International Journal for Quality in Health Care Advance Access originally published online on March 23, 2006
International Journal for Quality in Health Care 2006 18(3):211-219; doi:10.1093/intqhc/mzl003
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Adverse outcomes in Belgian acute hospitals: retrospective analysis of the national hospital discharge dataset
1 Center for Health Services and Nursing Research and 2 Biostatistical Center, Catholic University Leuven, Leuven, Belgium
Objective. The prevalence and variability of adverse outcome rates in Belgian acute hospitals is examined by using the national hospital discharge database.
Design, setting, and participants. Retrospective analysis based on administrative data of all Belgian acute hospitals, covering the full medical (n = 1 024 743) and surgical (n = 633 027) in-patients population for the year 2000.
Main outcome measures. For 11 adverse outcomes and failure-to-rescue, the rates and variability among hospitals were studied. The all patient refined diagnostic-related groups (APR-DRG) method was used for risk adjustment.
Results. The prevalence of adverse outcomes was 7.12% in the medical and 6.32% in the surgical group. Rates ranged from 6.25 (deep venous thrombosis) to 32.3 (urinary tract infection) outcomes per 1000 discharges in the medical group and from 3.39 (deep venous thrombosis) to 17.6 (urinary tract infection) outcomes per 1000 discharges in the surgical group. The failure-to-rescue rate was 240 and 211 per 1000 discharges, respectively. Except for pressure ulcers and hospital-acquired sepsis, the prevalence of adverse outcomes was significantly higher (P = 0.001) in the medical group. All adverse outcome rates varied substantially among the hospitals surveyed.
Conclusions. This study identifies the occurrence of adverse outcomes in a national population. It adds information to the growing body of knowledge in predominantly Anglo-Saxon countries about adverse outcomes. Striking variation exists in the risk-adjusted adverse outcome rates across Belgian acute hospitals, revealing a large potential for quality gains that encourage further action.
Keywords: administrative data, patient safety, quality measurement, risk adjustment
Address reprint requests to Prof. Dr. Arthur Vleugels, Center for Health Services and Nursing Research, Catholic University Leuven, Kapucijenvoer 3514, B-3000 Leuven, Belgium. Tel. ++32-16-33.69.71; Fax. ++32-16-33.69.70. E-mail: Arthur.vleugels{at}med.kuleuven.be
Accepted for publication February 15, 2006.
The Institute of Medicine report To Err is Human [1] provides compelling evidence that medical errors pose daily risks throughout the health care system in the United States. Comparable studies have been published for other countries: Australia [2], Canada [3], New Zealand [4], and the United Kingdom [5]. Apart from one Danish study [6], similar data for non-Anglo-Saxon countries are sparse.
Thorough review of medical records is considered to be the gold standard for monitoring adverse events [5,7]. Analyzing medical records, however, has proven to be an infeasible method for broad-based monitoring because of associated costs, access restrictions, privacy, and patient rights legislation that limit access to patient records, and because a meaningful analysis of medical records requires that the researcher has specific clinical contextual knowledge and understanding of the records being analysed [7,8]. Incident reporting and surveillance systems vary in their effectiveness. Results stemming from these systems are rarely accessible for research purposes and lead to substantial underestimations of adverse events [9]. Administrative databases, on the other hand, present a valuable alternative. They often cover large populations over consecutive years, are electronically accessible, and inexpensive to analyse [8,10]. The opportunity to analyse hospital discharge databases has been exploited by various authors [11,12]. This method has proven to be useful in assessing the frequency and distribution of adverse events [11,12], although results have to be interpreted with caution [13,14]. The objective of the present study is to describe the prevalence and variability of adverse outcome rates in Belgian acute hospitals through an analysis of the national hospital discharge database. It is questioned if the patient safety alerts, found in predominantly Anglo-Saxon countries, can be confirmed when studying a national in-hospital stay population of a health care system with a different culture, organization, and financing mechanism.
