Improving hospital quality risk-adjustment models using interactions identified by hierarchical group lasso regularisation.

BACKGROUND: Risk-adjustment (RA) models are used to account for severity of illness in comparing patient outcomes across hospitals. Researchers specify covariates as main effects, but they often ignore interactions or use stratification to account for effect modification, despite limitations due to rare events and sparse data. Three Agency for Healthcare Research and Quality (AHRQ) hospital-level Quality Indicators currently use stratified models, but their variable performance and limited interpretability motivated the design of better models. METHODS: We analysed patient discharge de-identified data from 14 State Inpatient Databases, AHRQ Healthcare Cost and Utilization Project, California Department of Health Care Access and Information, and New York State Department of Health. We used hierarchical group lasso regularisation (HGLR) to identify first-order interactions in several AHRQ inpatient quality indicators (IQI) - IQI 09 (Pancreatic Resection Mortality Rate), IQI 11 (Abdominal Aortic Aneurysm Repair Mortality Rate), and Patient Safety Indicator 14 (Postoperative Wound Dehiscence Rate). These models were compared with stratum-specific and composite main effects models with covariates selected by least absolute shrinkage and selection operator (LASSO). RESULTS: HGLR identified clinically meaningful interactions for all models. Synergistic IQI 11 interactions, such as between hypertension and respiratory failure, suggest patients who merit special attention in perioperative care. Antagonistic IQI 11 interactions, such as between shock and chronic comorbidities, illustrate that naïve main effects models overestimate risk in key subpopulations. Interactions for PSI 14 suggest key subpopulations for whom the risk of wound dehiscence is similar between open and laparoscopic approaches, whereas laparoscopic approach is safer for other groups. Model performance was similar or superior for composite models with HGLR-selected features, compared to those with LASSO-selected features. CONCLUSIONS: In this application to high-profile, high-stakes risk-adjustment models, HGLR selected interactions that maintained or improved model performance in populations with heterogeneous risk, while identifying clinically important interactions. The HGLR package is scalable to handle a large number of covariates and their interactions and is customisable to use multiple CPU cores to reduce analysis time. The HGLR method will allow scholars to avoid creating stratified models on sparse data, improve model calibration, and reduce bias. Future work involves testing using other combinations of risk factors, such as vital signs and laboratory values. Our study focuses on a real-world problem of considerable importance to hospitals and policy-makers who must use RA models for statutorily mandated public reporting and payment programmes.

The HCBS taxonomy: a new language for classifying home- and community-based services.

INTRODUCTION: As states make home- and community-based services (HCBS) more accessible, researchers have become more interested in understanding service use and spending. Because state Medicaid programs differ in the types of services they offer and in how they report these services, analyzing HCBS at the national level is challenging. OBJECTIVE: Describe the HCBS taxonomy and present findings on HCBS waiver expenditures and users. DATA: This brief analyzed fee-for-service claims from 28 approved states in 2010 Medicaid Analytic eXtract (MAX) files. We summed all expenditures and counted the unique number of users across each HCBS taxonomy service and category. METHODS: The taxonomy was developed jointly by Truven Health (at that time Thomson Reuters) and Mathematica Policy Research, with stakeholder input, and reviewed using procedure codes. Today, the taxonomy is organized by 18 categories and over 60 specific services. FINDINGS: For calendar year 2010, 28 states spent almost $23.6 billion on HCBS, with 80 percent of expenditures categorized as round-the-clock, home-based, and day services. Other services, such as case management, or equipment, modifications, and technology were widely used, but are not particularly costly and do not account for a large proportion of expenditures in every state. CONCLUSIONS: By providing a common language, the taxonomy presents detailed information on services and makes it easier to assess and identify state-level variation for HCBS.

The impact of stressful life events on excessive alcohol consumption in the French population: findings from the GAZEL cohort study.

