30-day readmission rates for individual hospitals in 2009, making readmissions a nationally recognized quality metric. Initially CMS focused on reporting readmissions data related medical conditions (myocardial infarction, heart failure, and pneumonia), and the initial financial penalties were related to these three conditions[1]. However, in the 2012 Medicare Hospital Quality Chartbook, CMS began to publish statistics on 30-day readmissions following primary total hip and knee arthroplasty, made publically available in 2013[1]. Total hip arthroplasty (THA) and total knee arthroplasty (TKA), are common elective
ACCEPTED MANUSCRIPT procedures and an obvious choice for CMS to begin analyzing readmissions and penalizing hospitals for excessive readmissions in surgical procedures.
T
Previous studies have found readmission rates around 4-4.5% following primary
RI P
TKA and THA; the reasons for readmission included wound infections, thromboembolic events, and cardiac and pulmonary related events[2-5]. Studies have looked at a variety of
SC
risk factors for readmissions in general surgery and orthopaedic surgery patients,
NU
including social factors, co-morbidities, intra-operative factors, post-operative complications, pain scores, factors relating to follow up, and length of stay. The findings in
MA
previous studies looking at these factors have been inconsistent, identifying a range of factors including age, gender, comorbidities, cancer history, and ASA class to be predictors
ED
of readmission[2, 3, 6-16]. In other studies, length of stay and discharge to rehabilitation
PT
centers or skilled nursing facilities have been associated with readmission risk[4, 5, 17].
been identified.
CE
Despite the range of variables studied, consistent predictors of readmission have not yet
AC
Surgical patients provide an opportunity to define points of intervention designed to lower readmission rates. While previous studies have elucidated possible risk factors for readmission, their inconsistent results and identification of non-modifiable risk factors for readmission have frustrated efforts at readmission prevention. We hypothesize that for THA and TKA patients there are clinical management practices that may be modified to reduce risk of readmission.
ACCEPTED MANUSCRIPT Methods: This is a retrospective case-control study of patients who underwent primary THA
T
and primary TKA at Tufts Medical Center during the years 2007 to 2012. Our sample
RI P
included readmissions that occurred within 30 days of hospital discharge. Controls were selected at random from the same surgeon’s patient population. A sample size of 160
SC
patients (40 re-admit cases and 120 controls, 1:3 matching) provided 85% power to detect
NU
an odds ratio of 2.0 for comparing the difference in a readmission risk factor between cases and controls, assuming the proportion of individuals with the risk factor in the control
MA
group was 30%. This used a two-sided test with alpha=0.05. Therefore, we performed 3:1 matching of controls to readmissions (adding additional controls provided only a marginal
ED
increase in statistical power). A large number of variables were analyzed, including
PT
descriptive demographics, multiple co-morbidities, intra-operative variables, post-
CE
operative complications, discharge information, and descriptive data on the readmissions. Once approved by the Tufts Medical Center Institutional Review Board, data were
AC
collected via chart review identifying controls and readmissions using ICD-9 codes. The outcome of interest was 30-day hospital readmission after primary total joint arthroplasty. Demographic variables included age, gender, insurance, and marital status. Comorbidities included hypertension, diabetes, coronary artery disease (CAD), chronic obstructive pulmonary disease (COPD), obstructive sleep apnea (OSA)/asthma, history of alcohol or drug abuse, and depression. Body mass index (BMI) was also recorded. Intra-operative variables included estimated blood loss, ASA class, and duration of procedure. Duration of procedure was calculated from the time anesthesia was administered to the time of extubation. Post-operative complications included tachycardia, bradycardia, fever, DVT/PE,
ACCEPTED MANUSCRIPT pneumonia, surgical site infection, and the need for transfusion, which were noted in the medical chart as complications during the hospital stay. Length of hospital stay was also
RI P
rehabilitation centers and skilled nursing facilities).
T
recorded. Discharge status included home, home with services, or inpatient care (both
The descriptive data collected on the readmission patients included interval from
SC
discharge to readmission, length of stay after readmission, and diagnosis upon
NU
readmission. Readmission diagnoses were categorized as surgical, gastrointestinal, genitourinary, nutrition, pain, cardiopulmonary, DVT/PE, and other. These classifications
MA
were established based on our preliminary research findings and literature review. Dislocations and fractures were categorized as orthopaedic diagnoses. Surgical diagnoses
ED
included dislocations and fractures, as well as surgical site infections and cellulitis.
