Magazines/Newspapers

A framework for quality improvement: An analysis of factors responsible for improvement at hospitals participating in the Can Rapid Risk Stratification of Unstable Angina Patients Suppress Adverse Outcomes with Early Implementation of the ACC/AH

Description
A framework for quality improvement: An analysis of factors responsible for improvement at hospitals participating in the Can Rapid Risk Stratification of Unstable Angina Patients Suppress Adverse Outcomes with Early Implementation of the ACC/AHA
Published
of 15
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
  Outcomes, Health Policy, and Managed Care  A framework for quality improvement: An analysis offactors responsible for improvement at hospitalsparticipating in the Can Rapid Risk Stratification ofUnstable Angina Patients Suppress Adverse Outcomes with Early Implementation of the ACC/AHA Guidelines(CRUSADE) quality improvement initiative Seth W. Glickman, MD, MBA, a,b  William Boulding, PhD,  b Richard Staelin, PhD,  b  Jyotsna Mulgund, MS, c Matthew T. Roe, MD, MHS, c Barbara L. Lytle, MS, c  John S. Rumsfeld, MD, d  W. Brian Gibler, MD, e E. Magnus Ohman, MD, c Kevin A. Schulman, MD, a,b and Eric D. Peterson, MD, MPH c  Durham, NC; Denver, CO; and Cincinnati, OH Background  Hospitals are under increasing pressure to improve their quality of care. However, a key questionremains: how can hospitals best design and implement successful quality improvement (QI) programs? Hospitals currently employ a variety of QI initiatives but have little empirical evidence on which to base their quality efforts. Methods  Wedesignedandappliedahospitalcross-sectionalsurveyto212hospitalsparticipatinginCRUSADE(CanRapidRisk Stratification of Unstable Angina Patients Suppress Adverse Outcomes with Early Implementation of the American College of Cardiology/American Heart Association Guidelines), a voluntary QI initiative of patients with non – ST-segment elevation acutecoronarysyndromes(NSTEACS).Wefactoranalysisandanordinaryleastsquaresregressionmodeltodeterminethekeyhospitalfactors most associated with unexpected improvements in institutional QI in the treatment of NSTE ACS. Results  From 2002 to 2004, the following factors had a significant association with unexpected increases in the 2004QI in NSTE ACS treatment: the use of CRUSADE QI tools, clinical commitment to quality by a cardiology coadvocate,institutional financial commitment to quality, and barriers to QI related to resource availability and cultural resistance to change(all  P   b  .10). Of these factors, optimal use of CRUSADE QI tools was associated with the highest absolute improvement inprocess adherence score relative to other factors. Conclusions  We identified several institutional factors associated with improved quality of care in the treatment of high- risk NSTE ACS. We hope that this evidence-based framework will help guide the development and implementation of future QIprograms in order to improve the institutional quality of care for NSTE ACS. (Am Heart J 2007;154:1206-20.) Hospitals are under increasing pressure to improvetheir quality of clinical care. This pressure follows in the wake of multiple studies documenting wide gapsbetween care recommendations and actual care deliv- ered for many conditions. 1,2 These quality gaps havedriven greater demand for accountability among provi- ders, as exemplified by public reporting and pay-for- performance movements. 3,4  Although these externalforces provide strong impetus for quality improvement(QI), a key question that remains is how to design andimplement a successful QI initiative.Industrial quality management techniques have beentouted for decades as a means to improving health carequality, but evidence supporting their effectiveness incommunity settings is limited. 5-7  As a result, hospitals facesignificant challenges in implementing quality initiativesrelated to: (1) identifying priorities, (2) developingsustainable processes, (3) identifying the appropriate From the   a  Center for Clinical and Genetic Economics, Duke University Medical Center,Durham, NC,  b  Fuqua School of Business, Duke University, Durham, NC,  c  Duke Clinical Research Institute, Duke University, Durham, NC,  d  Section of Cardiology, Denver VAMedical Center, Denver, CO, and   e  Department of Emergency Medicine, University of Cincinnati College of Medicine, Cincinnati, OH.Submitted April 24, 2007; accepted August 2, 2007.Reprint requests: Eric D. Peterson, MD, MPH, Duke Clinical Research Institute, PO Box 17969, Durham, NC 27715.E-mail: peter016@mc.duke.edu 0002-8703/$ - see front matter © 2007, Mosby, Inc. All rights reserved.doi:10.1016/j.ahj.2007.08.001  framework for successful implementation of QI initia- tives, and (4) demonstrating a concrete and positiveconnection between QI initiatives and clinical careperformance outcomes.Major factors thought to influence quality includeadministrative and clinical support for QI, organizationalculture, performance improvement initiatives, datafeedback, and contextual factors (such as hospitalownership status and academic affiliation). 