Abstract

Purpose of the Study: To examine the effects of the present on admission (POA) indicator on the prevalence of and factors associated with postsurgical adverse events in older patients. Design and Methods: This is a secondary data analysis of 82,898 surgical patients aged 65 years or older in 252 acute care hospitals in California in 2004. Four adverse events were counted using the Agency for Healthcare Research and Quality’s Patient Safety Indicator (PSI) definitions with and without using the POA indicator. We also examined the effects of the POA indicator on the relationships between patient- and hospital-level factors and adverse events, using generalized linear mixed models. Results: The use of the POA indicator resulted in a marked reduction in the estimated rates of all 4 adverse event rates. Adjustment for POA conditions also influenced factors associated with adverse events. Compared with those with newly occurring adverse events only, admissions with only POA conditions were more likely to be admitted through the emergency department, be unplanned, and belong to patients with one or more preceding admissions or those with multiple admissions within the same year. Implications: Adverse event rates estimated from discharge abstracts using PSI methodology could be overstated when the POA indicator was not used. The POA indicator could influence predictors of adverse events. Studies on geriatric safety and outcomes using large administrative data sets should consider using the POA indicator. Further studies are needed on how to determine POA conditions.

More than one third of all surgical procedures, or about 9 million, were performed on people aged 65 or older in 2005 (DeFrances & Hall, 2007), and this number is likely to increase as the U.S. population ages (U.S. Census Bureau, 2004). Compared with younger adults, older adults experience higher mortality and morbidity from surgical procedures (Johnson et al., 2007; Khuri et al., 2007; SooHoo, Lieberman, Ko, & Zingmond, 2006), which results in greater service use and cost (Kaye et al., 2009; Rivard et al., 2008). In spite of a broad awareness of older surgical patients as a high-risk group, a relatively small number of empirical studies have been conducted on adverse events in this group (Hamel, Henderson, Khuri, & Daley, 2005; Kaye et al., 2009; Turrentine, Wang, Simpson, & Jones, 2006). These studies are often limited to patients with a single diagnosis or receiving only one type of surgical procedure (Alexander et al., 2000; Lieberman, Romano, Mahendra, Keyzer, & Chilcott, 2006) or patients in a single hospital or veterans affairs hospitals only (Kaye et al., 2009; Turrentine et al.).

As of October 2008, the Centers for Medicare and Medicaid Services (CMS) has stopped paying for eight potentially preventable adverse events, so-called never events (DoBias, 2009). When implementing this new pay-for-performance policy, the CMS mandated the indication, in patient discharge abstracts, of whether or not a condition (diagnosis) was present on admission (POA). The rationale for the POA indicator is that if a diagnosis was POA, then the hospital possibly should not be penalized for its occurrence, unless it occurred during an immediately prior stay in the same hospital. The use of the POA indicator may affect reports of the prevalence of adverse events and their relationships with patient and hospital factors, which have not been examined in older patients.

The two aims of this study were to examine (a) the prevalence of four adverse events (decubitus ulcer, infections due to medical care, postoperative respiratory failure, and postoperative pulmonary embolism or deep vein thrombosis [DVT]) that are among the CMS’s “never events” in older surgical patients, with and without the use of the POA indicator and (b) whether the factors associated with adverse events were similar regardless of whether or not the POA indicator was used. We analyzed hospital discharge abstracts from California, which has been mandating POA coding since the mid-1990s.

Methods

Conceptual Framework

Patient safety literature emphasizes the importance of having a comprehensive systems view of various factors associated with adverse events (Kohn, Corrigan, & Donaldson, 2000; Vincent, Taylor-Adams, & Stanhope, 1998). In order to hold such a systems view, this study was based on Vincent and colleagues’ framework for analyzing risk and safety in clinical medicine. The framework was derived from Reason's (1995) model of organizational accidents and Hurst and Ratcliffe's (1994) socio-technical pyramid of safety management systems, and it was modified using medical literature on error and adverse outcomes. It is a framework for assessing the potential influences on patient safety, and it consists of two categories of influences: one includes patient, task, staff, and team characteristics and the other includes work environment, organizational context, and, more broadly, institutional context. By incorporating a wide range of factors influencing safety events, the framework guides formalized and extended analysis of how various factors affect adverse events in clinical practice. Based on this framework, this study adopted a systems approach examining the variations in adverse events among older surgical patients according to both patient and hospital factors, including work environment (e.g., nurse staffing level) and organizational factors (e.g., ownership and teaching status).

Empirical studies on surgical patient safety have examined a wide range of adverse events. In-hospital or 30-day mortality is the most widely observed (Alexander et al., 2000; Hamel et al., 2005; Kaye et al., 2009), but various morbidity measures such as infection or pulmonary embolism have been also used (Needleman, Buerhaus, Mattke, Stewart, & Zelevinsky, 2002; Poulose et al., 2005; Soohoo et al., 2006). Some studies focusing on patients having certain diagnoses or procedures have examined adverse events that are more specific to that population (Alexander et al., 2000); but many others adopted measures relevant to surgical patients in general, such as the Agency for Healthcare Research and Quality’s (AHRQ) Patient Safety Indicators (PSIs) that our study used (Hamel et al.; Khuri et al., 2005). Patient characteristics are often simply used as control variables or are variables in risk adjustment models (Mark, Harless, McCue, & Xu, 2004; Needleman et al.); but some studies have reported that adverse events are more likely to occur among the oldest of the old, those with higher comorbidities or medically underserved people, such as ethnic minorities or the uninsured (Cho, Ketefian, Barkauskas, & Smith, 2003; Hamel et al.). Empirical studies also support a potential link between work environment and organizational factors and adverse events (Yuan, Cooper, Einstadter, Randall, & Rimm, 2000; Zhan & Miller, 2003). For example, nurse skill mix and staffing have been shown to be associated with the occurrence of pressure ulcers, postoperative respiratory failure, and postoperative pulmonary embolism or DVT (Cho et al.; Kovner & Gergen, 1998; Needleman et al.); and organizational change has been associated with a decrease in catheter-related bloodstream infections (Pronovost et al., 2006).

Data Sources

Patient data were obtained from the 2004 California State Inpatient Data set (SID), which is part of the Healthcare Cost and Utilization Project (HCUP) sponsored by AHRQ (2009). The California SID includes uniform hospital discharge abstracts for all inpatient stays in California in calendar year 2004, and it provides clinical and nonclinical information on all patients regardless of payer. Among the states participating in the HCUP in 2004, California and New York were the only two states that provided the indicator of whether or not a diagnosis was POA (Coffey, Milenkovic, & Andrews, 2006), and the POA coding in the California data set was highly reliable (Houchens, Elixhauser, & Romano, 2008). The POA indicator is critical for increasing the validity of identification of the incidence of adverse events during hospitalization in studies analyzing secondary administrative data, which are often criticized for overcounting adverse events by ignoring POA conditions (Bahl, Thompson, Kau, Hu, & Campbell, 2008; Naessens, Campbell, Berg, Williams, & Culbertson, 2007).

