Adjustment For Patient-level Factors Research Articles

Mortality and Morbidity in Adults With Rheumatic Heart Disease

Rheumatic heart disease (RHD) remains a public health issue in low- and middle-income countries (LMICs). However, there are few large studies enrolling individuals from multiple endemic countries. To assess the risk and predictors of major patient-important clinical outcomes in patients with clinical RHD. Multicenter, hospital-based, prospective observational study including 138 sites in 24 RHD-endemic LMICs. The primary outcome was all-cause mortality. Secondary outcomes were cause-specific mortality, heart failure (HF) hospitalization, stroke, recurrent rheumatic fever, and infective endocarditis. This study analyzed event rates by World Bank country income groups and determined the predictors of mortality using multivariable Cox models. Between August 2016 and May 2022, a total of 13 696 patients were enrolled. The mean age was 43.2 years and 72% were women. Data on vital status were available for 12 967 participants (94.7%) at the end of follow-up. Over a median duration of 3.2 years (41 478 patient-years), 1943 patients died (15% overall; 4.7% per patient-year). Most deaths were due to vascular causes (1312 [67.5%]), mainly HF or sudden cardiac death. The number of patients undergoing valve surgery (604 [4.4%]) and HF hospitalization (2% per year) was low. Strokes were infrequent (0.6% per year) and recurrent rheumatic fever was rare. Markers of severe valve disease, such as congestive HF (HR, 1.58 [95% CI, 1.50-1.87]; P < .001), pulmonary hypertension (HR, 1.52 [95% CI, 1.37-1.69]; P < .001), and atrial fibrillation (HR, 1.30 [95% CI, 1.15-1.46]; P < .001) were associated with increased mortality. Treatment with surgery (HR, 0.23 [95% CI, 0.12-0.44]; P < .001) or valvuloplasty (HR, 0.24 [95% CI, 0.06-0.95]; P = .042) were associated with lower mortality. Higher country income level was associated with lower mortality after adjustment for patient-level factors. Mortality in RHD is high and is correlated with the severity of valve disease. Valve surgery and valvuloplasty were associated with substantially lower mortality. Study findings suggest a greater need to improve access to surgical and interventional care, in addition to the current approaches focused on antibiotic prophylaxis and anticoagulation.

Changes of hospital-level variation in use of optimal medical treatment for long-term follow-up from a multi-level methodological approach

Abstract Background Contemporary guidelines documented that optimal medical therapy (OMT) should be prescribed to improve clinical outcome in patients with acute myocardial infarction (AMI). However, the changes of hospital-level variations in prescription rates of OMT for long-term follow-up are unknown. Purpose We aimed to investigate (1) impact of hospital-level variation on the use of each medication and OMT and (2) changes of hospital-level variation in prescription rates of individual medical therapy and OMT during the follow-up. Methods We examined hospital medications of 13,516 AMI patients without documented contraindications to antiplatelet agents, β-blockers (BB), angiotensin-converting enzyme inhibitors (ACE) /angiotensin II receptor blockers (ARB), or statins from the Korean Acute Myocardial Infarction Registry (KAMIR) – National Institute of Health (NIH) database. OMT was defined as use of all 4 indicated medications. Hierarchical generalized linear mixed models estimated hospital-level variation in prescription rates of OMT and each medication after adjustment for patient-level factors. Variation was explored with a median rate ratio (MRR), which estimates the relative difference in risk ratios of 2 hypothetically identical patients at 2 different hospitals. Results The prescription rates of OMT and each effective cardiac medications are presented in Figure. There were substantial decreases in the prescription rates for OMT (from 66.7% to 51.6%) and dual antiplatelet agents (from 99.1% to 50.4%) for 2 years follow-up. After adjustment for patient-level covariates, hospital-level MRR for OMT was 1.47 at discharge, indicating an 47% likelihood that 2 random hospitals would differ in treating identical AMI patients. As the prescription rates of OMT and each effective cardiac medications decreased, adjusted hospital-level MRR for OMT (from 1.47 to 1.22), BB therapy (from 1.35 to 1.15), ACE/ARB therapy (from 1.68 to 1.16), and statin therapy (from 1.19 to 1.06) decreased for 2-year follow-up, whereas MRR for dual antiplatelet therapy increased at 1 year (MRR; 1.61) and 2 years (MRR; 1.63) follow-up (Figure). Among patient-level factors, age &amp;gt;70 years, Killip class&amp;gt;2, obesity, hypertension, previous coronary artery disease, renal dysfunction, reduced ejection fraction, percutaneous coronary intervention was independently associated with the use of OMT. Conclusion Hospital-level variations in the use of OMT and each effective cardiac medications were observed at discharge, but weakened overtime, except for use of dual antiplatelet therapy.

