Managing to Improve Quality

Managing to improve quality: The relationship between accreditation standards, safety practices, and patient outcomes

Deirdre K. Thornlow

Elizabeth Merwin

Background: Given the trend toward eliminating reimbursement for ‘‘never events,’’ hospital administrators are

challenged to implement practices designed to prevent their occurrence. Little evidence exists, however, that

patient safety practices, as evaluated using accreditation criteria, are related to the achievement of patient safety

outcomes.

Purpose: The aim of this study was to examine the relationship between patient safety practices, as measured by

accreditation standards, and patient safety outcomes as measured by hospital rates of infections, decubitus

ulcers, postoperative respiratory failure, and failure to rescue.

Methodology: Secondary data were used to examine relationships between patient-safety-related accreditation

standards and patient outcomes in U.S. acute care hospitals. Accreditation performance areas were reduced

into subscores to represent patient safety practices. Outcome rates were calculated using the Agency for

Healthcare Research and Quality Patient Safety Indicator software. Multivariate regression was performed to

determine the significance of the relationships.

Findings: Three of four multivariate models significantly explained variance in hospital patient safety indicator

rates. Accreditation standards reflecting patient safety practices were related to some outcomes but not others.

Rates of infections and decubitus ulcers occurred more frequently in hospitals with poorer performance in

utilizing patient safety practices, but no differences were noted in rates of postoperative respiratory failure or

failure to rescue.

Practice Implications: Certain adverse events, such as infections and decubiti, may be reduced by preventive

protocols that are reflected in accreditation standards, whereas other events, such as failure to rescue and

postoperative respiratory failure, may require multifaceted strategies that are less easily translated into protocols.

Our approach may have influenced the observed associations yet represents progress toward assessing whether

safety practices, as measured by accreditation standards, are related to patient outcomes.

July–September � 2009262

Deirdre K. Thornlow, PhD, RN, CPHQ, is Assistant Professor, School of Nursing, Duke University, Durham, North Carolina. E-mail: deirdre.thornlow@duke.edu. Elizabeth Merwin, PhD, RN, FAAN, is Associate Dean, Research, Madge M. Jones Professor of Nursing, and Director, Rural Health Care Research Center, School of Nursing, University of Virginia, Charlottesville. E-mail: eim5u@virginia.edu.

This study was approved by the institutional review board (UVA No. 2004-0255-00) and supported by Grant No. F31 NR009320-01 from the National Institute for Nursing Research. The contents of this article are solely the responsibility of the authors and do not necessarily represent the official views of the National Institute for Nursing Research.

Key words: adverse events, hospital accreditation, outcomes, patient safety, safety practices

Health Care Manage Rev, 2009, 34(3), 262-272 Copyright A 2009 Wolters Kluwer Health | Lippincott Williams & Wilkins

9Copyright @ 200 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

H ospitals should be motivated more than ever to commit resources and attention to patient safety, for beginning October 2008, the Centers for

Medicare and Medicaid Services (CMS) eliminated or reduced payments for pressure ulcers, hospital-acquired infections, and other ‘‘never events,’’ defined as prevent- able adverse events that should never occur in health care (CMS, 2007). This change in reimbursement follows earlier CMS initiatives that now require hospitals to submit evidence-based quality measures or suffer reduc- tions in their Medicare Annual Payment Updates. The CMS is expanding reimbursement models by reward- ing hospitals with higher Medicare payment for better mortality outcomes; poor performing hospitals will be penalized with reduced payments. Such reimbursement changes may be justified by research evidence demon- strating that most medical errors, or adverse events, are preventable (Lehman, Puopolo, Shaykevich, & Brennan, 2005; Thomas et al., 2000) and hospitalized patients who experience a medical error remain hospitalized longer and accrue greater costs when compared with controls (Nordgren, Johnson, Kirschbaum, & Peterson, 2004; Rojas, Silver, Llewellyn, & Rances, 2005).

The research approach in this study may inform hospitals about the influence of organizational character- istics and processes of care on patient safety outcomes. Although studies have shown associations between char- acteristics of hospital systems, such as teaching status, ownership status, nurse staffing, and patient safety out- comes (Ayanian, & Weissman, 2002; Devereaux et al., 2002; Kupersmith, 2005; Stanton, 2004), few studies have examined how these hospital characteristics influence utilization of patient safety practices, and even fewer studies have examined the impact of using patient safety practices on patient outcomes. In designing patient safeguards, it is essential to consider how patient safety practices, defined as types of care processes whose application reduces the probability of an adverse event (Shojania, Duncan, McDonald, & Wachter, 2002), con- tribute to safe care. This is challenging, as little evidence suggests that safety practices, such as those commonly used in non-health-care fields, confer benefit in acute care hospitals, especially on patient outcomes (Shojania et al., 2002). When evidence does exist regarding efficacy of safety processes, organizations have made attempts to translate such evidence into practice, to include incorpo- rating patient safety standards into the hospital accredi- tation process (Kizer & Blum 2005; Leape, Berwick, & Bates, 2002); however, the link between these practices and outcomes has not yet been clarified.

In this study, secondary data were used to examine relationships among hospital systems, utilization of patient safety practices, as measured by accreditation standards, and patient outcomes in acute care hospitals to determine whether the use of patient safety practices influences rates

of four patient safety indicators: infections due to medical care, decubitus ulcers, postoperative respiratory failure, and failure to rescue. We hypothesized that teaching hospitals, hospitals with higher nurse staffing levels, and hospitals using more patient safety practices would ex- perience lower rates of these patient safety incidents than would nonteaching hospitals, hospitals with lower levels of nurse staffing, and hospitals using fewer patient safety practices. Findings, implications for current practice, and suggestions for future research designed to improve patient safety in acute care hospitals will be presented.

