Integrating Social Drivers of Health into Hospital Ratings with Application to the 100 Top Hospitals Study.

The objective was to investigate the relationship between social drivers of health (SDOH) and hospital performance within the 100 Top Hospitals study, exploring methods to recognize hospitals serving marginalized communities. Publicly available data sourced from the Centers for Medicare and Medicaid Services and the 2023 100 Top Hospitals study was used. The study employed multivariable hierarchical generalized linear regression models to assess the association between an SDOH composite variable derived using principal component analysis and overall hospital performance measures within the 100 Top Hospitals study. The analysis revealed a statistically significant association between SDOH factors and study ranking results. The SDOH composite variable is a significant predictor of performance within the 100 Top Hospitals study. Accounting for SDOH is essential to recognize high-performing hospitals serving marginalized communities. The findings suggest a need for broader considerations of SDOH in hospital ranking methodologies across various industry programs.

A Data-Driven Approach to Defining Risk-Adjusted Coding Specificity Metrics for a Large U.S. Dementia Patient Cohort.

Medical coding impacts patient care quality, payor reimbursement, and system reliability through the precision of patient information documentation. Inadequate coding specificity can have significant consequences at administrative and patient levels. Models to identify and/or enhance coding specificity practices are needed. Clinical records are not always available, complete, or homogeneous, and clinically driven metrics to assess medical practices are not logistically feasible at the population level, particularly in non-centralized healthcare delivery systems and/or for those who only have access to claims data. Data-driven approaches that incorporate all available information are needed to explore coding specificity practices. Using N = 487,775 hospitalization records of individuals diagnosed with dementia and discharged in 2022 from a large all-payor administrative claims dataset, we fitted logistic regression models using patient and facility characteristics to explain the coding specificity of principal and secondary diagnoses of dementia. A two-step approach was produced to allow for the flexible clustering of patient-level outcomes. Model outcomes were then used within a Poisson binomial model to identify facilities that over- or under-specify dementia diagnoses against healthcare industry standards across hospitalizations. The results indicate that multiple factors are significantly associated with dementia coding specificity, especially for principal diagnoses of dementia (AUC = 0.727). The practical use of this novel risk-adjusted metric is demonstrated for a sample of facilities and geospatially via a U.S. map. This study's findings provide healthcare facilities with a benchmark for assessing coding specificity practices and developing quality enhancements to align with healthcare industry standards, ultimately contributing to better patient care and healthcare system reliability.

A Novel Method for Assessing Risk-Adjusted Diagnostic Coding Specificity for Depression Using a U.S. Cohort of over One Million Patients.

Depression is a prevalent and debilitating mental health condition that poses significant challenges for healthcare providers, researchers, and policymakers. The diagnostic coding specificity of depression is crucial for improving patient care, resource allocation, and health outcomes. We propose a novel approach to assess risk-adjusted coding specificity for individuals diagnosed with depression using a vast cohort of over one million inpatient hospitalizations in the United States. Considering various clinical, demographic, and socioeconomic characteristics, we develop a risk-adjusted model that assesses diagnostic coding specificity. Results demonstrate that risk-adjustment is necessary and useful to explain variability in the coding specificity of principal (AUC = 0.76) and secondary (AUC = 0.69) diagnoses. Our approach combines a multivariate logistic regression at the patient hospitalization level to extract risk-adjusted probabilities of specificity with a Poisson Binomial approach at the facility level. This method can be used to identify healthcare facilities that over- and under-specify diagnostic coding when compared to peer-defined standards of practice.

Risk Adjustment of ICD-10-CM Coded Potential Inpatient Complications Using Administrative Data.

