Readmission Rates for One Versus Two-Midnight Length of Stay for Primary Total Knee Arthroplasty: Analysis of the Michigan Arthroplasty Registry Collaborative Quality Initiative (MARCQI) Database

Generalizability of machine learning models predicting 30-day unplanned readmission after primary total knee arthroplasty using a nationally representative database

Unplanned readmission after primary total knee arthroplasty (TKA) costs an average of US $39,000 per episode and negatively impacts patient outcomes. Although predictive machine learning (ML) models show promise for risk stratification in specific populations, existing studies do not address model generalizability. This study aimed to establish the generalizability of previous institutionally developed ML models to predict 30-day readmission following primary TKA using a national database. Data from 424,354 patients from the ACS-NSQIP database was used to develop and validate four ML models to predict 30-day readmission risk after primary TKA. Individual model performance was assessed and compared based on discrimination, accuracy, calibration, and clinical utility. Length of stay (> 2.5 days), body mass index (BMI) (> 33.21 kg/m2), and operation time (> 93 min) were important determinants of 30-day readmission. All ML models demonstrated equally good accuracy, calibration, and discriminatory ability (Brier score, ANN = RF = HGB = NEPLR = 0.03; ANN, slope = 0.90, intercept =  - 0.11; RF, slope = 0.93, intercept =  - 0.12; HGB, slope = 0.90, intercept =  - 0.12; NEPLR, slope = 0.77, intercept = 0.01; AUCANN = AUCRF = AUCHGB = AUCNEPLR = 0.78). This study validates the generalizability of four previously developed ML algorithms in predicting readmission risk in patients undergoing TKA and offers surgeons an opportunity to reduce readmissions by optimizing discharge planning, BMI, and surgical efficiency.

The Impact of Preoperative Medical Evaluation in an Orthopaedic Perioperative Medical Clinic on Total Joint Arthroplasty Outcomes: An Observational Study

BACKGROUND

A preoperative medical evaluation (PME) in total joint arthroplasty (TJA) is routine despite considerable variation and uncertainty regarding its benefits. The orthopaedic department in our academic health system established a perioperative medical clinic (PMC) to standardize perioperative management and to study the effect of this intervention on total hip arthroplasty (THA) and total knee arthroplasty (TKA) outcomes. This observational study compared the impact of a PME within 30 days prior to surgery at the PMC (Periop30) versus elsewhere ("Usual Care") on postoperative length of stay (LOS), extended LOS (i.e., a stay of >3 days), and Patient-Reported Outcomes Measurement Information System-10 (PROMIS-10) Global Physical Health (GPH) score improvement in TJA.

METHODS

We stratified adult patients (≥18 years of age) who underwent primary TJA between January 2015 and December 2020 into Periop30 or Usual Care. We utilized univariate tests (a chi-square test for categorical variables and a t test for continuous variables) to assess for differences in patient characteristics. For both TKA and THA, LOS was assessed with use of multivariable negative binomial regression models; extended LOS, with use of binary logistic regression; and PROMIS-10 GPH score, with use of mixed-effects models with random intercept and slope. Interaction terms between the focal predictor (Periop30, yes or no) and year of surgery were included in all models.

RESULTS

Periop30 comprised 82.3% of TKAs (1,911 of 2,322 ) and 73.8% of THAs (1,876 of 2,541). For THA, the Periop30 group tended to be male (p = 0.005) and had a higher body mass index (p = 0.001) than the Usual Care group. The Periop30 group had a higher rate of staged bilateral THA (10.6% versus 7.5%; p = 0.028) and a lower rate of simultaneous bilateral TKA (5.1% versus 12.2%; p < 0.001) than the Usual Care group. Periop30 was associated with a lower mean LOS for both TKA (43.46 versus 54.15 hours; p < 0.001) and THA (41.07 versus 57.94 hours; p < 0.001). The rate of extended LOS was lower in the Periop30 group than in the Usual Care group for both TKA (15% versus 26.5%; p < 0.001) and THA (13.3% versus 27.4%; p < 0.001). There was no significant difference in GPH score improvement between Periop30 and Usual Care for either TKA or THA.

CONCLUSIONS

Periop30 decreased mean LOS and the rate of extended LOS for TJA without an adverse effect on PROMIS-10 GPH scores.

LEVEL OF EVIDENCE

Therapeutic Level III . See Instructions for Authors for a complete description of levels of evidence.

