Readmission to a non-index hospital following total joint replacement

Introduction

Consistent with trends observed in other developed countries,^1-3 Australia has seen a significant surge in the incidence of total hip arthroplasty (THA) and total knee arthroplasty (TKA) over the past decade.⁴ Although joint arthroplasty surgeries are considered cost-effective, their high-volume and high-cost nature lead to a significant economic burden on healthcare resources. The annual direct hospital cost for such procedures is estimated to exceed $1 billion Australian dollars.⁵ A major portion of these expenses is related to the primary procedure. However, postoperative complications that require readmission also significantly increase the overall cost.⁵ Moreover, readmissions following joint replacement are associated with poorer patient outcomes including mortality.⁶

Despite the potentially critical role of readmissions in managing complications resulting from complex surgical procedures like THA and TKA, factors that drive poor outcomes in readmitted patients remain insufficiently studied. Previous studies have found that readmission to a hospital different from the one where the initial procedure was performed — known as non-index readmission—is linked to a significantly increased risk of mortality following major surgeries.^7-9 It is hypothesized that lack of continuity of care during the post-discharge period following surgery might increase the risk of mortality because treating clinicians are unfamiliar with the patient and the details of their procedure.⁷ Only a single study of USA Medicare beneficiaries has specifically investigated this issue in joint arthroplasty patients, reporting a significant increase in risk of 90-day mortality for THA or TKA arthroplasty patients who were readmitted to non-index hospitals.⁷ There is very limited knowledge from other settings about the frequency of non-index readmissions among joint arthroplasty patients and the associated clinical ramifications.

Australia’s healthcare system caters to a geographically diverse population, and both public and private hospitals provide specialized services,¹⁰ potentially accounting for high rates of non-index readmission. Given this context, our study seeks to explore the incidence and mortality outcomes of non-index compared to index readmissions among Australian patients who have undergone THA or TKA. Additionally, we explore whether this association varies according to reason for readmission (orthopaedic vs other conditions).

Methods

Results

Between 2003 and 2022, a total of 394,248 joint arthroplasty surgeries were recorded, comprising 149,456 THA and 244,792 TKA (Table I). The mean age of these patients was 68.8 years (standard deviati0n (SD) 10.1), and a majority (55.4%) were female. Two-thirds of these patients were from major cities, and 57.5% (n = 226,668) had private health insurance. Most of the joint arthroplasty procedures (62.0%; n = 244,477) were performed in private hospitals. Compared to patients who underwent THA, those who underwent TKA were older (mean age 69.4 years vs 67.9 years) and more likely to be female (56.7% vs 53.2%) and to have certain comorbidities including diabetes, hypertension, chronic kidney disease, chronic pulmonary disease, and rheumatoid arthritis (Table I).

A total of 37,431 (9.5%) patients were readmitted within 90 days of total joint replacement. The 90-day readmission rate was higher among patients who had the procedure at a public hospital than those had the procedure at a private hospital (12.0% vs 8.0%) (Figure 1). Among patients who were readmitted within 90 days, 14,594 (39%) had orthopaedic complications. Supplementary table c lists the top 15 ICD-10 codes for orthopaedic and other conditions categories. More than half (62.8%) of the readmissions for other conditions were to non-index hospitals, compared to 39.5% (n = 14,594) of readmissions due to orthopaedic complications. The rate of non-index readmissions varied according to the type of the hospital where the initial procedure took place, particularly for patients readmitted with other conditions. Among the latter patients, those who had their primary procedure at a private hospital had the highest rate (72.6%) of non-index readmission, whereas those who had their primary procedures at a public hospital had the lowest rate (51.6%) of non-index readmission (Figure 1). Only a small proportion (3.8%) of those readmitted to a non-index hospital were transferred back to the index hospital during readmission. The median length of stay ranged from eight days (interquartile range (IQR) 6 to 14) in 2003 to four days (IQR 3 to 6) in 2022, whereas 90-day readmission rate ranged from 10.1% to 10.0% during the same period (Supplementary table iv).

Fig. 1

Flowchart showing readmission destination for patients undergoing joint arthroplasty surgery, according to index hospital type and readmission cause.

Table II provides a comparison of the characteristics of patients who were readmitted to index and non-index hospitals. Compared to patients readmitted to the index hospital, those readmitted to a non-index hospital were older (mean age 71.5 years vs 70.0 years), more likely to live in major cities (61.9% vs 58.1%), and were more likely to have THA (39.9% vs 31.8%) as the primary procedure and have private insurance (52.0% vs 40.9%). Patients readmitted to a non-index hospital also had a higher prevalence of comorbidities, such as various cardiovascular diseases, chronic kidney disease, and rheumatoid arthritis and collagen vascular disease (Table II).