OBJECTIVE: To risk-adjust the Potential Inpatient Complication (PIC) measure set and propose a method to identify large deviations between observed and expected PIC counts. DATA SOURCES: Acute inpatient stays from the Premier Healthcare Database from January 1, 2019 to December 31, 2021. STUDY DESIGN: In 2014, the PIC list was developed to identify a broader set of potential complications that can occur as a result of care decisions. Risk adjustment for 111 PIC measures is performed across 3 age-based strata. Using patient-level risk factors and PIC occurrences, PIC-specific probabilities of occurrence are estimated through multivariate logistic regression models. Poisson Binomial cumulative mass function estimates identify deviations between observed and expected PIC counts across levels of patient-visit aggregation. Area under the curve (AUC) estimates are used to demonstrate PIC predictive performance in an 80:20 derivation-validation split framework. DATA COLLECTION/EXTRACTION METHODS: We used N=3,363,149 administrative hospitalizations between 2019 and 2021 from the Premier Healthcare Database. PRINCIPAL FINDINGS: PIC-specific model predictive performance was strong across PICs and age strata. Average area under the curve estimates across PICs were 0.95 (95% CI: 0.93-0.96), 0.91 (95% CI: 0.90-0.93), and 0.90 (95% CI: 0.89-0.91) for the neonate and infant, pediatric, and adult strata, respectively. CONCLUSIONS: The proposed method provides a consistent quality metric that adjusts for the population's case mix. Age-specific risk stratification further addresses currently ignored heterogeneity in PIC prevalence across age groups. Finally, the proposed aggregation method identifies large PIC-specific deviations between observed and expected counts, flagging areas with a potential need for quality improvements.

A Novel Approach to Developing Disease and Outcome-Specific Social Risk Indices.

INTRODUCTION: A variety of industry composite indices are employed within health research in risk-adjusted outcome measures and to assess health-related social needs. During the COVID-19 pandemic, the relationships among risk adjustment, clinical outcomes, and composite indices of social risk have become relevant topics for research and healthcare operations. Despite the widespread use of these indices, composite indices are often comprised of correlated variables and therefore may be affected by information duplicity of their underlying risk factors. METHODS: A novel approach is proposed to assign outcome- and disease group-driven weights to social risk variables to form disease and outcome-specific social risk indices and apply the approach to the county-level Centers for Disease Control and Prevention social vulnerability factors for demonstration. The method uses a subset of principal components reweighed through Poisson rate regressions while controlling for county-level patient mix. The analyses use 6,135,302 unique patient encounters from 2021 across seven disease strata. RESULTS: The reweighed index shows reduced root mean squared error in explaining county-level mortality in five of the seven disease strata and equivalent performance in the remaining strata compared with the reduced root mean squared error using the current Centers for Disease Control and Prevention Social Vulnerability Index as a benchmark. CONCLUSIONS: A robust method is provided, designed to overcome challenges with current social risk indices, by accounting for redundancy and assigning more meaningful disease and outcome-specific variable weights.

Development of a Novel Prospective Model to Predict Unplanned 90-Day Readmissions After Total Hip Arthroplasty.

BACKGROUND: For hospitals participating in bundled payment programs, unplanned readmissions after surgery are often termed "bundle busters." The aim of this study was to develop the framework for a prospective model to predict 90-day unplanned readmissions after elective primary total hip arthroplasty (THA) at a macroscopic hospital-based level. METHODS: A national, all-payer, inpatient claims and cost accounting database was used. A mixed-effect logistic regression model measuring the association of unplanned 90-day readmissions with a number of patient-level and hospital-level characteristics was constructed. RESULTS: Using 427,809 unique inpatient THA encounters, 77 significant risk factors across 5 domains (ie, comorbidities, demographics, surgical history, active medications, and intraoperative factors) were identified. The highest frequency domain was comorbidities (64/100) with malignancies (odds ratio [OR] 2.26), disorders of the respiratory system (OR 1.75), epilepsy (OR 1.5), and psychotic disorders (OR 1.5), being the most predictive. Other notable risk factors identified by the model were the use of opioid analgesics (OR 7.3), Medicaid coverage (OR 1.8), antidepressants (OR 1.6), and blood-related medications (OR 1.6). The model produced an area under the curve of 0.715. CONCLUSION: We developed a novel model to predict unplanned 90-day readmissions after elective primary THA. Fifteen percent of the risk factors are potentially modifiable such as use of tranexamic acid, spinal anesthesia, and opioid medications. Given the complexity of the factors involved, hospital systems with vested interest should consider incorporating some of the findings from this study in the form of electronic medical records predictive analytics tools to offer clinicians with real-time actionable data.