PT
Univariate analysis was used to identify factors associated with readmission, using
CE
chi-squared or Fisher’s exact test and independent-samples T test. The factors identified through univariate analysis, along with several a priori factors, were used in a
AC
multivariable logistic regression model. A p value of 35, this did not reach statistical significance (p = 0.108). Table 2 shows the hospital and post-discharge factors. Both duration of procedure and high ASA class (class III or IV) were significantly associated with 30-day readmissions (p = 0.02 and p = 0.001, respectively). Estimated blood loss and post-operative complications (including tachycardia, bradycardia, fever, DVT/PE, pneumonia, surgical site infection, and the need for transfusion) were not associated with readmission. Neither hospital length of stay nor discharge status were associated with readmission risk. Using the factors that were found to be significantly associated with 30-day readmission in the univariate analysis and several a priori factors, including age, gender,
ACCEPTED MANUSCRIPT and insurance status, we developed a logistic regression model (Table 3). In the model, female gender as well as duration of procedure and high ASA class remained significantly
T
associated with readmission when controlling for the other factors in the model (OR 4.646
RI P
95% CI 1.818 to 11.874, OR 1.257 95% CI 1.023 to 1.545, and OR 5.323 CI 95% 1.872 to 15.132, respectively).
SC
Figure 1 shows that the majority of readmissions occurred within the first week
NU
from discharge. Gastrointestinal issues, including post-operative ileus and bowel obstruction, were among the most common readmission diagnoses. All of the readmissions
MA
for pain occurred within the first week. The next two most common reasons for readmission were wound related (surgical site infections) and cardiopulmonary issues. The
ED
cardiopulmonary issues were more likely to readmit in the second week from discharge.
PT
The average length of readmission stay was 5.39 days (SD 6.23), and the median length of
AC
Discussion:
CE
readmission stay was 4.0 days.
The objective of this study was to identify potential points within the pre- and postoperative management of primary total joint arthroplasty patients that can be modified in order to reduce 30-day readmissions and improve patient outcomes. Surprisingly, we did not find comorbidities to be significantly associated with 30-day readmissions individually or in aggregate. BMI >35 was trending toward significance (p =0.108) in the univariate analysis, but was not associated with readmission in the logistic regression model, which may be due to an interaction between BMI and ASA class. Previous studies have found increased rates of adverse events among obese surgical patients[7, 18, 19]. Perhaps these
ACCEPTED MANUSCRIPT patients would therefore benefit from pre-procedure weight loss, through diet and tailored exercises, in order to improve post-operative outcomes and return to normal daily
T
activities. Limited mobility is one of the difficulties with total joint replacement patients;
RI P
however, pre-procedure weight assessment and physical therapy consult could prove beneficial.
SC
Though comorbidities were not significantly associated with readmissions, a high
NU
ASA class of III or IV was significantly associated with 30-day readmissions in both univariate analysis and in the logistic regression model. ASA class takes into account the
MA
presence of systemic disease and the severity of the disease, and has been used for analysis in previous orthopaedic studies[3, 16, 20, 21]. The majority of the patients in our study were
ED
either ASA class II, with mild systemic disease, or ASA class III, with severe systemic
PT
disease[21]. Since many of the patients have similar comorbidities, such as diabetes and
CE
hypertension, the significance of the high ASA class may indicate that poor control of those comorbidities is the relevant factor. In future studies we can look to markers of control,
AC
such as HbA1C levels or blood pressure to determine their association with 30-day readmission. Further, we can modify how we manage patients with poorly controlled comorbidities, by altering pre- and post-operative medical management, and by delaying surgery for medical optimization[22]. Procedure duration is a proxy for procedure complexity, which includes anesthesia preparation and the actual procedure, both of which can have an effect on patient outcomes. Though duration and complexity of procedure are not modifiable, knowing that the procedure has been more complicated than the norm might alter post-operative medical management. Currently, most primary total joint arthroplasty patients follow a
ACCEPTED MANUSCRIPT similar critical pathway post-operatively[23]. However, for longer more complex procedures the pathway could be altered with attention to wound care education and pre-discharge
T
physical therapy, as well as social services assessment of safety in the home. Given the
RI P
penalties likely to be associated with readmission, prolongation of the initial length of stay by one or two days for selected high risk patients would seem appropriate and prudent.
SC
Female gender, while not a modifiable factor, was found to be a significant predictor
NU
of readmissions in both the univariate analysis and within the logistic regression model. The reason for this association is unclear. Previous studies have shown that female patients
MA
undergo more complex procedures possibly due to more severe disease from delaying surgical intervention[24-26]. In this study we did not look at the indication for joint
ED
replacement or at markers for severity of joint disease, but looking forward we may want
PT
to analyze how various diagnoses, indications, and delays in intervention affect
warranted.