Health-related quality of life associated with warfarin and direct oral anticoagulants in venous thromboembolism

IntroductionVenous thromboembolism (VTE) is commonly treated with oral anticoagulants, including warfarin or direct oral anticoagulants (DOACs). Although DOACs are associated with favorable treatment satisfaction, few studies have assessed whether quality of life differs between DOAC and warfarin users. Materials and methodsWe invited adults enrolled in two California-based integrated health care delivery systems and with a history of VTE between January 1, 2015 and June 30, 2018 to complete a survey on their experience with anticoagulants. Health-related quality of life (QOL) was assessed using the RAND 36-item Short Form Health Survey (SF-36), which measures QOL in 2 general component scores (physical and mental). We used multivariable linear regression to compare mean QOL component scores between DOAC-users and warfarin-users, adjusting for patient and clinical characteristics. ResultsOverall, 2230 patients (43.1 % women and 31.8 % >75 years of age) taking anticoagulants answered at least 1 question on the SF-36, 975 taking DOACs and 1255 taking warfarin. After adjustment for patient-level factors, there were no significant differences in either physical component scores (39.2 v 38.3, p = 0.24) or mental component scores (48.5 v 49.0, p = 0.42) between DOAC and warfarin users. ConclusionsHealth-related QOL did not significantly differ between DOAC and warfarin users with a history of VTE.

What factors influence differential uptake of NHS Health Checks, diabetes and hypertension reviews among women in ethnically diverse South London? Cross-sectional analysis of 63,000 primary care records

SummaryBackgroundUptake of health checks among women has not been examined in relation to patient and General Practitioner (GP) practice level factors. We investigated patient and practice level factors associated with differential uptake of health checks.MethodsPrimary care records from 44 practices in Lambeth for women aged 40-74 years old (N = 62,967) from 2000-2018 were analysed using multi-level logistic regression models. An odds ratio (OR) >1 indicates increased occurrence of no health check.FindingsThe mean age (IQR) of the included female sample (aged 40-74 years) was 52.9 years (45.0-59.0). Adjusted for patient-level factors (age, ethnicity, English as first language, overweight/obesity, smoking, attendance to GP practices, and co-morbidity), the odds of non-uptake of health checks were higher for Other White (OR 1.24, 95% confidence interval 1.17-1.33), and Other ethnicity (1.20, 1.07-1.35) vs. White British. It was also higher for 50-69 year olds (1.55, 1.47-1.62), 70-74 year olds (1.60, 1.49-1.72) vs. 40-49 year olds. These ORs did not change on adjustments for practice level factors (proportion of patients living in deprived areas, proportion of patients with ≥1 chronic condition, ≥3 emergency diabetes admissions annually, GP density/1000 patients, quality outcome framework score of ≥ 95%, and patient satisfaction scores of ≥80%). Non-uptake was lower for Black Caribbeans, Bangladeshis, overweight/obese patients, frequent practice attenders and comorbid patients.InterpretationDifferential uptake in health checks remained after adjustment for patient and practice level factors. Better measures of social determinants of health and of practice context are needed.FundingNIHR Research for Patient Benefit Programme (NIHR202769).