A novel approach to attribute responsible physicians using inpatient claims.

OBJECTIVES: More robust attribution methods are necessary to understand physician-level variation in quality of care across risk-adjusted inpatient measures. We address a gap in the literature involving attribution of physicians to inpatient stays using administrative claims data, in which rule-based methods often inadequately attribute physicians. STUDY DESIGN: Methodology comparison study using a cross-section of inpatient stays. METHODS: A novel approach is proposed in which physicians' relative degrees of responsibility for inpatient stays are expressed through physician-specific attribution ratios informed by existing patient characteristics and comorbidities. Attribution results are compared with the rule-based benchmark method for 7 CMS-defined clinical cohorts, including a COVID-19 cohort. RESULTS: Using 6,835,460 unique patient encounters during 2020 (n = 136,339 in out-of-sample cohort), the proposed approach favored specialists generally considered responsible for primary clinical conditions when compared with the benchmark. The most salient shift within the acute myocardial infarction (+17.0%), heart failure (+20.2%), and coronary artery bypass graft (+4.0%) cohorts was toward the cardiovascular diseases specialty, and the chronic obstructive pulmonary disease (+24.0%) and pneumonia (+16.2%) cohorts resulted in a shift toward the pulmonary diseases specialty. The COVID-19 cohort resulted in considerable shifts toward infectious diseases and pulmonary diseases specialties (+17.4% and +14.1%, respectively). The stroke cohort experienced a considerable shift toward the neurology specialty (+42.2%). CONCLUSIONS: We provide a robust method to attribute physicians to patients, which is a necessary tool to understand physician-level variation in quality of care within the inpatient acute care setting. The proposed method provides consistency across facilities and eliminates unattributed patients resulting from unsatisfied business rules.

Modeling Coding Intensity of Procedures in a U.S. Population-Based Hip/Knee Arthroplasty Inpatient Cohort Adjusting for Patient- and Facility-Level Characteristics.

Variations in procedure coding intensity, defined as excess coding of procedures versus industry (instead of clinical) standards, can result in differentials in quality of care for patients and have additional implications for facilities and payors. The literature regarding coding intensity of procedures is limited, with a need for risk-adjusted methods that help identify over- and under-coding using commonly available data, such as administrative claims. Risk-adjusted metrics are needed for quality control and enhancement. We propose a two-step approach to risk adjustment, using a zero-inflated Poisson model, applied to a hip-knee arthroplasty cohort discharged during 2019 (n = 313,477) for patient-level risk adjustment, and a potential additional layer for adjustment based on facility-level characteristics, when desired. A 21.41% reduction in root-mean-square error was achieved upon risk adjustment for patient-level factors alone. Furthermore, we identified facilities that over- and under-code versus industry coding expectations, adjusting for both patient-level and facility-level factors. Excess coding intensity was found to vary across multiple levels: (1) geographically across U.S. Census regional divisions; (2) temporally with marked seasonal components; (3) by facility, with some facilities largely departing from industry standards, even after adjusting for both patient- and facility-level characteristics. Our proposed method is simple to implement, generalizable, it can be used across cohorts with different sets of information available, and it is not limited by the accessibility and sparsity of electronic health records. By identifying potential over- and under-coding of procedures, quality control personnel can explore and assess internal needs for enhancements in their health delivery services and monitor subsequent quality improvements.

