Interpreting risk-adjusted length of stay patterns for VA hospitals

Comparing Outcomes in Patients Undergoing Colectomy at Veteran Affairs Hospitals and Non-Veteran Affairs Hospitals: A Multiinstitutional Study

INTRODUCTION

The Veteran Affairs Surgical Quality Improvement Program (VASQIP) and National Surgical Quality Improvement Program (NSQIP) are large databases designed to measure surgical outcomes for their respective populations. We sought to compare surgical outcomes in patients undergoing colectomies at Veterans Affairs (VA) hospitals versus non-VA hospitals.

METHODS

After institutional review baord approval, records for 271,523 colectomies from NSQIP and 11,597 from VASQIP between the years 2015 and 2019 were compiled. Demographics, comorbidity, 30-d mortality, and other outcomes were examined using Chi-squared, analysis of variance, Mann Whitney U, and Fisher's Exact Test within SPSS version 26.

RESULTS

VASQIP patients were more likely to be male (94.3% versus 48.4%, P < 0.001) and older (median 63, 52-72 versus 67, 60-72 P < 0.001). Veterans were also more likely to have diabetes (25.3% versus 15.8%, P < 0.001), chronic obstructive pulmonary disease (15.4% versus 5.5%, P < 0.001), and congestive heart failure (17.0% versus 1.3%, P < 0.001). Veterans had slightly better 30-d mortality (2.4% versus 2.8%, P = 0.003), less organ space infections (2.8% versus 5.8%, P < 0.001), or postoperative sepsis (3.4% versus 5.3%). Non-VA patients were more likely to be having emergent surgery (13.4% versus 9.6%, P < 0.001) or undergo a laparoscopic approach (57.9% versus 50.2%, P < 0.001). Non-VA patients had shorter postoperative length of stay (5.99 d versus 7.32 d, P < 0.001) and were less likely to return to the operating room (5.3% versus 8.4%, P < 0.001) CONCLUSIONS: Despite increased comorbidity, VA hospitals and hospitals enrolled in NSQIP have managed to achieve markedly similar rates of 30-d mortality following colectomy. Further study is needed to better understand the differences between both the populations and surgical outcomes between VA hospitals and non-VA hospitals.

Brief Intervention and Follow-Up for Suicidal Patients With Repeat Emergency Department Visits Enhances Treatment Engagement

We implemented an innovative, brief, easy-to-administer 2-part intervention to enhance coping and treatment engagement. The intervention consisted of safety planning and structured telephone follow-up postdischarge with 95 veterans who had 2 or more emergency department (ED) visits within 6 months for suicide-related concerns (i.e., suicide ideation or behavior). The intervention significantly increased behavioral health treatment attendance 3 months after intervention, compared with treatment attendance in the 3 months after a previous ED visit without intervention. The trend was for a decreasing hospitalization rate.

Enhanced clinical pharmacy service targeting tools: risk-predictive algorithms

RATIONALE, AIMS AND OBJECTIVES

This study aimed to determine the value of using a mix of clinical pharmacy data and routine hospital admission spell data in the development of predictive algorithms. Exploration of risk factors in hospitalized patients, together with the targeting strategies devised, will enable the prioritization of clinical pharmacy services to optimize patient outcomes.

METHODS

Predictive algorithms were developed using a number of detailed steps using a 75% sample of integrated medicines management (IMM) patients, and validated using the remaining 25%. IMM patients receive targeted clinical pharmacy input throughout their hospital stay. The algorithms were applied to the validation sample, and predicted risk probability was generated for each patient from the coefficients. Risk threshold for the algorithms were determined by identifying the cut-off points of risk scores at which the algorithm would have the highest discriminative performance. Clinical pharmacy staffing levels were obtained from the pharmacy department staffing database.

RESULTS

Numbers of previous emergency admissions and admission medicines together with age-adjusted co-morbidity and diuretic receipt formed a 12-month post-discharge and/or readmission risk algorithm. Age-adjusted co-morbidity proved to be the best index to predict mortality. Increased numbers of clinical pharmacy staff at ward level was correlated with a reduction in risk-adjusted mortality index (RAMI).

CONCLUSIONS

Algorithms created were valid in predicting risk of in-hospital and post-discharge mortality and risk of hospital readmission 3, 6 and 12 months post-discharge. The provision of ward-based clinical pharmacy services is a key component to reducing RAMI and enabling the full benefits of pharmacy input to patient care to be realized.