Abstract 13686: Variation in Time in Therapeutic Range for Cooling for Out-of-Hospital Cardiac Arrest

Introduction: For comatose survivors of out-of-hospital cardiac arrest (OHCA), targeted temperature management (TTM) is strongly recommended with a goal temperature of 32-36C for a period of at least 24 hours. However, adherence to this target in clinical practice remains unknown. We developed time-in-therapeutic range (TTR) as a treatment metric for patients receiving TTM and evaluated patient- and site-level variation in TTR. Methods: We used data from the Resuscitation Outcomes Consortium-CCC trial which included patients with OHCA across 10 North American sites during 2011-2015. We identified patients who underwent TTM for &gt;12 hours. Serial temperature measures were evaluated between hypothermia start and end times with temperatures between consecutive measures imputed using a linear interpolation method. TTR was defined as percent of time between 32C and 36C during TTM (Fig A). Site was defined based on trial clusters, which represented hospitals served by the same EMS agency. Site-level variation in TTR&lt;90% was evaluated in hierarchical logistic regression using median odds ratio (OR), after adjustment for patient-level factors. Results: A total of 2,695 patients across 49 clusters were included with a median of 45 (IQR: 34 - 52) patients per cluster. The median duration of hypothermia was 23 (IQR: 21 - 24) hours with a median time outside therapeutic range of 0.9 (IQR: 0.0 - 4.2) hours. The median TTR was 96.1% but 1,654 (61%) patients had at least one temperature outside the therapeutic range and 991 (37%) patients had a TTR &lt;90%. There was large variation across sites in the proportion of patients with TTR&lt;90%, ranging from 10% to 68%, with a median OR of 1.74 (Fig B). Conclusions: Within a large randomized controlled trial, more than 1 in 3 OHCA patients treated with TTM had a TTR &lt;90%, with large variation in TTR across sites. These findings highlight an urgent need to focus on improving quality of TTM in clinical practice.

Abstract 183: Association of Hospital-Level Factors With Variation in Survival After Out-Of-Hospital Cardiac Arrest Between Receiving Hospitals in Japan

Background: Patient outcomes after out-of-hospital cardiac arrest (OHCA) varies at multiple levels (geographic regions, emergency medical service agencies, and receiving hospitals). However, the contribution of hospital-level factors to the outcome variation is unclear. We aimed to investigate the association between hospital-level factors and variation in outcomes after OHCA between hospitals. Methods: We performed secondary analysis of the Japanese Association for Acute Medicine OHCA Registry, a prospective multi-center registry. We included adults (≥18 years) with OHCA of medical origin between June 2014 and December 2015. The primary outcome was 1-month survival and secondary outcome was 1-month survival with favorable functional status, defined as Cerebral Performance Category scale 1 or 2. We constructed two multivariable hierarchical logistic regression models for each outcome: (1) patient-level factors and (2) patient- and hospital-level factors. Hospital-level factors included the number of medical staff involved in resuscitation, proportion of patients who received targeted temperature management among those with return of spontaneous circulation, and proportion of patients who received extracorporeal cardiopulmonary resuscitation among those with shockable rhythm. We calculated adjusted outcomes for each hospital and median odds ratios (MORs) to evaluate the effects of hospital-level factors on between-hospital variation in outcomes. Results: We analyzed the data of 9,303 adults with OHCA of medical origin treated at 67 hospitals. After adjustment for patient-level factors, 1-month survival was 0.02-0.14 (adjusted MOR, 1.36; 95% credible interval (CI), 1.17-1.77). After adjustment for patient- and hospital-level factors, the adjusted MOR was 1.14 (95% CI, 1.05-1.33). Similarly, the adjusted MORs for favorable functional outcomes were 1.48 (95% CI 1.18-2.12) with adjustment for patient-level factors and 1.30 (95% CI 1.07-1.80) with adjustment for patient- and hospital-level factors. Conclusions: With adjustment for hospital-level factors, we noted attenuation of variation in patient outcomes after OHCA across hospitals, suggesting that hospital-level factors may partly explain outcome variations.