Conceptual Framework

The Quality Health Outcomes Model (QHOM; Mitchell, Ferketich, & Jennings, 1998) served as the conceptual framework for this study. In 1998, the Expert Panel on Quality Health Care of the American Academy of Nursing published the QHOM as a conceptual framework for quality and outcomes research, most specifically as a means to test relationships among the elements of struc- ture, process, and outcomes. The QHOM built on these three elements from Donabedian’s (1966) seminal work in assessing the quality of medical care and incorporated client, or patient, characteristics as a fourth construct. The QHOM realigned the constructs to capture their dynamic, rather than linear, relationships. The traditional ‘‘struc- ture’’ construct was renamed ‘‘system’’ in the QHOM, whereas the traditional ‘‘process’’ construct was renamed ‘‘intervention.’’ The QHOM posits reciprocal interactions among the four constructs (system, intervention, outcome, and client), thus serving as a useful conceptual guide for health care systems researchers. Several investigators have used the QHOM model in acute and community care to organize their choice of variables among the four con- structs and to build evidence regarding the quality of health care (Mayberry & Gennarro, 2001; Radwin & Fawcett, 2002; Sin, Belza, LoGerfo, & Cunningham, 2005). In this study, system variables included hospital characteristics such as teaching status, ownership status, size, location, and nurse staffing levels; intervention vari- ables were defined as utilization of patient safety practices; outcomes, or patient safety indicators, were defined by the Agency for Healthcare Research and Quality (AHRQ, 2007); and client characteristics were defined as risk- adjusted variables including diagnosis, age, and gender. The client variables were used to flag potentially pre- ventable complications and to create hospital-level risk- adjusted patient safety indicator rates (AHRQ, 2007). Measures for each of the constructs are described below.

Methods

Secondary data were analyzed from a stratified probability sample of acute care hospitals. Hospital-level data were

Safety Practices and Patient Outcomes 263

9Copyright @ 200 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

gathered from the 2002 American Hospital Association (AHA) annual survey data and 2002 Joint Commission (JC) accreditation performance reports. Hospital accred- itation performance reports were retrieved online from JC Quality Check (www.jointcommission.org) in 2005. Patient-level data were gathered from the 2002 Nation- wide Inpatient Sample (NIS), the largest all-payer inpatient database in the United States which contains patient-level clinical and resource data on hospital stays from states participating in the Healthcare Cost and Utilization Project and is designed to approximate a 20% stratified probability sample of U.S. community hospitals (AHRQ, 2008). Patient characteristics were risk adjusted by age, gender, diagnoses, and comorbidities using the AHRQ (2007) comorbidity software algorithm, as de- scribed in the ‘‘Measures’’ section.

Study Sample

General medical–surgical community hospitals, as clas- sified by AHA, served as the population for study. Specialty hospitals, such as children’s, psychiatric, and rehabilitation hospitals, were excluded because the se- lected patient safety indicators address incidents that are more likely to occur in general medical–surgical adult patients. The 2002 NIS inpatient discharge-level file, which contains data for 100% of the discharges (n = 7,853,982) from 995 hospitals in 35 participating states, was used. Of the 35 states in the 2002 NIS sample, 10 states prohibit release of AHA hospital identifiers; there- fore, these states’ hospitals were excluded. In addition, one state did not code patient comorbidities, data that are necessary to risk adjust; therefore, this state and its hospitals also were excluded from the sample. And fi- nally, to complete analyses, it was necessary to match the 2002 NIS data to 2002 JC survey data; therefore, only those hospitals surveyed by JC in 2002 were included in this study—four states had no surveyed hospitals in the 2002 NIS sample. The study sample was composed of 1,430,981 inpatient discharge records from 115 hospitals in 20 states. Sample size equaled less than 115 hospitals for two of four outcomes studied due to coding limitations in two states that precluded inclusion (e.g., missing ad- mission type).

Measures

Data were used to construct variables that measured hospital systems, utilization of patient safety practices, and risk-adjusted patient outcomes. Hospital system variables were constructed from NIS and AHA survey data. These descriptive structural measures included nurse staffing (ratio of RN full-time equivalents to adjusted average daily census [A-ADC]), teaching status, hospital location (urban or rural based on metropolitan statistical area

population standards), hospital size (A-ADC), and owner- ship. Certain variables were recoded to conserve degrees of freedom. First, the continuous variable A-ADC was used to measure hospital size rather than NIS-designated categories small, medium, and large. Second, hospital ownership was originally classified by NIS in five cate- gories: government nonfederal; private, not-for-profit; private, investor owned; and two additional categories into which smaller strata of hospitals were collapsed. Federal hospitals are not sampled in NIS. Due to the large numbers of hospitals within the two NIS-collapsed categories, precise ownership information for each of the 115 sample hospitals was obtained from the 2006 AHA directory so that all hospitals could be accurately categorized, without the need for collapsed categories. Ownership was then coded into two levels: for-profit and other.