Ethnic variations in unplanned readmissions and excess length of hospital stay: a nationwide record-linked cohort study

Background: Studies in the USA have shown ethnic inequalities in quality of hospital care, but in Europe, this has never been analysed. We explored variations in indicators of quality of hospital care by ethnicity in the Netherlands. Methods: We analysed unplanned readmissions and excess length of stay (LOS) across ethnic groups in a large population of hospitalized patients over an 11-year period by linking information from the national hospital discharge register, the Dutch population register and socio-economic data. Data were analysed with stepwise logistic regression. Results: Ethnic differences were most pronounced in older patients: all non-Western ethnic groups > 45 years had an increased risk for excess LOS compared with ethnic Dutch patients, with odds ratios (ORs) (adjusted for case mix) varying from 1.05 [95% confidence intervals (95% CI) 1.02–1.08] for other non-Western patients to 1.14 (95% CI 1.07–1.22) for Moroccan patients. The risk for unplanned readmission in patients >45 years was increased for Turkish (OR 1.24, 95% CI 1.18–1.30) and Surinamese patients (OR 1.11, 95% CI 1.07–1.16). These differences were explained partially, although not substantially, by differences in socio-economic status. Conclusion: We found significant ethnic variations in unplanned readmissions and excess LOS. These differences may be interpretable as shortcomings in the quality of hospital care delivered to ethnic minority patients, but exclusion of alternative explanations (such as differences in patient- and community-level factors, which are outside hospitals’ control) requires further research. To quantify potential ethnic inequities in hospital care in Europe, we need empirical prospective cohort studies with solid quality outcomes such as adverse event rates.

Using patient safety indicators to estimate the impact of potential adverse events on outcomes

The authors estimated the impact of potentially preventable patient safety events, identified by Agency for Healthcare Research and Quality (AHRQ) Patient Safety Indicators (PSIs), on patient outcomes: mortality, length of stay (LOS), and cost. The PSIs were applied to all acute inpatient hospitalizations at Veterans Health Administration (VA) facilities in fiscal 2001. Two methods-regression analysis and multivariable case matching- were used independently to control for patient and facility characteristics while predicting the effect of the PSI on each outcome. The authors found statistically significant (p < .0001) excess mortality, LOS, and cost in all groups with PSIs. The magnitude of the excess varied considerably across the PSIs. These VA findings are similar to those from a previously published study of nonfederal hospitals, despite differences between VA and non-VA systems. This study contributes to the literature measuring outcomes of medical errors and provides evidence that AHRQ PSIs may be useful indicators for comparison across delivery systems.

Length of stay and costs for asthma patients by hospital characteristics--a five-year population-based analysis

This study sets out to explore the relationship between hospital characteristics, asthma length of stay (LOS), and costs per discharge. The study adopts hospitalization data from the Taiwan National Health Insurance Research Database covering the period from 1997 to 2001. Study subjects were identified from the database by principal diagnosis of asthma or asthmatic bronchitis, with a total of 139,630 cases being included in the study. Multiple-regression analyses were performed to explore the relationship between LOS, costs per discharge and hospital characteristics, adjusting for age, gender, and discharge status of patients, as well as complications or comorbidities. The regression analyses showed that, compared with district hospitals, medical centers and regional hospitals have longer and more statistically significant LOS, as well as higher costs. Hospitals operating on a for-profit basis have shorter LOS and lower costs than public and not-for-profit hospitals. This study shows the existence of wide variations in LOS and costs per discharge for asthma hospitalizations, between the various types of hospitals in Taiwan.

Coordinating health care: lessons from Norway

OBJECTIVE

What influences the coordination of care between general practitioners and hospitals? In this paper, general practitioner satisfaction with hospital-GP interaction is revealed, and related to several background variables.

METHOD

A questionnaire was sent to all general practitioners in Norway (3388), asking their opinion on the interaction and coordination of health care in their district. A second questionnaire was sent to all the somatic hospitals in Norway (59) regarding formal routines and structures. The results were analysed using ordinary least squares regression.

RESULTS

General practitioners tend to be less satisfied with the coordination of care when their primary hospital is large and cost-effective with a high share of elderly patients. Together with the degree to which the general practitioner is involved in arenas where hospital physicians and general practitioners interact, these factors turned out to be good predictors of general practitioner satisfaction.

IMPLICATION

To improve coordination between general practitioners and specialists, one should focus upon the structural traits within the hospitals in different regions as well as creating common arenas where the physicians can interact.