Figure 2 shows the factors that were associated with non-index readmission. Among these, reason for readmission had the strongest association with non-index readmission: compared to patients who were readmitted for orthopaedic complications, those who were readmitted for other conditions were 2.5-times more likely to be readmitted to a non-index hospital (aOR 2.52; 95% CI 2.41 to 2.63). Other factors associated with non-index readmission included older age (aOR 75+ years vs < 65 years: 1.33; 95% CI 1.16 to 1.52), higher socioeconomic status (aOR least disadvantaged vs most disadvantaged: 1.21; 95% CI 1.12 to 1.31), and having private health insurance (aOR 1.59; 95% CI 1.52 to 1.66). Patients who underwent the initial procedure at a private hospital were more likely to be readmitted to a non-index hospital (aOR 1.71; 95% CI 1.65 to 1.79). Patients who underwent TKA procedures were less likely to have a non-index readmission compared to those with THA (aOR 0.71; 95% CI 0.68 to 0.74). Additionally, patients from regional/remote areas had lower odds of non-index readmission (aOR vs major cities 0.88; 95% CI 0.83 to 0.92) (Figure 2).

Fig. 2

Factors associated with non-index 90-day readmission following joint arthroplasty surgery. Odds ratios (ORs) with 95% confidence intervals (CIs) are estimated from logistic regression models with adjustment for age, sex, and comorbidities. ORs are shown as empty squares for reference groups and as black squares for other groups. Lines indicate corresponding 95% CIs.

Rates of 90-day and one-year mortality among readmitted patients were 1.4% and 3.3%, respectively (Table II). Patients who had non-index readmission had a higher crude rate of 90-day mortality than those who had index readmission (1.8% vs 0.9%) (Table II). After accounting for potential confounders using logistic models with IPW, non-index readmission was significantly associated with higher 90 day mortality (aOR 1.69; 95% CI 1.39 to 2.05) (Figure 3). Analysis of one-year mortality also showed a positive association between non-index readmission and mortality (aOR 1.31; 95% CI 1.16 to 1.47). When stratified by readmission cause, we found that the associations between non-index readmission and mortality were similar for patients readmitted with orthopaedic complications and other conditions (90-day mortality aOR 1.61; 95% CI 0.98 to 2.64, and aOR 1.67; 95% CI 1.35 to 2.06, respectively; p-heterogeneity = 0.64). Similar increases in odds of one-year mortality were observed for readmissions due to orthopaedic complications (aOR 1.40; 95% CI 1.06 to 1.84) and other conditions (aOR 1.27; 95% CI 1.11 to 1.45) (p-heterogeneity = 0.43).

Fig. 3

Associations of non-index readmission with 90-day and one-year mortality, overall and by cause of readmission. *Models were adjusted for age, sex, procedure type, socioeconomic status, regionality, private insurance, readmission cause, comorbidities, and hospital funding type, as appropriate. Odds ratios are shown as empty squares for reference groups and as black squares for other groups. Lines indicate corresponding 95% confidence intervals.

Discussion

In this population-wide study, we found that 9.5% of patients were readmitted within 90 days following total joint replacement surgery and among these, 53.7% were readmitted to a hospital other than the operating hospital (i.e. non-index readmission). Non-index readmission was more prevalent where the readmission was due to conditions other than orthopaedic complications, which constituted 61.0% of all readmissions. Factors associated with non-index readmission included older age, higher socioeconomic status, private index hospital, private health insurance, and residence in a rural or remote area. Non-index readmission following joint replacement was associated with a significant increase in the odds of 90-day and one-year all-cause mortality, after adjusting for observed confounders. When stratified by reason for readmission (orthopaedic complications vs other conditions), similar increases in odds of mortality were observed, although this did not attain statistical significance for 90-day mortality in patients readmitted for orthopaedic complications.

Our 90-day readmission rates following THA (9.1%) and TKA (9.8%) align with those of previous studies and with a systematic review,^7,18-20 which reported pooled 90-day readmission rates of 7.7% for THA and 9.7% for TKA. A recent study using data from NSW public hospitals between 2015 and 2018, reported a 60-day readmission rate of 11.9% for TKA and 10.6% for THA.¹⁸ The lower rates found in our study likely reflect its coverage of both public and private hospitals. Only 38.0% of all joint arthroplasty procedures were performed in public hospitals, and these patients had a much higher 90-day readmission rate (12.0%) compared to those who had procedures in private hospitals (8.0%). Additional analyses indicate that patients from private hospitals generally had fewer comorbidities compared to those from public hospitals (Supplementary table v). However, a more significant finding was the difference in median length of stay: five days (IQR 4 to 8) for public hospitals and nine days (IQR 5 to 16) for private hospitals. This extended period of inpatient care in private hospitals may facilitate more comprehensive postoperative management, potentially leading to fewer readmissions.

We reported that more than half (53.7%) of all readmissions following total joint arthroplasty were to a hospital different from the one where the procedure was performed. Our rate of 90 day non-index readmission for patients who had their procedure in a public hospital (42.2%) was higher than the rate of 60-day non-index readmission reported by the recent NSW study that included public hospitals only (36.0%).¹⁸ Few international studies have delved into the rate of non-index readmission following joint arthroplasty surgery.^7,21 The USA study of Medicare beneficiaries reported non-index readmission rates of 24% and 26.5% following THA and TKA, respectively.⁷ The much higher rates we observed can likely be explained Australia’s dual system of public and private hospitals serving distinct roles.^22,23 Private hospitals in Australia largely focus on elective procedures and most do not have emergency departments. Therefore, patients who undergo joint arthroplasty procedures at a private hospital are very unlikely to have an emergency readmission to the same private hospital.