Joint Commission Accreditation Performance Reports were used to construct a measure of each hospital’s utilization of patient safety practices. During accreditation surveys, 45 performance areas encompassing nearly 500 standards are evaluated and scored. It should be noted that accreditation performance areas were scored on a scale of 1 to 5, with 5 being the poorest score; thus, a higher score indicates poorer performance in using that safety practice. Because only half of the JC standards related to patient safety, as noted by the JC (2003), and little variation existed among our study hospitals in their overall ac- creditation scores (M = 92.3, SD = 3.68), we sought a parsimonious method to differentiate patient safety practices among hospitals. Using a 4-point scale (1 = ‘‘not relevant’’ to 4 = ‘‘very relevant’’), an expert panel composed of hospital quality improvement directors and nurse executives, excluding the authors, evaluated the 45 performance areas to determine, in their expert opinion, which performance areas most embodied patient safety. The expert panelists unanimously rated 12 per- formance measures as most relevant to patient safety (rating of 3 or greater). The content validity index was measured by percentage reviewer agreement for each item and for the total 45-item instrument. Across the in- strument, the content validity index equaled 0.74; a score of 0.78 or better indicates good content validity (Polit, Beck, & Owen, 2007). We then conducted principal components analysis, using an orthogonal rotation, to determine the underlying structure for the 12 retained measures; this produced a four-component solution that was evaluated as adequate using four criteria: eigenvalue >1, variance, scree plot, and residuals. In this study, factor loadings ranged from .414 to .798, with an average of .631, generally considered very good (Comrey & Lee, 1992; Fleury, 1998). Aggregate factor scores created parsimony in variables tested; likewise, an additional strength of this technique is that, because factors are orthogonal, the factor scores are nearly uncorrelated and can be used in

July–September � 2009264 Health Care Management REVIEW

9Copyright @ 200 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

regression analyses without producing multicollinearity among the subsets of variables (Tabachnick & Fidell, 2001). The four regression factor scores were used to represent patient safety practices in this study and were named ‘‘surveillance capacity,’’ ‘‘assessing patient needs,’’ ‘‘care procedures,’’ and ‘‘measuring processes.’’ Detailed information regarding this factor analysis has been published elsewhere (Thornlow, 2008).

Hospital rates of occurrence for each of four indicators, infections due to medical care, decubitus ulcers, failure to rescue, and postoperative respiratory failure, were calcu- lated by applying the Patient Safety Indicator (PSI) soft- ware (version 3.0a) to the NIS data set (http://www. qualityindicators.ahrq.gov/psi_download.htm). These indi- cators were selected because prior research has suggested that these outcomes are potentially attributable to organi- zational characteristics (AHRQ, 2007; Romano, Geppert, Davies, Miller, Elixhauser, & McDonald, 2003), includ- ing nurse staffing (Aiken, Clarke, Sloane, Sochalski, & Silber, 2002; Blegen, Goode, & Reed, 1998; Kovner & Gergen, 1998; Lichting, Knauf, & Milholland, 1999; Needleman, Buerhaus, Mattke, Stewart, & Zelevinsky, 2002; Unruh, 2003) and care processes (Danks, 2006; Frantz, 2004; Gastmeier & Geffers, 2006; Kovner & Gergen, 1998; Lyder, 2003). In calculating rates, the PSI software generates an algorithm that uses the ICD-9-CM diagnosis and procedure codes, date of procedure, and pa- tient characteristics, including age, gender, and diagnosis- related group, to flag potentially preventable complications. In running the software, hospital-level risk-adjusted ratios with smoothing were calculated for the four patient safety indicators. The smoothing process applies shrinkage fac- tors to reflect a reliability adjustment unique to each in- dicator. The less reliable the PSI is over time and across hospitals, the more the estimate shrinks the PSI toward the overall mean. The resulting rate appears ‘‘smoother,’’ or more conservative than the raw rate, and random year- to-year fluctuations in performance are likely to be re- duced (AHRQ, 2007).

Data Analysis

Univariate and multivariate regression analyses were conducted at the hospital level; each PSI outcome was analyzed separately. Moderate to strong statistically significant correlations existed among the raw, risk- adjusted, and smoothed rates for each PSI; therefore, all analyses were conducted on smoothed PSI outcome rates as smoothed rates had been reported in similar studies (Miller et. al., 2005; Thornlow & Stukenborg, 2006). Although we tested its association with patient outcomes, overall accreditation score was not significant in either univariate or multivariate regression analyses and was thus not included in further testing. To com- plete hypothesis testing, final regression models were

built using variables that were hypothesized to be significant a priori; variables found to be significantly associated with patient outcomes in preliminary univar- iate regression analyses ( p < .05) or in preliminary multivariate regression analysis ( p < .05) were included in the final multivariate models.

Findings

Most of the 115 hospitals included in the sample were classified as urban (n = 78), nonteaching (n = 88), and not-for-profit (n = 76); almost half were considered large institutions (n = 56) and approximately 43 hospitals (37%) were located in the South (Table 1). Study sample hospitals differed from the 2002 NIS sample in that the study sample had fewer small hospitals, fewer rural hospitals, fewer government nonfederal hospitals, and fewer hospitals located in the Midwest than the national NIS sample did. No differences in teaching status existed between the 2002 NIS sample and the study sample (Table 1). The A-ADC of study hospitals ranged from 10 to 1,397 patients per day (M = 265.59, SD = 247.68) with an average nurse ratio of 1.18 RN full-time equivalents (SD = 0.44). Overall accreditation scores ranged from 83 to 99 on a scale of 100 (M = 92.3, SD = 3.68). Hospital risk-adjusted (smoothed) rates for the patient safety indicators ranged from 1.8 cases of infection per 1,000 discharges (0.0018) to 9.8 cases of postoperative respiratory failure per 1,000 elective sur- gical discharges (0.0098) to 21.5 incidences of decubitus ulcers per 1,000 discharges (0.0215) to 133.9 cases of failure to rescue, or deaths, per 1,000 discharges among patients who developed potentially preventable compli- cations during their hospital stay (0.1339). These rates are comparable to 2002 national rates for these indi- cators (http://hcupnet.ahrq.gov/; Table 2).