Primary health care and hospital interactions: effects for hospital length of stay

AIMS

Norwegian healthcare services are divided between primary and secondary care providers. A growing problem is that every third patient of 75 years of age or more experiences an extended stay in a somatic hospital while waiting to be sent to primary healthcare services. The interaction between these two levels of healthcare services is analysed to examine the effect on a patient's length of stay in hospital.

METHODS

Recent studies have asserted that research on length of stay in hospital should include influential factors such as system variation and system characteristics, in addition to standardizing for case-mix. New organizational routines are identified in 50 Norwegian somatic hospitals. A multivariate linear regression is used in both a static and a dynamic model to explain variations in hospital length of stay and in additional length of stay (5% of stays are defined as outliers).

RESULTS

The study shows that newly specialized structures constructed to enhance the interaction between the two levels have had no effect. Length of stay is dependent on the capacity of the primary healthcare provider and on the share of elderly in the hospital catchment area, the type of patients, the procedure performed, and the characteristics of the hospital.

CONCLUSION

Variation in length of stay between hospitals is primarily explained by the capacity of primary healthcare providers. However, some support is found in the dynamic model that introduces the proposition that a hospital-owned hotel would decrease the length of stay of patients in hospital.

US hospital care for patients with HIV infection and pneumonia: the role of public, private, and Veterans Affairs hospitals in the early highly active antiretroviral therapy era

STUDY OBJECTIVES

We evaluated differences in processes and outcomes of HIV-related pneumonia care among patients in Veterans Affairs (VA), public, and for-profit and not-for-profit private hospitals in the United States. We compared the results of our current study (1995 to 1997) with those of our previous study that included a sample of patients receiving care during the years 1987 to 1990 to determine how HIV-related pneumonia care had evolved over the last decade.

SETTING/PATIENTS

The sample consisted of 1,231 patients with HIV infection who received care for Pneumocystis carinii pneumonia (PCP) and 750 patients with HIV infection who received care for community-acquired pneumonia (CAP) during the years 1995 to 1997.

MEASUREMENT

We conducted a retrospective medical record review and evaluated patient and hospital characteristics, HIV-related processes of care (timely use of anti-PCP medications, adjunctive corticosteroids), non-HIV-related processes of care (timely use of CAP treatment medications, diagnostic testing, ICU utilization, rates of endotracheal ventilation, placement on respiratory isolation), length of inpatient hospital stay, and inpatient mortality.

RESULTS

Rates of timely use of antibiotics and adjunctive corticosteroids for treating PCP were high and improved dramatically from the prior decade. However, compliance with consensus guidelines that recommend < 8 h as the optimal time window for initiation of antibiotics to treat CAP was lower. For both PCP and CAP, variations in processes of care and lengths of in-hospital stays, but not mortality rates, were noted at VA, public, private not-for-profit hospitals, and for-profit hospitals.

CONCLUSIONS

This study provides the first overview of HIV-related pneumonia care in the early highly active antiretroviral therapy era, and contrasts current findings with those of a similarly conducted study from a decade earlier. Quality of care for patients with PCP improved, but further efforts are needed to facilitate the appropriate management of CAP. In the third decade of the epidemic, it will be important to monitor whether variations in processes of care for various HIV-related clinical diagnoses among different types of hospitals persist.

Differences in length of stay in Veterans Health Administration and other United States hospitals: is the gap closing?

OBJECTIVES

Compare risk-adjusted length of stay (LOS) in VA and other United States (non-VA) hospitals and determine if relative differences in LOS have changed in recent years.

RESEARCH DESIGN

Retrospective cohort study.

PATIENTS

Patients with ten common medical diagnoses admitted to all VA hospitals and to non-VA hospitals included in the National Hospital Discharge Survey (NHDS) during 1996 through 1999.

DATA

Comparable data elements were obtained from VA administrative databases and the NHDS. LOS was adjusted for age, gender, marital status, and comorbidity. Comorbidity was assessed using a validated methodology that considers 30 conditions.

RESULTS

Unadjusted mean LOS was longer in VA than non-VA patient for all 4 years, in aggregate (7.1 vs. 4.9 days, respectively; P < 0.001), and for each year individually. However, the difference in mean LOS in VA and non-VA patients declined from 2.9 days in 1996 to 1.6 days in 1999. LOS in VA patients remained longer (P < 0.001) in linear regression analyses, adjusting for demographics and comorbidity. However, the difference in LOS declined from 28.5% (95% CI, 28.1%-29.0%) in 1996 to 17.0% (95% CI, 16.6%-17.4%) in 1999. These results were similar in analyses of individual geographic regions.