CE
readmission outcomes. Overall additional analysis for the reason for the gender affect is
AC
The data collected on readmissions is best analyzed when looking at the readmission diagnosis in relation to interval from discharge. The majority of the patients are readmitted within the first two weeks after discharge, however the most common first post-operative follow-up appointment is not until four to six weeks after surgery. This period represents a timeframe outside of which most readmissions occur and outside of the 30-day readmission period as defined by CMS. While there is significant rehabilitation and physical therapy following joint arthroplasty, it may be beneficial to consider a followup appointment or phone call with a nurse practitioner or physician assistant within the
ACCEPTED MANUSCRIPT first two weeks in an attempt to address any preventable readmissions. It would also be prudent to ensure close follow up care with the patient’s primary care physician.
T
When looking at the readmission diagnoses, the most common reason for
RI P
readmission is surgical complications (including dislocation, periprosthetic fracture, or surgical site infection), and the second most common reason for readmission was
SC
gastrointestinal related issues. Interestingly, patients with inadequate pain control or with
NU
medical complications were most frequently readmitted within two weeks of discharge, while those with surgical issues were most often readmitted later. Within the first week
MA
from discharge, our analysis indicated that readmissions tend to take the form of GI related problems and pain management issues. Narcotic induced and post-operative ileuses are
ED
predictable post-operative complications that present within this first week. A future study
PT
investigating these GI complications and the effect of pain management could be insightful,
CE
and may identify a need for improved peri-operative pain management that better balances pain control with the risk of ileus, and includes non-narcotic medications.
AC
Patients with DVT/PE were found to readmit throughout the 30-day period. All patients in our study received pre-operative chemical DVT prophylaxis (Coumadin), doseadjusted Coumadin in the hospital post-operatively, mechanical DVT prophylaxis, and were discharged on pharmacologic DVT prophylaxis, following national guidelines for preventing venous thromboembolic disease[27]. Yet several patients were still readmitted for either DVT or PE. Wound integrity is similarly monitored during the hospital stay, though patients continue to be readmitted for wound complications such as surgical site infection. Anticoagulation for DVT prophylaxis may increase the incidence of wound hematomas and subsequent infection. Our findings underscore the need to continuously re-
ACCEPTED MANUSCRIPT evaluate peri-operative complications in order to find the most effective and least harmful regimen for DVT prophylaxis.
T
There are three major limitations to our study. First of all, we only studied those
RI P
patients who received primary total joint replacements at Tufts Medical Center and were readmitted to Tufts. We were not able to collect data on patients who may have been
SC
readmitted elsewhere. Secondly, since our patients were drawn only from Tufts Medical
NU
Center, we cannot determine if the factors affecting readmission are unique to our patient selection, pre-operative preparation, surgical technique, and post-operative clinical
MA
pathway. Other institutions whose arthroplasty services employ different protocols may identify different factors. Finally, our study was retrospective and therefore subject to the
ED
limitations of retrospective analysis. A follow-up prospective study could provide more
PT
granular detail on factors associated with readmission.
CE
In conclusion, we have found that female gender, ASA class, and duration of procedure predict readmission in a multivariate regression analysis model. Similar to our
AC
findings in a study of readmission following bariatric surgery, the burden of comorbidities when measured individually or cumulatively was not significantly associated with readmissions. However, ASA class and duration of procedure, as markers of comorbidity control and procedure complexity, consistently relate to readmission risk[28]. It is likely that surgical complications are related to patient selection and surgical technique and may not be sensitive to changes in post-operative management and follow up. On the other hand, medical readmissions, which predominantly occur within the first two weeks postdischarge might be diminished by modification in the clinical pathway or follow up protocols. Based on our findings we can extend the post-operative critical care pathway to
ACCEPTED MANUSCRIPT post-discharge care, and we can create branch care pathways based on readmission risk. Our finding that pain and medical complication related readmissions tend to occur earlier
T
in the post-operative recovery period than those related to surgical complications suggests
RI P
that modified care paths, improved education, and earlier outpatient follow-up for high-
MA
NU
SC
risk of readmission patients might further reduce readmission rates.
Acknowledgements:
ED
The authors would like to thank the following individuals for their statistical assistance and
AC
CE
PT
support with this project: Jessica Paulus ScD, Janis Breeze MPH, and David Tybor MS, MPH.
ACCEPTED MANUSCRIPT References:
8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18.
T
RI P
SC
NU
7.
MA
6.
ED
5.
PT
4.
CE
3.
