Risk-adjusted Models Research Articles

P-1807. Using The Modified Delphi Method to Create Expert Consensus for Candidate Variables for Risk-Adjusted Benchmarking Metrics of Hospital Antimicrobial Consumption

Abstract Background Studies have shown that both patient- and hospital-level factors can influence the evaluation of antimicrobial consumption, and appropriate risk adjustments are necessary for reliable benchmarking. The curation of candidate variables through experts can be helpful in gaining the trust of users. We conducted a modified two-stage Delphi method to build consensus on candidate variables that should be considered in developing a risk-adjustment model for benchmarking metrics of hospital antimicrobial consumption. Summary of Expert Consensus for Candidate Variables for Risk-Adjusted Benchmarking Metrics Methods In preparation, we listed potential variables for consideration from past literature (8 facility characteristics, 6 patient demographics, 4 care processes, 56 comorbidities, 42 procedure categories). We distributed a survey to a 9-member expert panel to rate variables independently to “can be considered” or “should not be included” with comments. Consensus was defined as agreement by 7 or more out of 9 members, and the results with comments were presented to the panel at a conference call for further discussions. The second survey was distributed, focusing on items that did not reach consensus or the panel suggested reconsideration. Results At the first survey, 8 facility characteristics, 3 patient demographics, 4 care processes, 27 comorbidities, and 19 procedure categories reached a consensus. The panel discussed the remaining 29 comorbidities and 23 procedure categories and whether any item that reached consensus required reconsideration on a follow-up conference call. In the second survey, 3 facility characteristics, 2 patient demographics, 27 comorbidity categories, and 4 procedure categories were included in the second survey based on the discussion. At the end of the process, 7 facility characteristics, 6 patient demographics, 4 care processes, 49 comorbidities, and 41 procedure categories reached a consensus (Figure). Conclusion In this modified two-stage Delphi process, we successfully classified the majority of variables as either to be considered or not to be included in statistical model development. These variables, curated by subject matter experts, can serve as a foundation for further model development for objective evaluation and benchmarking of hospital antimicrobial consumption. Disclosures Michihiko Goto, MD MSCI, Merck: Grant/Research Support Daniel J. Livorsi, MD, Merck: Grant/Research Support

Ranking Surgeon Performance After ACL Reconstruction Using the Knee Injury and Osteoarthritis Outcome Score (KOOS) Subscales.

Patient-reported outcomes (PROs) are considered the gold standard for evaluating value-based care in orthopaedics. However, there is little evidence to guide implementation of PROs for surgeon performance evaluation. To develop a risk-adjusted surgeon performance measure using the Knee injury and Osteoarthritis Outcome Score (KOOS) for patients undergoing anterior cruciate ligament reconstruction (ACLR). Cross-sectional study; Level of evidence, 3. Patients (N = 1248; 662 men; mean age, 30 ± 13 years) who underwent ACLR performed by 40 surgeons between 2010 and 2018 were identified from a large, nationally representative sports medicine clinical data registry. Linear regression was used to predict change scores for each KOOS subscale (Pain, Symptoms, Activities of Daily Living [ADL], Function in Sports and Recreation, and Knee-Related Quality of Life) while adjusting for patient baseline characteristics. A risk-adjusted performance measure was calculated for each KOOS subscale as the difference between the unadjusted and the risk-adjusted predicted change score across all patients treated by a single surgeon. Surgeon-relative quartile ranking was compared across outcome subscale scores. One-third of the patients (34%) displayed acute cartilage damage, and 56% had a meniscal injury. In the risk adjustment models, older age, presence of diabetes, current smoking status, acute cartilage damage, concurrent cartilage treatment, lower baseline Veterans RAND 12-Item Health Survey mental and physical component scores, and lower baseline Marx and KOOS subscale values all had a significant negative influence on the predicted KOOS subscale change values (P < .05 for all). Surgeon performance, ranked in quartile groups, was the same for 10 surgeons but varied by 1 to 2 quartiles for the other 30 surgeons across the different KOOS subscales. These results showed that surgeon performance varies widely when evaluated using different KOOS subscales for patients undergoing ACLR. Based on the preliminary results and clinical perspective, the authors recommend the ADL and Symptoms subscales as the best options to differentiate surgeon performance for these patients. However, evaluation of surgeon performance may require consideration or use of a set of PROs or the development of a single index PRO that is sensitive to the range of outcome dimensions important to patients.