BACKGROUND: Major life changes may play a causative role in health through lifestyle factors, such as alcohol. The objective was to examine the impact of stressful life events on heavy alcohol consumption among French adults. METHODS: Trajectories of excessive alcohol consumption in 20,625 employees of the French national gas and electricity company for up to 5 years before and 5 years after an event, with annual measurements from 1992. We used repeated measures analysis of time series data indexed to events, employing generalized estimating equations. RESULTS: For women, excessive alcohol use increased before important purchase (p = 0.021), children leaving home (p<0.001), and death of loved ones (p = 0.03), and decreased before widowhood (p = 0.015); in the year straddling the event, increased consumption was observed for important purchase (p = 0.018) and retirement (p = 0.002); at the time of the event, consumption decreased for marriage (p = 0.002), divorce, widowhood, and death of loved one (all p<0.001), and increased for retirement (p = 0.035). For men, heavy alcohol consumption increased in the years up to and surrounding the death of loved ones, retirement, and important purchase (all p<0.001), and decreased after (all p<0.001, except death of loved one: p = 0.006); at the time of the event, consumption decreased for all events except for children leaving home and retirement, where we observed an increase (all p<0.001). For women and men, heavy alcohol consumption decreased prior to marriage and divorce and increased after (all p<0.001, except for women and marriage: p = 0.01). CONCLUSION: Stressful life events promote healthy and unhealthy alcohol consumption. Certain events impact alcohol intake temporarily while others have longer-term implications. Research should disentangle women's and men's distinct perceptions of events over time.

Understanding older adults' motivators and barriers to participating in organized programs supporting exercise behaviors.

Little is known about older adults' perceptions of organized programs that support exercise behavior. We conducted semi-structured interviews with 39 older adults residing in King County, Washington, who either declined to join, joined and participated, or joined and then quit a physical activity-oriented program. We sought to explore motivators and barriers to physical activity program participation and to elicit suggestions for marketing strategies to optimize participation. Two programs supporting exercise behavior and targeting older persons were the source of study participants: Enhance(®)Fitness and Physical Activity for a Lifetime of Success. We analyzed interview data using standard qualitative methods. We examined variations in themes by category of program participant (joiner, decliner, quitter) as well as by program and by race. Interview participants were mostly females in their early 70s. Approximately half were non-White, and about half had graduated from college. The most frequently cited personal factors motivating program participation were enjoying being with others while exercising and desiring a routine that promoted accountability. The most frequent environmental motivators were marketing materials, encouragement from a trusted person, lack of program fees, and the location of the program. The most common barriers to participation were already getting enough exercise, not being motivated or ready, and having poor health. Marketing messages focused on both personal benefits (feeling better, social opportunity, enjoyability) and desirable program features (tailored to individual needs), and marketing mechanisms ranged from traditional written materials to highly personalized approaches. These results suggest that organized programs tend to appeal to those who are more socially inclined and seek accountability. Certain program features also influence participation. Thoughtful marketing that involves a variety of messages and mechanisms is essential to successful program recruitment and continued attendance.

How are the costs of care for medical falls distributed? The costs of medical falls by component of cost, timing, and injury severity.

PURPOSE OF THE STUDY: To examine the components of cost that drive increased total costs after a medical fall over time, stratified by injury severity. DESIGN AND METHODS: We used 2004-2007 cost and utilization data for persons enrolled in an integrated care delivery system. We used a longitudinal cohort study design, where each individual provides 2-3 years of administrative data grouped into 3-month intervals relative to an index date. We identified 8,969 medical fallers through International Classification of Diseases, 9th Revision, codes and E-Codes and used 8,956 nonfaller controls, identified through age and gender frequency matching. Total costs were partitioned into 7 components: inpatient, outpatient, emergency, radiology, pharmacy, postacute care, and "other." RESULTS: The large increase in costs after a hospitalized fall is mainly associated with inpatient and postacute care components. The spike in costs after a nonhospitalized fall is attributable to outpatient and "other" (e.g., ambulatory surgery or community health services) components. Hospitalized fallers' inpatient, emergency, postacute care, outpatient, and radiology costs are not always greater than those for nonhospitalized fallers. IMPLICATIONS: Components associated with increased costs after a medical fall vary over time and by injury severity. Future studies should compare if delivering certain acute and postacute health services improve health and reduce cost trajectories after a medical fall more than others. Additionally, since the older adult population and the problem of falls are growing, health care delivery systems should develop standardized methodology to monitor medical fall rates.

Older adults' perceptions of clinical fall prevention programs: a qualitative study.

