Association of Do-Not-Resuscitate Patient Case Mix With Publicly Reported Risk-Standardized Hospital Mortality and Readmission Rates.

Importance: The Centers for Medicare and Medicaid Services's (CMS's) 30-day risk-standardized mortality rate (RSMR) and risk-standardized readmission rate (RSRR) models do not adjust for do-not-resuscitate (DNR) status of hospitalized patients and may bias Hospital Readmissions Reduction Program (HRRP) financial penalties and Overall Hospital Quality Star Ratings. Objective: To identify the association between hospital-level DNR prevalence and condition-specific 30-day RSMR and RSRR and the implications of this association for HRRP financial penalty. Design, Setting, and Participants: This cross-sectional study obtained patient-level data from the Medicare Limited Data Set Inpatient Standard Analytical File and hospital-level data from the CMS Hospital Compare website for all consecutive Medicare inpatient encounters from July 1, 2015, to June 30, 2018, in 4484 US hospitals. Hospitalized patients had a principal diagnosis of acute myocardial infarction (AMI), heart failure (HF), stroke, pneumonia, or chronic obstructive pulmonary disease (COPD). Incoming acute care transfers, discharges against medical advice, and patients coming from or discharged to hospice were among those excluded from the analysis. Exposures: Present-on-admission (POA) DNR status was defined as an International Classification of Diseases, Ninth Revision diagnosis code of V49.86 (before October 1, 2015) or as an International Statistical Classification of Diseases and Related Health Problems, Tenth Revision diagnosis code of Z66 (beginning October 1, 2015). Hospital-level prevalence of POA DNR status was calculated for each of the 5 conditions. Main Outcomes and Measures: Hospital-level 30-day RSMRs and RSRRs for 5 condition-specific cohorts (mortality cohorts: AMI, HF, stroke, pneumonia, and COPD; readmission cohorts: AMI, HF, pneumonia, and COPD) and HRRP financial penalty status (yes or no). Results: Included in the study were 4 884 237 inpatient encounters across condition-specific 30-day mortality cohorts (patient mean [SD] age, 78.8 [8.5] years; 2 608 182 women [53.4%]) and 4 450 378 inpatient encounters across condition-specific 30-day readmission cohorts (patient mean [SD] age, 78.6 [8.5] years; 2 349 799 women [52.8%]). Hospital-level median (interquartile range [IQR]) prevalence of POA DNR status in the mortality cohorts varied: 11% (7%-16%) for AMI, 13% (7%-23%) for HF, 14% (9%-22%) for stroke, 17% (9%-26%) for pneumonia, and 10% (5%-18%) for COPD. For the readmission cohorts, the hospital-level median (IQR) POA DNR prevalence was 9% (6%-15%) for AMI, 12% (6%-22%) for HF, 16% (8%-24%) for pneumonia, and 9% (4%-17%) for COPD. The 30-day RSMRs were significantly higher for hospitals in the highest quintiles vs the lowest quintiles of DNR prevalence (eg, AMI: 12.9 [95% CI, 12.8-13.1] vs 12.5 [95% CI, 12.4-12.7]; P < .001). The inverse was true among the readmission cohorts, with the highest quintiles of DNR prevalence exhibiting the lowest RSRRs (eg, AMI: 15.3 [95% CI, 15.1-15.5] vs 15.9 [95% CI, 15.7-16.0]; P < .001). A 1% absolute increase in risk-adjusted hospital-level DNR prevalence was associated with greater odds of avoiding HRRP financial penalty (odds ratio, 1.06; 95% CI, 1.04-1.08; P < .001). Conclusions and Relevance: This cross-sectional study found that the lack of adjustment in CMS 30-day RSMR and RSRR models for POA DNR status of hospitalized patients may be associated with biased readmission penalization and hospital-level performance.

Can a Convolutional Neural Network Classify Knee Osteoarthritis on Plain Radiographs as Accurately as Fellowship-Trained Knee Arthroplasty Surgeons?

BACKGROUND: Osteoarthritis (OA) is the leading cause of disability among adults in the United States. As the diagnosis is based on the accurate interpretation of knee radiographs, use of a convolutional neural network (CNN) to grade OA severity has the potential to significantly reduce variability. METHODS: Knee radiographs from consecutive patients presenting to a large academic arthroplasty practice were obtained retrospectively. These images were rated by 4 fellowship-trained knee arthroplasty surgeons using the International Knee Documentation Committee (IKDC) scoring system. The intraclass correlation coefficient (ICC) for surgeons alone and surgeons with a CNN that was trained using 4755 separate images were compared. RESULTS: Two hundred eighty-eight posteroanterior flexion knee radiographs (576 knees) were reviewed; 131 knees were removed due to poor quality or prior TKA. Each remaining knee was rated by 4 blinded surgeons for a total of 1780 human knee ratings. The ICC among the 4 surgeons for all possible IKDC grades was 0.703 (95% confidence interval [CI] 0.667-0.737). The ICC for the 4 surgeons and the trained CNN was 0.685 (95% CI 0.65-0.719). For IKDC D vs any other rating, the ICC of the 4 surgeons was 0.713 (95% CI 0.678-0.746), and the ICC of 4 surgeons and CNN was 0.697 (95% CI 0.663-0.73). CONCLUSIONS: A CNN can identify and classify knee OA as accurately as a fellowship-trained arthroplasty surgeon. This technology has the potential to reduce variability in the diagnosis and treatment of knee OA.

The Clinical and Financial Consequences of the Centers for Medicare and Medicaid Services' Two-Midnight Rule in Total Joint Arthroplasty.

BACKGROUND: To lessen the financial burden of total joint arthroplasty (TJA) and encourage shorter hospital stays, the Centers for Medicare and Medicaid Services (CMS) recently removed TKA from the inpatient-only list. This policy change now requires providers and institutions to apply the two-midnight rule (TMR) to short-stay (1-midnight) inpatient hospitalizations (SSIH). METHODS: The National Inpatient Sample from 2012 through 2016 was used to analyze trends in length of stay following elective TJA. Using publically-available policy documentation, published median Medicare payments, and National Inpatient Sample hospital costs, we analyzed the application of the TMR to SSIHs and compared the results to the previous policy environment. Specifically, we modeled 3 scenarios for all 2016 Medicare SSIHs: (1) all patients kept an extra midnight to satisfy the TMR, (2) all patients discharged as an outpatient, and (3) all patients discharged as an inpatient. RESULTS: The overall percentage of Medicare SSIHs increased significantly from 2.7% in 2012 to 17.8% in 2016 (P < .0001). Scenario 1 resulted in no change in out-of-pocket (OOP) costs to patients, no change in CMS payments, and hospital losses of $117.0 million. Scenario 2 resulted in no change in patient OOP costs, reduction in payments from CMS of $181.8 million, and hospital losses of $357.3 million. Scenario 3 resulted in no change in patient OOP costs, no change in CMS payments, and an estimated $1.71 billion of SSIH charges at risk to hospitals for audit. CONCLUSION: The results of this analysis reveal the conflict between length of stay trends following TJA and the imposition of the TMR.