Development and Application of New Risk-Adjustment Models to Improve the Current Model for Hospital Standardized Mortality Ratio in South Korea

Development and Application of New Risk-Adjustment Models to Improve the Current Model for Hospital Standardized Mortality Ratio in South Korea

Private-Sector Readmissions for Inpatient Surgery in Veterans Health Administration Hospitals

The Veterans Health Administration (VHA) reports multiple indicators of hospital surgical performance, including hospital risk-standardized 30-day readmission rates (RSRRs). Currently, most routinely reported measures do not include readmissions that occur outside VHA hospitals. The impact of readmissions outside the VHA on hospital RSRR is not known. To measure the impact of including non-VHA readmissions on VHA hospital performance rankings for 30-day readmission. This retrospective cohort study included patients aged at least 65 years from 2013 to 2019 from the Veterans Affairs Surgical Quality Improvement Program linked to patient-level data from the VHA and Medicare. Data were limited to patients with VHA and Medicare enrollment during the year prior to surgery. Data were analyzed from November 2023 through July 2024. The main outcome was readmissions to acute care VHA or non-VHA hospitals within 30 days of discharge. VHA hospital-level RSRRs were estimated using separate generalized linear mixed-effects risk adjustment models that alternatively included VHA-only or VHA plus non-VHA readmissions. VHA hospitals were then stratified into quintiles based on RSRRs derived using VHA-only or VHA plus non-VHA readmissions. Changes in hospital performance quintiles with the addition of non-VHA readmissions were calculated, and characteristics of VHA hospitals most impacted by including non-VHA readmissions were evaluated. The eligible cohort included 108 265 patients (mean [SD] age, 72.2 [6.5] years; 105 661 [97.6%] male) who underwent surgery in 104 VHA hospitals. The combined readmission rate was 14.0%. The proportion of readmissions occurring outside the VHA ranged from 0% to 55.3% across the 104 VHA hospitals (median, 20.9%). Using VHA and non-VHA readmissions, 24 VHA hospitals (23.1%) improved performance and 23 hospitals (22.1%) worsened performance, defined as a decrease or increase, respectively, of 1 or more RSRR quintiles. Improvements in hospital performance rank were associated with larger surgical volume (-7.48; 95% CI, -11.33 to 03.64; P < .001), urban location, greater surgical complexity (-9.86; 95% CI, -16.61 to -3.11; P = .005), and lower proportion of readmissions outside the VHA (-8.15; 95% CI, -12.75 to -3.55; P < .001). In this cohort study, VHA hospitals whose readmission performance metric improved by including non-VHA readmissions had higher patient volume, higher complexity, and lower proportion of care outside the VHA. Thus, improving continuity of care may have a paradoxical effect of worsening VHA performance metrics.

Equity and efficiency effects of flat premiums.

This paper analyzes the impact of flat premiums on equity and efficiency within a regulated market. We examine the consequences of shifting from a risk-adjusted premium model to a flat premium system, particularly focusing on how this shift affects different income groups and market efficiency. Using a combination of theoretical modeling and empirical analysis, we find that flat premiums may lead to increased cross-subsidization among participants, with notable effects on both equity and efficiency. Our results suggest that while flat premiums simplify the regulatory framework, they also introduce trade-offs that policymakers must carefully consider.