Report to Congress: Medicare Payment Policy March 2013. Adelani M, Keeney J, Nunley R, Clohisy J, Barrack R. Readmission following total knee arthroplasty: venous thromboembolism as a "never event" is a counterproductive misnomer. J Arthroplasty 2013;28:747-50. Avram V, Petruccelli D, Winemaker M, Beer Jd. Total Joint Arthroplasty Readmission Rates and Reasons for 30-Day Hospital Readmission. J Arthroplasty 2013. Barsoum W, TG TM, Klika A, et al. Predicting patient discharge disposition after total joint arthroplasty in the United States. J Arthroplasty 2010;25:885-92. Bini S, Fithian D, Paxton L, Khatod M, Inacio M, Namba R. Does discharge disposition after primary total joint arthroplasty affect readmission rates? J Arthroplasty 2010;25:114-7. Dallal R, Trang A. Analysis of perioperative outcomes, length of hospital stay, and readmission rate after gastric bypass. Surg Endosc 2012;26:754-8. Dowsey M, Liew D, Choong P. Economic burden of obesity in primary total knee arthroplasty. Arthritis Care Res 2011;63:1375-81. Feigenbaum P, Neuwirth E, Trowbridge L, et al. Factors contributing to all-cause 30day readmissions: a structured case series across 18 hospitals. Med Care 2012;50:599-605. Fox J, Suter L, Wang K, Wang Y, Krumholz H, Ross J. Hospital-based, acute care use among patients within 30 days of discharge after coronary artery bypass surgery. Ann Thorac Surg 2013;96:96-104. Hansen D, Fox J, Gross C, Bruun J. Hospital readmissions and emergency department visits following laparoscopic and open colon resection for cancer. Dis Colon Rectum 2013;56:1053-61. Kassin M, Owen R, Perez S, et al. Risk factors for 30 day hospital readmissions among general surgery patients. J Am Coll Surg 2012;215:322-30. Lawrence J, Keller D, Samia H, et al. Discharge within 24 to 72 hours of colorectal surgery is associated with low readmission rates when using enhanced recovery pathways. . J Am Coll Surg 2013;216:390-4. Li Z, Armstrong E, Parker J, Danielsen B, Romano P. Hospital variation in readmission after coronary artery bypass surgery in California. Circ Cardiovasc Qual Outcomes 2012;5:729-37. Shulan M, Gao K, Moore C. Predicting 30-day all-cause hospital readmissions. Health Care Manag Sci 2013;16:167-75. Singh J, Kwoh C, Richardson D, Chen W, Ibrahim S. Sex and surgical outcomes and mortality after primary total knee arthroplasty: a risk-adjusted analysis. Arthritis Care Res 2013;65:1095-102. Pugely A, Callaghan J, Martin C, Cram P, Gao Y. Incidence of and risk factors for 30day readmission following elective primary total joint arthroplasty: analysis from the ACS-NSQIP. J Arthroplasty 2013;28:1499-504. Riggs R, Roberts P, Aronow H, Younan T. Joint replacement and hip fracture readmission rates: impact of discharge destination. PM R 2010;2:806-10. Reinke CE, Kelz RR, Zubizarreta JR, et al. Obesity and readmission in elderly surgical patients. Surgery 2012;152:355-62.
AC
1. 2.
ACCEPTED MANUSCRIPT
26. 27. 28.
T
RI P
SC
NU
25.
MA
24.
ED
23.
PT
21. 22.
CE
20.
Huddleston JI, Wang Y, Uquillas C, Herndon JH, Maloney WJ. Age and obesity are risk factors for adverse events after total hip arthroplasty. Clinical Orthopaedics & Related Research 2012;470:490-6. Bjorgul K, Novicoff WM, Saleh KJ. Evaluating comorbidities in total hip and knee arthroplasty: available instruments. J Orthopaed Traumatol 2010;11:203-9. ASA physical classification system American Society of Anesthesiologists 2013. Tsai T, Joynt K, Orav E, Gawande A, Jha A. Variation in surgical-readmission rates and quality of hospital care. N Engl J Med 2013;369:1134-42. Müller M, Dedes K, Dindo D, Steiner S, Hahnloser D, Clavien P. Impact of clinical pathways in surgery. Langenbecks Arch Surg 2009;394:31-9. Srikanth V, Fryer J, Zhai G, Winzenberg T, Hosmer D, Jones G. A meta-analysis of sex differences prevalence, incidence and severity of osteoarthritis. Osteoarthritis Cartilage 2005;13:769-81. O'Connor M, Hooten E. Breakout session: Gender disparities in knee osteoarthritis and TKA. Clin Orthop Relat Res 2011; 469:1883-5. Tosi L, BD BB, Boskey A. Does sex matter in musculoskeletal health? The influence of sex and gender on musculoskeletal health. J Bone Joint Surg Am 2005;87:1631-47. Jacobs JJ, Mont MA, Boggio LN, et al. Preventing venous thromboembolic disease in patients undergoing elective hip and knee arthroplasty: evidence-based guideline and evidence report. American Academy of Orthopedic Surgeons 2011.28. Tayne, S, Merrill CA, Smith EL, Mackey WC. Risk Factors for 30-day Readmissions and Modifiying Post-operative Care Following Gastric Bypass Surgery. Under Submission at Journal of the American College of Surgeons 2014.