CONCLUSIONS

Risk-adjusted LOS was longer in VA hospitals than in other United States hospitals. However, differences in LOS narrowed between 1996 and 1999. These findings suggest that changes in the organization and delivery of VA health care in the mid-1990s may be closing the gap between the VA and other healthcare systems in hospital utilization.

Mortality and length of stay in a veterans affairs hospital and private sector hospitals serving a common market

OBJECTIVE

To compare severity-adjusted in-hospital mortality and length of stay (LOS) in a Veterans Administration (VA) hospital and private sector hospitals serving the same health care market.

DESIGN

Retrospective cohort study.

SETTING

A large VA hospital and 27 private sector hospitals in the same metropolitan area.

PATIENTS

Consecutive VA (N = 1,960) and private sector (N = 157,147) admissions in 1994 to 1995 with 9 high-volume diagnoses.

MEASUREMENTS

Severity of illness was measured using validated multivariable models that were based on data abstracted from medical records. Outcomes were adjusted for severity and compared in VA and private sector patients using multiple logistic or linear regression analysis.

MAIN RESULTS

Unadjusted mortality was similar in VA and private sector patients (5.0% vs 5.6%, respectively; P =.26), although mean LOS was longer in VA patients (12.7 vs 7.0 days; P <.001). Adjusting for severity, the odds of death in VA patients was similar (odds ratio [OR] 1.07; 95% confidence interval [95% CI], 0.74 to 1.54; P =.73). However, a larger proportion of deaths in VA patients occurred later during hospitalization (P <.001), and the odds of death in VA patients were actually lower (P <.05) in analyses limited to deaths during the first 7 (OR, 0.56) or 14 (OR, 0.63) days. Adjusted LOS was longer (P <.001) in VA patients for all 9 diagnoses.

CONCLUSIONS

If the current findings generalizable to other markets, hospital mortality, a widely used performance measure, may be similar or lower in VA and private sector hospitals serving the same markets. The longer LOS of VA patients may reflect differences in practice patterns and may be an important source of bias in comparisons of VA and private sector hospitals.

Do different case-mix measures affect assessments of provider efficiency? Lessons from the Department of Veterans Affairs

Although case-mix adjustment is critical for provider profiling, little is known regarding whether different case-mix measures affect assessments of provider efficiency. We examine whether two case-mix measures, Adjusted Clinical Groups (ACGs) and Diagnostic Cost Groups (DCGs), result in different assessments of efficiency across service networks within the Department of Veterans Affairs (VA). Three profiling indicators examine variation in resource use. Although results from the ACGs and DCGs generally agree on which networks have greater or lesser efficiency than average, assessments of individual network efficiency vary depending upon the case-mix measure used. This suggests that caution should be used so that providers are not misclassified based on reported efficiency.

Applying diagnostic cost groups to examine the disease burden of VA facilities: comparing the six "Evaluating VA Costs" study sites with other VA sites and Medicare

OBJECTIVES

To compare the disease burden of Veterans Health Administration (VA) patients at six study sites with all other VA patients and the Medicare population.

DESIGN

A 60% random sample of all VA veteran patients during federal fiscal year 1997 was obtained from administrative databases. A split-sample technique provided a 40% sample (n = 1,046,803) for development and a 20% sample (n = 524,461) for validation. We selected the six study sites from the 40% sample, yielding a total of 50,080 patients in those sites.

METHODS

We used Diagnostic Cost Groups to classify patients into clinical groupings based on age, gender, and International Classification of Diseases, Ninth Revision, Clinical Modification diagnoses. The Diagnostic Cost Group model produces relative risk scores that describe patients' expected resource use normalized to the Medicare population. We compared the severity of the six sites with each other and with all other VA facilities and the severity of VA patients with that of Medicare beneficiaries.

RESULTS

There were minor statistically significant differences between the study sites and all other VA facilities. Compared with the Medicare population, VA's population was younger and had lower expected resource use (relative risk scores were 1.0 and 0.76, respectively).

CONCLUSIONS

Disease burden of the six study sites is representative of all other VA facilities. Although lower relative risk scores suggest that VA patients are healthier than Medicare beneficiaries, when age is taken into account, scores are more comparable. Interpreting the expected resource utilization of the VA population against other benchmarks should be performed carefully.