Overall, three of the four multivariate models attained significance. Results are shown for the preliminary multivariate regression models rather than for the final multivariate regression models because no appreciable differences were noted in the strength or direction of relationships or amount of variance explained between the preliminary and final multivariate models (Table 3). Hospital system characteristics and patient safety practices accounted for 21.9% of the adjusted variance in hospital rates of infection ( p = .000), 13.0% of the adjusted variance in hospital rates of postoperative respiratory failure ( p = .011), and 8.9% of the adjusted variance in hospital rates of decubitus ulcers ( p = .029). None of the models significantly explained variance in hospital rates of failure to rescue ( p = .436).

Hospital system characteristics were not consistently associated with patient outcomes in either univariate or multivariate regression analyses (Table 3). Although not hypothesized to be significant a priori, larger hospitals

Safety Practices and Patient Outcomes 265

9Copyright @ 200 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

had higher rates of infections due to medical care and postoperative respiratory failure than did smaller hos- pitals, but no differences were noted by hospital size in either failure to rescue or decubitus ulcers. And when compared with not-for-profit and government nonfed- eral hospitals, for-profit hospitals exhibited higher rates of decubitus ulcers and postoperative respiratory fail- ure, but no differences were noted by ownership for rates for failure to rescue or infections due to medical care. No differences in hospital patient safety indicator rates were noted by teaching status or levels of nurse staf- fing (Table 3).

Utilization of patient safety practices, as measured by accreditation standards, was significantly associated with two of the four patient safety outcomes in both univariate and multivariate analyses. Hospitals with poorer performance using the patient safety practice ‘‘assessing patient needs’’ (Subscore 2) had higher rates of infection due to medical care than did hospitals with better performance using that practice, and hospitals with poorer performance using ‘‘care procedures’’ (Sub- score 3) had higher rates of decubitus ulcers than did

hospitals with better performance using that patient safety practice. Utilization of patient safety practices was not associated with hospital rates of postoperative respiratory failure or failure to rescue in either univariate or multivariate analyses.

Discussion

In this study, hospital system characteristics did not consistently explain patient outcomes, echoing previous findings in which associations varied depending on the outcome measured (Baker et al., 2002; Romano et al., 2003; Thornlow & Stukenborg, 2006). In this study, larger hospitals demonstrated higher rates of adverse events than smaller hospitals did, but for only two of the four indicators analyzed: infections due to medical care and postoperative respiratory failure. Perhaps, increased con- tact from a larger number of staff increases the probability of cross-contamination and infection; likewise, the need for interdisciplinary communication among the many pro- viders in a large hospital may predispose such institutions to higher rates of postoperative respiratory failure than in a

Table 1

Comparison of study sample hospitals to NIS sample hospitals

Study hospitals (sample, n = 115)

National sample hospitals (NIS, n = 995)

Difference between study and national samples

Hospital characteristics n % n % �2

Hospital teaching status 2.76 Nonteaching 88 76.5 729 82.8 Teaching 27 23.5 151 17.2

Hospital ownership (five categories) 17.34** Government or private, collapsed 49 42.6 290 33.0 Government, nonfederal (public) 9 7.8 193 21.9 Private, collapsed 27 23.5 168 19.1 Private, not-for-profit 19 16.5 103 11.7 Private, investor owned 11 9.6 126 14.3

Hospital location 8.68** Rural 37 32.2 411 46.7 Urban 78 67.8 469 53.3

Hospital bed size 22.13*** Small 20 17.4 343 39.0 Medium 39 33.9 255 29.0 Large 56 48.7 282 32.1

Hospital region 9.75* Northeast 24 20.9 112 12.7 Midwest 22 19.1 262 29.8 South 43 37.4 340 38.6 West 26 22.6 166 18.9

Note. NIS = Nationwide Inpatient Sample.

*Significant at p < .05.

**Significant at p < .01.

***Significant at p < .001.

July–September � 2009266 Health Care Management REVIEW

9Copyright @ 200 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

Table 2

Descriptive statistics for study sample hospitals (n = 115)

Variable Operational definition Mean SD 2002 National PSI rate

Nurse staffing Measured as the ratio of RN full-time equivalents to A-ADC

1.18 0.44 –

A-ADC Hospital size was measured as A-ADC. A-ADC reflects the average number of both inpatients and outpatients treated at the hospital on a daily basis.

265.59 247.68 –

Overall score This score is derived from an assessment of an organization’s compliance with all applicable Joint Commission standards at the time of the full triennial accreditation survey. Score is based on a scale of 0 to 100, with 100 representing the highest score.

92.30 3.68 –

Decubitus ulcer (n = 115 hospitals)

Decubitus ulcer development per 1,000 discharges in lengths of stay of five or more days. Excludes patients with paralysis, diseases of the skin, subcutaneous tissue, and breast. Excludes obstetrical admissions and admissions from long-term care.

0.0215 (21.50) 0.01 (23.63)

Failure to rescue (n = 99 hospitals)

Deaths per 1,000 discharges among patients who develop potentially preventable complications during their hospital stay. Excludes patients transferred in or out, patients admitted from long-term care, neonates, and patients over 74 years.

0.1339 (133.90) 0.02 (129.37)

Selected infections due to medical care (n = 115 hospitals)

Rate per 1,000 discharges of infections due to medical care, primarily those related to intravenous lines and catheters. Defined by including cases based on secondary diagnosis associated with the same admission. Excludes patients with potentially immunocompromised states (e.g., AIDS and cancer).