Hospital data were obtained from the California hospital cost report (California Office of Statewide Health Planning and Development [COSHPD], 2004), a mandatory report submitted by all licensed California hospitals to California’s Department of Health. It includes detailed data on hospital organizational characteristics such as teaching status, ownership, and staffing. The California hospital cost report is known to be a more reliable data source for nurse staffing than other sources, such as the American Hospital Association hospital survey data (Jiang, Stocks, & Wong, 2006).

Sample

The sample included 82,898 discharges of older adults aged 65 or older who underwent a surgical procedure in 252 short-term general acute hospitals in California. We selected older adults in 20 common surgical diagnosis-related groups (DRGs), who had undergone major general, orthopedic, or vascular surgery procedures, similar to the selection procedure used by others (Aiken, Clarke, Sloane, Sochalski, & Silber, 2002). These procedural DRGs have relatively large eligible older adult populations (denominators) and a higher incidence of one or more of the four adverse events (numerators). We excluded hospitals that had fewer than 20 total older surgical patient discharges for the 20 DRGs selected in this study and those with missing values in variables included in the analysis or with reporting errors. The excluded hospitals were more likely to be smaller nonteaching, public hospitals.

Measures

Adverse events, the outcome variables of this study, were identified using AHRQ’s (2008b) PSI software version 3.2. The software algorithms were developed by the University of California Stanford Evidence-Based Practice Center with funding from AHRQ, through extensive literature review, consultation with clinicians and coding experts, and empirical testing using HCUP data (Miller, Elixhauser, Zhan, & Meyer, 2001). AHRQ’s PSIs have been widely used to examine safety and quality of care in hospitals (Miller et al.; Thornlow & Stukenborg, 2006), although there are concerns about their validity as tools to compare quality of care across hospitals (Bahl et al., 2008; Miller et al.; Rivard et al., 2008).

The PSI algorithm searches the ICD-9-CM codes related to adverse events in up to 25 secondary diagnoses in the California SID and indicates in the discharge record whether or not an adverse event occurred (AHRQ, 2008b). The PSI algorithm takes a conservative approach in counting adverse events in order to improve validity and specificity of case identifications based on administrative data. First, the algorithm includes an indicator to better screen out cases with a diagnosis related to an adverse event that may be POA, if known. Second, the PSI algorithm does not include in the population (denominator), all cases with relevant ICD-9-CM codes related to adverse events. Rather, the algorithm uses certain exclusion criteria in identifying the eligible risk pool. For example, the risk pool for decubitus ulcers that occur during hospitalization excludes patients who were transferred from another acute care facility, were admitted from a long-term care facility, stayed fewer than 5 days, etc. The four adverse events observed in this study were decubitus ulcers (PSI 3), infections due to medical care (PSI 7), postoperative respiratory failure (PSI 11), and postoperative pulmonary embolism or DVT (PSI 12), which are common postoperational complications in older patients (Hamel et al., 2005; Thomas & Brennan, 2000). The prevalence of the other 16 PSIs were all less than 1.0 in 1,000 discharges in our sample after adjusting for POA conditions, which limits their reliability as safety measures (West, Weeks, & Bagian, 2007), so we did not adopt them. The definitions of the PSIs and their detailed inclusion and exclusion criteria in identifying eligible cases can be found in the PSI technical specifications document provided by AHRQ (2008b).

A key patient characteristic observed in this study was number of chronic conditions. Studies have reported that multiple chronic conditions (MCC) significantly increase health care cost and mortality in older adults (Vogeli et al., 2007), but few studies have examined the impacts of MCC on safety-related events in hospitalized older patients. In this study, the number of chronic conditions was calculated using the chronic conditions index (CCI) algorithm (AHRQ, 2008a). The CCI algorithm categorizes approximately 13,600 current ICD-9-CM diagnosis codes into two groups: chronic and nonchronic conditions. A chronic condition is defined as a condition that lasts 12 months or longer and meets one or both of the following conditions: (a) it places limitations on self-care, independent living, and social interactions; and/or (b) it results in the need for ongoing intervention with medical products, services, and special equipment (AHRQ, 2008a). After categorizing the primary and up to 25 secondary diagnoses as chronic or nonchronic conditions, we divided the ICD-9-CM diagnosis codes into 18 specific body system categories using 18 major diagnostic categories (MDCs; e.g., diseases and disorders of the musculoskeletal system). A patient might have had more than one chronic condition in a certain MDC, but we regarded the MDC as the unit of comorbidity and counted that patient as having one chronic condition that affected a specific body system (Wolff, Starfield, & Anderson, 2002). The risk for experiencing adverse events during hospitalization may also vary by patients’ other demographic and clinical characteristics, and we adjusted for the following variables, selected based on the literature: age, sex, race, primary payer, source of admission (emergency department [ED] admission or not), and primary DRG (Cho et al., 2003; Kovner, Jones, Zhan, Gergen, & Basu, 2002; Wolff et al.).

We also examined several hospital characteristics that have been widely examined in patient safety literature (Aiken et al., 2002; Kovner et al., 2002; SooHoo et al., 2006). Hospitals were categorized as nonprofit, public, or for-profit hospitals according to ownership, following the COSHPD (2004) definitions of type of care. The size of the hospitals was categorized based on the average number of staffed beds: small (1–99 beds), medium (100–299 beds), or large (300+ beds; Aiken et al.). Teaching hospitals were defined based on the American Medical Association’s Graduate Medical Education Directory, and rural hospitals were defined by Section 124840 of the California Health and Safety Codes (COSHPD). Total nursing hours per patient day is the sum of the adjusted productive nursing hours of registered nurses, licensed vocational nurses, and nurse aides delivering direct nursing care to patients, divided by total inpatient days. We excluded nonproductive hours—paid-non-working hours such as sick-leave or paid vacations—and adjusted the total nursing hours by multiplying it by the ratio of inpatient revenues to total revenues (Kovner et al.; Mark et al., 2004) because only aggregated nursing hours including both inpatient and outpatient hours were reported in the California financial report.

Analytic Approaches

The unit of analysis of this study was not the patient but the discharge as the PSI algorithm we adopted is discharge based. Using descriptive statistics, we described patient and hospital characteristics and summarized adverse events rates calculated with and without adjustment for POA conditions. We developed two generalized linear mixed models to examine the relationships of patient and hospital factors to adverse events. The models were identical except for whether or not POA conditions were adjusted for. The generalized linear mixed models took into account the two levels (patient and hospital) of data structure and simultaneously examined the effects of patient- and hospital-level predictors on adverse events. Lastly, we compared the characteristics of admissions (e.g., route, type, and existence of preceding hospitalization) with POA conditions with those with non-POA conditions using the chi-square test. Data management and analysis were conducted with SAS 9.1.2. The SAS GLIMMIX was used for hierarchical modeling of the four dichotomous outcome variables.