Global mapping of institutional and hospital-based (Level II-IV) arthroplasty registries: a scoping review

PURPOSE

Four joint arthroplasty registries (JARs) levels exist based on the recorded data type. Level I JARs are national registries that record primary data. Hospital or institutional JARs (Level II-IV) document further data (patient-reported outcomes, demographic, radiographic). A worldwide list of Level II-IV JARs must be created to effectively assess and categorize these data.

METHODS

Our study is a systematic scoping review that followed the PRISMA guidelines and included 648 studies. Based on their publications, the study aimed to map the existing Level II-IV JARs worldwide. The secondary aim was to record their lifetime, publications' number and frequency and recognise differences with national JARs.

RESULTS

One hundred five Level II-IV JARs were identified. Forty-eight hospital-based, 45 institutional, and 12 regional JARs. Fifty JARs were found in America, 39 in Europe, nine in Asia, six in Oceania and one in Africa. They have published 485 cohorts, 91 case-series, 49 case-control, nine cross-sectional studies, eight registry protocols and six randomized trials. Most cohort studies were retrospective. Twenty-three per cent of papers studied patient-reported outcomes, 21.45% surgical complications, 13.73% postoperative clinical and 5.25% radiographic outcomes, and 11.88% were survival analyses. Forty-four JARs have published only one paper. Level I JARs primarily publish implant revision risk annual reports, while Level IV JARs collect comprehensive data to conduct retrospective cohort studies.

CONCLUSIONS

This is the first study mapping all Level II-IV JARs worldwide. Most JARs are found in Europe and America, reporting on retrospective cohorts, but only a few report on studies systematically.

Modern robotics platform associated with reduced 90-day complications in hospital setting

BACKGROUND

Early complications in total knee arthroplasty (TKA) associated with modern robotics platforms integrated with digital balancing technology have not been investigated.

OBJECTIVE

The objective was to compare 90-day complication rates between a manual technique and a modern robotic-assisted ligament balancing TKA platform.

METHODS

895 primary TKA procedures from a single surgeon were retrospectively reviewed (614 manual TKA, 281 using a modern robotics platform with an integrated digitally controlled ligament balancing device). Post-operative complications within the 90-day episode of care were recorded by the Michigan Arthroplasty Registry Collaborative Quality Initiative. Differences in complication rates between techniques were further divided into inpatient hospital, outpatient hospital, and ambulatory surgery center (ASC) cohorts.

RESULTS

In the pooled hospital cohort and inpatient hospital cohort, 'Return to OR' was significantly lower for the robotic assisted group (1.3% vs 5.2% and 0% vs 4.9%, respectively, p< 0.0370). Specifically, the need for manipulation under anesthesia (MUA) was reduced for the robotics group (0.0% vs 2.5%, p= 0.0352). No differences in overall complication rate or cause of complication was found between manual and robotics cases for the hospital outpatient and ASC cohorts (p⩾ 0.68).

CONCLUSION

Within the hospital setting, robotic assisted ligament balancing technology was associated with reduced 90-day postoperative complications for 'Return to OR' and MUA.

Preoperative Pain Catastrophizing and Neuropathic Pain Do Not Predict Length of Stay and Early Post-Operative Complications following Total Joint Arthroplasty

BACKGROUND

Both pain catastrophizing and neuropathic pain have been suggested as prospective risk factors for poor postoperative pain outcomes in total joint arthroplasty (TJA).

OBJECTIVE

We hypothesized that pain catastrophizers, as well as patients with pain characterized as neuropathic, would exhibit higher pain scores, higher early complication rates and longer lengths of stay following primary TJA.

METHODS

A prospective, observational study in a single academic institution included 100 patients with end-stage hip or knee osteoarthritis scheduled for TJA. In pre-surgery, measures of health status, socio-demographics, opioid use, neuropathic pain (PainDETECT), pain catastrophizing (PCS), pain at rest and pain during activity (WOMAC pain items) were collected. The primary outcome measure was the length of stay (LOS) and secondary measures were the discharge destinations, early postoperative complications, readmissions, visual analog scale (VAS) levels and distances walked during the hospital stay.