Diagnostic Coding Intensity among a Pneumonia Inpatient Cohort Using a Risk-Adjustment Model and Claims Data: A U.S. Population-Based Study.

Hospital payments depend on the Medicare Severity Diagnosis-Related Group's estimated cost and the set of diagnoses identified during inpatient stays. However, over-coding and under-coding diagnoses can occur for different reasons, leading to financial and clinical consequences. We provide a novel approach to measure diagnostic coding intensity, built on commonly available administrative claims data, and demonstrated through a 2019 pneumonia acute inpatient cohort (N = 182,666). A Poisson additive model (PAM) is proposed to model risk-adjusted additional coded diagnoses. Excess coding intensity per patient visit was estimated as the difference between the observed and PAM-based expected counts of secondary diagnoses upon risk adjustment by patient-level characteristics. Incidence rate ratios were extracted for patient-level characteristics and further adjustments were explored by facility-level characteristics to account for facility and geographical differences. Facility-level factors contribute substantially to explain the remaining variability in excess diagnostic coding, even upon adjusting for patient-level risk factors. This approach can provide hospitals and stakeholders with a tool to identify outlying facilities that may experience substantial differences in processes and procedures compared to peers or general industry standards. The approach does not rely on the availability of clinical information or disease-specific markers, is generalizable to other patient cohorts, and can be expanded to use other sources of information, when available.

Assessing Hospital Resource Utilization with Application to Imaging for Patients Diagnosed with Prostate Cancer.

Resource utilization measures are typically modeled by relying on clinical characteristics. However, in some settings, those clinical markers are not available, and hospitals are unable to explore potential inefficiencies or resource misutilization. We propose a novel approach to exploring misutilization that solely relies on administrative data in the form of patient characteristics and competing resource utilization, with the latter being a novel addition. We demonstrate this approach in a 2019 patient cohort diagnosed with prostate cancer (n = 51,111) across 1056 U.S. healthcare facilities using Premier, Inc.'s (Charlotte, NC, USA) all payor databases. A multivariate logistic regression model was fitted using administrative information and competing resources utilization. A decision curve analysis informed by industry average standards of utilization allows for a definition of misutilization with regards to these industry standards. Odds ratios were extracted at the patient level to demonstrate differences in misutilization by patient characteristics, such as race; Black individuals experienced higher under-utilization compared to White individuals (p < 0.0001). Volume-adjusted Poisson rate regression models allow for the identification and ranking of facilities with large departures in utilization. The proposed approach is scalable and easily generalizable to other diseases and resources and can be complemented with clinical information from electronic health record information, when available.

County-Level Sociodemographic Factors Associated with COVID-19 Incidence and Mortality in North and South Carolina.

BACKGROUND There is limited research regarding associations between county-level factors and COVID-19 incidence and mortality. While the Carolinas are geographically connected, they are not homogeneous, with statewide political and intra-state socioeconomic differences leading to heterogeneous spread between and within states.METHODS Infection and mortality data from Johns Hopkins University during the 7 months since the first reported case in the Carolinas was combined with county-level socioeconomic/demographic factors. Time series imputations were performed whenever county-level reported infections were implausible. Multivariate Poisson regression models were fitted to extract incidence (infection and mortality) rate ratios by county-level factor. State-level differences in filtered trends were also calculated. Geospatial maps and Kaplan-Meier curves were constructed stratifying by median county-level factor. Differences between North and South Carolina were identified.RESULTS Incidence and mortality rates were lower in North Carolina than South Carolina. Statistically significant higher incidence and mortality rates were associated with counties in both states with higher proportions of Black/African American populations and those without health insurance aged < 65 years. Counties with larger populations aged ≥ 75 years were associated with increased mortality (but decreased incidence) rates.LIMITATIONS COVID-19 data contained multiple inconsistencies, so imputation was needed, and covariate-based data was not synchronous and potentially insufficient in granularity given the epidemiology of the disease. County-level analyses imply within-county homogeneity, an assumption increasingly breached by larger counties.CONCLUSION While statewide interventions were initially implemented, inter-county racial/ethnic and socioeconomic variability points to the need for more heterogeneous interventions, including policies, as populations within particular counties may be at higher risk.