Incorporating clinical and demographic data into the Elixhauser Comorbidity Model: deriving and validating an enhanced model in a tertiary hospital’s internal medicine department

Background and objectivesThe Elixhauser Comorbidity Model is a prominent, freely-available risk adjustment model which performs well in predicting outcomes of inpatient care. However, because it relies solely on diagnosis codes, it may not capture the full extent of patient complexity. Our objective was to enhance and validatethe Elixhauser Model by incorporating additional clinical and demographic data to improve the accuracy of outcome prediction.MethodsThis retrospective observational cohort study included 55,945 admissions to the internal medicine service of a large tertiary care hospital in Jerusalem. A model was derived and validated to predict four primary outcomes. The four primary outcomes measured were length of stay (LOS), in-hospital mortality, readmission within 30 days, and increased care.ResultsInitially, the Elixhauser Model was applied using standard Elixhauser definitions based on diagnosis codes. Subsequently, clinical variables such as laboratory test results, vital signs, and demographic information were added to the model. The expanded models demonstrated improved prediction compared to the baseline model. For example, the R2 for log LOS improved from 0.101 to 0.281 and the c-statistic to predict in-hospital mortality improved from 0.711 to 0.879.ConclusionsAdding readily available clinical and demographic data to the base Elixhauser model improves outcome prediction by a considerable margin. This enhanced model provides a more comprehensive representation of patients’ health status. It could be utilized to support decisions regarding admission and to what setting, determine suitability for home hospitalization, and facilitate differential payment adjustments based on patient complexity.

Supplementing risk adjustment with high‐risk pooling using historical data for identifying the high risks

AbstractMany regulated health insurance markets with community‐rated premiums rely on risk adjustment (RA) to mitigate insurer‐incentives to risk select. However, insurers remain typically undercompensated for chronically ill enrollees. We use historical data on health spending and risk adjuster information to identify individuals undercompensated by the Dutch RA model of 2021 and find a selective group (1% of the population) with an average annual undercompensation of €6,050. We supplement the RA model with a risk sharing modality called high‐risk pooling (HRP) to organize residual‐based compensations towards insurers for the identified group to reduce the mean undercompensation to zero. The effects are evaluated on subgroups defined by chronic disease, finding a 42% reduction of their average undercompensation. Therefore, through compensating 1% of the population, the insurer‐incentives to select against chronically ill individuals substantially diminish. These results are compared to outlier‐risk sharing (reinsurance), proving HRP to be more effective at reducing selection incentives.

Donor and Recipient Factors Associated with Primary Graft Dysfunction Following Lung Transplantation: A DMG Registry Analysis

Current risk-adjusted models to predict primary graft dysfunction (PGD) following lung transplantation (LTx) do not include bedside donor critical care data. Donor management goals (DMG) represent predefined critical care endpoints aimed at optimizing multi-organ donor management. Here, we sought to identify novel predictors to better understand the relationship between donor management and PGD following LTx. We used the national DMG registry to identify a cohort of LTx recipients linked to their respective donors between January 1st, 2015 and March 1st, 2023. Grade 3 PGD (PGD3) was defined according to modified ISHLT criteria. Multivariable modeling was performed to identify risk factors for the development of PGD3. A total of 2704 eligible patients were identified of which 643 (23.8%) developed PGD3. After multivariable modeling, the likelihood of PGD3 was greater with increased donor age (OR 1.06 [1.02, 1.10] per 5 year change, p=0.003), increased donor serum pH at the time of authorization (OR 1.14 [1.02, 1.25] per 0.1 increase, p=0.016), donor history of cocaine use (OR 1.34 [1.05, 1.71], p=0.020), and increased recipient central venous pressure (1.03 [1.01, 1.06], p=0.005). Recipients who received donor lungs in which the DMG for PF ratio was met had a lower likelihood of developing PGD3 (OR 0.63 [0.46, 0.86], p=0.006). This study leverages a novel detailed donor management database to identify factors associated with the development of PGD3. These factors may be used to recognize donors and recipients that may benefit from early interventions to improve short-term outcomes.