0.00181 (1.81) 0.00 (1.53)

Postoperative respiratory failure (n = 100 hospitals)

Rates of postoperative respiratory failure per 1,000 elective surgical discharges. Limits code to secondary diagnoses to eliminate respiratory failure that was present on admission. Excludes patients who have major respiratory or circulatory disorders. Limits the population at risk to elective surgery patients.

0.0098 (9.80) 0.00 (7.97)

Variable Operational definition Median SD Cronbach’s �

Patient safety practices/ subscores

Subscore 1 (surveillance capacity) Reassessment procedures Implementation of patient safety plans �0.2754 1 .510 Orientation, training staff Assessing staff competency

(continues)

Safety Practices and Patient Outcomes 267

9Copyright @ 200 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

smaller hospital. Additional research is needed to examine why larger hospitals demonstrated higher rates for these patient outcomes.

For-profit hospitals had higher rates of adverse events than did not-for-profit and nonfederal government hospitals, but again, for only two of four indicators: decubitus ulcer and postoperative respiratory failure.

Reasons for these findings are unclear yet support those of other studies where for-profit hospitals had higher rates of postoperative pneumonia, pulmonary compro- mise (Kovner & Gergen, 1998), postoperative respira- tory failure, and decubitus ulcers (Romano et al., 2003) than other hospital types did. For-profit and not-for- profit hospitals may differ in the types of resources used

Table 3

Multivariate regression analysis: relationship of hospital systems and utilization of patient safety practices to patient safety outcomes

Standardized � coefficients

Decubitus ulcer Failure to rescue

Infection due to medical care

Postoperative respiratory failure

Hospital characteristics Western region .09 .13 .16 .00 Urban location .07 �.01 .04 .11 Hospital size (A-ADC) .17 .17 .30* .36* Teaching hospital .05 �.11 .08 �.02 For-profit owner .24* .04 .02 .32* RN staffing (FTE/A-ADC) �.05 �.08 .12 �.16

Patient safety practices Subscore 1: surveillance capacity .03 .10 �.13 �.09 Subscore 2: assessing patients �.04 .06 .25* .07 Subscore 3: care procedures .27** �.12 �.10 .07 Subscore 4: measuring process .11 .19 �.15 .11

F test 2.12 1.11 4.20 2.48 Model significance .03* .44 .00*** .01* R2 .17 .10 .29 .22 Adjusted R2 .09 .00 .22 .13

Note. A-ADC = adjusted average daily census; FTE = full-time equivalent.

*Significant at p < .05.

**Significant at p < .01.

***Significant at p < .001.

Table 2

Continued

Variable Operational definition Median SD Cronbach’s �

Subscore 2 (assessing patient needs) Initial assessment �0.1212 1 .533 Availability of patient-specific information Medication use

Subscore 3 (care procedures) Infection control 0.0611 1 .096 Planning and providing care Operative procedures

Subscore 4 (measuring processes) Patient and family education �0.2055 1 .024 Measurement of processes and outcomes

Note. A-ADC = adjusted average daily census; PSI = Patient Safety Indicator.

July–September � 2009268 Health Care Management REVIEW

9Copyright @ 200 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

to detect early signs of respiratory distress in postoper- ative patients, such as pulse oximetry or capnography, or in the equipment used to prevent pressure sores, such as specialty beds and/or mattresses. Further study may illuminate these differences.

Similarly, accreditation standards specific to patient safety processes did not appear related to all four outcomes analyzed in this study. Rates of infections and decubitus ulcers occurred more frequently in hospitals with poorer performance in utilizing patient safety practices, but no differences were noted in rates of postoperative respiratory failure or failure to rescue. One possible explanation suggests that infections and decubitus ulcers may be more amenable to the implementation of policies and proce- dures targeted to prevent their occurrence. Additional research is needed to determine whether patient safety practices geared toward the more complicated problems, failure to rescue and postoperative respiratory failure, are equally as effective in reducing these events.

Our measurement approach constitutes only one way of conceptualizing patient safety practices and assumes that such practices can be measured by JC accreditation standards. In reality, these standards may not measure those practices and procedures that are most important for ensuring safe patient care. That we found associations for only two of the four indicators and that previous stud- ies reported no association between JC survey assessments and other measures of quality and patient outcomes (Chen, Rathore, Radford, & Krumholz, 2003; Miller et al., 2005) suggest the limitation in this approach. It may be that variables crafted to measure nursing sur- veillance, critical thinking, leadership, communication, and collaboration would more effectively gauge the ef- fects of patient safety practices, especially for the more complicated outcomes failure to rescue and postoperative respiratory failure. Research examining the relationship of these patient safety attributes to adverse events is war- ranted and, if found to be significant, suggests the possible need to modify accreditation standards.

Certain limitations exist in this study, including that most administrative data used for secondary analysis consist of data initially collected for other purposes (Zhan & Miller, 2003) and may thus carry potential bias in detecting certain types of patient safety events due to ICD-9-CM coding limitations. Additional concerns exist when calculating severity of illness (Lawthers et al., 2000; Zhan & Miller, 2003); however, the AHRQ Comorbidity Method, which was used in this study to risk adjust patient characteristics, was formulated to be used in conjunction with the PSI software and is designed to be used with administrative data (Elixhauser, Steiner, Harris, & Coffey, 1998). As noted earlier, the hospitals in this study are smaller, less rural, less likely to be government or nonfederal hospitals, and less likely to be located in the Midwest than are

hospitals in the national sample. These differences probably related to the need to match medical–surgical hospitals in the 2002 NIS database with hospitals that were surveyed by the JC in 2002. It has been reported that smaller and more rural hospitals are less likely to participate in JC surveys (Miller et al., 2005). The factor analysis was effective in reducing the number of accreditation items determined to be most relevant to patient safety by the external panelists; however, the low reliability coefficients for Subscores 3 and 4 may indicate low intercorrelation among factored items (Table 2). The significant influence of Subscore 3 on decubitus ulcers was likely due to one item, ‘‘planning and providing patient care,’’ which was the only item that attained significance when the three individual items replaced the aggregate factor score in post hoc regression analysis. When individual items, rather than aggregate scores, were included in post hoc regression models, these items continued to demonstrate nonsig- nificant relationships with outcomes.