Results

The majority of older adults in our sample were White (79.1%); female (61.9%); Medicare patients (88.9%); and in nonprofit hospitals (59.1%; Table 1). The older adults were 77.2 years old with 3.2 chronic conditions on average (SD = 1.5, range = 1–11). Overall, 65.0% of patients had three or more chronic conditions; about 17.4% of patients had five or more types of chronic conditions; and about 2.1% of patients had seven or more types of chronic conditions.

Table 1.

Characteristics of Older Surgical Patients in California Acute Care Hospitals, 2004

Characteristic n (%) M (SD
Patient discharges (n = 82,898) 
    Age (in years)  77.2 (7.6) 
    Sex 
        Male 31,581 (38.1)  
        Female 51,317 (61.9)  
    Race/ethnicity 
        White 65,589 (79.1)  
        Black 2,799 (3.4)  
        Hispanic 9,338 (11.3)  
        Other 5,172 (6.2)  
    Primary payer 
        Medicare 73,702 (88.9)  
        Medicaid 2,319 (2.8)  
        Private insurance 5,852 (7.1)  
     Other 1,025 (1.2)  
    ED admission (yes = 1) 29,958 (36.1)  
    Number of chronic conditions  3.2 (1.5) 
    Surgery typea 
        General 23,542 (28.4)  
        Orthopedic 51,010 (61.5)  
        Vascular 8,346 (10.1)  
    Selected comorbid conditions 
        Hypertension 45,206 (54.5)  
        Diabetes with/without chronic complications 15,022 (18.1)  
        Chronic pulmonary disease 13,855 (16.7)  
        Hypothyroidism 10,057 (12.1)  
        Congestive heart failure 8,106 (9.8)  
Hospitals (n = 252) 
    Ownership 
        Profit 63 (25.8)  
        Public 38 (15.1)  
        Nonprofit 145 (59.1)  
    Size 
        Small (fewer than 99 beds) 55 (21.8)  
        Medium (100–299 beds) 139 (55.2)  
        Large (300+ beds) 58 (23.0)  
    Teaching hospital (yes = 1) 22 (8.7)  
    Rural hospital (yes = 1) 33 (13.1)  
    Total nursing hours per patient day  10.2 (2.5) 
Characteristic n (%) M (SD
Patient discharges (n = 82,898) 
    Age (in years)  77.2 (7.6) 
    Sex 
        Male 31,581 (38.1)  
        Female 51,317 (61.9)  
    Race/ethnicity 
        White 65,589 (79.1)  
        Black 2,799 (3.4)  
        Hispanic 9,338 (11.3)  
        Other 5,172 (6.2)  
    Primary payer 
        Medicare 73,702 (88.9)  
        Medicaid 2,319 (2.8)  
        Private insurance 5,852 (7.1)  
     Other 1,025 (1.2)  
    ED admission (yes = 1) 29,958 (36.1)  
    Number of chronic conditions  3.2 (1.5) 
    Surgery typea 
        General 23,542 (28.4)  
        Orthopedic 51,010 (61.5)  
        Vascular 8,346 (10.1)  
    Selected comorbid conditions 
        Hypertension 45,206 (54.5)  
        Diabetes with/without chronic complications 15,022 (18.1)  
        Chronic pulmonary disease 13,855 (16.7)  
        Hypothyroidism 10,057 (12.1)  
        Congestive heart failure 8,106 (9.8)  
Hospitals (n = 252) 
    Ownership 
        Profit 63 (25.8)  
        Public 38 (15.1)  
        Nonprofit 145 (59.1)  
    Size 
        Small (fewer than 99 beds) 55 (21.8)  
        Medium (100–299 beds) 139 (55.2)  
        Large (300+ beds) 58 (23.0)  
    Teaching hospital (yes = 1) 22 (8.7)  
    Rural hospital (yes = 1) 33 (13.1)  
    Total nursing hours per patient day  10.2 (2.5) 

Notes: ED = emergency department.

a

Selected diagnosis-related groups and some frequent procedures: general surgeries (146, 148, 150, 154, 159, 164, 170, 191, 197, and 493; rectal resection, hemicolectomy, sigmoidectomy, small bowel resection, astrectomy, hernia repair, appendectomy, and excision; lysis peritoneal adhesions); orthopedic surgeries (209, 210, 217, 471, 496, 497, and 499; arthroplasty knee, hip replacement, treatment; fracture or dislocation of hip and femur, debridement of wound, spinal fusion, and laminectomy); and vascular surgeries (110, 113, and 120; aortic resection, peripheral vascular bypass, knee amputation, and amputation through foot).

Without adjusting for POA indicators, about 26 patients per 1,000 discharges had decubitus ulcers (Table 2). The next most prevalent adverse event was postoperative pulmonary embolism or DVT (15.6 per 1,000 discharges), followed by postoperative respiratory failure (12.3). Infections due to medical care was the least prevalent (4.1). After adjusting for POA indicators—in other words, excluding adverse events POA—the rates of all adverse events dropped and the ranks were changed. Postoperative respiratory failure had the highest incidence at 9.7 per 1,000 discharges, followed by decubitus ulcers (5.8). The largest decrease in adverse events rate after adjusting for the POA indicators was in the decubitus ulcer rate, which decreased by 77.4%, from 25.6 to 5.8, followed by postoperative pulmonary embolism or DVT (−69.9%) and infection due to medical care (−53.7%).

Table 2.

Adverse Event Rates

 Unadjusted for POA
 
Adjusted for POA
 
% Decrease of adverse events 
AHRQ’s PSI Adverse event rates/1,000 Adverse event rates/1,000 
PSI 03 decubitus ulcer 41,562 1,062 25.6 40,737 237 5.8 77.4 
PSI 07 infections due to medical care 57,973 240 4.1 57,844 111 1.9 53.7 
PSI 11 postoperative respiratory failure 39,792 488 12.3 39,689 385 9.7 21.1 
PSI 12 postoperative pulmonary embolism or deep vein thrombosis 82,372 1,288 15.6 81,467 383 4.7 69.9 
 Unadjusted for POA
 
Adjusted for POA
 
% Decrease of adverse events 
AHRQ’s PSI Adverse event rates/1,000 Adverse event rates/1,000 
PSI 03 decubitus ulcer 41,562 1,062 25.6 40,737 237 5.8 77.4 
PSI 07 infections due to medical care 57,973 240 4.1 57,844 111 1.9 53.7 
PSI 11 postoperative respiratory failure 39,792 488 12.3 39,689 385 9.7 21.1 
PSI 12 postoperative pulmonary embolism or deep vein thrombosis 82,372 1,288 15.6 81,467 383 4.7 69.9 

Note: AHRQ = Agency for Healthcare Research and Quality; D = denominator; N = numerator; POA = present on admission; PSI = Patient Safety Indicator.

The relationships between several patient and hospital factors and adverse events changed according to whether or not POA conditions were adjusted for. When counting all adverse events regardless of whether they were POA (Table 3), being female (odds ratio [OR] = 1.15, compared with being male), being Black (OR = 1.51, compared with being White), and receiving surgery from a public hospital (OR = 1.49, compared with nonprofit hospitals) increased the odds for decubitus ulcer; but all these were not significant when we counted only adverse events that were newly occurred during hospitalization (Table 4). Similarly, payer type and total nursing hours were no longer significantly associated with postoperative pulmonary embolism or DVT after adjusting for POA conditions (Table 4).