A review of socioeconomic factors associated with acute myocardial infarction-related mortality and hospital readmissions.

BACKGROUND: Risk-adjustment models are widely used methodological approaches within the healthcare industry to measure hospital performance and quality of care. However, the Centers for Medicare and Medicaid Services (CMS) do not fully adjust for socioeconomic status (SES) in acute myocardial infarction (AMI) models. A review and evidence synthesis was conducted to identify associations of SES factors with hospital readmission and mortality in AMI patients. METHODS: Multiple electronic databases were queried to identify studies assessing risk for AMI-related mortality or hospital readmissions and SES factors. Identified studies were screened by title and abstract. Full-text reviews followed for articles meeting the inclusion criteria, including quality assessments. Data were extracted from all included studies, and evidence synthesis was performed to identify associations between SES factors and outcome variables. RESULTS: Ten studies were included in the review. One study showed that Black patients had higher AMI-related readmission rates compared to White patients (mean difference 4.3% [SD 1.4%], p < 0.001). Another study showed that income inequality was associated with increased risk of AMI-related readmissions (RR 1.18 [95% CI], 1.13-1.23). One study found that unemployed individuals experienced significantly greater rates of AMI-related mortality than those working full-time (HR 2.08, 1.51-2.87). According to another study, lack of health insurance was associated with worse rates for in-hospital AMI-related mortality (OR 1.77, 1.72-1.82). Based on one study, AMI-related mortality was higher in those with <8 years of education compared to those with >16 years (17.5% vs. 3.5%, p < 0.0001). Five of six studies found a significant association between ZIP code/neighborhood/location and AMI-related readmission or mortality. CONCLUSION: Race, ZIP code/neighborhood/location, insurance status, income/poverty, and education comprise SES factors found to be associated with AMI-related mortality and/or readmission outcomes. Including these SES factors in future updates of CMS's risk-adjusted models has the potential to provide more appropriate compensation mechanisms to hospitals.

Turning up the volume to address biases in predicted healthcare-associated infections and enhance U.S. hospital rankings: A data-driven approach.

OBJECTIVES: To examine potential biases in standardized infection ratio (SIR) metrics due to static U.S. Centers for Disease Control and Prevention (CDC) parameters and non-linearity of infection outcomes with volume. Correspondingly, to enhance the CDC predictions by incorporating additional information from volume metrics and explore an alternative approach to more fairly rank hospitals to address the SIR=0 problem. METHODS: This population-based study uses publicly available 2019 healthcare-associated infections (HAI) data from 3096 acute care U.S. hospitals. HAI-specific Poisson generalized additive models illustrate the recalibration of CDC predictions, using volume-based spline functions to adjust for biases. Implied cumulative distribution functions (CDF) were derived, and HAI-facility-specific probabilities were calculated. Hospital rankings implied from these HAI-stratified probabilities were calculated. RESULTS: Calibration plots demonstrate existing biases associated with CDC infection over-predictions. Volume-based spline functions were significant for all HAIs (P<.0004). CDF-based rankings resulted in larger discrimination across hospitals based on strength of evidence, especially among SIR=0 facilities. National maps depict ranking differences by HAI and state. CONCLUSION: Adjustment of SIR biases, which differ by facility volume, is needed to produce more accurate and fairer hospital rankings.

County-Level Socioeconomic Status Adjustment of Acute Myocardial Infarction Mortality Hospital Performance Measure in the U.S.