Variation in survival after out-of-hospital cardiac arrest between receiving hospitals in Japan: an observational study

ObjectivesPatient outcomes after out-of-hospital cardiac arrest (OHCA) varies at multilevel (geographical regions, emergency medical service agencies and receiving hospitals) in the USA. However, it remains unclear whether there is a...

Association of Herpes Simplex Virus Testing with Hospital Length of Stay for Infants ≤60 Days of Age Undergoing Evaluation for Meningitis.

Although neonatal herpes simplex virus (HSV) causes significant morbidity, utilization of the cerebrospinal fluid (CSF) HSV polymerase chain reaction (PCR) test remains variable. Our objective was to examine the association of CSF HSV PCR testing with length of stay (LOS) in a 20-center retrospective cohort of hospitalized infants aged ≤60 days undergoing evaluation for meningitis after adjustment for patient-level factors and clustering by center. Of 20,496 eligible infants, 7,399 (36.1%) had a CSF HSV PCR test performed, and 46 (0.6% of those tested) had a positive test. Infants who had a CSF HSV PCR test performed had a 23% longer hospital LOS (incident rate ratio 1.23; 95% CI: 1.14-1.33). Targeted CSF HSV PCR testing may mitigate the impact on LOS for low-risk infants.

Why do ambulance services have different non-transport rates? A national cross sectional study.

BackgroundSome patients calling ambulance services (known as Emergency Medical Services internationally) are not transported to hospital. In England, national ambulance quality indicators show considerable variation in non-transport rates between the ten large regional ambulance services. The aim of this study was to explain variation between ambulance services in two types of non-transport: discharge at scene and telephone advice.MethodsMixed model logistic regressions using one month of data (November 2014) from the Computer Aided Despatch systems of the ten large regional ambulance services in England.Results41% (251 677/615 815) of patients calling ambulance services were not transported to hospital. Most were discharged at scene after attendance by an ambulance (29% n = 182 479) and a small percentage were given telephone advice (7% n = 40 679). Discharge at scene rates varied by patient-level factors e.g. they were higher for elderly patients, where the reason for calling was falls, and for patients attended by paramedics with extended skills. These patient-level factors did not explain variation between ambulance services. After adjustment for patient-level factors, the following ambulance service level factors explained variation in discharge at scene rates: proportion of patients attended by paramedics with extended skills (odds ratio 1.05 (95% CI 1.04, 1.07)), the perception of ambulance service staff that paramedics with extended skills were established and valued within the workforce (odds ratio 1.84 (1.45, 2.33), and the perception of ambulance service staff that senior management viewed non-transport as risky (odds ratio 0.78 (0.63, 0.98)). Variation in telephone advice rates could not be explained.ConclusionsVariation in discharge at scene rates was explained by differences in workforce configuration and managerial motivation, factors that are largely modifiable by ambulance services.

Understanding variation in ambulance service non-conveyance rates: a mixed methods study