Table 3.

Estimated Relationships of Factors Associated With Adverse Events Without Adjustment for POA Conditions

 Decubitus ulcer (PSI 03) Infections due to medical care (PSI 07) Postoperative respiratory failure (PSI 11) Postoperative pulmonary embolism or deep vein thrombosis (PSI 12) 
 OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) 
Age 1.03*** (1.02–1.04) 1.00 (0.98–1.02) 1.01 (1.00–1.03) 1.00 (0.99–1.01) 
Female (vs. male) 1.15* (1.01–1.32) 0.92 (0.70–1.20) 0.55*** (0.46–0.66) 1.02 (0.91–1.15) 
Race (vs. White) 
    Black 1.51** (1.16–1.98) 1.18 (0.65–2.16) 0.71 (0.38–1.32) 0.96 (0.73–1.26) 
    Hispanic 1.08 (0.89–1.32) 0.91 (0.60–1.39) 1.05 (0.76–1.44) 0.87 (0.72–1.04) 
    Race: other 0.92 (0.70–1.22) 0.83 (0.46–1.47) 1.26 (0.86–1.84) 0.70** (0.55–0.90) 
No. of chronic conditions 1.88*** (1.80–1.97) 1.22*** (1.12–1.33) 1.60*** (1.51–1.71) 1.13*** (1.09–1.17) 
Primary payer (vs. Medicare) 
    Medicaid 1.34 (0.97–1.84) 1.38 (0.70–2.69) 0.94 (0.45–1.96) 1.57** (1.18–2.08) 
    Private insurance 1.11 (0.83–1.47) 0.92 (0.55–1.54) 1.07 (0.76–1.51) 1.02 (0.82–1.27) 
    Payer: other 0.98 (0.50–1.94)  0.73 (0.30–1.79) 0.94 (0.54–1.64) 
ED admission (1 = yes) 1.76*** (1.51–2.05) 3.12*** (2.32–4.19)  1.41*** (1.24–1.60) 
Hospital size (vs. medium) 
    Size: small 0.90 (0.69–1.18) 0.77 (0.41–1.46) 0.73 (0.46–1.16) 0.67* (0.49–0.91) 
    Size: large 0.87 (0.75–1.01) 1.29 (0.96–1.71) 0.86 (0.69–1.06) 1.16* (1.02–1.32) 
Teaching (vs. nonteaching) 1.12 (0.90–1.40) 1.34 (0.91–1.96) 0.89 (0.66–1.22) 1.87*** (1.60–2.19) 
Ownership (vs. nonprofit) 
    Profit 1.41*** (1.19–1.68) 0.96 (0.66–1.40) 1.56*** (1.22–2.00) 1.17 (0.99–1.37) 
    Public 1.49*** (1.20–1.83) 0.69 (0.41–1.16) 1.82*** (1.32–2.50) 0.92 (0.74–1.14) 
   Rural (vs. nonrural) 1.04 (0.74–1.45) 0.92 (0.44–1.92) 0.87 (0.50–1.54) 0.77 (0.52–1.13) 
Total nursing HPPD 1.00 (0.97–1.03) 0.98 (0.92–1.04) 1.05 (1.00–1.10) 1.05** (1.02–1.08) 
 Decubitus ulcer (PSI 03) Infections due to medical care (PSI 07) Postoperative respiratory failure (PSI 11) Postoperative pulmonary embolism or deep vein thrombosis (PSI 12) 
 OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) 
Age 1.03*** (1.02–1.04) 1.00 (0.98–1.02) 1.01 (1.00–1.03) 1.00 (0.99–1.01) 
Female (vs. male) 1.15* (1.01–1.32) 0.92 (0.70–1.20) 0.55*** (0.46–0.66) 1.02 (0.91–1.15) 
Race (vs. White) 
    Black 1.51** (1.16–1.98) 1.18 (0.65–2.16) 0.71 (0.38–1.32) 0.96 (0.73–1.26) 
    Hispanic 1.08 (0.89–1.32) 0.91 (0.60–1.39) 1.05 (0.76–1.44) 0.87 (0.72–1.04) 
    Race: other 0.92 (0.70–1.22) 0.83 (0.46–1.47) 1.26 (0.86–1.84) 0.70** (0.55–0.90) 
No. of chronic conditions 1.88*** (1.80–1.97) 1.22*** (1.12–1.33) 1.60*** (1.51–1.71) 1.13*** (1.09–1.17) 
Primary payer (vs. Medicare) 
    Medicaid 1.34 (0.97–1.84) 1.38 (0.70–2.69) 0.94 (0.45–1.96) 1.57** (1.18–2.08) 
    Private insurance 1.11 (0.83–1.47) 0.92 (0.55–1.54) 1.07 (0.76–1.51) 1.02 (0.82–1.27) 
    Payer: other 0.98 (0.50–1.94)  0.73 (0.30–1.79) 0.94 (0.54–1.64) 
ED admission (1 = yes) 1.76*** (1.51–2.05) 3.12*** (2.32–4.19)  1.41*** (1.24–1.60) 
Hospital size (vs. medium) 
    Size: small 0.90 (0.69–1.18) 0.77 (0.41–1.46) 0.73 (0.46–1.16) 0.67* (0.49–0.91) 
    Size: large 0.87 (0.75–1.01) 1.29 (0.96–1.71) 0.86 (0.69–1.06) 1.16* (1.02–1.32) 
Teaching (vs. nonteaching) 1.12 (0.90–1.40) 1.34 (0.91–1.96) 0.89 (0.66–1.22) 1.87*** (1.60–2.19) 
Ownership (vs. nonprofit) 
    Profit 1.41*** (1.19–1.68) 0.96 (0.66–1.40) 1.56*** (1.22–2.00) 1.17 (0.99–1.37) 
    Public 1.49*** (1.20–1.83) 0.69 (0.41–1.16) 1.82*** (1.32–2.50) 0.92 (0.74–1.14) 
   Rural (vs. nonrural) 1.04 (0.74–1.45) 0.92 (0.44–1.92) 0.87 (0.50–1.54) 0.77 (0.52–1.13) 
Total nursing HPPD 1.00 (0.97–1.03) 0.98 (0.92–1.04) 1.05 (1.00–1.10) 1.05** (1.02–1.08) 

Notes: Diagnosis-related groups were adjusted for in all the models; and because of a very low prevalence of adverse events among patients with Payer: other in the model for PSI 07 and patients with an ED admission in the model for PSI 11, the estimates of the two variables were unstable, so we omitted them from the final models reported in the table. Regardless of this, estimations of the other variables in the models were consistent. CI = confidence interval; ED = emergency department; HPPD = hours per patient day; OR = odds ratio; POA = present on admission; PSI = Patient Safety Indicator.