Associations of smoking status with long-term cardiovascular events in different arterial territories: the EPIC Norfolk prospective population cohort

Abstract Background Smoking cessation is associated with a reduced risk of cardiovascular disease (CVD) events, mainly in ischemic heart disease (IHD) and stroke. However, the impact of smoking status (never, former or current smoking) on long-term CVD events across different arterial territories remains understudied. Knowledge of such associations is important, as it may assist in risk communication and smoking cessation motivation. We therefore investigated associations between smoking status and territory-specific CVD events in a large cohort with ≥20 years follow-up. Methods We included participants without baseline CVD from the European Prospective Investigation of Cancer (EPIC) Norfolk prospective population cohort. The strengths of associations between smoking, former smoking and never smoking with the first occurrence of a CVD event were evaluated using a competing risk-adjusted regression model. CVD events were defined as hospitalisation or death due to IHD, ischemic stroke, haemorrhagic stroke, peripheral arterial disease (PAD) or aortic aneurysm (AA). Patients were followed-up until the first event, non-CVD death (competing risk) or end of study (15-03-2018). The model was adjusted for sex, age, systolic blood pressure, low-density lipoprotein cholesterol (LDL-c), high-density lipoprotein cholesterol (HDL-c), diabetes, renal function and body mass index. We compared former and current smoking (at baseline) with never smoking. Results We included 23,581 participants (56% women) with a median baseline age of 58 years [interquartile range (IQR) 51-66]. During median follow up of 21.3 years (IQR 19.0-22.8), 26.4% experienced ≥1 CVD event. The first event of occurrence was IHD in 17.3%, ischemic stroke in 3.6%, haemorrhagic stroke in 1.4%, PAD in 3.1% and AA in 1.4%. Current (vs. never) smoking was associated with IHD [adjusted hazard rate (aHR) 1.53, 95% confidence interval (CI) 1.36-1.72], PAD (aHR 1.29, 95%CI 1.07-1.56) and AA (aHR 4.81, 95%CI 3.54-6.54). Current smoking was not significantly associated with ischemic stroke (aHR 1.20, 95% CI 0.95-1.52) and haemorrhagic stroke (aHR 1.18, 95% CI 0.81-1.70). In former (vs. never) smoking individuals, the association with AA remained statistically significant but less pronounced [aHR 1.69, 95%CI 1.29-2.23]. Conversely, former smoking was not associated with IHD (aHR 1.07, 95%CI 0.99-1.16) or PAD (aHR 1.09, 95%CI 0.95-1.26) (Figure 1). Conclusion After 20-years of follow-up, former smokers, compared to never-smokers, remain at considerably higher risk for CVD events, particularly for aortic aneurysm. This knowledge may assist in treatment decisions and risk communication.

SCORE2-ASIA risk prediction algorithms: revised models to estimate 10-year risk of cardiovascular disease in the Asia Pacific Region

Abstract Background Most current cardiovascular disease (CVD) prediction tools have been developed and validated solely in Western countries, while there are significant differences between Asian and Western populations in risk factor distributions and CVD incidence patterns. This study outlines the recalibration and validation of the SCORE2-ASIA model to estimate 10-year fatal and non-fatal CVD risk among adults aged 40-69 years with no previous CVD or diabetes in the Asia Pacific Region. Methods The sex-specific and competing risk-adjusted SCORE2 models were systematically recalibrated to reflect CVD incidence in the Asian clinical practice. We defined four risk regions in Asia according to country-specific CVD mortality rates (low risk (&amp;lt;100 CVD deaths per 100,000), moderate risk (100 to &amp;lt;150 CVD deaths per 100,000), high risk (150 to &amp;lt;300 CVD deaths per 100,000), and very high risk (≥300 CVD deaths per 100,000)). Models were recalibrated using region representative data on expected incidences and risk factor distributions, utilising the methodological approaches developed in recalibration of the SCORE2 models for Europe. The SCORE-ASIA model was externally validated in all risk regions using population-based cohort data from independents cohort data. To ensure the European coefficients well applied to Asian populations, the SCORE2 model derivation process was repeated in external validation cohorts. Results Region-specific incidence was estimated using country-specific data on CVD mortality and data on the ratio of fatal to total events on 9,496,212 individuals from 5 Asian countries (496,212 CVD events). For external validation, we analysed data from 12 additional cohorts in 13 Asian countries (4,775,976 individuals, 176,011 CVD events). After applying the derived risk prediction models to external validation cohorts, the pooled external validation C-index was 0.709 (95%CI 0.676-0.743). In the nationally representative CDMD cohort from Singapore (population completely separate from recalibration data) there was adequate agreement between predicted and observed risks. Risk factor coefficients derived in Asian populations were largely similar to the original SCORE2 coefficients, indicating that the original SCORE2 coefficients apply well to the Asian population. Predicted CVD risk varied several-fold across Asian risk regions. For example, the estimated 10-year CVD risk for a 50-year-old non-smoker, with a systolic blood pressure of 140mmHg, total cholesterol of 5.5mmol/L, and HDL-cholesterol of 1.3mmol/L, ranged from 12% for men in low-risk countries to 26% for men in very-high-risk countries, and from 7% for women in low-risk countries to 26% for women in very-high-risk countries. Conclusions The SCORE2-ASIA model accurately predicts 10-year risk of CVD for apparently healthy people in Asian countries after systematic recalibration, thereby enhancing the identification of individuals at higher risk of developing CVD across the Asia Pacific Region.