 |
Methods |
|---|
 Top Methods ResultsDiscussionConclusionAppendixAcknowledgementsReferences |
|---|
Data source
In Belgium, the collection of hospital discharge data has been compulsory since 1990 for all in-patients in all acute hospitals. The Belgian Hospital Discharge Dataset was commissioned by the Belgian Ministry of Public Health via the Royal Decree of 6 December 1994. The same decree directed the appointment of a commission to control the content and format of patient registration, the data collecting procedures, and the completeness, validity, and reliability of the collected data. The quality of the data is audited by the Ministry of Public Health in two ways. Firstly, a software program checks the data for missing, illogical, and outlier values. Secondly, by regular hospital visits, a random selection of patient records is reviewed to ensure that data were recoded correctly [15]. The Belgian Hospital Discharge Dataset contains patient demographics, data about the hospital stay (date and type of admission and discharge, referral data, admitting department, and destination after discharge), and clinical data (primary and secondary diagnoses as described in the ICD-9-CM, diagnostic and therapeutic procedures as described in the ICD-9-CM) [16].
Measurements of adverse outcomes
In this study, we adopted the definition of adverse events described by the Institute of Medicine as ‘injuries caused by medical management rather than by underlying disease or condition of the patient’ [1]. On the basis of the work by Needleman et al. [17], we identified 11 adverse outcomes that can be coded, together with in-hospital deaths, on the basis of the Belgian Hospital Discharge Dataset. Eight of these outcomes apply to both medical and surgical discharges (urinary tract infections, pressure ulcers, hospital-acquired pneumonia, shock/cardiac arrest, upper-gastrointestinal bleeding, hospital-acquired sepsis, deep venous thrombosis, and central nervous system complications) and three apply to only surgical discharges (surgical wound infection, pulmonary failure, and metabolic derangement). For the medical and surgical group, we also examined failure-to-rescue, a concept that refers to clinical conditions in which early identification and medical and nursing interventions influence the risk of death [17]. Failure-to-rescue is operationally defined as patient deaths resulting from hospital-acquired pneumonia, shock/cardiac arrest, upper-gastrointestinal bleeding, hospital-acquired sepsis, or deep venous thrombosis. Two types of validated coding rules are used (see Appendix): ICD-codes for secondary diagnoses that point towards the (adverse) outcome(s) and ICD-codes used to identify patients to be excluded from the pool of patients to be analysed [17–19]. The latter is necessary to exclude patients for whom a given diagnosis is, with great probability, not the expression of an adverse outcome but part of the patients’ original clinical condition. We used this approach to substitute for the lack of information about the actual time during the patients’ hospital stay at which the secondary diagnosis or adverse outcome occurred. This method also excludes very high risk patients in whom the adverse outcome is probably a consequence of their illness rather than of insufficient care [18].
Study population
We obtained data on all Belgian acute in-patient hospitalizations from all 123 Belgian acute hospitals for the year 2000. Patients admitted to 1-day clinics (923 098), psychiatric wards, and extended-care facilities (77 175) were excluded. We also excluded patients (20 897) from the APR-DRG (version 15) 950–956 (950, extensive procedure unrelated to principal diagnosis; 951, prostatic procedure unrelated to principal diagnosis; 952, non-extensive procedure unrelated to principal diagnosis; 955, principal diagnosis invalid as discharge diagnosis; and 956, ungroupable) that are not assigned to a Major Diagnostic Category (MDC) [20]. The final sample included in-hospital stay data from 1 657 770 patients. Using the APR-DRG, we sorted the sample into two groups: medical (n = 1 024 743) and surgical (n = 633 027) discharges.
Risk adjustment
Adverse outcomes result from a complex mix of factors. Controlling for patient risk enables researchers to adjust for a recognizable source of outcome variation and to identify more accurately actual quality differences [8,21,22]. Numerous risk-adjustment systems have been used, but none are clearly superior; moreover, different systems may give different results [21,23,24].
In the present study, we used two approaches to deal with differences in patient risks. Firstly, by applying strict exclusion rules to each adverse outcome, we increased the homogeneity of the pool of patients assessed for the various adverse outcomes, as described by Needleman et al. [18]. Secondly, we used the APR-DRG to adjust for differences in case mix and severity of illness or risk of mortality [20]. The risk adjustment for the 11 adverse outcomes was made by using the severity of illness (SOI) groups according to the APR-DRG. For risk adjustments related to failure-to-rescue rates, we used the risk of mortality (ROM) groups according to the APR-DRG.