The U.S. Centers for Medicare and Medicaid Services' (CMS's) Hospital Compare (HC) data provides a collection of risk-adjusted hospital performance metrics intended to allow comparison of hospital-provided care. However, CMS does not adjust for socioeconomic status (SES) factors, which have been found to be associated with disparate health outcomes. Associations between county-level SES factors and CMS's risk-adjusted 30-day acute myocardial infarction (AMI) mortality rates are explored for n = 2462 hospitals using a variety of sources for county-level SES information. Upon performing multiple imputation, a stepwise backward elimination model selection approach using Akaike's information criteria was used to identify the optimal model. The resulting model, comprised of 14 predictors mostly at the county level, provides an additional 8% explanatory power to capture the variability in 30-day risk-standardized AMI mortality rates, which already account for patient-level clinical differences. SES factors may be an important feature for inclusion in future risk-adjustment models, which will have system and policy implications for distributing resources to hospitals, such as reimbursements. It also serves as a stepping stone to identify and address long-standing SES-related inequities.

Associations between Hospital Quality Outcomes and Medicare Spending per Beneficiary in the USA.

The cost of healthcare in the United States has increased over time. However, patient health outcomes have not trended with spending. There is a need to better comprehend the association between healthcare costs in the United States and hospital quality outcomes. Medicare spending per beneficiary (MSPB), a homogeneous metric across providers, can be used to evaluate the association between episodic Medicare spending and quality of care. Fifteen inpatient outcome measures were selected from Hospital Compare data among all (n = 4758) facilities and transformed to quintiles to ensure comparability across measures and to reduce the influence of outliers on the analysis. Both univariate and multiresponse multinomial ordered probit regression models were utilized across outcome domains to quantify associations between outcomes and spending. We found that MSPB was not associated with quality of care in most cases, adding evidence of a lack of outcome accountability among Medicare-funded facilities. Furthermore, worse outcomes were found to be associated with increased spending for some metrics. Policies are needed to align quality of care outcomes with the increasing costs of U.S. healthcare.

CRISPRclassify: Repeat-Based Classification of CRISPR Loci.

Detection and classification of CRISPR-Cas systems in metagenomic data have become increasingly prevalent in recent years due to their potential for diverse applications in genome editing. Traditionally, CRISPR-Cas systems are classified through reference-based identification of proximate cas genes. Here, we present a machine learning approach for the detection and classification of CRISPR loci using repeat sequences in a cas-independent context, enabling identification of unclassified loci missed by traditional cas-based approaches. Using biological attributes of the CRISPR repeat, the core element in CRISPR arrays, and leveraging methods from natural language processing, we developed a machine learning model capable of accurate classification of CRISPR loci in an extensive set of metagenomes, resulting in an F1 measure of 0.82 across all predictions and an F1 measure of 0.97 when limiting to classifications with probabilities >0.85. Furthermore, assessing performance on novel repeats yielded an F1 measure of 0.96. Although the performance of cas-based identification will exceed that of a repeat-based approach in many cases, CRISPRclassify provides an efficient approach to classification of CRISPR loci for cases in which cas gene information is unavailable, such as metagenomes and fragmented genome assemblies.

Predicting Hospital Overall Quality Star Ratings in the USA.

The U.S. Centers for Medicare and Medicaid Services (CMS) assigns quality star ratings to hospitals upon assessing their performance across 57 measures. Ratings can be used by healthcare consumers for hospital selection and hospitals for quality improvement. We provide a simpler, more intuitive modeling approach, aligned with recent criticism by stakeholders. An ordered logistic regression approach is proposed to assess associations between performance measures and ratings across eligible (n = 4519) U.S. hospitals. Covariate selection reduces the double counting of information from highly correlated measures. Multiple imputation allows for inference of star ratings when information on all measures is not available. Twenty performance measures were found to contain all the relevant information to formulate star rating predictions upon accounting for performance measure correlation. Hospitals can focus their efforts on a subset of model-identified measures, while healthcare consumers can predict quality star ratings for hospitals ineligible under CMS criteria.