Heterogeneous associations of traditional atherosclerotic risk factors and long-term events in different arterial territories: the EPIC Norfolk prospective population cohort

Abstract Background Atherosclerotic risk factors have been associated with cardiovascular disease (CVD) events in specific arterial territories. However, these observations are limited by relatively short-term follow-up (i.e. ten years) and a focus on a limited number of outcome events. We therefore investigated associations between atherosclerotic risk factors and territory-specific cardiovascular events in a large cohort with ≥20 years follow-up. Methods We included participants without baseline CVD from the European Prospective Investigation of Cancer (EPIC) Norfolk prospective population cohort. The strengths of associations between CVD risk factors (smoking, low-density lipoprotein cholesterol (LDL-c), and systolic blood pressure (SBP)) and the first occurrence of a CVD event were evaluated using a competing risk-adjusted regression model. CVD events were defined as hospitalisations or death due to ischemic heart disease (IHD), ischemic stroke, haemorrhagic stroke, peripheral arterial disease (PAD) or aortic aneurysm (AA). Patients were followed-up until the first event, non-CVD death (competing risk) or end of study (15-03-2018). The model was adjusted for sex, age, diabetes, HDL-c, kidney function and body mass index. LDL-c and SBP were analysed by comparing highest versus lowest quartiles. Smoking (at baseline) was compared to never smoking. Results We included 23,581 participants (56% women) with a median baseline age of 58 years [interquartile range (IQR) 51-66]. During median follow up of 21.3 years (IQR 19.0-22.8), 26.4% of individuals experienced ≥1 CVD event. The first event of occurrence was IHD in 17.3%, ischemic stroke in 3.6%, haemorrhagic stroke in 1.4%, PAD in 3.1% and AA in 1.4%. LDL-c was strongest associated with IHD [adjusted hazard rate (aHR) 1.7, 95% confidence interval (CI) 1.5-1.9], SBP with ischemic stroke (aHR 1.9, 95%CI 1.5-2.5) and smoking with AA (aHR 4.8, 95%CI 3.5-6.5) (Figure 1). Conclusions There is considerable heterogeneity in the associations between atherosclerotic risk factors and long-term, first, territory-specific cardiovascular events. Knowledge of these differences may assist in personalised treatment decisions and risk communication.

Effect of disability, homelessness, and neighborhood marginalization on risk-adjustment for hospital performance measurement.