Objective. To investigate motivational factors and barriers to participating in fall risk assessment and management programs among diverse, low-income, community-dwelling older adults who had experienced a fall. Methods. Face-to-face interviews with 20 elderly who had accepted and 19 who had not accepted an invitation to an assessment by one of two fall prevention programs. Interviews covered healthy aging, core values, attributions/consequences of the fall, and barriers/benefits of fall prevention strategies and programs. Results. Joiners and nonjoiners of fall prevention programs were similar in their experience of loss associated with aging, core values they expressed, and emotional response to falling. One difference was that those who participated endorsed that they "needed" the program, while those who did not participate expressed a lack of need. Conclusions. Interventions targeted at a high-risk group need to address individual beliefs as well as structural and social factors (transportation issues, social networks) to enhance participation.

A longitudinal analysis of total 3-year healthcare costs for older adults who experience a fall requiring medical care.

OBJECTIVES: To compare longitudinal changes in healthcare costs between fallers admitted to the hospital at the time of the fall (admitted), those not admitted to the hospital (nonadmitted), and nonfaller controls; test hypotheses related to differences in mean costs between and within these groups over time; and estimate the costs attributable to falling. DESIGN: Longitudinal cohort. SETTING: Group Health Cooperative of Puget Sound. PARTICIPANTS: Seven thousand nine hundred ninety-three nonadmitted fallers, 976 admitted fallers, and 8,956 nonfallers aged 67 and older enrolled in an integrated healthcare delivery system. Fallers were identified according to fall-related E-Codes and International Classification of Diseases, Ninth Revision codes recorded between January 1, 2004, and December 31, 2006. Nonfallers were frequency matched on age group and sex. MEASUREMENTS: Quarterly costs during a 3-year period were modeled using generalized estimating equations. Covariates included index age, sex, RxRisk (a comorbidity adjuster), fall status, time, and interactions between fall status and time. RESULTS: Cost differences between the faller cohorts and nonfallers were greatest in quarters closest to the fall (all P<.01) and persisted throughout the entire year of follow-up. Although nonfaller costs increased with time, faller cohort costs increased more quickly (all P<.01). For admitted fallers, 92% of costs incurred in the quarter of the fall were estimated to be attributable to falling ($27,745 of $30,038, P<.001). CONCLUSION: Falls for which medical attention are sought resulted in higher costs than for nonfallers for up to 12 months after a fall, particularly for falls requiring hospitalization. Prevention efforts should focus on reducing fall-related injuries requiring hospitalization because they produce the highest excess costs and have a higher likelihood of 1-year mortality.

A protocol of dual prophylaxis for venous thromboembolism prevention in gynecologic cancer patients.

OBJECTIVE: To evaluate a quality improvement protocol for venous thromboembolism prevention in postoperative gynecologic cancer patients. METHODS: On January 1, 2006, we initiated a universal protocol of dual prophylaxis with sequential compression devices and three times daily heparin (or daily low molecular weight heparin) until discharge in gynecologic cancer patients having major surgery. Patients with both malignancy and age over 60 years (or history of prior clot) were discharged on 2 weeks of anticoagulant. Before January 2006, all patients were given sequential compression devices starting before the induction of anesthesia, continuing until discharge from the hospital. Records of gynecologic cancer service patients admitted in 2005 and 2006 were reviewed, excluding patients with a history of heparin-induced thrombocytopenia or those admitted on an anticoagulant. Any pulmonary embolism or deep vein thrombosis diagnosed within 6 weeks of surgery was identified. We performed chi2 and Wilcoxon rank sum tests as well as multivariable regression analysis for confounders. RESULTS: Six of the 311 women meeting inclusion criteria in 2006 (1.9%) and 19 of 294 (6.5%) in 2005 had venous thromboembolism (odds ratio 0.33, 95% confidence interval 0.12-0.88, multivariable analysis adjusting for baseline differences between the groups). Heparin was given to 98.1% of patients in the hospital in 2006, and 91.1% of those meeting high-risk criteria were discharged on an anticoagulant. No differences in major bleeding complications were seen between years. CONCLUSION: A protocol of dual prophylaxis with prolonged prophylaxis in high-risk patients was successfully implemented and was associated with a significant reduction in the rate of venous thromboembolism without increasing bleeding complications. LEVEL OF EVIDENCE: II.