*p < .05. **p < .01. ***p < .001.

Table 4.

Estimated Relationships of Factors Associated With Adverse Events With Adjustment for POA Conditions

 Decubitus ulcer (PSI 03) Infections due to medical care (PSI 07) Postoperative respiratory failure (PSI 11) Postoperative pulmonary embolism or deep vein thrombosis (PSI 12) 
 OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) 
Age 1.02* (1.00–1.04) 0.99 (0.97–1.02) 1.01 (1.00–1.03) 1.00 (0.98–1.01) 
Female (vs. male) 0.97 (0.74–1,27) 0.89 (0.60–1.32) 0.46*** (0.38–0.57) 1.00 (0.81–1.23) 
Race (vs. White) 
    Black 1.64 (0.98–2.73) 1.59 (0.75–3.36) 0.95 (0.51–1.76) 0.87 (0.51–1.47) 
    Hispanic 0.84 (0.53–1.31) 0.40* (0.17–0.94) 1.12 (0.79–1.59) 0.94 (0.67–1.32) 
    Race: other 1.12 (0.67–1.88) 0.78 (0.33–1.82) 1.64* (1.10–2.44) 0.52* (0.31–0.89) 
No. of chronic conditions 1.87*** (1.71–2.04) 1.14* (1.01–1.30) 1.58*** (1.47–1.69) 1.12*** (1.04–1.19) 
Primary payer (vs. Medicare) 
    Medicaid 1.13 (0.56–2.29) 1.42 (0.50–4.06) 1.26 (0.61–2.63) 0.70 (0.33–1.52) 
    Private insurance 1.05 (0.61–1.83) 0.95 (0.46–1.96) 1.06 (0.73–1.55) 0.92 (0.61–1.37) 
    Payer: other 0.48 (0.07–3.44)  0.67 (0.25–1.83) 1.17 (0.48–2.86) 
ED admission (1 = yes) 1.73*** (1.29–2.32) 4.03*** (2.59–6.25)  1.48*** (1.17–1.87) 
Hospital size (vs. medium) 
    Size: small 1.09 (0.65–1.84) 1.13 (0.48–2.63) 0.73 (0.43–1.25) 0.40* (0.19–0.83) 
    Size: large 0.81 (0.60–1.10) 1.46 (0.96–2.24) 0.89 (0.70–1.13) 1.26 (1.00–1.59) 
Teaching (vs. nonteaching) 1.44 (0.95–2.18) 1.21 (0.69–2.13) 1.11 (0.79–1.54) 1.73*** (1.31–2.29) 
Ownership (vs. nonprofit) 
    Profit 1.50* (1.07–2.11) 1.13 (0.67–1.91) 1.52** (1.14–2.01) 1.09 (0.81–1.48) 
    Public 1.10 (0.68–1.79) 0.79 (0.38–1.65) 1.82*** (1.28–2.59) 0.60* (0.38–0.96) 
Rural (vs. nonrural) 1.07 (0.54–2.11) 0.60 (0.17–2.08) 0.72 (0.37–1.41) 0.59 (0.25–1.39) 
Total nursing HPPD 1.01 (0.54–2.11) 1.02 (0.93–1.12) 1.04 (0.99–1.09) 1.05 (0.99–1.10) 
 Decubitus ulcer (PSI 03) Infections due to medical care (PSI 07) Postoperative respiratory failure (PSI 11) Postoperative pulmonary embolism or deep vein thrombosis (PSI 12) 
 OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) 
Age 1.02* (1.00–1.04) 0.99 (0.97–1.02) 1.01 (1.00–1.03) 1.00 (0.98–1.01) 
Female (vs. male) 0.97 (0.74–1,27) 0.89 (0.60–1.32) 0.46*** (0.38–0.57) 1.00 (0.81–1.23) 
Race (vs. White) 
    Black 1.64 (0.98–2.73) 1.59 (0.75–3.36) 0.95 (0.51–1.76) 0.87 (0.51–1.47) 
    Hispanic 0.84 (0.53–1.31) 0.40* (0.17–0.94) 1.12 (0.79–1.59) 0.94 (0.67–1.32) 
    Race: other 1.12 (0.67–1.88) 0.78 (0.33–1.82) 1.64* (1.10–2.44) 0.52* (0.31–0.89) 
No. of chronic conditions 1.87*** (1.71–2.04) 1.14* (1.01–1.30) 1.58*** (1.47–1.69) 1.12*** (1.04–1.19) 
Primary payer (vs. Medicare) 
    Medicaid 1.13 (0.56–2.29) 1.42 (0.50–4.06) 1.26 (0.61–2.63) 0.70 (0.33–1.52) 
    Private insurance 1.05 (0.61–1.83) 0.95 (0.46–1.96) 1.06 (0.73–1.55) 0.92 (0.61–1.37) 
    Payer: other 0.48 (0.07–3.44)  0.67 (0.25–1.83) 1.17 (0.48–2.86) 
ED admission (1 = yes) 1.73*** (1.29–2.32) 4.03*** (2.59–6.25)  1.48*** (1.17–1.87) 
Hospital size (vs. medium) 
    Size: small 1.09 (0.65–1.84) 1.13 (0.48–2.63) 0.73 (0.43–1.25) 0.40* (0.19–0.83) 
    Size: large 0.81 (0.60–1.10) 1.46 (0.96–2.24) 0.89 (0.70–1.13) 1.26 (1.00–1.59) 
Teaching (vs. nonteaching) 1.44 (0.95–2.18) 1.21 (0.69–2.13) 1.11 (0.79–1.54) 1.73*** (1.31–2.29) 
Ownership (vs. nonprofit) 
    Profit 1.50* (1.07–2.11) 1.13 (0.67–1.91) 1.52** (1.14–2.01) 1.09 (0.81–1.48) 
    Public 1.10 (0.68–1.79) 0.79 (0.38–1.65) 1.82*** (1.28–2.59) 0.60* (0.38–0.96) 
Rural (vs. nonrural) 1.07 (0.54–2.11) 0.60 (0.17–2.08) 0.72 (0.37–1.41) 0.59 (0.25–1.39) 
Total nursing HPPD 1.01 (0.54–2.11) 1.02 (0.93–1.12) 1.04 (0.99–1.09) 1.05 (0.99–1.10) 

Notes: Diagnosis-related groups were adjusted for in all the models; and because of a very low prevalence of adverse events among patients with Payer: other in the model for PSI 07 and patients with an ED admission in the model for PSI 11, the estimates of the two variables were unstable, so we omitted them from the final models reported in the table. Regardless of this, estimations of the other variables in the models were consistent. CI = confidence interval; ED = emergency department; HPPD = hours per patient day; OR = odds ratio; POA = present on admission; PSI = Patient Safety Indicator.

*p < .05. **p < .01. ***p < .001.