Background It is not known how disability, homelessness, or neighborhood marginalization influence risk-adjusted hospital performance measurement in a universal health care system. Methods We evaluated the effect of including these equity-related factors in risk-adjustment models for in-hospital mortality, and 7- and 30-day readmission in 28 hospitals in Ontario, Canada. We compared risk-adjustment with commonly-used clinical factors to models that also included homelessness, disability, and neighborhood indices of marginalization. We evaluated models in historical data using internal-external cross-validation. We calculated risk-standardized outcome rates for each hospital in a recent reporting period using mixed-effects logistic regression. Results The cohort included 544,805 admissions. Adjustment for disability, homelessness, and neighborhood marginalization had little impact on discrimination or calibration of risk-adjustment models. However, it influenced comparative hospital performance on risk-standardized 30-day readmission rates, resulting in 5 hospitals being reclassified between below-average, average, and above-average groups. No hospitals were reclassified for mortality and 7-day readmission. Conclusion In a system with universally insured hospital services, adjustment for disability, homelessness, and neighborhood marginalization influenced estimates of hospital performance for 30-day readmission but not 7-day readmission or in-hospital mortality. These findings can inform researchers and policymakers as they thoughtfully consider when to adjust for these factors in hospital performance measurement.

Investigating the Impact of Additional Congenital Malformations on Speech Outcomes at age Five in Children with a Cleft Palate.

To determine whether children born with a cleft palate ± lip (CP ± L) and additional congenital differences (ACDs - including 'Congenital malformations and deformations' as coded in ICD-10), are less likely to meet the three national speech outcome standards at age five compared to children with CP ± L and no ACDs. An observational study, utilizing national data from the UK Cleft Registry and Audit NEtwork (CRANE) Database linked to national administrative data of hospital admissions. National Health Service, England. 2191 children (993 female, 1198 male) with CP ± L, born 2006-2012 in England, with complete speech data and consent to data linkage. Perceptual speech analysis utilised the Cleft Audit Protocol for Speech - Augmented (CAPS-A) rating and United Kingdom National Speech Outcome Standards (Speech Standard 1 (SS1), Speech Standard 2a (SS2a) and Speech Standard 3 (SS3)). Of 2191 children, 759 (35%) had at least one ACD. Presence of one ACD did not significantly impact speech outcomes but two or more ACDs reduced the odds of achieving all three speech standards: SS1 aOR 0.602 (CI 0.45-0.82, P = .002), SS2a aOR 0.563 (CI 0.41-0.77, P = .001), SS3 aOR 0.606 (0.43-0.84, P = .003). When exploring ACDs by ICD-10 groupings, congenital malformations of the 'Eye, ear, face and neck', 'Circulatory system', 'Digestive system', 'Musculoskeletal system' and 'Other congenital malformations' reduced a child's odds of achieving the speech standards. ACDs, in the absence of a known syndrome, have a significant impact on speech outcome at age five. Incorporating these factors into risk-adjustment models for service level outcome reporting is recommended.

Social Determinants of Health and Risk-Adjusted Sepsis Mortality in the Nationwide Veterans Affairs Healthcare System.

Traditional risk prediction and risk adjustment models have focused on clinical characteristics, but accounting for social determinants of health (SDOH) and complex health conditions could improve understanding of sepsis outcomes and our ability to predict outcomes, treat patients, and assess quality of care. To evaluate the impact of SDOH and health scales in sepsis mortality risk prediction and hospital performance assessment. Observational cohort study. One hundred twenty-nine hospitals in the nationwide Veterans Affairs (VA) Healthcare System between 2017 and 2021. Veterans admitted through emergency departments with community-acquired sepsis. Individual- and community-level SDOH (race, housing instability, marital status, Area Deprivation Index [ADI], and rural residence) and two health scales (the Care Assessment Need [CAN] score and Claims-Based Frailty Index [CFI]). The primary outcome was 90-day mortality from emergency department arrival; secondary outcomes included 30-day mortality and in-hospital mortality. Among 144,889 patients admitted to the hospital with community-acquired sepsis, 139,080 were men (96.0%), median (IQR) age was 71 (64-77) years, and median (IQR) ADI was 60 (38-81). Multivariable regression models had good calibration and discrimination across models that adjusted for different sets of variables (e.g., AUROC, 0.782; Brier score, 1.33; and standardized mortality rate, 1.00). Risk-adjusted hospital performance was similar across all models. Among 129 VA hospitals, three hospitals shifted from the lowest or highest quintile of performance when comparing models that excluded SDOH to models that adjusted for all variables. Models that adjusted for ADI reported odds ratios (CI) of 1.00 (1.00-1.00), indicating that ADI does not significantly predict sepsis mortality in this cohort of patients. In patients with community-acquired sepsis, adjusting for community SDOH variables such as ADI did not improve 90-day sepsis mortality predictions in mortality models and did not substantively alter hospital performance within the VA Healthcare System. Understanding the role of SDOH in risk prediction and risk adjustment models is vital because it could prevent hospitals from being negatively evaluated for treating less advantaged patients. However, we found that in VA hospitals, the potential impact of SDOH on 90-day sepsis mortality was minimal.