BackgroundIn England in 2015/16, ambulance services responded to nearly 11 million calls. Ambulance Quality Indicators show that half of the patients receiving a response by telephone or face to face were not conveyed to an emergency department. A total of 11% of patients received telephone advice only. A total of 38% of patients were sent an ambulance but were not conveyed to an emergency department. For the 10 large ambulance services in England, rates of calls ending in telephone advice varied between 5% and 17%. Rates of patients who were sent an ambulance but not conveyed to an emergency department varied between 23% and 51%. Overall non-conveyance rates varied between 40% and 68%.ObjectiveTo explain variation in non-conveyance rates between ambulance services.DesignA sequential mixed methods study with five work packages.SettingTen of the 11 ambulance services serving &gt; 99% of the population of England.Methods(1) A qualitative interview study of managers and paramedics from each ambulance service, as well as ambulance commissioners (totalling 49 interviews undertaken in 2015). (2) An analysis of 1 month of routine data from each ambulance service (November 2014). (3) A qualitative study in three ambulance services with different published rates of calls ending in telephone advice (120 hours of observation and 20 interviews undertaken in 2016). (4) An analysis of routine data from one ambulance service linked to emergency department attendance, hospital admission and mortality data (6 months of 2013). (5) A substudy of non-conveyance for people calling 999 with breathing problems.ResultsInterviewees in the qualitative study identified factors that they perceived to affect non-conveyance rates. Where possible, these perceptions were tested using routine data. Some variation in non-conveyance rates between ambulance services was likely to be due to differences in the way rates were calculated by individual services, particularly in relation to telephone advice. Rates for the number of patients sent an ambulance but not conveyed to an emergency department were associated with patient-level factors: age, sex, deprivation, time of call, reason for call, urgency level and skill level of attending crew. However, variation between ambulance services remained after adjustment for patient-level factors. Variation was explained by ambulance service-level factors after adjustment for patient-level factors: the percentage of calls attended by advanced paramedics [odds ratio 1.05, 95% confidence interval (CI) 1.04 to 1.07], the perception of ambulance service staff and commissioners that advanced paramedics were established and valued within the workforce of an ambulance service (odds ratio 1.84, 95% CI 1.45 to 2.33), and the perception of ambulance service staff and commissioners that senior management was risk averse regarding non-conveyance within an ambulance service (odds ratio 0.78, 95% CI 0.63 to 0.98).LimitationsRoutine data from ambulance services are complex and not consistently collected or analysed by ambulance services, thus limiting the utility of comparative analyses.ConclusionsVariation in non-conveyance rates between ambulance services in England could be reduced by addressing variation in the types of paramedics attending calls, variation in how advanced paramedics are used and variation in perceptions of the risk associated with non-conveyance within ambulance service management. Linking routine ambulance data with emergency department attendance, hospital admission and mortality data for all ambulance services in the UK would allow comparison of the safety and appropriateness of their different non-conveyance rates.FundingThe National Institute for Health Research Health Services and Delivery Research programme.

Mortality risks associated with emergency admissions during weekends and public holidays: an analysis of electronic health records

SummaryBackgroundWeekend hospital admission is associated with increased mortality, but the contributions of varying illness severity and admission time to this weekend effect remain unexplored.MethodsWe analysed unselected emergency admissions to four Oxford University National Health Service hospitals in the UK from Jan 1, 2006, to Dec 31, 2014. The primary outcome was death within 30 days of admission (in or out of hospital), analysed using Cox models measuring time from admission. The primary exposure was day of the week of admission. We adjusted for multiple confounders including demographics, comorbidities, and admission characteristics, incorporating non-linearity and interactions. Models then considered the effect of adjusting for 15 common haematology and biochemistry test results or proxies for hospital workload.Findings257 596 individuals underwent 503 938 emergency admissions. 18 313 (4·7%) patients admitted as weekday energency admissions and 6070 (5·1%) patients admitted as weekend emergency admissions died within 30 days (p<0·0001). 9347 individuals underwent 9707 emergency admissions on public holidays. 559 (5·8%) died within 30 days (p<0·0001 vs weekday). 15 routine haematology and biochemistry test results were highly prognostic for mortality. In 271 465 (53·9%) admissions with complete data, adjustment for test results explained 33% (95% CI 21 to 70) of the excess mortality associated with emergency admission on Saturdays compared with Wednesdays, 52% (lower 95% CI 34) on Sundays, and 87% (lower 95% CI 45) on public holidays after adjustment for standard patient characteristics. Excess mortality was predominantly restricted to admissions between 1100 h and 1500 h (pinteraction=0·04). No hospital workload measure was independently associated with mortality (all p values >0·06).InterpretationAdjustment for routine test results substantially reduced excess mortality associated with emergency admission at weekends and public holidays. Adjustment for patient-level factors not available in our study might further reduce the residual excess mortality, particularly as this clustered around midday at weekends. Hospital workload was not associated with mortality. Together, these findings suggest that the weekend effect arises from patient-level differences at admission rather than reduced hospital staffing or services.FundingNIHR Oxford Biomedical Research Centre.

Is There an Increasing Regionalization of Surgical Repair of Craniosynostosis Procedures Into Teaching Hospitals? Implications of Regionalization.