In our study, higher socioeconomic status and having private insurance were both associated with higher rates of non-index readmission. Patients with a higher socioeconomic status are more likely to have private insurance and therefore to undergo a joint arthroplasty procedure at a private hospital or as a privately-funded patient in a public hospital. However, Brooke et al⁷ reported that the proportion of publicly-funded USA Medicaid joint arthroplasty patients was similar between index and non-index readmissions, which may reflect structural differences between the USA and Australian health systems. Patients with a THA were more likely to be admitted to a non-index hospital than those with TKA. This likely reflects the acute nature of THA complications which include dislocation and periprosthetic fracture, which would be treated at the nearest hospital that has an emergency department.²⁰

Our research found that patients readmitted to a non-index hospital experienced higher odds of mortality within both 90-day and one-year periods. This aligns with previous studies,^7,21 which also reported an increased risk of mortality following major operations, including THA and TKA, when patients were readmitted to non-index hospitals. However, only Brooke et al⁷ has previously reported a joint replacement-specific association between non-index readmission and 90 day mortality. Their 90-day crude mortality rates (index: 4.8%; non-index: 5.5%) were much higher than ours (index: 0.9%; non-index: 1.8%), likely reflecting their older patient population (mean age 74.5 years vs. 68.8 years). However, their IPW odds ratio for 90 day mortality (aOR 1.27; 95% CI 1.19 to 1.33), was comparable with, although lower than, ours (aOR 1.69; 95% CI 1.39 to 2.05).

Brooke et al⁷ also compared outcomes of non-index readmission according to whether the readmission was coded to a surgical or a medical diagnosis related group (DRG). They reported that patients coded to surgical DRGs were less likely to readmit to non-index hospitals, and that the mortality disadvantage associated with non-index readmission was greater for these patients than for those who were coded to medical DRGs. Our study was the first to explore how non-index readmission in joint arthroplasty patients was influenced by whether the readmission was for specific orthopaedic complications, or for other conditions. We found that patients readmitted to a non-index hospital for orthopaedic complications had higher odds of 90-day and one-year mortality than counterparts who were readmitted to the index hospital, although the former association did not attain statistical significance. Returning to the index hospital could offer several advantages for patients with orthopaedic complications, as they could be managed by the original surgical team who are already familiar with the intricacies of the patient’s procedure, including specific techniques, instruments, and challenges encountered.^7,21 A prior study demonstrated that the advantages of being readmitted to an index hospital persisted, irrespective of whether the same physicians were involved in the clinical management.²⁴

We found that the majority (61.0%) of all readmissions within 90 days of a joint arthroplasty procedure were due to conditions other than orthopaedic complications, and that the majority (62.8%) of these patients were admitted to non-index hospitals. Compared to patients with orthopaedic complications, patients who were readmitted for other conditions were 2.5-times more likely to be readmitted to a non-index hospital, and also had much higher 90-day and one-year mortality rates. As many of these patients presented with acute cardiovascular conditions or infections (see Supplementary table iii), which might require urgent care, it could potentially be beneficial for them to seek care at a hospital closer to their residence. Nonetheless, these patients experienced elevated odds of 30-day and obe-year mortality compared with counterparts who were readmitted to the index hospital, even after adjustment using IPWs for potential confounders. In contrast to the findings of Brooke et al¹⁰ comparing patients coded to surgical and medical DRGs, the magnitude of our IPW odds ratios was similar for patients admitted with orthopaedic complications and other conditions. Our findings suggest that the benefits of returning to the operating hospital extend beyond familiarity with the details of the surgical procedure, and perhaps reflect other positive aspects of continuity of care.

Findings from our study using routinely collected data, must be interpreted within the context of certain limitations. The APDC dataset, an administrative database, may be influenced by disparities in coding practices among clinical coders and healthcare institutions.²⁵ Although we used multivariable modelling incorporating propensity matching to control for a wide range of patient-level and hospital-level confounders, there remains a potential for residual confounding due to unmeasured variables that are linked with mortality. We could not investigate the role of provider continuity because provider details were not available in our datasets. Despite these limitations, we used a large, population-wide dataset from Australia’s largest state and employed best-practice statistical analysis to minimize bias and enhance the generalizability of our results.

Over half of the patients in this cohort, who needed readmission within 90 days post-THA and TKA, were readmitted to a different hospital. Factors influencing readmission location included the reason for readmission and the hospital performing the initial surgery. After controlling for confounding, readmission to a non-index hospital linked to higher mortality risk. These findings underscore the need for improved care coordination and clinical integration after joint arthroplasty to boost patient outcomes.