Practice Implications

The relationship between the utilization of patient safety practices and patient outcomes has not been well studied, which makes it difficult for hospitals to identify mod- ifications needed to improve patient outcomes. Our find- ings suggest that certain adverse events, such as infections and decubitus ulcers, may be reduced by preventive protocols that are reflected in accreditation standards, whereas other events, such as failure to rescue and postoperative respiratory failure, may require multifaceted strategies that are less easily translated into protocols. To illustrate, research has shown that decubitus ulcers can be prevented or treated successfully when evidence-based care, such as identifying patients at risk and reducing pressure over bony prominences, is provided (Armstrong & Bortz, 2001; Cullum, McInnes, & Bell-Syer, 2004; Frantz, 2004; Lyder, 2003). Employing procedures such as the Guidelines for the Prevention and Treatment of Pressure Ulcers (Wound, Ostomy, Continence Nursing, 2003) in acute care hospitals is crucial because length of stay is almost twice as long for the patients who develop pressure ulcers as compared with patients who are at risk but do not develop pressures ulcers (Loan, Jennings, Brosch, Depaul, & Hildreth, 2003). Decubitus ulcers are significant in- dependent predictors of hospital costs, with the estimated cost of managing a single full-thickness pressure ulcer as high as $70,000 and the total cost for treatment of pressure ulcers in the United States estimated at $11 billion per year (Reddy, Gill, & Rochon, 2006; Redelings, Lee, & Sorvillo, 2005).

Similarly, policies and procedures such as the Guide- lines for the Prevention of Intravascular Catheter-Related In- fections (O’Grady et. al., 2002) have been found effective

Safety Practices and Patient Outcomes 269

9Copyright @ 200 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

in reducing catheter-related bloodstream infection rates (Behrenholtz, Pronovost, & Lipsett, 2004; Theaker, 2005). However, despite evidence that guidelines are effective, researchers have repeatedly demonstrated prob- lems with health care workers’ compliance to infection control policies and procedures, including compliance to hand hygiene measures (Larson, Albrecht, & O’Keefe, 2005; McKinley et al., 2005; Peterson & Walker, 2006). As many as 250,000 hospital-acquired venous catheter- related bloodstream infections occur annually in the United States, and treatment cost averages as much as $56,000 (O’Grady et al., 2002). Hospital-acquired in- fections not only increase costs but also increase mor- bidity, extend hospital stays, and are associated with significant increases in in-hospital mortality (Coffin & Zaoutis, 2005).

In this study, hospitals using fewer patient safety practices demonstrated higher rates of infection and decubitus ulcers, suggesting that requisite policies and procedures were not followed. Applying evidence-based guidelines provides an efficient and cost-effective method for preventing these types of adverse events. Alternatively, failure to rescue and postoperative re- spiratory failure may reflect situations that require mul- tifaceted approaches that are less easily translated into protocols. Recently, however, hospitals have begun to institute practices such as rapid response teams that are designed to assist bedside staff to intervene upon early signs of patient deterioration. This study preceded the large-scale implementation of such protocols yet high- lights the importance of applying patient safety practices to prevent the occurrence of adverse events.

The AHRQ quality and patient safety indicators have been identified as possible sources for expanded com- parative public reporting. Our measurement approach represents a potential method for assessing patient safety practices, using accreditation standards, and associating these practices with patient outcomes. Hospitals can employ a similar approach by running the AHRQ PSI software on their own inpatient discharge billing data to screen for potentially preventable adverse events and never events; to identify opportunities for improvement that may justify review of practices, policies, or procedures designed to prevent occurrence; and to benchmark their hospital rates to the national rates posted on the AHRQ Web site (http://hcupnet.ahrq.gov).

Rates are also categorized by hospital type so that administrators can ascertain whether their hospital’s characteristics (e.g., ownership status and hospital size) may have influenced their adverse event rates. Such rate comparisons allow for introspection so that preventive measures, such as pressure relieving devices, improved infection control procedures, and interdisciplinary com- munication tools, can be instituted and evaluated for their impact on patient safety.

Conclusion

This study examined the association between hospital characteristics and the use of patient safety practices on patient outcomes. We found that certain hospital charac- teristics were significantly associated with some patient outcomes, but not others, and accreditation standards specific to patient safety practices did not appear related to all four outcomes studied. Certain adverse events, such as infections and decubitus ulcers, may be reduced through preventive procedures, whereas other events, such as failure to rescue and postoperative respiratory failure, may reflect situations that require multifaceted strategies that are less easily translated into protocols or less easily measured by accreditation standards.

This study provides a useful reflection on the challenge of determining relationships between organizational characteristics and patient outcomes and highlights the need for additional inquiry regarding appropriate mea- surement of acute care processes, especially patient safety practices. Our measurement approach may have influ- enced the observed associations between protective safety practices and patient outcomes; however, it represents progress toward examining this relationship. With the continual addition of patient safety-related standards and National Patient Safety Goals into the JC accreditation process, more research is needed to determine whether these practices have impacted patient outcomes. Such research has the potential to be not only cost saving due to changes in reimbursement but also life saving for hospitalized patients.