Objective The objective of the present study is to examine whether surgical repairs for craniosynostosis have been regionalized to teaching hospitals over the 8-year period from 2003 to 2010. Design Retrospective analysis of hospital discharge database. Setting Nationwide Inpatient Sample for years 2003 to 2010. All patients aged up to 3 years who had a surgical repair for craniosynostosis were selected. Interventions Surgical repair for craniosynostosis. Main Outcome Measures Performance of surgery in a teaching hospital. Results During the study period (years 2003 to 2010), a total of 19,417 patients aged up to 3 years underwent a surgical repair for craniosynostosis. The number of surgical procedures increased during the study period. It ranged from 1628 procedures in year 2003 to 3001 procedures during 2010. Data show that 83.3% of all procedures in 2003 were performed in teaching hospitals; whereas, 97.5% of procedures in 2010 were performed in teaching hospitals. Following adjustment for patient-level factors, year 2010 was associated with increased odds of having the surgical procedures performed in a teaching hospital as opposed to a nonteaching hospital when compared with year 2003 (odds ratio = 10.43, 95% confidence interval, 1.10 to 98.98; P = .04). Conclusions An increasing proportion of surgical repairs of craniosynostosis are performed in teaching hospitals, suggesting there is an increasing concentration of these complex surgical procedures in select centers. As more longitudinal data become available, the relative benefits and drawbacks associated with regionalization of surgical repairs of craniosynostosis should be examined.

Hospital variation in thrombolysis times among patients with acute ischemic stroke: the contributions of door-to-imaging time and imaging-to-needle time.

Given the limited time window available for treatment with tissue plasminogen activator (tPA) in patients with acute ischemic stroke, guidelines recommend door-to-imaging time (DIT) within 25 minutes of hospital arrival and door-to-needle (DTN) time within 60 minutes for patients with acute ischemic stroke. Despite improvements in DITs, DTN times for tPA treatment in patients with acute ischemic stroke remain suboptimal. To examine the contributions of DIT and imaging-to-needle (ITN) time to delays in timely delivery of tPA to patients with acute ischemic stroke and to assess between-hospital variation in DTN times. A cohort analysis of 1193 patients having acute ischemic stroke treated with intravenous tPA between January 2009 and December 2012. Multilevel linear regression models included random effects for 25 Michigan hospitals participating in the Paul Coverdell National Acute Stroke Registry. The primary outcome was a continuous measure of DTN time, in minutes, from emergency department arrival to thrombolytic delivery. The mean age was 68.1 years, the median National Institutes of Health Stroke Scale score was 11.0 (interquartile range, 6-17), 51.4% were female, and 37.5% were of nonwhite race/ethnicity.The mean (SD) DTN time was 82.9 (35.4) minutes, the mean (SD) DIT was 22.8 (15.9) minutes, and the mean (SD) ITN time was 60.1 (32.3) minutes. Most patients (68.4%) had DIT within 25 minutes, while 28.7% had DTN time within 60 minutes. Hospital variation accounted for 12.7% of variability in DTN times. Neither annual stroke volume nor primary stroke center designation was a significant predictor of shorter DTN time. Patient factors (age, sex, race/ethnicity, arrival mode, onset-to-arrival time, and stroke severity) explained 15.4% of the between-hospital variation in DTN times. After adjustment for patient-level factors, DIT explained 10.8% of the variation in hospital risk-adjusted DTN times, while ITN time explained 64.6%. Compared with DIT, ITN time is a much greater source of variability in hospital DTN times and is a more common contributor to delays in timely tPA therapy for acute ischemic stroke. More attention is needed to determine systems changes that can decrease ITN time for patients with acute ischemic stroke.