Statistical analysis
For the overall descriptive statistics, we used the number of adverse outcomes per 1000 discharges for the population at risk to represent the crude adverse outcome rate.
We used indirect standardization [25] with APR-DRG and either SOI or ROM to calculate the expected adverse outcome rates. The hierarchical Bayesian approach [26] was used to calculate the variability of adverse outcomes among hospitals. A two-stage hierarchical model was used where the first level corresponds to the sampling distribution of the observed data (Oi) and the second level corresponds to the (unobserved) parameters (ëi). The gamma–Poisson model and the lognormal-Poisson models were fitted. Both models were compared using the Deviance Information Criterion (DIC). The DIC allows choosing between two non-nested statistical models (when the second model is not a generalization of the first model), because a lower DIC value indicates a better fit of the model to the data. The gamma–Poisson model was leading to the smallest DIC values and thus adopted in this study. This model assumes that the true risk ratio 
i is drawn from a gamma distribution with parameters 
and ß and that the observed counts Oi follows a Poisson distribution with mean iEi. That is, i 
gamma (,ß) and Oi Poisson (iEi). The hyper-parameters and ß are assumed to follow a gamma distribution, that is, gamma(10–3, 10–3) and gamma(10–3, 10–3) which are non-informative priors. The rate per 1000 discharges is calculated as ratei = 1000 x i x Ei/ni (Ei represents the expected number of adverse outcomes and ni represents the number of individuals at risk). The variability among hospitals is expressed by a P90/P10 ratio (the 90th centile divided by the 10th centile) with corresponding 95% credibility intervals. The 90th centile (P90) and the 10th centile (P10) values are sampled from distributions of i. The ‘turnip’ plots illustrate the variability of the risk-adjusted adverse outcome rates among Belgian hospitals in a graphical way [27]. Finally, the clinical importance of variations among hospitals is quantified by calculating the improvement that would occur if the overall mean could be shifted to the 20th centile. This measure represents a realistic estimate of the ‘best practice’ rate that could be achieved [28]. The number of outcomes that could be prevented if the overall rate were reduced to the 20th centile, that is the centile gains, can then be calculated as: centile gains = (total patients in the risk pool in all hospitals) x (rateoverall – rate20th centile) [28].
The analysis in this article was done using SAS software version 9.1 (SAS Institute, Cary, NC, USA), R, and WinBUGS 1.4.1.
|
Results |
|---|
|
TopMethodsResultsDiscussionConclusionAppendixAcknowledgementsReferences |
|---|
The mean age of patients in this study was 49.85 years [standard deviation (SD) = 25.43]; 45.25% of patients were male. The average length of stay was 7.6 days for the medical (SD = 10.7) as well as the surgical patients (SD = 12.4). The crude in-hospital death rate for medical and surgical discharges was 38.4 and 14.3 per 1000 discharges, respectively. The mean number of discharges per hospital was 13 477 (range: 3351–54 701).
Crude rates of adverse outcomes in Belgian acute hospitals
We analysed medical in-patient stays for 8 adverse outcomes and surgical in-patient stays for 11 adverse outcomes. We identified at least one adverse outcome in 72 958 (7.12%) medical hospital stays and in 40 027 (6.32%) surgical hospital stays. Table 1 presents the crude adverse outcome rates before APR-DRG and SOI/ROM adjustment. The adverse outcome rates ranged from 6.25 (deep venous thrombosis) to 32.3 (urinary tract infection) per 1000 discharges in the medical group and from 3.39 (deep venous thrombosis) to 17.6 (urinary tract infection) per 1000 discharges in the surgical group. The crude failure-to-rescue rate for medical and surgical discharges was 240 and 211 per 1000 discharges, respectively. All adverse outcome rates, except pressure ulcers and hospital-acquired sepsis, were significantly higher (P = 0.001) among the medical discharges than among the surgical discharges.