Applying the 3M All Patient Refined Diagnosis Related Groups Grouper to measure inpatient severity in the VA

OBJECTIVES

To assess the severity level of acute inpatient care in the Veterans Health Administration (VA) using the 3M All Patient Refined Diagnosis Related Groups (APR-DRGs) Grouper and compare severity levels in the six study sites with other Veterans Affairs Medical Centers.

METHODS

Acute inpatient stays were generated based on bedsection movement information in VA Inpatient Medical SAS data sets from federal fiscal years 1997 and 1998. All nonacute bedsections were excluded. The APR-DRG Grouper generated APR-DRG and severity level for each acute inpatient stay using relevant VA data in a fixed format. Severity and length of stay (LOS) within each major APR-DRG (those accounting for at least 0.5% of all acute inpatient stays or days) were compared between study sites and other centers using z scores.

RESULTS

Of 315 APR-DRGs, 63 major groups accounted for more than two thirds of all stays and days of care in both years. The study sites were similar in average patient severity and LOS to other centers for most APR-DRGs. For those with significant differences, the six centers had shorter LOS and higher severity. The magnitude of differences was large in LOS and small in severity.

CONCLUSIONS

The study sites are generally representative of the overall VA acute inpatient stays. Some adjustments were needed to reflect that the six sites had relatively sicker patients and lower LOS in some of APR-DRGs when resource utilization estimations in the six sites were generalized to the entire VA system. The severity measure of the 3M APR-DRG Grouper can be adapted to the VA controlling for the complicated nature of VA inpatient care.

Methods for estimating private sector payments for VA acute inpatient stays

OBJECTIVES

To describe methods for estimating hypothetical private sector payments for Veterans Health Administration (VA) acute inpatient stays.

METHODS

We assumed all VA hospitalizations would have occurred under a hypothetical VA system that paid private sector providers but had the current benefit package for VA patients. We compared aggregate budgets for VA inpatient care (less physician salaries) at six VA hospitals over federal fiscal year 1999 to aggregated hypothetical private sector payments developed using VA diagnosis-related groups matched to metropolitan-based average Medicare payments. Counts of care came from the VA's statistical analysis system (SAS) inpatient files. Inpatient stays with both medical or surgical and psychiatric or rehabilitation care were counted as two stays. An external auditor conducted three reviews of VA coding practices during the study year, and the appropriateness of admissions was examined using a commercial utilization review tool.

RESULTS

For 30,518 inpatient discharges, hypothetical payments were $188 million, compared with the VA budget of $171 million. Fifteen of the 25 most frequent diagnosis-related groups in the VA were also in the top 25 for Medicare in 1998 and 1999. Audits established that the overall financial impact of VA coding problems was similar to that in the private sector.

DISCUSSION

Differences in organization, practice, and incentives limit estimates of the financial impact of shifting VA acute inpatient care to the private sector.

In-hospital mortality following coronary artery bypass graft surgery in Veterans Health Administration and private sector hospitals

OBJECTIVES

Compare severity-adjusted in-hospital mortality in patients undergoing coronary artery bypass graft surgery (CABG) in VA and private sector hospitals in two geographic regions.

RESEARCH DESIGN

Retrospective Cohort Study.

SUBJECTS

Consecutive male patients undergoing CABG from October 1993 to December 1996 in: 43 VA hospitals with cardiac surgery programs (n = 19,266); 32 hospitals in New York (NY) State (n = 44,247); and 10 hospitals in Northeast (NE) Ohio (n = 9696).

METHODS

Demographic and clinical data were abstracted from medical records. Logistic regression analysis identified 10 independent patient-level predictors (P <0.01) of in-hospital mortality: age, prior CABG, angioplasty before CABG, ejection fraction, diabetes, peripheral vascular disease, congestive heart failure (CHF), cerebrovascular disease, renal insufficiency, and chronic obstructive pulmonary disease (COPD).

RESULTS

Unadjusted mortality was higher in VA patients than in NY or NE Ohio patients (3.5% vs. 2.0%, and 2.2%, respectively). Mortality decreased (P <0.001) with increasing volume (3.6% in low [<500 cases], 3.0% in moderate [500-1000 cases], and 2.0% in high [>1000 cases] volume hospitals). Median volume was lower in VA than private sector hospitals (410 vs. 1520), and no VA hospitals were classified as high volume. Adjusting for patient-level predictors and volume, the odds of death was higher in VA patients, relative to private sector patients (OR, 1.34; 95% CI, 1.11-1.63; P <0.001). In stratified analyses, the odds of death in VA patients was similar in low volume hospitals (OR, 0.86; P = 0.39), but higher in moderate volume hospitals (OR, 1.50; P = 0.01).