RESULTS

The prevalence of pain catastrophizing (PCS ≥ 30) and neuropathic pain (PainDETECT ≥ 19) was 45% and 20.4%, respectively. Preoperative PCS correlated positively with PainDETECT (rs = 0.501, p = 0.001). The WOMAC positively correlated more strongly with PCS (rs = 0.512 p = 0.01) than with PainDETECT (rs = 0.329 p = 0.038). Neither PCS nor PainDETECT correlated with the LOS. Using multivariate regression analysis, a history of chronic pain medication use was found to predict early postoperative complications (OR 38.1, p = 0.47, CI 1.047-1386.1). There were no differences in the remaining secondary outcomes.

CONCLUSIONS

Both PCS and PainDETECT were found to be poor predictors of postoperative pain, LOS and other immediate postoperative outcomes following TJA.

Length of Stay Increases 90-day Readmission Rates in Patients Undergoing Primary Total Joint Arthroplasty

Postoperative length of stay (LOS) of 1 day compared with that of 2 to 4 days after primary total knee arthroplasty (TKA) or total hip arthroplasty (THA) and 90-day readmissions.

Hospital volume-outcome relationship in total knee arthroplasty: a systematic review and dose-response meta-analysis

PURPOSE

This systematic review and dose-response meta-analysis aimed to investigate the relationship between hospital volume and outcomes for total knee arthroplasty (TKA).

METHODS

MEDLINE, Embase, CENTRAL and CINAHL were searched up to February 2020 for randomised controlled trials and cohort studies that reported TKA performed in hospitals with at least two different volumes and any associated patient-relevant outcomes. The adjusted effect estimates (odds ratios, OR) were pooled using a random-effects, linear dose-response meta-analysis. Heterogeneity was quantified using the I²-statistic. ROBINS-I and the GRADE approach were used to assess the risk of bias and the confidence in the cumulative evidence, respectively.

RESULTS

A total of 68 cohort studies with data from 1985 to 2018 were included. The risk of bias for all outcomes ranged from moderate to critical. Higher hospital volume may be associated with a lower rate of early revision ≤ 12 months (narrative synthesis of k = 7 studies, n = 301,378 patients) and is likely associated with lower mortality ≤ 3 months (OR = 0.91 per additional 50 TKAs/year, 95% confidence interval [0.87-0.95], k = 9, n = 2,638,996, I² = 51%) and readmissions ≤ 3 months (OR = 0.98 [0.97-0.99], k = 3, n = 830,381, I² = 44%). Hospital volume may not be associated with the rates of deep infections within 1-4 years, late revision (1-10 years) or adverse events ≤ 3 months. The confidence in the cumulative evidence was moderate for mortality and readmission rates; low for early revision rates; and very low for deep infection, late revision and adverse event rates.

CONCLUSION

An inverse volume-outcome relationship probably exists for some TKA outcomes, including mortality and readmissions, and may exist for early revisions. Small reductions in unfavourable outcomes may be clinically relevant at the population level, supporting centralisation of TKA to high-volume hospitals.

LEVEL OF EVIDENCE

III.

REGISTRATION NUMBER

The study was registered in the International Prospective Register of Systematic Reviews (PROSPERO CRD42019131209 available at: https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=131209 ).

Impact of Length of Stay on Readmission in Hospitalized Patients

Background Readmission and length of stay (LOS) are two hospital-level metrics commonly used to assess the performance of hospitalist groups. Healthcare systems implement strategies aimed at reducing both. It is possible that tactics aimed at improving one measure in individual patients may adversely impact the other.  Objective We sought to analyze the impact of length of stay on readmission risk in an inpatient general medical population to assess whether patients with a lower length of stays were readmitted more frequently to the hospital. Methods We performed a retrospective analysis of inpatient adult patients admitted to our institution between January 2016 and December 2019. We recorded demographic variables and the outcomes of LOS and 30-day readmission. We excluded patients who expired, left against medical advice, or were transferred to other hospitals. We performed both univariate and multivariate analyses. Results There were 91,723 patients included in the study of which 10,598 (11.6%) were readmitted. The geometric LOS for all patients was 5.37 days and was higher in readmitted patients (6.87 vs 5.18 days, respectively, p < 0.001). Patients with higher readmission rates were older, had a higher proportion of male gender, African-American ethnicity, and were more likely to have Medicare or Medicaid payors. After performing a multivariate regression analysis, we found that a high LOS was associated with a higher likelihood of readmission (P < 0.001). Conclusion Contrary to our initial hypothesis, we found that general medical patients with a higher LOS had a higher likelihood of being readmitted to the hospital after adjusting for other variables. It is possible that factors not captured in the current dataset may help explain both the increase in LOS and readmission risk.