Acknowledgments

The authors wish to thank other members of the dis- sertation committee, including Patricia Hollen, PhD, RN, FAAN; Arlene Keeling, PhD, RN; and William Knaus, MD. Special thanks to Ruth Anderson, PhD, RN, FAAN, who reviewed this manuscript.

References

Agency for Healthcare Research and Quality. (2007). Quality indicators: Guide to patient safety indicators. Rockville, MD: Agency for Healthcare Research and Quality. Retrieved October 10, 2007, from http://www.qualityindicators.ahrq. gov/psioverview

Agency for Healthcare Research and Quality. (2008). Overview of the Nationwide Inpatient Sample (NIS). Retrieved September 13, 2008, from http://www.hcup.ahrq.gov/nisoverview

Aiken, L., Clarke, S., Sloane, D., Sochalski, J., & Silber, J. (2002). Hospital nurse staffing and patient mortality, nurse burnout, and job dissatisfaction. JAMA, 288, 1987–1993.

Armstrong, D., & Bortz, P. (2001). An integrative review of pressure relief in surgical patients. AORN Journal, 73(3), 645–648.

Ayanian, J. Z., & Weissman, J. S. (2002). Teaching hospitals

July–September � 2009270 Health Care Management REVIEW

9Copyright @ 200 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

and quality of care: A review of the literature. Milbank Quarterly, 80(3), 569–593.

Baker, C. M., Messmer, P. L., Gyurko, C. C., Domagala, S. E., Conly, F. M., Eads, T. S., et al. (2002). Hospital owner- ship, performance, and outcomes: Assessing the state- of-the-science. Journal of Nursing Administration, 30(5), 227–240.

Behrenholtz, S., Pronovost, P., & Lipsett, P. (2004). Eliminating catheter-related blood stream infections in the intensive care unit. Critical Care Medicine, 32(10), 2014–2020.

Blegen, M., Goode, C., & Reed, L. (1998). Nurse staffing and patient outcomes. Nursing Research, 47(1), 43–50.

Centers for Medicare and Medicaid Services. (2007). HHS reports to Congress on Value-Based Purchasing of Hospital Services by Medicare. Retrieved November 2007, from http://www.hhs.gov.news/press/2007/11/pr20071126a

Chen, J., Rathore, S. S., Radford, M. J., & Krumholz, H. M. (2003). JCAHO accreditation and quality of care for acute myocardial infarction. Health Affairs, 22(2), 243–254.

Coffin, S., & Zaoutis, T. E. (2005). Infection control, epi- demiology, and patient safety. Infectious Disease Clinics of North America, 19, 647–665.

Comrey, A. L., & Lee, H. B. (1992). A first course in factor analysis. (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum Asso- ciates Publishers.

Cullum, N., McInnes, E., & Bell-Syer, S. E. M. (2004). Support surfaces for pressure ulcer prevention. Update of Cochrane Database of Systematic Reviews 2004(3):CD001735.

Danks, L. A. (2006). Central venous catheters: A review of skin cleansing and dressings. British Journal of Nursing, 15(12), 650–654.

Devereaux, P. J., Choi, P., Lachetti, C., Weaver, B., Schunemann, H., Haines, T., et al. (2002). A systematic review and meta-analysis of studies comparing mortality rates of private for-profit and private not-for profit hospitals. Canadian Medical Association Journal, 166, 1399–1406.

Donabedian, A. (1966). Evaluating the quality of medical care. Milbank Memorial Fund Quarterly, 44(3, Suppl. 1), 66–206.

Elixhauser, A., Steiner, C., Harris, D. R., & Coffey, R. M. (1998). Comorbidity measures for use with administrative data. Medical Care, 36(1), 8–27.

Fleury, J. (1998). The index of self-regulation; Development and psychometric analysis. Journal of Nursing Measurement, 6(1), 3–17.

Frantz, R. A. (2004). Evidence-based protocol: Prevention of pressure ulcers. Journal of Gerontological Nursing, 30(2), 4–11.

Gastmeier, P., & Geffers, C. (2006). Prevention of catheter- related bloodstream infections: Analysis of studies pub- lished between 2002 and 2005. Journal of Hospital Infection, 64(4), 326–335.

Joint Commission on Accreditation of Healthcare Organiza- tions. (2003). Facts about patient safety. Retrieved August 18, 2004, from http://www.jcaho.org

Kizer, K. W., & Blum, L. N. (2005). Safe practices for better health care. In Advances in patient safety: From research to implementation. Volumes 1–4 (AHRQ Publication No. 050021; 1–4). Rockville, MD: Agency for Healthcare Research and Quality.

Kovner, C., & Gergen, P. J. (1998). Nurse staffing levels and adverse events following surgery in U.S. hospitals. Image: Journal of Nursing Scholarship, 30(4), 315–321.

Kupersmith, J. (2005). Quality of care in teaching hospitals: A literature review. Academic Medicine, 80(5), 458–466.

Larson, E. L., Albrecht, S., & O’Keefe, M. (2005). Hand hygiene behavior in a pediatric emergency department and a pediatric intensive care unit: Comparison of use of 2 dispenser systems. American Journal of Critical Care, 14(4), 304–312.

Lawthers, A., McCarthy, E., Davis, R., Peterson, L., Palmer, H., & Iezzoni, L. I. (2000). Identification of in-hospital com- plications from claims data: Is it valid? Medical Care, 38(8), 785–795.

Leape, L., Berwick, D., & Bates, D. (2002). What practices will most improve safety? JAMA, 288(4), 501–507.

Lehman, L., Puopolo, A., Shaykevich, S., & Brennan, T. (2005). Iatrogenic events resulting in intensive care admission: Frequency, cause, and disclosure to patients and institutions. American Journal of Medicine, 118, 409–413.