CONCLUSIONS

VA hospitals had lower CABG volume than private sector hospitals in NY and NE Ohio, and higher in-hospital mortality. However, the difference in mortality was limited to moderate-volume hospitals. These findings suggest that hospital volume is an important modifier in comparisons of CABG mortality in VA and private sector hospitals. The higher mortality in VA hospitals may, in part, be caused by differences in surgical capacity and patient demand that lead to lower volume cardiac surgery programs.

The best of both worlds? Potential of hybrid prospective/concurrent risk adjustment

BACKGROUND

There remains considerable uncertainty about whether prospective or concurrent risk adjustment (RA) is preferable. Although concurrent models have better predictive power than prospective models, the large payments associated with concurrent RA create incentives for fraudulent coding. A hybrid strategy--in which prospective payments were used for patients with low expected costs and concurrent payments were available upon the diagnosis of a small number of common, expensive conditions--might improve predictive performance while requiring less auditing than fully concurrent RA. In addition, within-condition RA (using clinical data) for the selected conditions could further improve predictive power.

OBJECTIVES

To assess how such a hybrid strategy might perform, focusing on a small number of chronic, expensive conditions that are verifiable (hence auditable).

SUBJECTS AND MEASURES

All patients from seven health plans who had two complete years of utilization data were considered. RA models were estimated among patients younger than 65 (n = 319,209) using the Hierarchical Coexisting Conditions (HCC) model with and without stratification of the sample based on the presence of one or more of 100 verifiable, expensive, predictive conditions (VEP100). R2 and predictive ratios were calculated for each model studied.

RESULTS

Patients with a VEP100 condition (9.3% of the population) accounted for 84.3% of the variation in cost. R2 was 0.08 using a prospective HCC model on the entire population, but increased to 0.26 for a hybrid using prospective HCCs on the 90.7% of the sample without a VEP100 condition and a simple concurrent model consisting of dummy variables for each of the VEP100 conditions.

CONCLUSION

Combined with targeted auditing, a hybrid approach to RA could improve our ability to match payments to costs. However, because this would require additional, costly data collection, more research is needed to determine whether this benefit justifies the data collection and auditing burden.

Length of stay as a source of bias in comparing performance in VA and private sector facilities: lessons learned from a regional evaluation of intensive care outcomes

OBJECTIVES

Compare intensive care unit (ICU) mortality and length of stay (LOS) in a VA hospital and private sector hospitals and examine the impact of hospital utilization on mortality comparisons.

RESEARCH DESIGN

Retrospective cohort study.

SUBJECTS

Consecutive ICU admissions to a VA hospital (n = 1,142) and 27 private sector hospitals (n = 51,249) serving the same health care market in 1994 to 1995.

MEASURES

Mortality and ICU LOS were adjusted for severity of illness using a validated method that considers physiologic data from the first 24 hours of ICU admission. Mortality comparisons were made using two different multivariable techniques.

RESULTS

Unadjusted in-hospital mortality was higher in VA patients (14.5% vs. 12.0%; P = 0.01), as was hospital (28.3 vs. 11.3 days; P <0.001) and ICU (4.3 vs. 3.9 days; P <0.001) LOS. Using logistic regression to adjust for severity, the odds of death was similar in VA patients, relative to private sector patients (OR 1.16, 95% CI 0.93-1.44; P = 0.18). However, a higher proportion of VA deaths occurred after 21 hospital days (33% vs. 13%; P <0.001). Using proportional hazards regression and censoring patients at hospital discharge, the risk for death was lower in VA patients (hazard ratio 0.70; 95% CI 0.59-0.82; P <0.001). After adjusting for severity, differences in ICU LOS were no longer significant (P = 0.19).

CONCLUSIONS

Severity-adjusted mortality in ICU patients was lower in a VA hospital than in private sector hospitals in the same health care market, based on proportional hazards regression. This finding differed from logistic regression analysis, in which mortality was similar, suggesting that comparisons of hospital mortality between systems with different hospital utilization patterns may be biased if LOS is not considered. If generalizable to other markets, our findings further suggest that ICU outcomes are at least similar in VA hospitals.