8-10 Hospitalscurrently employ a variety of these QI initiatives buthave little empirical evidence on which to basetheir efforts.The purpose of this paper is to investigateorganizational, cultural, and operational featuresassociated with successful performance among a largeand diverse cohort of hospitals. Specifically, wedeveloped and applied a hospital cross-sectionalsurvey to 284 hospitals participating in the CRUSADE(Can Rapid Risk Stratification of Unstable AnginaPatients Suppress Adverse Outcomes with Early Implementation of the American College of Cardiol- ogy/American Heart Association Guidelines) QI initia- tive. We then determined key hospital factors mostassociated with institutional quality improvement inthe treatment of patients with non – ST-segmentelevation acute coronary syndromes (NSTE ACS). Theresults of this study are intended to help hospitalsfocus future quality improvement efforts. Methods Data sources Clinical process of care and patient characteristic informa- tion was obtained from the CRUSADE database. CRUSADE isan ongoing voluntary observational data collection and QIinitiative, which began January 1, 2001. 2,11-13 CRUSADEcenters collect and submit clinical information regardinginhospital care and outcomes of patients with NSTE ACS withhigh-risk clinical features. Data are abstracted by a trained datacollector at each hospital, using standardized definitions. Variables collected include demographic and clinical informa- tion, medical therapies and associated major contraindications,use and timing of cardiac procedures, laboratory results,hospital characteristics, inhospital outcomes, and dischargetherapies and interventions.Each CRUSADE hospital is asked to designate a cardiology coadvocate (MD), an emergency medicine coadvocate (MD),and a QI coordinator. These coadvocates are identified by hospitals as  “ champions ”  of improving high-risk acutecoronary syndrome (ACS) care, who lead local efforts andutilize the CRUSADE platform to identify gaps and promoteimprovement in ACS care. A variety of QI tools are providedto sites, including regional educational meetings and tele- conferences, site feedback and reports (including bench- marking against national standards and other CRUSADEhospitals), and numerous educational interventions (includingtreatment algorithms, standing order sets, and risk stratifica- tion tools). Survey   We developed a proposed framework for hospital QI effortsbased on previous medical and business literature. 14-17 Thisframework incorporated both key structural (ie, administrative,clinical, and financial commitment to quality; culture) andprocess (ie, QI tools, operations, etc) features. This framework  was incorporated into the Quality Improvement Environment(QIE) Interview, a multipart telephone interview survey conducted at284hospitals fromJanuary toNovember2004.Theinterviewwasadministered byasingle trained surveyresearcher and administered to the CRUSADE study coordinator at eachhospital. Any question that the study coordinator was unable toanswer was referred to more appropriate individuals at eachhospital. The answers to the survey questions were coded andstored in a database. Study population The initial population consisted of 96649 NSTE ACS patientstreated at 284 hospitals that participated in the QIE survey between January 1, 2002, and January 1, 2004. Of 284 sites, 212(75%) submitted at least 10 patient records in both 2002 and2004. These 212 sites, providing a total of 83155 NSTE ACSpatients, formed our analysis population. Study definitions Nine individual American College of Cardiology/AmericanHeart Association class I (useful and effective) guideline- recommended therapies (among patients eligible to receivethem) were used to calculate composite process adherencescores for the treatment of NSTE ACS at each hospital.These included 4 process-of-care measures (aspirin, β -blockers, heparin, and intravenous glycoprotein IIb/IIIainhibitors) used within the first 24 hours, as well as5 discharge regimens (aspirin,  β -blockers, clopidogrel,angiotensin-converting enzyme inhibitors, and lipid-loweringagents). Patient eligibility for relevant measures was deter- mined according to defined American College of Cardiology/  American Heart Association guideline indications andreported contraindications. Patients who died during thefirst 24 hours were excluded from the denominator for assessment of acute care processes, and those dying at any time during their hospital stay were excluded from thedischarge care assessment. Transfer patients were includedin this analysis. For patients transferred into a hospital(n = 16373), medications assessment within the first24 hours and at discharge were included in the analysisfor the receiving hospital. Patients transferred out of ahospital (n = 8693) were included for assessment of acutecare processes but excluded from discharge care assessment. Statistical analysis Dependent variable.  