Invited Commentary

Invited Commentary

Treatment Intensity in Child and Adolescent Mental Health Services and Health Care Reform in Norway, 1998–2006

Treatment Intensity in Child and Adolescent Mental Health Services and Health Care Reform in Norway, 1998–2006

An Assessment of the Quality of Mammography Care at Facilities Treating Medically Vulnerable Populations

Women in medically vulnerable populations, including racial and ethnic minorities, socioeconomically disadvantaged, and residents of rural areas, experience higher breast cancer mortality than do others. Whether mammography facilities that treat vulnerable women demonstrate lower quality of care than other facilities is unknown. To assess the quality of mammography women receive at facilities characterized as serving a high proportion of medically vulnerable populations. We prospectively collected self-reported breast cancer risk factor information, mammography interpretations, and cancer outcomes on 1,579,929 screening mammography examinations from 750,857 women, aged 40-80 years, attending any of 151 facilities in the Breast Cancer Surveillance Consortium between 1998 and 2004. To classify facilities as serving medically vulnerable populations, we used 4 criteria: educational attainment, racial/ethnic minority, household income, and rural/urban residence. After adjustment for patient-level factors known to affect mammography accuracy, facilities serving vulnerable populations had significantly higher mammography specificity than did other facilities: ie, those serving a higher proportion of women who were minorities [odds ratio (OR): 1.32; 95% confidence interval (CI): 1.01-1.73], living in rural areas (1.45; 1.15-1.73), and with lower household income (1.33; 1.05-1.68). We observed no statistically significant differences between facilities in mammography sensitivity. Facilities serving high proportions of vulnerable populations provide screening mammography with equal or better quality (as reflected in higher specificity with no corresponding decrease in sensitivity) than other facilities. Further research is needed to understand the mechanisms underlying these findings.

Delays in Suspicion and Isolation Among Hospitalized Persons With Pulmonary Tuberculosis at Public and Private US Hospitals During 1996 to 1999

While prior studies have shown that public and private hospitals differ in their rates of suspicion and isolation of patients who are at risk for tuberculosis (TB), no study has investigated whether this variation is due to differences in the impact of patient case-mix on hospitals or to variations attributable to specific hospital practice patterns. To investigate patient-level and hospital-level factors associated with delays in TB suspicion and isolation among inpatients with pulmonary TB disease. Retrospective cohort study of patients hospitalized with culture-positive pulmonary TB during 1996 to 1999. Patients with culture-proven pulmonary TB treated at three public hospitals (765 patients) and seven not-for-profit private hospitals (172 patients) in Chicago, Los Angeles, and southern Florida that provided care for five or more patients with TB per year during the study period. Two-day rates (within 48 h from admission) of acid-fast bacilli (AFB) smear orders and 1-day rates (within 24 h from admission) of TB isolation. Two-day rates of ordering AFB smears were > 80% at three public and two private hospitals vs 65 to 75% at five private hospitals. One-day rates of TB isolation at the three public hospitals were 64%, 79%, and 86%, respectively, vs 39 to 58% at the seven private hospitals. Delays of > 2 days in ordering AFB smears were associated with patient-level factors: absence of cough (adjusted odds ratio [AOR], 6.02; 95% confidence interval [CI], 3.82 to 9.52), cavitary lung lesion (AOR, 5.17; 95% CI, 1.98 to 13.50), night sweats (AOR, 3.38; 95% CI, 1.90 to 5.99), chills (AOR, 1.70; 95% CI, 1.01 to 2.88), and female gender (AOR, 1.66; 95% CI, 1.06 to 2.60). Delays of > 1 day in ordering pulmonary isolation were associated with patient-level factors: absence of cough (AOR, 3.40; 95% CI, 2.31 to 5.03), cavitary lung lesion (AOR, 2.66; 95% CI, 1.57 to 4.50), night sweats (AOR, 1.98; 95% CI, 1.35 to 2.92), and history of noninjecting drug use (AOR, 1.86; 95% CI, 1.16 to 2.99) and one hospital-level factor: receiving care at a nonpublic hospital. Even after adjustment for patient-level factors, TB patients at private hospitals were half as likely as those at public hospitals to be placed in pulmonary isolation (AOR, 0.47; 95% CI, 0.30 to 0.72), while odds of suspecting TB in these same patients were similar at both hospitals. Private hospitals should order TB isolation for all patients for whom AFB smears are ordered, a policy that has been instituted previously at public hospitals in our study.