When adjusting for POA conditions, having a higher number of chronic conditions increased the risks for all four adverse events we observed (Table 4). On average, an increase in the number of chronic conditions corresponded to increases of 87% in the likelihood of having decubitus ulcers, 58% in the likelihood of postoperative pulmonary embolism or DVT, 14% in the odds of infections due to medical care, and 12% in the odds of postoperative pulmonary embolism or DVT. Patients admitted via the ED were more likely to experience three of the four adverse events: infections due to medical care (OR = 4.03), decubitus ulcers (OR = 1.73), and postoperative pulmonary embolism or DVT (OR = 1.48). Among the hospital characteristics, patients in profit hospitals were more likely to experience decubitus ulcers (OR = 1.50) and also postoperative pulmonary embolism or DVT (OR = 1.52) than those in nonprofit hospitals.

We compared the characteristics of admissions with a POA condition with those with adverse events newly occurring during hospitalization (Table 5). The admissions with POA decubitus ulcers were more likely to be admitted through the ED and to be unplanned compared with those with non-POA decubitus ulcers. Patients with POA postoperative pulmonary embolism or DVT were more likely to have had one or more preceding admissions and to have been hospitalized more than once compared with their counterparts. Lastly, compared with those with only non-POA conditions, the admissions with one or more adverse events that were all POA conditions were more likely to be admitted through the ED, to be unplanned admissions and to have a preceding admission and multiple admissions, during 2004.

Table 5.

Comparison of Admissions With POA Conditions With Admissions With Non-POA Conditions

 Decubitus ulcer (PSI 03) Infections due to medical care (PSI 07) Postoperative respiratory failure (PSI 11) Postoperative pulmonary embolism or deep vein thrombosis (PSI 12) All 
 POA Non-POA χ2 POA Non-POA χ2 POA Non-POA χ2 POA Non-POA χ2 POAa Non-POAb χ2 
 
Admission through ED 
    Yes 72.2 65.1 4.40* 60.6 66.7 0.93 0.0 0.3 0.27 47.1 44.2 0.92 56.3 35.7 110.73*** 
    No 27.8 34.9  39.4 33.3  100.0 99.7  52.9 55.8  43.7 64.3  
Planned admission 
    Yes 14.3 26.0 17.51*** 22.0 16.2 1.29 100.0 100.0  36.7 40.0 1.23 29.3 55.3 187.78*** 
    No 85.7 74.0  78.0 83.8  0.0 0.0  63.3 60.0  70.7 44.7  
First admission in 2004c 
    Yes 52.2 57.0 1.70 56.7 68.5 3.49 68.0 65.7 0.18 59.7 65.9 4.30* 57.4 65.1 16.60** 
    No 47.8 43.0  43.3 31.5  32.0 34.3  40.3 34.1  42.6 34.9  
Total number of admissions in 2004c 
    1 26.6 29.8 0.94 34.6 45.0 2.68 32.0 40.3 2.32 34.9 42.3 6.37* 31.6 39.8 19.42*** 
    1+ 73.4 70.2  65.4 55.0  68.0 59.7  65.1 57.7  68.4 60.2  
Total 100.0 100.0  100.0 100.0  100.0 100.0  100.0 100.0  100.0 100.0  
 Decubitus ulcer (PSI 03) Infections due to medical care (PSI 07) Postoperative respiratory failure (PSI 11) Postoperative pulmonary embolism or deep vein thrombosis (PSI 12) All 
 POA Non-POA χ2 POA Non-POA χ2 POA Non-POA χ2 POA Non-POA χ2 POAa Non-POAb χ2 
 
Admission through ED 
    Yes 72.2 65.1 4.40* 60.6 66.7 0.93 0.0 0.3 0.27 47.1 44.2 0.92 56.3 35.7 110.73*** 
    No 27.8 34.9  39.4 33.3  100.0 99.7  52.9 55.8  43.7 64.3  
Planned admission 
    Yes 14.3 26.0 17.51*** 22.0 16.2 1.29 100.0 100.0  36.7 40.0 1.23 29.3 55.3 187.78*** 
    No 85.7 74.0  78.0 83.8  0.0 0.0  63.3 60.0  70.7 44.7  
First admission in 2004c 
    Yes 52.2 57.0 1.70 56.7 68.5 3.49 68.0 65.7 0.18 59.7 65.9 4.30* 57.4 65.1 16.60** 
    No 47.8 43.0  43.3 31.5  32.0 34.3  40.3 34.1  42.6 34.9  
Total number of admissions in 2004c 
    1 26.6 29.8 0.94 34.6 45.0 2.68 32.0 40.3 2.32 34.9 42.3 6.37* 31.6 39.8 19.42*** 
    1+ 73.4 70.2  65.4 55.0  68.0 59.7  65.1 57.7  68.4 60.2  
Total 100.0 100.0  100.0 100.0  100.0 100.0  100.0 100.0  100.0 100.0  

Notes: ED = emergency department; POA = present on admission; PSI = Patient Safety Indicator.

a

This group includes admissions triggered with one or more PSIs, all of which were POA conditions.

b

This group includes admissions triggered with one or more PSIs, all of which were non-POA conditions.

c

The unit of analysis of this study was the discharge (admission), but, in order to examine the pattern of hospital service use of a patient with a discharge with one or more PSIs triggered, we created these two patient-based variables by merging additional data including scrambled patient identification numbers into our analytic data set.

*p < .05. **p < .01. ***p < .001.

Discussion

To our best knowledge, our study is the first population-based secondary data analysis reporting a large decrease of adverse event rates after adjustment for POA conditions in hospitalized older patients. This study highlights the importance of adjusting for conditions that were POA in evaluating the extent of adverse events and suggests that adverse event rates could be overstated when POA conditions are not adjusted for in geriatric patient safety studies using large administrative data. Only approximately one fifth of the decubitus ulcers (PSI 03) in secondary diagnoses were events newly occurring during hospitalization. Similarly, less than half of the infections due to medical care (PSI 07, 46.3%) and postoperative pulmonary embolism or DVT (PSI 12, 30.1%) were reported as adverse events newly occurring during hospitalization.

We adopted the widely used definitions of AHRQ’s PSI algorithms in identifying the adverse events. When adjusting for POA conditions, the incidence of decubitus ulcers dropped from 25.6 to 5.8 per 1,000 patients in the study, which was slightly lower than the 8.8 reported in one single-hospital study using a different version of PSI software (Bahl et al., 2008) and the 7.9 reported in another single-hospital study based on a chart review of 123 patients (Horwitz, Cuny, Cerese, & Krumholz, 2007). The four adverse events we observed were all never events for which CMS has discontinued reimbursement (DoBias, 2009), but we could not adopt CMS’s algorithm because it was not fully developed when this study was conducted. Empirical studies evaluating the algorithms could provide critical information regarding the success of the new CMS policy.

Factors associated with adverse events were also not consistent when the POA indicator was used versus when it was not used. Several patient and hospital characteristics (e.g., being Black, having Medicaid as the primary payer, total nursing hours) were no longer significant risk factors for one or more adverse events when POA conditions were adjusted for, which implies prediction models of adverse events without the use of the POA indicator could be biased. Further studies on the influences of the POA indicator on geriatric patient outcomes with different samples and different measures are necessary.

Unlike other factors, chronic illness burden and ED admission were two key patient characteristics consistently positively associated with adverse events regardless of adjustment for POA conditions. Older patients with a greater number of chronic conditions were at a higher risk for all four adverse events we observed. These findings are consistent with existing studies reporting that patients with MCC are more likely to experience negative outcomes, such as preventable complications, death, and high health care costs (Vogeli et al., 2007; Wolff et al., 2002). About one third of the older patients who underwent major surgical procedures in our sample were admitted through the ED, and they were more likely to experience three of the four adverse events. By targeting these high comorbid patients admitted through the ED for early, organization-wide safety interventions, hospitals’ safety initiatives would be more effective (Khaliq & Broyles, 2006). Compared with patients in nonprofit hospitals, those in profit hospitals were consistently more likely to experience two of the four adverse events—decubitus ulcer and postoperative respiratory failure—but not the two other events. Similar to our study, other studies have reported inconsistent findings between hospital ownership and other characteristics and patient safety (Thornlow & Stukenborg, 2006), and further studies are necessary.

Compared with patient discharges with newly occurring adverse events, those with POA conditions showed complex utilization patterns of hospital care—more unplanned admissions, more admissions via the ED, and more preceding hospitalizations. This supports our contention that POA conditions are related to (and potentially caused by) prior hospital encounters and should not be attributed to the current hospital stay. On the other hand, this finding indicates that an episode of illness can span several health care encounters (including more than one hospital stay) and that limiting PSI events to those that occur during a single-hospital stay excludes events that could have been caused by poor preceding hospital care, which is a study limitation. Thus, future research is needed to classify whole episodes of illness, including all related hospitalizations, with regard to quality of care. For this purpose, future electronic medical record systems that support the tracing of an entire episode-of-care for each patient and have the capacity for real-time analysis of POA-adjusted PSI events would enhance such patient safety research and practice.

As with any secondary analysis, incomplete reporting and coding errors may exist in the administrative data analyzed in this study, even though data were cleaned, audited, and standardized by COSHPD and HCUP. The PSI and POA data are subject to measurement error, and differences in the coefficients between the POA-adjusted and -unadjusted models may have occurred by chance. We did not use unit-level nurse staffing data. We also did not have nurse skill-mix or education-level data, or surgeon-level data. There could be organizational or market characteristics that were not observed in this study but are important for the safety of older surgical patients. We adjusted for patient case mix using chronic conditions, DRGs, and other patient characteristics to the degree possible using this administrative data, similar to other authors (Aiken et al., 2002; Cho et al., 2003; Kovner et al., 2002). Nevertheless, it is possible that there were unmeasured confounders. Therefore, as in most observational studies, our observed relationships between patient and hospital factors and adverse events are not proof of a causal link. This limitation should be considered when interpreting the findings. Finally, the findings may not be generalized to patients with conditions other than those included in this study or patients hospitalized in other states.

Based on our findings, reported adverse event rates in hospitalized older adults should decrease now that the reporting of POA conditions for Medicare beneficiaries has been mandated, although this has not been sufficiently examined yet. More studies are needed examining hospital quality using administrative data with adjustment for POA conditions. The impact of the new CMS policy on the prevalence of adverse events should be also evaluated. To increase the usefulness of secondary administrative data in improving geriatric safety, more research should also be conducted on the accuracy of reporting of POA conditions and the reliability of algorithms in identifying POA conditions from administrative data.

Funding

This study was supported by the Research Challenge Fund at New York University.

The authors thank Dr. Robert Norman, director of the Biostatistics Department at New York University College of Dentistry, for his support in data management and statistical analysis.

References

Aiken
LH
Clarke
SP
Sloane
DM
Sochalski
J
Silber
JH
Hospital nurse staffing and patient mortality, nurse burnout, and job dissatisfaction
Journal of American Medical Association
 , 
2002
, vol. 
288
 (pg. 
1987
-
1993
)
The Agency for Healthcare Research and Quality
Chronic care indicators for ICD-9-CM
 , 
2008
 
The Agency for Healthcare Research and Quality
Patient safety indicators: technical specifications
 , 
2008
 
The Agency for Healthcare Research and Quality
Introduction to the HCUP State Inpatient Databases (SID)
 , 
2009
 
Alexander
KP
Anstrom
KJ
Muhlbaier
LH
Grosswald
RD
Smith
PK
Jones
RH
, et al.  . 
Outcomes of cardiac surgery in patients age ≥80 years: Results from the National Cardiovascular Network
Journal of the American College of Cardiology
 , 
2000
, vol. 
35
 (pg. 
731
-
738
)
Bahl
V
Thompson
MA
Kau
TY
Hu
HM
Campbell
DA
Jr.
Do the AHRQ patient safety indicators flag conditions that are present at the time of hospital admission?
Medical Care
 , 
2008
, vol. 
46
 (pg. 
516
-
522
)
California Office of Statewide Health Planning and Development
Hospital annual financial data: selected data file documentation
 , 
2004
 
Cho
S-H
Ketefian
S
Barkauskas
VH
Smith
DG
The effects of nurse staffing on adverse events, morbidity, mortality, and medical costs
Nursing Research
 , 
2003
, vol. 
52
 (pg. 
71
-
79
)
Coffey
R
Milenkovic
M
Andrews
RM
The case for the present-on-admission (POA) indicator (HCUP methods series report No. 2006-01)
 , 
2006
 
Retrieved December 22, 2009, from http://www.hcup-us.ahrq.gov/reports/2006_1.pdf
DeFrances
CJ
Hall
MJ
2005 National Hospital Discharge Survey. Advance data from vital and health statistics (No. 385)
 , 
2007
Hyattsville, MD
National Center for Health Statistics
DoBias
M
CMS makes ‘never events’ nonpayment rules official
 , 
2009
 
Hamel
MB
Henderson
WG
Khuri
SF
Daley
J
Surgical outcomes for patients aged 80 and older: morbidity and mortality from major noncardiac surgery
Journal of American Geriatric Society
 , 
2005
, vol. 
53
 (pg. 
424
-
429
)
Horwitz
LI
Cuny
JF
Cerese
J
Krumholz
HM
Failure to rescue: Validation of an algorithm using administrative data
Medical Care
 , 
2007
, vol. 
45
 (pg. 
283
-
297
)
Houchens
RL
Elixhauser
A
Romano
PS
How often are potential patient safety events present on admission?
The Joint Commission Journal on Quality and Patient Safety
 , 
2008
, vol. 
34
 (pg. 
154
-
163
)
Hurst
NW
Ratcliffe
K
Development and application of a Structured Audit Technique for the Assessment of Safety Management Systems (STATAS)
 , 
1994
Rugby
Institute of Chemical Engineers
Jiang
HJ
Stocks
C
Wong
CJ
Disparities between two common data sources on hospital nurse staffing
Journal of Nursing Scholarship
 , 
2006
, vol. 
38
 (pg. 
187
-
193
)
Johnson
RG
Wittgen
CM
Hutter
M
Henderson
WG
Mosca
C
Khuri
SF
Comparison of risk-adjusted 30-day postoperative mortality and morbidity in Department of Veterans Affairs hospitals and selected university medical centers: vascular surgical operations in women
Journal of American College of Surgeons
 , 
2007
, vol. 
204
 (pg. 
1137
-
1146
)
Kaye
KS
Anderson
DJ
Sloane
R
Chen
LF
Choi
Y
Link
K
, et al.  . 
The effect of surgical site infection on older operative patients
Journal of American Geriatrics Society
 , 
2009
, vol. 
57
 (pg. 
46
-
54
)
Khaliq
AA
Broyles
RW
Hospital admissions: Who is admitted through the emergency department?
Health Services Management Research
 , 
2006
, vol. 
19
 (pg. 
13
-
22
)
Khuri
SF
Henderson
WG
Daley
J
Jonasson
O
Jones
RS
Campbell
DA
Jr.
, et al.  . 
The patient safety in surgery study: Background, study design, and patient populations
Journal of the American College of Surgeons
 , 
2007
, vol. 
204
 (pg. 
1089
-
1102
)
Khuri
SF
Henderson
WG
DePalma
RG
Mosca
C
Healey
NA
Kumbhani
DJ
Determinants of long-term survival after major surgery and the adverse effect of postoperative complications
Annals of Surgery
 , 
2005
, vol. 
242
 (pg. 
326
-
341
)
Kohn
LT
Corrigan
JM
Donaldson
MS
To err is human: Building a safer health system
 , 
2000
Washington, DC
National Academy Press
Kovner
C
Gergen
PH
Nurse staffing levels and adverse events following surgery in U.S. hospitals
Journal of Nursing Scholarship
 , 
1998
, vol. 
30
 (pg. 
315
-
321
)
Kovner
C
Jones
C
Zhan
C
Gergen
PJ
Basu
J
Nurse staffing and postsurgical adverse events: An analysis of administrative data from a sample of U.S. hospitals, 1990–1996
Health Services Research
 , 
2002
, vol. 
37
 (pg. 
611
-
629
)
Lieberman
JR
Romano
PS
Mahendra
G
Keyzer
J
Chilcott
M
The treatment of hip fractures: variations in care
Clinical Orthopaedics snd Related Research
 , 
2006
, vol. 
442
 (pg. 
239
-
244
)
Mark
BA
Harless
DW
McCue
M
Xu
Y
A longitudinal examination of hospital registered nurse staffing and quality of care
Health Services Research
 , 
2004
, vol. 
39
 (pg. 
279
-
300
)
Miller
MR
Elixhauser
A
Zhan
C
Meyer
GS
Patient safety indicators: Using administrative data to identify potential patient safety concerns
Health Services Research
 , 
2001
, vol. 
36
 (pg. 
11
-
132
)
Naessens
JM
Campbell
CR
Berg
B
Williams
AR
Culbertson
R
Impact of diagnosis-timing indicators on measures of safety, comorbidity, and case mix groupings from administrative data sources
Medical Care
 , 
2007
, vol. 
45
 (pg. 
781
-
788
)
Needleman
J
Buerhaus
PI
Mattke
S
Stewart
M
Zelevinsky
K
Nurse staffing levels and the quality of care in hospitals
New England Journal of Medicine
 , 
2002
, vol. 
346
 (pg. 
1715
-
1722
)
Risk factors for post-operative mortality in bariatric surgery
The Journal of Surgical Research
 , 
2005
, vol. 
127
 (pg. 
1
-
17
)
Pronovost
P
Needham
D
Berenholtz
S
Sinopoli
D
Chu
H
Cosgrove
S
, et al.  . 
An intervention to decrease catheter-related bloodstream infections in the ICU
New England Journal of Medicine
 , 
2006
, vol. 
355
 (pg. 
2725
-
2732
)
Reason
JT
Vincent
CA
Understanding adverse events: Human factors
Clinical risk management
 , 
1995
London
BMJ Publications
(pg. 
391
-
410
)
Rivard
PE
Luther
SL
Christiansen
CL
Shibei
Z
Loveland
S
Elixhauser
A
, et al.  . 
Using patient safety indicators to estimate the impact of potential adverse events on outcomes
Medical Care Research and Review
 , 
2008
, vol. 
65
 (pg. 
67
-
87
)
SooHoo
NF
Lieberman
JR
Ko
CY
Zingmond
DS
Factors predicting complication rates following total knee replacement
The Journal of Bone and Join Surgery
 , 
2006
, vol. 
88A
 (pg. 
480
-
485
)
Thomas
EJ
Brennan
TA
Incidence and types of preventable adverse events elderly patients: Population based review of medical records
British Medical Journal
 , 
2000
, vol. 
320
 (pg. 
741
-
742
)
Thornlow
DK
Stukenborg
GJ
The association between hospital characteristics and rates of preventable complications and adverse events
Medical Care
 , 
2006
, vol. 
44
 (pg. 
265
-
269
)
Turrentine
FE
Wang
H
Simpson
VB
Jones
RS
Surgical risk factors, morbidity, and mortality in elderly patients
Journal of American College of Surgeons
 , 
2006
, vol. 
203
 (pg. 
865
-
877
)
U.S. Census Bureau
U.S. interim projections by age, sex, race, and Hispanic origin: 2000–2050
 , 
2004
 
Vincent
C
Taylor-Adams
S
Stanhope
N
Framework for analyzing risk and safety in clinical medicine
British Medical Journal
 , 
1998
, vol. 
316
 (pg. 
1154
-
1157
)
Vogeli
C
Shields
AE
Lee
TA
Gibson
TB
Marder
WD
Weiss
KB
Multiple chronic conditions: Prevalence, health consequences, and implications for quality, care management, and costs
Journal of General Internal Medicine
 , 
2007
, vol. 
22
 
Suppl. 3
(pg. 
391
-
395
)
West
AN
Weeks
WB
Bagian
JP
Rare adverse medical events in VA inpatient care: Reliability limits to using patient safety indicators as performance measures
Health Services Research
 , 
2007
, vol. 
43
 (pg. 
249
-
266
)
Wolff
JL
Starfield
B
Anderson
G
Prevalence, expenditures, and complications of multiple chronic conditions in the elderly
Archives of Internal Medicine
 , 
2002
, vol. 
162
 (pg. 
2269
-
2276
)
Yuan
Z
Cooper
G
Einstadter
D
Randall
C
Rimm
AA
The association between hospital type and mortality and length of stay: A study of 16.9 million hospitalized Medicare beneficiaries
Medical Care
 , 
2000
, vol. 
38
 (pg. 
231
-
245
)
Zhan
C
Miller
MR
Excess length of stay, charges, and mortality attributable to medical injuries during hospitalization
Journal of the American Medical Association
 , 
2003
, vol. 
290
 (pg. 
1868
-
1874
)

Author notes

Decision Editor: William J. McAuley, PhD