Importance of risk adjusting central line-associated bloodstream infection rates in children.

Central line-associated bloodstream infection (CLABSI) is one of the most prevalent pediatric healthcare-associated infections and is used to benchmark hospital performance. Pediatric patients have increased in acuity and complexity over time. Existing approaches to risk adjustment do not control for individual patient characteristics, which are strong predictors of CLABSI risk and vary over time. Our objective was to develop a risk adjustment model for CLABSI in hospitalized children and compare observed to expected rates over time. We conducted a prospective cohort study using electronic health record data at a quaternary Children's Hospital. We included hospitalized children with central catheters. Risk factors identified from published literature were considered for inclusion in multivariable modeling based on association with CLABSI risk in bivariable analysis and expert input. We calculated observed and expected (risk model-adjusted) annual CLABSI rates. Among 16,411 patients with 520,209 line days, 633 patients experienced 796 CLABSIs. The final model included age, behavioral health condition, non-English speaking, oncology service, port catheter type, catheter dwell time, lymphatic condition, total parenteral nutrition, and number of organ systems requiring ICU level care. For every organ system receiving ICU level care the odds ratio for CLABSI was 1.24 (95% CI 1.12-1.37). Although not statistically different, observed rates were lower than expected rates for later years. Failure to adjust for patient factors, particularly acuity and complexity of disease, may miss clinically significant differences in CLABSI rates, and may lead to inaccurate interpretation of the impact of quality improvement efforts.

External validation of the parsimonious EuroLung risk models: analysis of the Brazilian Lung Cancer Registry.

The purpose of this study was to assess performance in the Brazilian Lung Cancer Registry Database by using the parsimonious EuroLung risk models for morbidity and mortality. The EuroLung1 and EuroLung2 models were tested and evaluated through calibration (calibration plot, Brier score, and the Hosmer-Lemeshow test) and discrimination (ROC AUCs), in a national multicenter registry of 1,031 patients undergoing anatomic lung resection. The evaluation of performance in Brazilian health care facilities utilizing risk-adjustment models, specifically EuroLung1 and EuroLung2, revealed substantial miscalibration, as evidenced by calibration plots and Hosmer-Lemeshow tests in both models. In terms of calibration, EuroLung1 exhibited a calibration plot with overlapping points, characterized by a slope of 1.11 and a Brier score of 0.15; the Hosmer-Lemeshow test yielded a statistically significant p-value of 0.015; and the corresponding ROC AUC was 0.678 (95% CI: 0.636-0.721). The EuroLung2 model displayed better calibration, featuring fewer overlapping points in the calibration plot, with a slope of 1.22, with acceptable discrimination, as indicated by a ROC AUC of 0.756 (95% CI: 0.670-0.842). Both models failed to accurately predict morbidity and mortality outcomes in this specific health care context. Discrepancies between the EuroLung model predictions and outcomes in Brazil underscore the need for model refinement and for a probe into inefficiencies in the Brazilian health care system.