Our dependent measure is based onthe institution's quality performance, that is, adherence toclinical guidelines, over the period 2002 to 2004. Previousstudy shows that these adherence performance scores arerelated to patient outcomes (mortality). 2  We calculated acomposite process measure score for NSTE ACS processes yearly from 2002 to 2004 for each hospital. We did this by aggregating the number of times each therapy was adminis- tered divided by the sum of total eligible opportunities for all Glickman et al  1207  American Heart JournalVolume 154, Number 6  patients at a hospital in a given period. This is the samemethod used by the Center for Medicare and MedicaidServices to calculate hospital level and composite processadherence scores in the Hospital Quality ImprovementDemonstration (pay-for-performance) program. 18 To assess the impact of various QI factors on performance, we defined our outcome of interest as the  change  inadherence at a site  in excess of that site ’   s expected change ,given its baseline (2002) score. To obtain this  “ unexpected ” change (ie, change greater than or less than that expected), we used the residuals obtained from the ordinary leastsquares regression of a site's adherence score in 2004 on itsscore in 2002. By controlling for the 2002 score and by imputing the expected increase in the 2004 score, we are notonly able to adjust for any outside effects that affect allhospitals but also are able to conduct  “  within-hospital ” analyses. This within-hospital analysis allows us to circumventmany of the issues associated with straight cross-sectionalstudies. In other words, by taking this approach, we are ableto remove all fixed effects, such as hospital size, patientcharacteristics outside influences that apply to a givenhospital, and others, that would likely have a direct effect onQI score. Thus, our results can be interpreted in terms of theeffects of the relevant independent variable on QI after adjusting for hospital effects that remain fixed during theperiod of analysis. Independent variables.  We formed the independent vari- ables used in our analysis, as follows. First, questions from theQIE survey were reviewed and organized by the research teaminto broad factors to form a taxonomy for classifying QI effortsfor NSTE ACS ( Table II ). Most survey answers were coded on anumerical scale; for each factor, the answers to individualquestions were summated to create a final aggregate score. For  variables that were missing values for any hospitals, mean valuesfrom available data were imputed. The imputed values werethen used in creating the factors. We then evaluated theCronbach  α  coefficients to ascertain that variables within afactor were positively correlated and that the component variables were consistent with the latent factor purported(see Table II for the percentage missing data and Cronbach  α for each factor).The question pertaining to major hospital barriers to QI wasan open-ended one. In order to provide insight regardingthese barriers, we coded the barrier that each hospitalreported to be the most restrictive in hampering QI efforts.The answers to this question were categorized according to 4broad-based categories (resource availability, operational inef- fectiveness, cultural factors (including organizational andphysician resistance to change), and answers that werenondiagnostic for any specific categorization. The responsesand category definitions were given to 2 independent coders, who organized answers into 1 of the 4 categories. There wasapproximately 90% consistency in coding between theseindividuals; final categorization of remaining discrepancies wascompleted by the research team. As a starting point, we hypothesized that hospitals engagedin few or no QI efforts were unlikely to identify any barriers,our logic being that, without any activity, the hospital wouldbe unlikely to encounter any barriers. In contrast, thosehospitals that were hampered by operational barriers (thestarting point for QI efforts) were considered more advancedthan those sites that could not quantify any barriers but werestill felt to be distant from the final QI barrier of resourceconstraints. Using similar logic, we hypothesized thathospitals that had surmounted operational barriers and hadencountered cultural barriers would show more unexpectedimprovement than those hampered by operational barriersbut still less than those that now indicated a resourceavailability barrier. In other words, there is a hierarchy of quality barriers and implied quality efforts. Hospitals addressthe easiest barriers first before moving on to more difficultones. Presumably, hospitals that report facing more difficultquality barriers have overcome easier barriers with their QIefforts and, therefore, are more advanced than those of hospitals facing easier barriers. This hierarchy ranges from nobarriers identified, to operational barriers, to cultural andphysician resistance to change, to resource constraints.Factors with Cronbach  α  N 0.40 were then retained in our analysis model. In general, a Cronbach  α  N 0.70 is thought torepresent reliable internal consistency. 