AC
19.
ACCEPTED MANUSCRIPT
MA
ED
PT
CE
AC
T
P-value 0.726 0.386* Sig. 0.108 0.486
SC
RI P
Std Dev 14.17 7.429 % 26.0% 48.8% 51.2%
NU
Table 1: Demographic and Preoperative Data Demographic & Preoperative Readmissions Control (n = 38) (n = 129) Mean Std Dev Mean Age, y, mean (SD) 63.079 11.9986 63.969 BMI, mean (SD) 32.625 9.59001 31.1464 n % n BMI >35 15 39.0% 33 Procedure THA 21 55.3% 63 TKA 17 44.7% 66 Sex Male 10 26.3% 61 Female 28 73.7% 68 Race White 32 84.2% 102 Non-White 6 15.8% 32 Insurance Private 26 68.4% 95 Public 12 31.6% 34 Marital Status Companion 16 42.1% 61 Comorbidities (present) Diabetes 10 26.3% 36 Smoking 3 7.9% 11 COPD 6 15.8% 18 HTN 25 65.8% 82 CAD 5 13.2% 20 Depression 9 23.7% 39 OSA/Asthma 9 23.7% 43 Alcohol/Drug abuse 5 13.2% 10 Number of Comorbidities CM0 0 0.0% 1 CM1 13 34.2% 34 CM2 12 31.6% 34 CM 3+ 10 26.3% 42 *: Equal variances not assumed **: Fisher’s Exact
0.022
47.3% 52.7% 0.484
79.1% 24.8% 0.527 73.6% 26.4% 47.3%
0.573
27.9% 8.5% 14.0% 63.6% 15.5% 30.2% 33.3% 7.8%
0.847 1.000** 0.777 0.802 1.000** 0.433 0.259 0.336**
0.8% 26.4% 26.4% 32.6%
0.586** 0.344 0.527 0.465
ACCEPTED MANUSCRIPT
MA
ED
PT
CE AC
T
P-value 0.02
RI P
Std Dev 38.165
SC
1.2989 %
0.145* Sig.
37.2% 39.5% 18.6% 4.7% 52.7%
0.525 0.995 0.736 0.427** 0.001
41.9%
0.144 0.443
NU
Table 2: Intraoperative and Post-Operative Data Intraoperative Readmissions Control (n = 119) (n = 239) Mean Std Dev Mean Duration, mean 199.57 46.72 181.97 (SD) LOS, mean 5.053 5.4671 3.721 n % n EBL 1000 3 7.9% 6 ASA High (III-IV) 31 81.6% 68 Post-Operative Complications 21 55.3% 54 D/C status Home 1 2.6% 7 Home w/ Services 13 34.2% 32 SNF 24 63.2% 90 *: Equal variances not assumed **: Fisher’s Exact
5.4% 24.8% 69.8%
ACCEPTED MANUSCRIPT Table 3: Logistic Regression Model Variable B S.E. Wald
Sig.
Exp(B) .996 4.646 1.179
.015 .479 .449
.071 10.291 .134
1 1 1
.790 .001 .714
-.025 1.727 .232
.025 .533 .102
.962 10.511 5.152
1 1 1
.327 .001 .023
.976 5.621 1.261
-4.647
1.736
7.170
1
.007
.010
RI P
SC
NU
MA ED PT CE
T
-.004 1.536 .164
AC
Age Female Public Insurance BMI ASA High Duration per 20 min Interval Constant
df
95% C.I. for Exp(B) Lower Upper .996 1.029 1.818 11.874 .489 2.842 .928 1.979 1.032
1.025 15.964 1.540
CE
Legends to Figures:
PT
ED
MA
NU
SC
RI P
T
ACCEPTED MANUSCRIPT
AC
Fig. 1 Readmission Diagnosis by Days from Discharge. A: Days 0 – 7, B: Days 8 – 14, C: Days 15 – 21, D: Days 22 – 30.