Variational Bayes machine learning for risk adjustment of general outcome indicators with examples in urology

Risk adjustment is often necessary for outcome quality indicators (QIs) to provide fair and accurate feedback to healthcare professionals. However, traditional risk adjustment models are generally oversimplified and not equipped to disentangle complex factors influencing outcomes that are out of a healthcare professional’s control. We present VIRGO, a novel variational Bayes model trained on routinely collected, large administrative datasets to risk-adjust outcome QIs. VIRGO uses detailed demographics, diagnosis, and procedure codes to provide individualized risk adjustment and explanations on patient factors affecting outcomes. VIRGO achieves state-of-the-art on external datasets and features capabilities of uncertainty expression, explainable features, and counterfactual analysis capabilities. VIRGO facilitates risk adjustment by explaining how patient factors led to adverse outcomes and expresses the uncertainty of each prediction, allowing healthcare professionals to not only explore patient factors with unexplained variance that are associated with worse outcomes but also reflect on the quality of their clinical practice.

Adjusting the risk-adjustment: Accounting for variation between organisations in the responsiveness of their expenditure to need

There is concern that basing healthcare budgets on risk adjustment estimates derived from historical utilisation data may reinforce patterns of unmet need. We propose a method to avoid this, based on a measure of how closely local health organisations align resources to the needs of their populations. We refer to this measure as the ‘responsiveness of expenditure to need’ and estimate it using national person-level data on use of acute hospital and secondary mental health services in England. We find large variation in responsiveness in both services and show that higher expenditure responsiveness in mental health is associated with fewer suicides. We then re-estimate the national risk-adjustment model removing the data from the organisations with the lowest expenditure responsiveness to need. As expected, higher need individuals are estimated to have higher expenditure needs when less responsive organisations are removed from the estimation of the risk-adjustment. Removal of organisations with below-average responsiveness results in the neediest deciles of individuals having an extra £163 (7%) annual need for acute hospital care and an additional £79 (27%) annual need for mental health services. The application of this approach to risk adjustment would result in more resources being directed towards organisations serving higher-need populations.

A Framework for the Design of Risk-Adjustment Models in Health care Provider Payment Systems.

Prospective payments for health care providers require adequate risk adjustment (RA) to address systematic variation in patients' health care needs. However, the design of RA for provider payment involves many choices and difficult trade-offs between incentives for risk selection, incentives for cost control, and feasibility. Despite a growing literature, a comprehensive framework of these choices and trade-offs is lacking. This article aims to develop such a framework. Using literature review and expert consultation, we identify key design choices for RA in the context of provider payment and subsequently categorize these choices along two dimensions: (a) the choice of risk adjusters and (b) the choice of payment weights. For each design choice, we provide an overview of options, trade-offs, and key references. By making design choices and associated trade-offs explicit, our framework facilitates customizing RA design to provider payment systems, given the objectives and other characteristics of the context of interest.

A high hospital frailty risk score indicates an increased risk for complications following surgical treatment of proximal humerus fractures

BackgroundApproximately 70 % of proximal humerus fractures (PHF) occur after the age of 60. High complication rates have been described in correlation with the treatment of PHF. Major risk factors for the outcome might be frailty, mobility and comorbidities of patients at the time of hospital admission. The aim of this study was to create risk adjusted quality indicators for surgical treatment of proximal humerus fractures based on German claims data and to evaluate the impact of the Hospital Frailty Risk Score (HFRS) on risk adjustment. MethodsRetrospective claims data (2015–2021) were used to create risk adjusted quality indicators for eight outcomes by clustered multivariable logistic regression. The comparison of different risk adjustment model performances was done by ROC-AUC and Standardized Mortality/Morbidity Ratios. ResultsIn total, N = 34,912 patients (median age 75 years, 80.3 % female) were included. The most common surgical procedure was open reduction and internal fixation with plate osteosynthesis with 39.7 %, followed by reverse shoulder arthroplasty with 25.3 %. The most influential risk factor for all outcomes was a high HFRS with an Odds Ratio of 2.0 (95 %-Confidence Interval 1.8–2.3) for any secondary surgery (365 days) up to an Odds Ratio of 17.6 (95 %-Confidence Interval 14.9–20.8) for general complications during the index stay. ConclusionComparative quality reporting for the surgical treatment of PHF appears feasible with the developed models for risk adjustment using claims data. Preoperative evaluation of HFRS in PHF can contribute to risk assessment, and individual patient management. It therefore enables personalized treatment decisions.