19 For the purposes of performing exploratory analyses, we included several factorsin our model below this cutoff. In addition, we included afew standalone variables ( Table I ). Instead of dropping fromthe analysis any variable or factor not found to havesignificant impact on unexpected performance outcome, weretained them to show the dissociation between thesefactors and unexpected performance improvement. Ordinary least squares regression techniques were used for all modelsand 1-tailed tests with  α  = 0.1 were established as the levelof significance. The regression coefficients (estimates) Table I.  Hospital characteristics and process adherence scores Characteristic QIE hospitals Facility characteristicsNo. of sites 212No. of beds (mean ± SD) 377 ± 209 Academic facility (n [%]) 59 (27.8)Region (n [%])Midwest 59 (27.8)Northeast 55 (25.9)South 74 (34.9) West 24 (11.3)Facility type (n [%])No services 20 (9.4)Catheterization laboratory only 20 (9.4)Percutaneous coronary intervention, no surgery 15 (7.1)Surgery 157 (74.1)Process adherence score percentage (median [25th, 75th percentile])2002 73.7 (68.3, 78.0)2003 77.7 (72.6, 81.6)2004 82.5 (78.3, 86.4)Improvement in score, absolute percentage increase, median (25th, 75thpercentile)2002-2003 3.9 (0.9, 7.6)2003-2004 4.7 (1.9, 7.5)2002-2004 8.5 (4.9, 12.9)No. of patient records submitted per hospital, median (25th, 75thpercentile)2002 86 (52, 174)2003 117 (73, 179)2004 106 (58, 171) 1208  Glickman et al  American Heart JournalDecember 2007  represent the effect on the dependent variable (unexpectedimprovement in the adherence score) of an increase of 1 unit in the independent variable (survey questions and/or factors) after controlling for all other variables — the larger the coefficient, the greater the effect. We also calculated theabsolute improvement in adherence score attributed to eachfactor (by moving from a minimum to a maximum survey score) by multiplying the regression coefficient by the totalnumber of points possible from the survey for each factor.For descriptive purposes, we compared patient characteris- tics of patients admitted in 2002 with those admitted in 2004. We also compared patient characteristics and QI factors acrossquartiles of the 2002 adherence score, of the 2004 adherencescore,andacrossquartilesof deltascoresandoftheresiduals. Allanalyses were performed using SAS version 9.2 (SAS Institute,Cary, NC). Results Table I shows the basic characteristics and processadherence scores of the 212 CRUSADE hospitals thatparticipated in the QIE survey and met our criterionof   ≥ 10 patient records in both 2002 and 2004.Median hospital process adherence scores demon- strated significant improvement (   P  b  .01) over theperiod studied. Table II lists the factors and their component survey questions (and percentage of missing data for each factor) retained in our model(Chronbach  α  N 0.40), as well as stand-alone variablesused for our analysis.The results of both univariate and multivariateordinary least squares regression models are listed inTable III. In the multivariate model, the followingfactors and questions had a significant association withunexpected QI improvement: the use of CRUSADE QItools, clinical commitment to quality by the cardiology coadvocate, barriers to QI related to resource avail- ability, cultural and physician resistance to change, andfinancial commitment to quality. The following factors were not associated with this unexpected improve- ment: clinical commitment to quality by the emergency  Table II.  Quality improvement factors used in this analysis FactorsSurvey questions(question no.) Cronbach  α  Missing data (%) QI tools How helpful has CRUSADEbeen in providing tools for QI? (14).57 1.7 How helpful has CRUSADE beenin identifying aspects needingimprovement? (21)Clinical commitment to quality (cardiology)Is the Ccoa highly involved? (3) .79 2.5Does the Ccoa review data? (4)Does the Ccoa disseminate data? (5)Does the Ccoa propose initiatives? (6)Clinical commitment to quality (emergency medicine)Is the EMcoa highly involved? (7) .73 8.4Does the EMcoa review data? (8)Does the EMcoa disseminate data? (9)Does the EMcoa propose initiatives? (10) Administrative commitment to quality Is the hospital administration highly committed to QI efforts? (23).79 18.1Is there consensus in evidence basedguideline goals? (28a)Is there sustained support to pursueevidence based guideline goals? (28b)Forum of feedback In what forum has CRUSADE data beenpresented at your hospital to personnel who are not designated site report recipients? (7 response choicesfor direct vs indirect feedback mechanisms) (1b).43 0.0Extent of feedback Who is CRUSADE data presented to at your hospital? (11 response choices) (1a).75 0.0Culture: employee satisfaction How satisfied are nurses as employees? (31) .69 7.3How satisfied are physicians as employees?