Lichting, L. K., Knauf, R., & Milholland, D. K. (1999). Some impacts of nursing on acute care hospital outcomes. Journal of Nursing Administration, 29(2), 25–33.

Loan, L. A., Jennings, B. M., Brosch, L. R., Depaul, D., & Hildreth, P. (2003). Indicators of nursing care quality: Findings from a pilot study. Outcomes Management, 7(2), 51–58

Lyder, C. H. (2003). Pressure ulcer prevention and manage- ment. JAMA, 289(2), 223–226.

Mayberry, L. J., & Gennarro, S. (2001). A quality of health outcomes model for guiding obstetrical practice. Journal of Nursing Scholarship, 33, 141–146.

McKinley, R., Gillepsie, W., Kraus, J., Harrison, S., Medeiros, R., Hawkins, et al. (2005). Focus group data as a tool in assessing effectiveness of a hand hygiene campaign. American Journal of Infection Control, 33(6), 368–373.

Miller, M. R., Pronovost, P., Donithan, M., Zeger, S., Zhan, C., Morlock, L., et al. (2005). Relationship between performance measurement and accreditation: Implications for quality of care and patient safety. American Journal of Medical Quality, 20(5), 239–252.

Mitchell, P., Ferketich, S., & Jennings, B. (1998). Quality Health Outcomes Model: American Academy of Nursing Expert Panel on Quality Health Care. Journal of Nursing Scholarship, 30, 43–46.

Needleman, J., Buerhaus, P., Mattke, S., Stewart, M., & Zelevinsky, K. (2002). Nurse-staffing levels and the quality of care in hospitals. New England Journal of Medicine, 346, 1715–1722.

Nordgren, L. D., Johnson, T., Kirschbaum, M., & Peterson, M. L. (2004). Medical errors: Excess hospital costs and lengths of stay. Journal for Healthcare Quality, 26(2), 42–48.

O’Grady, N. P., Alexander, M., Dellinger, P., Gerberding, J., Heard, S., Maki, D., et al. (2002). Guidelines for the pre- vention of intravascular catheter-related infections. Infection Control and Hospital Epidemiology, 23(12), 759–769.

Peterson, A. M., & Walker, P. H. (2006). Hospital-acquired infections as patient safety indicators. Annual Review of Nursing Research, 24, 75–99.

Polit, D. F., Beck, C. T., & Owen, S. V. (2007). Is the CVI an acceptable indicator of content validity? Appraisal and recommendations. Research in Nursing and Health, 30(4), 459–467.

Radwin, L., & Fawcett, J. (2002). A conceptual model-based program of nursing research: Retrospective and prospective applications. Journal of Advanced Nursing, 40, 355–360.

Safety Practices and Patient Outcomes 271

9Copyright @ 200 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

Reddy, M., Gill, S. S., & Rochon, P. A. (2006). Preventing pressure ulcers: A systematic review. JAMA, 296, 974–984.

Redelings, M. D., Lee, N. E., & Sorvillo, F. (2005). Pressure ulcers: More lethal than we thought? Advances in Skin & Wound Care, 18(7), 367–372.

Rojas, M., Silver, A., Llewellyn, C., & Rances, L. (2005). Study of adverse occurrences and major functional impairment following surgery. In Advances in patient safety: From research to implementation. Volumes 1–4 (AHRQ Publication No. 050021; 1–4). Rockville, MD: Agency for Healthcare Research and Quality.

Romano, P. S., Geppert, J. J., Davies, S., Miller, M. R., Elixhauser, A., & McDonald, K. M. (2003). A national profile of patient safety in US hospitals. Health Affairs, 22(2), 154–165.

Shojania, K., Duncan, B., McDonald, K., & Wachter, R. (2002). Safe but sound: Patient safety meets evidence- based medicine. JAMA, 288(4), 508–513.

Sin, M., Belza, B., LoGerfo, J., & Cunningham, S. (2005). Eval- uation of a community-based exercise program for elderly Korean immigrants. Public Health Nursing, 22(5), 407–413.

Stanton, M. (2004). Hospital nurse staffing and quality of care (Research in Action Issue 14. AHRQ Publication

No. 04-0029). Rockville, MD: Agency for Healthcare Research and Quality.

Tabachnick, B. G., & Fidell, L. S. (2001). Using multivariate statistics (4th ed.). Boston, MA: Allyn & Bacon.

Theaker, C. (2005). Infection control issues in central venous catheter care. Intensive & Critical Care Nursing, 21(2), 99–109.

Thomas, E., Studdert, D., Burstin, H., Orav, E., Zeena, T., Williams, E., et al. (2000). Incidence and types of adverse events and negligent care in Utah and Colorado. Medical Care, 38(3), 261–271.

Thornlow, D. (2008). Relationship of patient safety practices to patient outcomes. Saarbrucken, Germany: VDM Publishing.

Thornlow, D. K., & Stukenborg, G. J. (2006). The association between hospital characteristics and rates of preventable com- plications and adverse events. Medical Care, 44(3), 265–269.

Unruh, L. (2003). Licensed nurse staffing and adverse events in hospitals. Medical Care, 41(1), 142–152.

Wound, Ostomy, Continence Nursing (WOCN). (2003). Guidelines for the prevention and treatment of pressure ulcers. Retrieved April 1, 2005, from www.wocn.org

Zhan, C., & Miller, M. (2003). Administrative data based patient safety research: A critical review. Quality and Safety in Health Care, 12(Suppl. 2), ii58–ii63.

July–September � 2009272 Health Care Management REVIEW

9Copyright @ 200 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.