|
Variability of risk-adjusted adverse outcome rates among Belgian acute hospitals
Figures 1–
3 show the variability of risk-adjusted adverse outcome rates among the 123 Belgian acute hospitals. These graphs illustrate that, even after risk adjustment, all adverse outcome rates varied substantially among acute hospitals. In Table 2, it is illustrated that the P90/P10 ratio tended to be larger and varied more in the surgical group than in the medical group. For surgical discharges, P90/P10 ranged from 2.4 [2.0;2.8] (shock or cardiac arrest) to 7.9 [4.9;11.0] (upper-gastrointestinal bleeding). For medical discharges, P90/P10 ranged from 1.74 [1.5;1.8] (deep venous thrombosis) to 5.71 [4.2;7.3] (central nervous system complications). The variability of failure-to-rescue across Belgian hospitals was small (P90/P10 = 1.67 [1.5;1.8]) in the medical group as well as in the surgical group (P90/P10 = 1.34 [1.2;1.5]).
|
|
|
|
The potential gain of shifting the mean adverse outcome rates to the 20th centile are substantially higher for all adverse outcome indicators (Table 2). The three adverse outcomes with the largest potential gains in the medical group are urinary tract infections (n = 15 539), central nervous system complications (n = 5945), and hospital-acquired sepsis (n = 4226). In the surgical group, the three adverse outcomes with the largest potential gains are pulmonary failure (n = 5983), urinary tract infections (n = 5938), and hospital acquired pneumonia (n = 5026). The failure-to-rescue measure indicates the deaths of patients with complications. The potential gain of a shift of the overall mean to the 20th centile is 1555 in the medical group and 588 in the surgical group, respectively.
|
Discussion |
|---|
|
TopMethodsResultsDiscussionConclusionAppendixAcknowledgementsReferences |
|---|
Because the national Belgian Hospital Discharge Dataset included comprehensive data of all patients discharged from all Belgian acute hospitals in 2000, by definition, it is representative of care performed at Belgian acute hospitals. Our analysis revealed substantial adverse outcome rates. At least, one adverse outcome occurred in 7.12% of medical patients and 6.32% of surgical patients. These figures fall within the 3.7–16.6% range reported by other authors [2–5,29]. However, vigilance is required when comparing these results. The adverse outcomes reported in this study, namely, do not consider the criteria of prolonged hospitalization or disability at time of discharge that were used by other authors [2–5,29]. What’s more, an administrative database was used as opposed to medical record reviews [2–5,29]. As a consequence, the measures used in this study should be considered as general screens allowing to identify potential quality problems, which can be further investigated and studied.
The three most frequently reported adverse outcomes were the same in both medical and surgical groups: urinary tract infections, hospital-acquired pneumonia, and hospital-acquired sepsis. The least reported adverse outcome, deep venous thrombosis, was the same for both groups. With the exception of the frequencies of pressure ulcers and hospital-acquired sepsis, adverse outcome rates were generally higher in medical than in surgical patients. As suggested by Needleman et al. [18], also in our database, the surgical patients may have been, in general, healthier than the medical patients and therefore had a lower risk of developing adverse outcomes. Other studies, using medical record review, have found that there were more adverse events in the surgical patients: in particular, the Australian study reported that the surgical rate was 21.9% and the medical rate was 13.3% [30]. Fifty-two per cent of surgical adverse events were associated with technical errors, where the procedure was appropriate and indicated. These are not covered using in-patient data.
The mean length of stay in the Belgian hospitals is 1.5 times higher in the medical group and 1.6 times higher in the surgical group than the study sample of Needleman et al. [17]. As a result, one could expect the Belgian adverse outcome rates to be higher. However, our observations were similar to those reported by Needleman et al. [17] for hospitals in the United States, with the exception of frequency of pressure ulcers, urinary tract infections, and metabolic derangement, which were two to eight times lower, and of failure-to-rescue rates, which were 7–29% higher. We cannot exclude the possibility that the first three outcomes were under-reported in the Belgian Hospital Discharge Dataset. One possible explanation for this apparent discrepancy is that, in Belgium, pressure ulcers, for example, are primarily considered to be a nursing rather than a medical problem, and thus would not be recorded in medical data. Another possible explanation is that the medical and surgical groups, in this study, are assigned based on the APR-DRG system. This implies that the study sample also includes the risk pools that were not studied by Needleman et al. [18] (minor surgery, endoscopy procedures, and interventional cardiology). Furthermore, Needleman et al. [18] restricted the surgical risk pool to patients who undergo the procedure on the first or second day after admission, whereas we included all surgical patients.
Adverse outcome rates vary widely across Belgian acute hospitals, as they do across hospitals in other countries [3,31]. The large number of calculated potential quality gains, when shifting the mean to the 20th centile, illustrates the potential impact on public health. The variability in adverse outcome rates, however, does not necessarily reflect similar variation in the quality of care across Belgian hospitals. Firstly, we derived the adverse outcome rates from an administrative database. Differences in quality of registration among hospitals could have contributed to the variability in the adverse outcome rates we observed. Concerns with coding accuracy and coding variation [32], with the limited insight into the timing of the outcomes and with the financial incentives of coding [33], are common. The Belgian Ministry of Public Health reports that 37.7% of the audited patient records (from 1995 to 1998) contain secondary diagnoses that are not registered in the Belgian Hospital Discharge Dataset [34]. On the other hand, the audit reports show that the coding quality improves every year [34]. It is assumed that this tendency continues into the year 2000 data. Furthermore, the Belgian Hospital Discharge Dataset is generally accepted as the basis for the Belgian hospital financing system and can thus be considered to be stable after 10 years use.
Secondly, many different patient attributes affect risks, including age, sex, acute physiological stability, principal diagnosis and its severity, the extent and complexity of comorbid illnesses, functional status, psychosocial and cultural factors, socio-economic characteristics, and preferences for specific outcomes [21]. We acknowledge that the APR-DRG and SOI/ROM adjustment cannot completely control for all levels of risk leading to the development of adverse outcomes, nor are the studies available, or likely, to tell us which risk adjustor is best. The APR-DRG system, as well as SOI and ROM classes, has been evaluated as a tool for risk adjustment [21–23]. Its value, like that of other risk-adjustment measurements, depends on the disease group under study.
Thirdly, we used a conservative approach to overcome some of the shortcomings associated with using administrative data to measure adverse outcomes. By using adverse-outcome-specific exclusion rules, we increased the homogeneity of the groups analysed. Applying this method increased the likelihood that true adverse outcomes were identified from the discharge abstracts.
|
Conclusion |
|---|
|
TopMethodsResultsDiscussionConclusionAppendixAcknowledgementsReferences |
|---|
Adverse outcomes do not, per se, necessarily signal poor-quality care, nor does their absence necessarily indicate good-quality care [29]. Our results warrant serious consideration, because they underscore the potential benefit of using a convenient tool to identify adverse outcomes in whole populations. Regardless of the results from any future validation work on the use of administrative data, we highly recommend further investigating this approach to understand quality problems in Belgian acute hospitals. It is not the aim of this study to detect adverse outcomes at the individual patient level. Rather, our primary aim was to screen the rates of adverse outcomes at the hospital level to examine potential systematic quality problems in greater depth. Our findings reveal the need to improve patient outcomes at some institutions and in given areas of care; they reveal a substantial potential for quality gains. Hopefully, these findings will lead to improvement of the quality of patient care.
One unanswered question in this study that needs to be addressed in future research is whether the quality of care is consistent across clinical disciplines or services within hospitals. More flexible exclusion rules should be applied to adequately address this question [35]. Another point for future research involves the improvement of risk-adjustment methods. Some have suggested that, for patient safety monitoring purposes, APR-DRGs should be calculated without taking into account secondary complications that emerge during the hospital stay [22,23]. Thus, we recommend that future registration guidelines require these cases to be flagged as they are registered in the hospital discharge dataset. This approach will decrease the number of comorbidities existing before hospitalization that are wrongly denoted as adverse outcomes in subsequent analyses and will thus increase the specificity of the used method.
Despite the different study approach and the different culture, hospital financing and organization system of the generally studied Anglo-Saxon countries [1–5], also in this study the applied quality screens, illustrate a substantial amount of potential quality problems and inter-hospital variation that warrant further action.
|
Appendix |
|---|
|
TopMethodsResultsDiscussionConclusionAppendixAcknowledgementsReferences |
|---|
|
|
Acknowledgements |
|---|
|
TopMethodsResultsDiscussionConclusionAppendixAcknowledgementsReferences |
|---|
We gratefully acknowledge the SAS programming support of Martine Beullens, Catholic University Leuven. This study was performed in preparation for a lecture on ‘nurse staffing and patient outcomes’ on the conference organized by the Ministry of Public Health in collaboration with the US Embassy in Belgium on 19 May 2004. We would also like to thank the Ministry of Public Health for providing access to the anonymized Belgian Hospital Discharge Data Set.
|
References |
|---|
|
TopMethodsResultsDiscussionConclusionAppendixAcknowledgementsReferences |
|---|
- Kohn LT, Corrigan JM, Donaldson M et al. To Err Is Human: Building a Safer Health System. Washington DC: National Academy Press, 1999.
- Wilson RM, Runciman WB, Gibberd RW et al. The quality in Australian health care study. Med J Aust 1995; 163: 458–471.[ISI][Medline]
- Baker GR, Norton PG, Flintoft V et al. The Canadian Adverse Outcomes Study: the incidence of adverse outcomes among hospital patients in Canada. CMAJ 2004; 170: 1678–1686.
[Abstract/Free Full Text] - Davis P, Lay-Yee R, Briant R et al. Adverse outcomes in New Zealand public hospitals I: occurrence and impact. N Z Med J 2002; 115: U271.
- Vincent C, Neale G, Woloshynowych M. Adverse outcomes in British hospitals: preliminary retrospective record review. Br Med J 2001; 322: 517–519.
[Abstract/Free Full Text] - Schiøler T, Lipczak H, Pedersen BL et al. Incidence of adverse outcomes in hospitalized patients: the Danish Adverse Outcome Study (DAES). Ugeskr Læger 2001; 163: 5370–5378.
- Michel P, Quenon JL, de Sarasqueta AM et al. Comparison of three methods for estimating rates of adverse outcomes and rates of proutcomeable adverse outcomes in acute care hospitals. Br Med J 2004; 328: 199.
- Zhan C, Miller MR. Excess length of stay, charges, and mortality attributable to medical injuries during hospitalization. JAMA 2003; 290: 1868–1874.
[Abstract/Free Full Text] - Cullen DJ, Bates DW, Small SD et al. The incident reporting system does not detect adverse drug outcomes: a problem for quality improvement. Jt Comm J Qual Improv 1995; 21: 541–548.[Medline]
- Needleman J, Buerhaus PI, Mattke S et al. Measuring hospital quality: can medicare data substitute for all-payer data? Health Serv Res 2003; 38: 1487–1508.[CrossRef][ISI][Medline]
- Iezzoni LI, Daley J, Heeren T et al. Identifying complications of care using administrative data. Med Care 1994; 32: 700–715.[ISI][Medline]
- Romano PS, Geppert JJ, Davies S et al. A national profile of patient safety in U.S. hospitals. Health Aff 2003; 22: 154–166.
[Abstract/Free Full Text] - Weingart SN, Iezzoni LI, Davis RB et al. Use of administrative data to find substandard care: validation of the complications screening program. Med Care 2000; 38 (8): 796–806.[CrossRef][ISI][Medline]
- Naessens JM, Huschka TR. Distinguishing hospital complications of care from pre-existing conditions. Int J Qual Health Care 2004; 16 (suppl. 1): i27–i35.[CrossRef]
- Vander Meersch E. M.K.G. audits. Nieuws over ziekenhuisregistratiesystemen, 2002.
- France RFH, Mertens I. Case mix in Belgium: a rapid coverage and a progressive AP-DRG use. In Case Mix: Global Views, Local Actions: Evolution in Twenty Countries. Amsterdam: IOS Press, 2001.
- Needleman J, Buerhaus P, Mattke S et al. Nurse-staffing levels and the quality of care in hospitals. N Engl J Med 2002; 346: 1715–1722.
[Abstract/Free Full Text] - Needleman J, Buerhaus P, Mattke S et al. Nurse Staffing and Patient Outcomes in Hospitals. 230-99-0021. Boston: Harvard School of Public Health, 2001.
- Iezzoni LI. Complications Screening Program for the Health Care Financing Administration (CSP-HCFA): CSP Updated to FY 1993 Coding Guidelines. Boston: Beth Israel Hospital, 1994.
- Averill RF, Goldfield N, Steinbeck B et al. Development of the All Patient Refined DRGs (APR-DRGs). 3 M HIS Research Report. Wallingford, Connecticut: 3M Health Information Systems, 1997.
- Iezzoni LI. The risks of risk adjustment. JAMA 1997; 278: 1600–1607.[Abstract]
- Romano PS, Chan BK. Risk-adjusting acute myocardial infarction mortality: are APR-DRGs the right tool? Health Serv Res 2000; 34: 1469–1489.[ISI][Medline]
- Iezzoni LI, Ash AS, Shwartz M et al. Judging hospitals by severity-adjusted mortality rates: the influence of the severity-adjustment method. Am J Public Health 1996; 86: 1379–1387.
[Abstract/Free Full Text] - Iezzoni LI, Ash AS, Shwartz M et al. Predicting in-hospital deaths from coronary artery bypass graft surgery. Do different severity measures give different predictions? Med Care 1998; 36: 28–39.[CrossRef][ISI][Medline]
- Fleiss JL, Levin B, Cho Paik M. Statistical Methods for Rates and Proportions, third edition. Hoboken, NJ: John Wiley & Sons, Inc., 2003.
- Gelman A, Carlin JB, Stern HS, Rubin, DB. Bayesian Data Analysis. London: Chapman & Hall, 1995.
- The Center for the Evaluative Clinical Sciences at Dartmouth. The Quality of Medical Care in the United States: A Report on the Medicare Program. The Dartmouth Atlas of Health Care. In Wennberg JE, Cooper MM, eds. Chicago, IL: American Hospital Publishing, 1999.
- Gibberd R, Hancock S, Howley P, Richards K. Using indicators to quantify the potential to improve the quality of health care. Int J Qual Health Care 2004; 16 (suppl. 1): i37–i43.
- Brennan TA, Leape LL, Laird NM et al. Incidence of adverse outcomes and negligence in hospitalized patients. Results of the Harvard Medical Practice Study I. N Engl J Med 1991; 324: 370–376.[Abstract]
- Kable AK, Gibberd RW, Spigelman AD. Adverse events in surgical patients in Australia. Int J Qual Health Care 2002; 14: 269–276.
[Abstract/Free Full Text] - Brennan TA, Hebert LE, Laird NM et al. Hospital characteristics associated with adverse outcomes and substandard care. JAMA 1991; 265: 3265–3269.[Abstract]
- Romano PS, Mark DH. Bias in the coding of hospital discharge data and its implications for quality assessment. Med Care 1994; 32: 81–90.[ISI][Medline]
- Simborg DW. DRG creep: a new hospital-acquired disease. N Engl J Med 1981; 304: 1602–1604.[ISI][Medline]
- Deroubaix AN, Hauzeur C, Simoens-Desmet A. Statistieken van het Ministerie van Volksgezondheid over de MKG-audits. Nieuws over ziekenhuisregistratiesystemen, 2002.
- Mattke S, Needleman J, Buerhaus P et al. Evaluating the role of patient sample definitions for quality indicators sensitive to nurse staffing patterns. Med Care 2004; 42: II21–II33.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  T J B Kline, C Willness, and W A Ghali Predicting patient complaints in hospital settings Qual. Saf. Health Care, October 1, 2008; 17(5): 346 - 350. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Minerva BMJ, July 22, 2006; 333(7560): 208 - 208. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||