(32)Stand-alone variablesFinancial commitment to quality Are financial resources dedicated to CRUSADE? (25) NA 3.8Barriers to quality improvement (resource availability, cultural/physician resistance to change,operational ineffectiveness) What is the biggest barrier to quality improvement at your hospital? (15)NA 0.0 EMcoa  , Emergency Medicine coadvocate;  Ccoa  , Cardiology coadvocate. Glickman et al  1209 American Heart JournalVolume 154, Number 6  medicine coadvocate, forum and extent of clinicalquality feedback, administrative commitment to quality,employee satisfaction, and barriers to QI related tooperational ineffectiveness.Figure 1 shows the relative impact of significantfactors from the multivariate model on QI. Thepercentages calculated are the absolute improvementin adherence score if a hospital moved from theminimum to maximum factor score possible on theQIE survey. Of these factors, maximal use of CRUSADEQI tools led to the largest absolute percent improve- ment in quality score, followed by barriers related toresource availability, clinical commitment by cardiology coadvocate, barriers related to culture, and financialcommitment to quality. Discussion  After developing a survey instrument to evaluatehospital QI efforts, we examined the hospital factorsmost responsible for demonstrated improvement inthe treatment of ACS from 2002 to 2004. As shownin Table III, we found that a hospital's use of theCRUSADE initiative and its associated QI tools,clinical commitment to quality by a cardiology coadvocate, dedication of financial resources to a QIprogram, and barriers related to resource availability and hospital culture were positively associated withsignificantly improved performance in high-risk ACScare from 2002 to 2004, in comparison to thehospital cohort.The QI tools provided by CRUSADE includedetailed benchmarking and feedback in the form of quarterly reports, as well as feedback on numerouscare processes not reported by other initiatives, risk stratification tools, and standing orders. The use of these tools had a significant association withimprovement in adherence scores from 2002 to 2004and were associated with the biggest improvement inadherence scores relative to other factors ( Figure 1 ).On average, a hospital that moved from theminimum to the maximum factor score in use of CRUSADE QI tools demonstrated a 6.7% absoluteincrease in process adherence score for treating NSTE ACS. Such improvement is consistent with a reduc- tion in mortality through incorporation of evidence- based clinical care tools in the Guidelines Applied in Table III.  Factor model: the effect of QI factors on hospital process improvement score (2002 to 2004) Factor (maximum points possible)Univariate analysis Multivariate analysisEstimate (95% CI)  P   Estimate (95% CI)  P  CRUSADE quality improvement tools (10) 0.71 (0.32-1.11)  b .01 0.67 (0.27-1.10)  b .01Clinical commitment to quality: cardiology (8) 0.58 (0.20-0.96)  b .01 0.50 (0.07-0.91) .03*Barrier to QI: resource availability (1) 1.96 ( − 0.62 to 4.53) .11 5.08 (0.84-9.46) .03*Financial commitment to quality (1) 2.06 (0.63-3.49)  b .01 1.58 (0.20-3.27) .04*Barrier to QI: cultural and physician resistance to change (1) 0.81 ( − 0.60 to 2.22) .17 2.98 ( − 0.44 to 7.09) .10*Clinical commitment to quality: emergency medicine (8) 0.11 ( − 0.26 to 0.49) .31  − 0.33 ( − 0.74 to 0.07) .09Forum of feedback (2) 1.39 (0.32-2.45) .02 0.90 ( − 0.50 to 2.16) .13Barrier to QI: nondiagnostic (1) 0.07 ( − 1.77 to 1.90) .48 2.70 ( − 1.21 to 6.61) .13 Administrative commitment to quality (12) 0.35 (0.03-0.66) .04 0.14 ( − .012 to 0.64) .26Barrier to QI: operational ineffectiveness (1)  − 1.58 ( − 3.17 to 0.01) .05 1.50 ( − 2.07 to 5.65) .26Culture: employee satisfaction (15) 0.23 ( − 0.36 to 0.83) .26  − 0.18 ( − .78 to 0.53) .32Extent of feedback 0.29 (0.05-0.53) .02  − 0.05 ( − 0.33 to 0.29) .39 *Indicates significant positive association with QI at the  P   = .10 level. Figure 1 Relative factor weightings: moving from minimum score to maximumscore in each of these factors results in the following absolute percent improvement in adherence score. For each of the factors associated with quality improvement in the multivariate model, the percent improvement in adherence score was determined if a hospital movedfrom the minimum to maximum factor score possible (from the QIEsurvey). These percentages, indicating absolute improvement inadherence scores, and their 95% CIs are displayed.   Figure 11210  Glickman et al  American Heart JournalDecember 2007
Search
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks